Weather Conditions and Disparities in Physical Activity and Obesity in the United States

Presenter:

Daniel Eisenberg

Authors:

Daniel Eisenberg, Edward Okeke

Chair: Michael Hagan; Discussant: Chad Meyerhoefer Wed June 7, 2006 9:45-11:15 Room 121

Context: The adult obesity rate in the U.S. is now over 30 percent. Exercise has been shown to be an effective weight control device in carefully controlled randomized trials, but it is less clear whether interventions designed to encourage exercise are effective in real-world community settings. One environmental factor that may be important in real-world settings is the weather. Little is known about how sensitive exercise participation rates are to outdoor weather conditions. For many people, particularly those in lower socioeconomic status (SES) groups, perceived exercise options may become very limited during unpleasant (unusually hot, cold, or wet) weather. Information regarding the relationship between weather conditions, exercise, and obesity across SES groups would be valuable for policymakers, community planners, health professionals, and ordinary citizens.

Objectives: We analyze the effect of weather conditions on exercise participation and the effect of exercise participation on body weight.

Statistical Methods: We employ linear regression models, including models with year, state, and state*month fixed effects, to test how exercise participation responds to unusual weather conditions. We then employ a two-stage instrumental variable (IV) model to test the causal effect of exercise on body weight, with weather conditions serving as IVs for exercise.

Data: We examine nationally representative data from the Behavioral Risk Factor Surveillance System (BRFSS) during 1993-2001 in the United States (total N = approximately 1.5 million). Individual level BRFSS data include measures of body mass index (BMI), exercise activities, diet, and socioeconomic and demographic characteristics. Using interview dates in BRFSS, we link these data to recent local weather conditions, available from the National Climatic Data Center.

Results: Our analysis is ongoing. Preliminary results suggest a strong first stage relationship between weather conditions and physical activity.

Discussion: The results of this study will help inform current public health efforts to increase exercise and reduce obesity. We will improve understanding of whether interventions must account for the diverse weather conditions in this country across regions and the seasons, and whether such interventions should focus on certain SES groups.

Permalink

Cost Effectiveness of Interventions for Teen Diet and Exercise

Presenter:

Marilyn Frenn

Authors:

Marilyn Frenn, Evelyn Kuhn, Hua Liu, Ramesh Sachdeva

Chair: Michael Hagan; Discussant: Michael Hagan Wed June 7, 2006 9:45-11:15 Room 121

PURPOSE: To test the effectiveness of a school-based intervention on dietary fat intake and physical activity, determine which intervention was more effective in reducing dietary fat and increasing physical activity in middle school youth, and to assess the cost of each intervention in relation to effectiveness.

BACKGROUND: Obesity in children results in many health problems during childhood and later in life. Healthy People 2010 goals include reducing overweight among adolescents aged 12-19 through sound dietary practices and increased physical activity. African Americans, Hispanics, Native Americans and children from low-income families in any race are most at risk for obesity and associated health problems. Project FUN was developed using the Health Promotion/Transtheoretical Model to tailor computerized interventions. Tailored interventions, used in the computerized model, modify feedback based on characteristics previously demonstrated to improve nutrition and exercise behaviors. Individualized interventions, used in the Email model, modify feedback based on many characteristics evident in clients. The third model utilized peers to provide group support for healthy behaviors.

METHODS: Culturally diverse, low-middle income, 6th and 7th grade youth (N=447) were assigned by classroom to the control or one of the three intervention groups to avoid diffusion of the interventions to the control condition. The first intervention group received education from a computerized model, the second group had the computerized model plus Emailed feedback, and the third group had the computerized model plus access to peer feedback. Trained nursing students gave individual Emailed responses to students’ answers to questions posted to a discussion board. The peers were trained with three additional computer modules and a _ day session on campus to provide support for improved diet and exercise behaviors in their class. Costs of intervention were calculated for each intervention group. The outcome variables were percentage dietary fat and physical activity measured by exercise log.

RESULTS: Costs for computerized model alone, were $39.89 per student. For computerized model plus Email feedback the costs were $53.12. And finally, computerized model plus peer feedback costs were $73.40. General Linear Model demonstrated a difference among groups for reducing dietary fat (p<0.037) with Email being the most effective, followed by peer feedback. Changes in physical activity showed significant interactions with gender (p=0.042), grade (p=0.036), and with gender, grade, and race (p=0.046) again with the Email approach being most effective. Computerized model and computerized model plus peer feedback approaches were effective for some, but not all subgroups.

CONCLUSION: Although computerized interventions were least expensive, those assisted by individualized Email most effectively decreased dietary fat intake and increased physical activity behaviors among middle school youth. Further refinement of the computer-tailored tool, based on gender, race, and grade, may improve its effectiveness.

Permalink

The Effect of Physical Activity on Short Run Medical Costs and Lost Work Days among US Adults

Presenter:

Eric Keuffel

Authors:

Eric Keuffel

Chair: Michael Hagan; Discussant: Michael Hagan Wed June 7, 2006 9:45-11:15 Room 121

Author: Eric Keuffel (ekeuffel@wharton.upenn.edu), Wharton, University of Pennsylvania.

Title: The Effect of Physical Activity on Short Run Medical Costs and Lost Work Days among US Adults

Background: Regular physical activity is frequently defined as rigorous activity sustained for at least 30 minutes three times per week. While the epidemiological implications of regular physical activity are well-established, the association between activity level and economic medical costs is not as well understood in US adults. The incentives for offering effective primary prevention programs promoting physical activity are influenced by anticipated cost savings and work days saved. Given the high rate of managed care and employee turnover in the US, short run returns are particularly important. Prior studies offer varying conclusions; some find short-run returns to physical activity others find no effect.

Objective: This study estimates the effect of regular physical activity on medical costs and days work lost after accounting for demographics, medical conditions, health insurance status and other covariates. I examine whether current year physical activity, prior year physical activity and persistent two-year physical activity result in lower medical costs and work days lost.

Data: The study data are from the 2001-2002 Medical Expenditure Panel Survey (MEPS). There are 15,269 adults (> age 16) in panel six of MEPS with data recorded in both years. MEPS records physical activity with a bivariate measure (1-rigourous physical activity at least 3 times per week/30 minutes per session; 0-not physically active) for both 2001 and 2002.

Methods: Year 2002 medical costs are modeled with a two part approach using STATA 9.1. Part I uses a probit regression to estimate the probability of positive costs. Part II estimates total medical costs conditional on positive costs with a generalized linear model (GLM gamma). Marginal effects accounting for both stages were estimated. A negative binomial model estimated the percent change in work days associated with physical activity. In each case, three separate specifications test whether current year physical activity (2002), prior year physical activity (2001) or persistent physical activity (2001 and 2002) significantly effect costs or days work missed. The base case uses robust standard errors and clusters on household.

Results: Mean 2002 medical costs in the sample are $3,025 (SD=$8,207) and 18 percent of the sample have no costs. After controlling for covariates; current, prior and persistent physical activity significantly increases (p<.10) the probability having positive costs, but only by 0.8 -1.1 percent. Physical activity is associated with lower conditional costs. Overall, point estimates for the marginal effects of current and prior year physical activity were -$32 and -$203, respectively. Mean days of work missed due to illness were 4.7 (SD=15.3 n=8,661). Persistent physical activity results in 17% fewer days missed due to sickness (p<.01).

Conclusions: Returns to physical activity in terms of medical costs are relatively modest over two years and most of the returns appear to accrue as a result of activity in the first year. Missed work days are a potentially important component when calculating the benefits of physical activity.

Permalink

Are there differential effects of Medicaid and SCHIP managed care on children with chronic conditions?

Presenter:

Amy Davidoff

Authors:

Amy Davidoff, Brigette Courtot, Emerald Adams

Chair: Ciaran Phibbs; Discussant: Ciaran Phibbs Wed June 7, 2006 9:45-11:15 Room 213

Rationale: Results of recent studies examining effects of Medicaid managed care have been mixed, with various studies reporting small reductions in ER use and hospitalizations, and increases in outpatient visits. The effects of managed care may be particularly strong for children with chronic health conditions, for whom managed care organizations have strong incentives to manage care and control costs. Alternatively, baseline use by these children may be appropriate, and managed care may exert its effects by disrupting established provider relationships. The use of behavioral health or specialty carveouts to capitated plans may further interfere with care coordination.

Objectives: To examine effects of different types of mandatory managed care programs, including use of carveouts, on children with and without chronic health conditions enrolled in Medicaid or SCHIP .

Methodology: Data on child characteristics, health status, and access and use of healthcare services are from pooled National Health Interview Survey data (1997-2002.) Data on Medicaid and SCHIP managed care program types, areas served, populations covered, and use of carve outs were collected from annual CMS Medicaid Managed Care Enrollment Reports, state specific SCHIP plans, and an Urban Institute survey of managed care implementation. A county and year specific database was created on type of managed care and whether enrollment was mandatory for children generally, and for SSI recipients or other children with chronic conditions. Managed care data were linked to children eligible for Medicaid or SCHIP, with program specific eligibility determined by application of state and year specific eligibility rules. Linear probability models were estimated for all publicly insured children, with a vector of managed care program types, child health status, and interactions between them, with controls for child, family, area characteristics, state and year.

Results: For children without chronic conditions, few managed care plan effects were significant relative to FFS. Mandatory capitated or mixed PCCM/capitated programs were associated with increased likelihood of using prescription medication. In contrast, for children with chronic conditions, mandatory PCCM and mixed mandatory PCCM/capitated programs without carve outs were associated with fewer physician visits, any specialist visits, ER use, hospital stays, and prescription drug use. When carveouts were present, effect sizes were smaller, and only the negative effects on physician visits, specialist visits and prescription drug use were significant.

Conclusions: Although managed care delivery systems may affect perceived access and use of preventive and acute services for generally healthy children, the results of this analysis suggest that the effects operate primarily on children with chronic health conditions. Mandatory enrollment in managed care, both PCCM and capitated plans, was associated with reductions in use of a variety of services. It is not possible to tell whether these changes resulted in more appropriate use of services. However, we did not observe a corresponding increase in reported unmet need for medical care or prescription drugs, thus, the net change may represent an improvement in care. Additional focus on quality and appropriateness effects of managed care on children with chronic conditions is warranted.

Permalink

Effects of the State Children's Health Insurance Program (SCHIP) on Access to Dental Care and Use of Dental Services

Presenter:

Hua Wang

Authors:

Hua Wang, Edward Norton, Gary Rozier

Chair: Ciaran Phibbs; Discussant: Gabriel Picone Wed June 7, 2006 9:45-11:15 Room 213

Rationale: Lack of dental insurance is one of the main barriers to access to dental care for many low-income children in the U.S. The State Children’s Health Insurance Program (SCHIP), created by Congress in 1997, expands eligibility for public dental insurance to uninsured low-income children in almost all states. SCHIP may have extensive influence because it is also designed to facilitate Medicaid enrollment and it gives states flexibility in experimenting with new models that may overcome historical obstacles to access to dental care in public programs. Yet the extent to which SCHIP has improved children’s access to and use of dental services is largely unknown, especially at the national level.

Objective: To provide national estimates of the total implementation effects of SCHIP on dental care access and use for low-income children.

Methodology: Two separate analyses are conducted to estimate (1) the effect of program availability on dental care access and use for low-income children (regardless of eligibility or enrollment); and (2) more specifically, changes in dental care access and use for children who gained public insurance as a result of SCHIP implementation. In the first analysis, we consider SCHIP implementation a natural experiment, which has considerable variation in the timing of program implementation across states. We use the variation to identify the effect of SCHIP availability on dental care access (unmet need for dental care due to cost in the past year) and dental services use (time since last dental visit) for any low-income (<300% Federal Poverty Level) child in county and time fixed effects models. In the second analysis, we employ the instrumental variables method to “identify” children who had public insurance due to SCHIP implementation in addition to deal with endogeneity of insurance. State-level SCHIP program features are selected as instruments for public coverage, including program availability, eligibility thresholds, and waiting periods. By focusing on SCHIP’s overall effects, both analyses avoid the difficulty of imputing program eligibility or misreporting of SCHIP enrollment in the data. Both analyses estimate linear probability models adjusted for survey designs. The data source is the National Health Interview Survey 1997-2002 (N = 40,000+).

Results: SCHIP availability for more than one year reduced the likelihood of experiencing unmet dental care need for any low-income child by 2.8 percentage point; increased the probability of having a dental visit within 6 months or in the past 6-12 months by 2.2 and 0.9 percentage point, respectively. Compared with their uninsured counterparts, those who obtained public coverage from SCHIP implementation were less likely to report unmet need for dental care by 11.6 percentage point, and more likely to have visited a dentist within 6 months or in the past 6-12 months by 31 and 35.7 percentage point, respectively. School-aged children (6-17 years) fared better than younger children. Type of SCHIP program had no differential effects.

Conclusions: Consistent results from two analytical approaches provide solid evidence that SCHIP implementation has significantly reduced financial barriers to dental care and increased use of dental services for low-income children in the U.S.

Permalink

The Role of Race and Ethnicity in Children's Health Insurance Coverage

Presenter:

Thomas Selden

Authors:

Thomas Selden, Yuriy Pylypchuk,

Chair: Ciaran Phibbs; Discussant: Kosali Simon Wed June 7, 2006 9:45-11:15 Room 213

The Role of Race and Ethnicity in Children’s Health Insurance Coverage: A Decomposition Analysis Using the Medical Expenditure Panel Survey

Large racial and ethnic disparities exist with respect to the insurance of children. In 2002, 77 percent of white (non-Hispanic) children age 0-18 had private coverage during the year, versus only 41% of Hispanic children and 47% of non-Hispanic black children. For non-Hispanic black children, this gap is closed by far higher rates of public coverage: 48% versus 17% among white children. Among Hispanic children public coverage fails to close the gap, and the prevalence of full-year uninsurance at 15% is more than twice the average of all other children.

What explains these large disparities? In this paper, we pool data from the 2000-2002 Medical Expenditure Panel Survey (MEPS) to implement Oaxaca-Blinder decomposition analysis. Pooling yields a sample of nearly 30,000 observations. Factors considered include age, sex, geographic location, family composition, family poverty, parent education, parent employment, the nativity and citizenship of children and their parents, and the language used to administer the survey. Because having a parent who is eligible for employment-related insurance is a key factor in children obtaining private insurance, we also conduct a decomposition analysis of disparities in parent eligibility rates and child take-up rates.

Whereas Oaxaca-Blinder decomposition was developed for the analysis of continuous dependent variables, the dependent variables in our analysis are discrete. This raises issues regarding econometric specification. We implement two solutions. The first is simply to apply Oaxaca-Blinder methods to linear probability regressions. The second is to decompose estimates from a multinomial logit model (MNL). MNL decomposition is a recently-developed technique that requires matching of children across groups. We explore a range of strategies for matching, including a method we develop that more flexibly incorporates sample weighting schemes applicable to most household survey data.

Perhaps not surprisingly, the key determinant of racial and ethnic coverage disparities among children is poverty. Secondary, but nevertheless important, factors include citizenship, language, parent education, and Census division. Together, the factors we examine explain nearly 80 percent of the racial and ethnic coverage disparities in our sample.

Permalink

A Multi-linear Multi-Attribute Utility Function for the Health Utilities Index Mark 3 System

Presenter:

David Feeny

Authors:

David Feeny, William Furlong, George Torrance, Charles Goldsmith, Sonja DePauw, JoAnn Kingston-Riechers

Chair: Albert Okunade; Discussant: Albert Okunade Wed June 7, 2006 9:45-11:15 Room 225

Rationale and Objectives. Estimated multi-attribute utility functions have relied on linear additive or multiplicative functional forms that assume respectively a lack of preference interactions among attributes or only one type of preference interaction. Are there quantitatively important and statistically significant interactions in preferences among attributes in the Health Utilities Index Mark 3 (HUI3) system? How would the performance of the less restrictive multi-linear model compare to the performance of the multiplicative model?

Methodology. HUI3 has 8 attributes, vision, hearing, speech, ambulation, dexterity, emotion, cognition, and pain, with 5 or 6 levels per attribute. A preference survey was conducted of a random sample of the general population (n = 256) using a one-half 2 to the eight power fractional factorial design plan. The same survey provided scores for the estimation of a multiplicative multi-attribute utility function. A parallel survey (n = 248) provided directly measured standard gamble utility scores for 73 HUI3 health states. The fractional factorial design permits the identification of all 8 main effects, 26 of 28 two-way interactions, and 4 of 56 three-way interactions terms. The estimated equation was forced to pass through 0 (for the health state with all attributes at lowest functional level) and 1 (all attributes at highest level). Agreement between directly measured scores from the second sample and scores from the multi-linear and multiplicative utility functions was assessed using an intra-class correlation coefficient.

Results. For the multi-linear model, the adjusted R-squared was 0.63. All 8 main effects were quantitatively important (coefficient >0.024) and statistically significant (p < 0.10). Two-way interaction terms indicating preference complementarity were quantitatively important and statistically significant in 18 cases and insignificant in 2 cases. Two-way interaction terms indicating preference substitutes were important and significant in 4 cases and insignificant in 2 cases. All 4 three-way interaction terms were important and significant. Agreement between directly measured scores and scores from the multiplicative function was much higher than agreement between directly measured scores and scores from the multi-linear function.

Conclusions. There are quantitatively important and statistically significant interactions in preferences among attributes of health status. These results call into question the use of linear additive multi-attribute utility functions. The multiplicative function out performed the multi-linear function in out-of-sample prediction. The omnibus interaction term of the multiplicative function indicates preference complementarity and appears to handle the preference interactions more than adequately.

Permalink

Construct Validity of the Health Utilities Index Mark 3 (HUI3): Alzheimer Disease, Arthritis and Cataract

Presenter:

Keiko Asakawa

Authors:

Keiko Asakawa, David Feeny

Chair: Albert Okunade; Discussant: Albert Okunade Wed June 7, 2006 9:45-11:15 Room 225

BACKGROUND: HUI3 is a prominent measure of health-related quality of life (HRQL) widely used in population health surveys, clinical studies and cost-utility analyses. HUI3 includes 8 attributes (vision, hearing, speech, ambulation, dexterity, emotion, cognition and pain), with 5 or 6 levels for each attribute. HUI3 has been used to assess health status in a number of chronic conditions. However, many of these analyses are based on samples that may not be representative of the entire population and there is little evidence for some important chronic conditions.

AIM: To assess the population-health construct validity of the HUI3 system in people with Alzheimer disease (AD), arthritis (AR) and cataract (CA), all three of which are conditions prevalent among the elderly.

METHODS: 1996/97 Canadian National Population Health Survey community and institutions cross-sectional Microdata files were used for the analyses. Data for those aged 40 and over were analyzed. Overall and single-attribute HUI3 scores for the 5 groups were compared to assess construct validity: AD only, AR only, CA only, AACA (at least two of the three conditions), and REF (none of the three conditions). Community and institutional population were analyzed separately. A total of 18 multiple linear regressions, 9 each for community (sample size is approximately 36,000) and institutional (sample size is approximately 1,100) data, were conducted to estimate effects of AD, AR and/or CA on overall and single-attribute utility scores. In all models, analyses controlled for individual characteristics, socio-economic status, health risk factors and the number of co-morbidities. A total of 76 a priori hypotheses, 38 each for community and institutional samples, were specified; adjusted-mean differences among 5 groups were obtained from each equation for hypothesis testing. For HUI3, differences of 0.03 (0.05) or more in overall (single-attribute) utility scores are interpreted as clearly clinically important. To take account of the complex survey design, bootstrap and Taylor linearization methods were used for variance estimations for community and institutional data, respectively. Statistical analyses were conducted using SUDAAN 9.0.1 and SAS 9.0.

RESULTS: For community and institutional samples, 23 and 28 hypothesis tests were consistent with a priori hypotheses, respectively. In both community and institutional samples, overall scores were lower for those with AD, AR and/or CA than for reference group; cognition scores were lower for AD than those for other groups; pain scores for AR were lower than those for other groups. Speech scores were lower for AD than for others residing in institutions. Adjusted-mean differences were all statistically significant (p<0.05) and clinically important. Consistent with our expectations, there was no statistically significant (at 5%) and clinically important difference in mean hearing scores among the 5 groups.

CONCLUSIONS: HUI3 was able to discriminate various aspects of burden associated with AD, AR and CA in both community and institutional samples. HUI3 was able to describe differences in overall HRQL levels as well as burden associated with vision, speech, cognition and pain. HUI3 is useful in assessing HRQL of AD, AR and CA for those in the community and in institutions.

Permalink

Consistency between willingness to pay for reducing the risk of adverse drug events and the associated health-related utility gain

Presenter:

Rosa Rodriguez-Monguio

Authors:

Rosa Rodriguez-Monguio

Chair: Albert Okunade; Discussant: Albert Okunade Wed June 7, 2006 9:45-11:15 Room 225

Rationale: Adverse drug events (ADE) have an impact on health utility, and individuals are willing to pay to reduce the risk of suffering such events. Objectives: The objective of this study is to assess the consistency between the willingness to pay (WTP) for reducing the risk of adverse drug events and the associated health-related utility gain. The study quantifies the monetary value of reducing the risk of incidence of mild ADE and measure the health-related utility gain for such a reduction.

Methodology: Personal interviews were completed for 174 people in Madrid (Spain). Upon definition of a payment card, individuals were asked for the amount of money they would be willing to pay to reduce the risk of incidence of ADE. Using Standard Gamble, individuals were also asked for the health-related utility gain they would derive from a reduction in the probability (p) of occurrence of ADE. Consistency among the preference relation, the willingness to pay (WTP) to reduce the risk of ADE and the health-related utility gains was assessed. Consistency between the WTP for different risk reductions and the level of risk reduction of the same ADE was also evaluated.

Results: Gastrointestinal distress and heartburn (GI) was ranked as less preferred than skin rash (SR) by 99.4% of interviewees. Interviewees were willing to pay annually an average of 379.51_ to avoid GI (CI 325.40_ - 433.62) and 148.40 (CI 118.58_ - 178.22) to avoid SR. Interviewees were willing to pay 223.49 (CI 187.34_ - 259.64) for an 80% GI risk reduction, and 79.37 (CI 57.20_ - 101.55) for an 80% SR risk reduction. Interviewees were willing to pay 162.45 (CI 128.94_ - 195.96) for a 50% GI risk reduction and 49.25 (CI 30.47_ - 68.03_) for a 50% SR risk reduction. Interviewees considered the health-related utility gain (1-p) derived from the avoidance of GI 0.67 (CI 0.28- 0.35) higher than the health-related utility gain attributed to the avoidance of SR 0.09 (CI 0.07- 0.12). The WTP and the stated health-related utility were consistent for SR (Pearson correlation coefficient = -0.283, p-value <0.01). The Pearson correlation coefficient was not statistically significant in the case of GI.

Conclusions: The willingness to pay amount was consistent with the stated preference relation, and it was also consistent for different levels of risk reduction in the incidence of ADE. The willingness to pay was not proportional to the size of the risk reduction. The value of a health gain may differs depending on the method applied for each assessment.

Permalink

A study of income-motivated behavior among general practitioners

Presenter:

Tor Iversen

Authors:

Tor Iversen

Chair: Melayne McInnes; Discussant: Melayne McInnes Wed June 7, 2006 9:45-11:15 Room 226

Background: Regulated fee-for-service payment is likely to result in excessive number of services under monopolistic competition in the physician market. On the other hand, pure capitation payment is likely to result in underprovision of services. Hence, optimal payment systems for physicians are likely to be a mix of several components. We study the effect of a mixed capitation and fee-for-service system on the amount of services provided by physicians. In particular, we study to what extent the variation in service intensity among general practitioners (GPs) may be explained by an observation that some physicians have fewer regular patients than they would like to have (they experience a shortage of patients), combined with fee-for-service payment. If physicians, who experience a patient shortage, increase their intensity of service provision as a means to increase their income, we call it income-motivated behavior.

Theory: By means of non-linear programming a GP’s optimal practice style (described by the length of his list of patients and the level of service intensity) is derived under a mix of capitation and fee for service. We find that a GP, who experiences a shortage of patients, is likely to increase the number of services he provides to his patients if the marginal utility of leisure is less than the marginal utility of income from the extra services.

Empirical strategy: The sample consists of all GPs (3650) in the nationwide list patient system in Norway. Panel data at the level of individual GPs are available for the period 2001-2004. With panel data, unobserved heterogeneity is likely to occur. The assumptions of ordinary least squares regression are then violated since errors terms of different periods are correlated. We also suspect that patient shortage is not a random event. Unobserved heterogeneity is handled by generalized least squares estimation. Self selection is adjusted for by means of a Difference-in-differences estimator.

Results: We find that patient shortage increases a GP’s intensity of service provision (in particular the length of each visit) and hence, the income per listed person with 10 - 15 per cent. We also find that a GP’s income per listed person is influenced by the composition of the list according to indicators of need for services, and of accessibility according to the GP density in the municipality. These results are also valid when possible selection bias is accounted for, although the magnitude of the effects is then somewhat smaller.

Conclusion: Patient shortage is costly to the insurer because of income-motivated behavior related to the fee-for-service component of the payment system. An alternative would be to drop the fee-for-service component and let the payment system be based on the capitation fee only. But under capitation payment (with imperfect risk adjustment) not all patients are equally attractive because of variation in need for services. The present study may therefore demonstrate the classical trade off between selection and inefficiency in health care. We roughly calculate the cost of avoiding patient selection to be 3.3 % of total fee-for-service paid by the National Insurance.

Permalink

Factors Affecting Physician Productivity in a Proceduralist Specialty, Radiology

Presenter:

Cristian Meghea

Authors:

Jonathan Sunshine, Cristian Meghea

Chair: Melayne McInnes; Discussant: Melayne McInnes Wed June 7, 2006 9:45-11:15 Room 226

Rationale: There is only a small literature on the production function of physicians. We add to this literature, making a number of new contributions, methodological and substantive. Curtailing health expenses is a recurrent concern in almost all developed nations, and identifying methods to improve productivity may help ease the problem. Also, with a shortage of specialist physicians believed to be impending, finding means to enhance productivity is critical to good patient care.

Objectives: To empirically investigate the determinants of productivity of radiology practices, exploring the effect of physician labor input, physician characteristics, purportedly productivity-enhancing technologies and techniques, and other practice characteristics.

• We measure the productivity of the physician group, the production unit of medical services, while the focus of the previous literature was the individual physician.
• This study examines the effect of technology, unlike previous work.
• We account for the measurement error generally present in inputs data.
• This is the first study of a procedure-centered medical specialty. Previous literature focused on physician visits.
• We control for case mix in more detail than previous studies.

Data and Methods: Data are from the American College of Radiology’s (ACR’s) 2003 Survey of Radiologists, a nationally representative sample survey of radiologists in the United States containing information on both the respondents and the practice in which they work. We estimate the production function via OLS, the outcome being the logarithm of practice’s procedures per year. For flexibility, the physician labor inputs — full-time equivalent (FTE) radiologists, weekly hours, weeks worked annually — enter in both linear and logarithmic forms allowing for the possibility of non-constant input elasticities. To address the downward bias introduced by measurement error in labor inputs, we made various estimates of the error size and then used Monte Carlo methods to find what true elasticity combined with each plausible estimate of measurement error yields the elasticity observed in the regression.

Results: The FTE-radiologists elasticity of output is 0.8 directly measured (true elasticity between 0.85-0.90 if accounting for bias), the weekly hours elasticity is 0.4 (0.5-0.7, bias accounted) and the annual weeks elasticity is 0.4 (0.5-0.7, bias accounted). Only three of eight techniques used in radiology practices have a positive independent impact on productivity. Surprisingly, practices where individual radiologists work in more locations have higher productivity. Government owned practices are 18 percent less productive than practices owned solely by members. Practices in the Northeast and West census regions are more productive than practices in the South.

Conclusions: Due to unaccounted measurement error, previous studies probably underestimated the input elasticities in the production function of medical services. To increase output it is more efficient to add radiologists to the practice than to increase the hours or weeks worked. Some techniques used to improve radiologist productivity have less than the generally believed effect.

Permalink

Nobody does it better? The impact of surgeon specialty on outcomes for carotid endarterectomy.

Presenter:

Christopher Hollenbeak

Authors:

Christopher Hollenbeak, Adam Bowman, David Han

Chair: Melayne McInnes; Discussant: Melayne McInnes Wed June 7, 2006 9:45-11:15 Room 226

Rationale: Carotid endarterectomy (CEA) is a surgical procedure that is used as a prophylaxis for stroke. It is one of the most frequently performed surgical procedures in the United States. Traditionally, this procedure has been performed by surgeons in at least four specialties: vascular surgery, cardiothoracic surgery, general surgery, and neurosurgery. Although a few studies address the effect of surgeon specialty on outcomes, these studies have focused on inpatient mortality, which is not a common outcome, and no study has examined the impact of surgical specialty on long-term rates of stroke and other cerebrovascular events that the procedure is given to prevent.

Objectives: The purpose of this study was to determine whether surgeon specialty was associated with long-term outcomes of patients undergoing CEA in Pennsylvania.

Methodology: Data were from the Pennsylvania Health Care Cost Containment Council (PHC4) and included 17,635 patients admitted for CEA between 1995 and 1997. Long term outcomes were taken from patient readmission data for the 5-year follow-up period ending in 2002. The primary outcomes studied were mortality, stroke, combined stroke and mortality, transient ischemic attack (TIA), and re-occlusion of the ipsilateral artery. Time to event for these outcomes was modeled using a Cox proportional hazards model. Secondary outcomes included length of stay and total charges. These were modeled using a generalized linear model assuming a gamma family and a log link function.

Results: Using general surgeon as the reference group, and controlling for age, race, severity, and admission type, we found no significant difference across surgical specialties in overall mortality at 5 years after CEA. Patients treated by vascular surgeons were found to have significantly fewer (P=.005) strokes at 5 years than those treated by general surgeons. Patients of cardiothoracic (P=.085) and neurosurgeons (P=.969) did not have significantly fewer strokes than those treated by general surgeons. We found no significant difference across surgeon specialties for combined 5 year stroke and death rate. Patients of vascular surgeons had fewer TIAs than those of general surgeons (P=.041) at 5 years. Patients treated by cardiothoracic (P=.508) and neurosurgeons (P=.890) did not have significantly different 5 year TIA rates from those treated by general surgeons. Finally, patients of vascular surgeons were found to have a significantly lower re-occlusion rate (P=.009) when compared to patients of general surgeons. Patients of cardiothoracic (P=.247) and neurosurgeons (P=.284) did not have a significantly different re-occlusion rate than those treated by general surgeons.

Conclusions: These results suggest that there are significant differences in outcomes for carotid endarterectomy associated with the training of the operating surgeon. One explanation for this is the human capital differences that are fostered by the surgical specialties. There may also be learning-by-doing dimensions as surgeons who perform operations on vessels and on the neck may have an advantage in terms of outcomes.

Permalink

Employment-Contingent Health Insurance, Illness, and Labor Supply: Evidence from Married Women with Breast Cancer

Presenter:

Zhehui Luo

Authors:

Cathy Bradley, David Neumark, Zhehui Luo, Heather Bednarek

Chair: Timothy McBride; Discussant: TBA Wed June 7, 2006 9:45-11:15 Room 235

For the majority of non-elderly Americans, health insurance is either contingent upon their own employment or dependent upon the employment of a family member such as a spouse or parent. Past research has examined the labor supply behavior of individuals who have employment-contingent health insurance and of individuals dependent upon another’s policy. The effects of health on labor supply have also been studied. Absent from this literature, however, is an assessment of how the two-health and employment-contingent health insurance-interact to alter labor supply after an adverse health shock is experienced by an otherwise healthy employed individual. Such information would help policy makers understand some of the incentives and possible pitfalls of employment-contingent health insurance following illness. Women newly diagnosed with breast cancer were identified, shortly after diagnosis, from the Metropolitan Detroit Cancer Surveillance System (MDCSS), a population-based registry. Study eligibility criteria were age range of 30 to 64, English-speaking, and either employed or with an employed spouse at the time of diagnosis. We selected from this sample only women who were married and who were employed in the period just before diagnosis with cancer, and who were either insured through their own employer or through their spouse’s employer. Our empirical analysis of health, health insurance, and labor supply led to a clear finding that a negative health shock, as reflected in a diagnosis of breast cancer, decreases labor supply to a greater extent among women insured by their spouse’s health insurance policy than among women with health insurance through their employer. Moreover, the difference in responses associated with source of health insurance is greater for women with advanced-stage diseases-suggesting that even women who required aggressive treatment were sensitive to employment-contingent health insurance (ECHI) in making their labor supply decisions in the post-diagnosis period. We assessed whether women with ECHI were more attached to their jobs or had more desirable jobs. What we see is that there are only small and insignificant differences between those with insurance through the spouse, who were or were not offered ECHI. These results suggest, again, that the key difference is whether or not health insurance is contingent on employment, rather than job characteristics that might be associated with ECHI. The evidence from this analysis provided additional confidence in a causal interpretation of the findings rather than one attributable to selection. Employment-contingent insurance appears to be an incentive to remain working and to work at a greater intensity when faced with an adverse health shock. To our knowledge, this is the first study to prospectively and longitudinally examine how the labor supply responses of individuals experiencing a health shock depend on the source of health insurance. The findings underscore the labor supply incentives posed by ECHI. There may be potential benefits to employers from creating an incentive for employees to remain working after a health shock. However, the principal effect may be a health toll on individuals who remain working because of the incentives posed by their health insurance.

Permalink

Adverse selection in a voluntary-based RMHC insurance scheme

Presenter:

Hong Wang

Authors:

Hong Wang, Licheng Zhang, Winnie Yip, William Hsiao

Chair: Timothy McBride; Discussant: TBA Wed June 7, 2006 9:45-11:15 Room 235

This study examines adverse selection in a subsidized voluntary-based health insurance, the Rural Mutual Health Care (RMHC), in the poor rural area of China. This study was made possible by a unique longitudinal data set, which combines the measures of health status and other socio-economic and demographics variables from baseline survey (before RMHC is established) with the measures of enrollment status from first year evaluation survey (after RMHC is implemented one year) from the RMHC, which is a social experimental project in rural China. Total sample of this study includes 3492 rural residents from1020 households. Multinomial Logit model is employed for the data analysis.

The results of this study show that 70% rural residents enrolled in RMHC scheme. In general, enrolled individuals have worse health status than non-enrolled health status. Although household is set as the enrolment unit for the RMHC for the purpose of reducing adverse selection, nearly 1/3 enrolled household is partial-enrolled household. The residents in the group of non-enrolled individuals in partial-enrolled households have the best health status, while the residents in the group of enrolled individuals in partial-enrolled households have the worst health status. The residents in the group of non-enrolled individuals in non-enrolled households have the second best health status, while the residents in the group of enrolled individuals in full-enrolled households have the second worst health status. Pre-RMHC medical expenditure for enrolled individual in partial-enrolled households is 206.6 Yuan per capita per year, which is 1.7 times as much as the pre-RMHC medical expenditure for non-enrolled individual in partial-enrolled households. The study also reveals that pre-enrolled medical expenditure per capita per year of enrolled individuals is 9.6% higher than pre-enrolled medical expenditure of overall residents, including both enrolled and non-enrolled individuals.

In conclusion, although RMHC scheme reached a very high enrollment rate, adverse selection still exists, especially within the partial-enrolled households. Voluntary-based RMHC would not be financially sustainable if the adverse selection were not fully taken into account.

Permalink

Does Extending Health Insurance Coverage to the Uninsured Improve Population Health Outcomes?

Presenter:

Jennifer Rice

Authors:

Jennifer Rice, James Thornton

Chair: Timothy McBride; Discussant: TBA Wed June 7, 2006 9:45-11:15 Room 235

The large number of Americans without health insurance has produced a vigorous debate about whether the U.S. should adopt universal health insurance coverage. To assess whether extending health insurance coverage to a larger segment of the population is socially beneficial, it is necessary to measure the associated benefits and costs. Measuring the benefits requires a good estimate of the contribution of health insurance to measures of population health. Prior research suggests that uninsured individuals tend to have worse health outcomes than the insured; however, establishing a causal relationship and obtaining a reliable estimate of the causal effect has been difficult. In general, previous research has not adequately controlled for confounding factors associated with both health insurance coverage and health outcomes resulting in potential omitted variable bias. Many studies have also failed to account for possible reverse causation from health status to insurance coverage leading to simultaneity bias. Moreover, past research provides no information on the dynamics of the process by which changes in health insurance coverage affect mortality over time.

The objective of this paper is to investigate the aggregate relationship between health insurance and health outcomes for the general U.S. population, and extend previous work in this area in an attempt to obtain a more reliable estimate of the effect of insurance coverage on mortality that is largely purged of the influence of non-insurance determinants of health, and reverse causation. Our study addresses the following questions. Does increased insurance coverage lead to improved mortality outcomes in aggregate populations, and if so what is the size of the effect? Do different types of health insurance, such as employment, non-employment, and government financed insurance have different effects on mortality outcomes? What is the time path of the effect of insurance coverage on health outcomes? Specifically, does extending health insurance coverage to the uninsured affect aggregate health outcomes both contemporaneously and with a lag, and if so what is the size of the short-term and longer term effects?

The approach adopted in this investigation uses a panel of aggregate data on all 50 states for the period 1987-2003 to estimate a health insurance augmented, aggregate health production function for the U.S. Instrumental variable, fixed-effects, static and dynamic models are estimated that control for unobserved state heterogeneity and reverse causation. State-specific time trends are also included to allow unobservable determinants of health to vary within states over time. The effect of health insurance on mortality is therefore identified by relative within-state variation in insurance coverage. In addition, a number of observed covariates, such as income, education, unemployment, cigarette and alcohol consumption, and population demographic characteristics are also included in the aggregate health production function to control for potential determinants of mortality that may be correlated with insurance coverage. The data on health insurance coverage by state come from the U.S. Census Bureau. We also collected data on a large number of additional variables from a variety of sources to construct a unique and rich dataset for use in this study.

Permalink

More Bad News About High Gasoline Prices: A Look at Adult Cigarette Demand

Presenter:

John Tauras

Authors:

John Tauras, Sara Markowitz

Chair: Rosalie Pacula; Discussant: Don Kenkel Wed June 7, 2006 9:45-11:15 Room 309

This paper is about the spending choices of youth, with a particular focus on how the demand for cigarettes and alcohol are influenced by changes in the prices of other products. Youth tend to have small incomes and limited needs, with the result that many students spend the bulk of their income on only a few items. Fast food, clothing and entertainment make up the majority of products purchased by teenagers. The hypothesis to be tested in this project is that changes in the prices of the other goods commonly bought by teenagers and young adults will affect budget allocations and thereby affect the demand for alcohol and cigarettes.

Permalink

Does the economy affect teenage substance use?

Presenter:

Jeremy Arkes

Authors:

Jeremy Arkes

Chair: Rosalie Pacula; Discussant: Gulcin Gumus Wed June 7, 2006 9:45-11:15 Room 309

This research examines how teenage drug and alcohol use responds to changes in the economy. In contrast to the recent literature confirming procyclical alcohol use among adults, this research concludes that a weaker economy leads to greater marijuana and hard-drug use among teenagers. The results for alcohol use suggest no relationship. The findings are based on logistic models with state and year fixed effects, using teenagers from the NLSY-1997. The evidence also indicates that teenagers are more likely to sell drugs in weaker economies. This suggests one mechanism for countercyclical drug use—that access to illicit drugs is easier when the economy is weaker. These results also suggest that the strengthening economy in the 1990s mitigated what would otherwise have been much larger increases in teenage drug use.

Permalink

MJ use in late adolescence and its affect on delinquency and crime

Presenter:

Catherine MacLean

Authors:

J. Catherine MacClean, Michael T. French, Rosalie Liccardo Pacula

Chair: Rosalie Pacula; Discussant: Sara Markowitz Wed June 7, 2006 9:45-11:15 Room 309

Despite small declines in recent years, marijuana use remains high, particularly among adolescents for whom marijuana is the “drug of choice”. Unlike other illicit drugs, whether marijuana use generates crime remains fiercely debated as the criminogenic effects of marijuana have not been definitively determined. In this study, we use data from the National Epidemiologic Survey on Alcohol and Related Conditions to explore the effects of adolescent marijuana use on a range of delinquency and criminal activities. After controlling for the endogeneity of marijuana use in all specifications, we find strong evidence that various measures of marijuana consumption are related both to delinquency and criminal activity. These results have interesting policy and public health implications regarding marijuana use during adolescence.

Permalink

Behavioral Economics and Conflicts of Interest

Presenter:

George Loewenstein

Authors:

George Loewenstein

Chair: Mark Schlesinger; Discussant: Mark Schlesinger Wed June 7, 2006 9:45-11:15 Room 325

Permalink

Market Responses to Consumer Self Control problems

Presenter:

Botond Kosegi

Authors:

Botond Kosegi

Chair: Mark Schlesinger; Discussant: Mark Schlesinger Wed June 7, 2006 9:45-11:15 Room 325

Permalink

Choice, Price Competition, and Complexity in Health Insurance Markets

Presenter:

Richard Frank

Authors:

Richard Frank

Chair: Mark Schlesinger; Discussant: Mark Schlesinger Wed June 7, 2006 9:45-11:15 Room 325

Permalink

Methods for Comparing Health Disparities Over Time and Space

Presenter:

Tom McGuire

Authors:

Thomas G. McGuire, Ben Le Cook, Jeanne Miranda

Chair: Albert A. Okunade; Discussant: Christopher J. Ruhm Wed June 7, 2006 9:45-11:15 Room 326

Permalink

Assessing Socio-economic Disparities in Household Health Expenditures for Pre- and Post- 1997 Economic Crisis in Thailand

Presenter:

Chutima Suraratdecha

Authors:

Chutima Suraratdecha, Albert A. Okunade

Chair: Albert A. Okunade; Discussant: Christopher J. Ruhm Wed June 7, 2006 9:45-11:15 Room 326

Disparity research on health care access, process, outcome, cost and expenditures are rich and substantial for the developed countries (Kirby, Taliaferro, and Zuvekas, 2006). Similar studies are sparse for most Asian countries, including Thailand, mainly due to the lack of good micro databases. Reliable national household survey data observed at biennial intervals recently became available for Thailand, however. Therefore, consistent with the falling purchasing power of the bhat currency, high costs of imported drugs and medical devices, and reduction of public and private health services emanating from the 1997 economic crisis, Thai households are hypothesized to differentially adjust health care spending across socioeconomic strata. The goals of this paper are to measure the impact of the sudden economic shift emanating from the economic crisis on pre- and post- crisis household health expenditures, and to provide insights into the relationship of variations in health care spending across income quintiles and disparities in economic, regional (or provincial), and demographic factors including proximity to death. The household data for this research are the 1994, 1996, 1998, and 2000 national Socio-economic Surveys (SES), covering 98,632 Thailand households and included detailed information on household income, expenditures and demographics. Findings from Tobit regression models of health expenditure indicate that the 1997 economic crisis negatively affected household health expenditure differentially across income quintiles. Moreover, health expenditures are significantly influenced by household median age and proximity to death. Some implications for policy and further studies are explored.

Permalink

Health Insurance Disparities in Traditional and Contingent/Alternative Employment

Presenter:

Shelley White-Means

Authors:

Shelley I. White-Means, Joni Hersch

Chair: Albert A. Okunade; Discussant: TBA Wed June 7, 2006 9:45-11:15 Room 326

Relative to whites, Hispanics and blacks are less likely to have employer health insurance coverage. We examine whether ethnicity or race affects employment in traditional jobs or in contingent and alternative work arrangements, and whether ethnicity or race affects insurance offer, eligibility, and/or enrollment, conditional on employment sector. Health insurance disparities relative to whites are more pronounced for Hispanics, primarily due to disparities in employment by firms that offer coverage. Eliminating racial/ethnic disparities in offers, eligibility, and takeup would increase insurance coverage rates of Hispanics in traditional jobs and of both Hispanics and blacks in contingent and alternative jobs.

Permalink

Testing for Statistical Discrimination: Lessons from NHANES III, 1988-1994

Presenter:

Danielle Rose Ash

Authors:

Danielle Rose Ash

Chair: Albert A. Okunade; Discussant: TBA Wed June 7, 2006 9:45-11:15 Room 326

Permalink

The Effect of Health on Changing Labor Outcomes in Transition China

Presenter:

Will Dow

Authors:

Will Dow, Deokhee Yi

Chair: Paul Schultz; Discussant: John Mullahy Wed June 7, 2006 9:45-11:15 Room 332

Permalink

Long Term Consequences of Family Planning and Reproductive Health Intervention in Matlab, Bangladesh

Presenter:

Paul Schultz

Authors:

Paul Schultz, Shareen Joshi

Chair: Paul Schultz; Discussant: John Mullahy Wed June 7, 2006 9:45-11:15 Room 332

Permalink

Immediate and Longer-Term Effects of Health on Socio-economic Success

Presenter:

Duncan Thomas

Authors:

Duncan Thomas, Elizabeth Frankenberg, Jed Friedman, Jean-Pierre Habicht, Nick Ingwersen, Nathan Jones, Christopher McKelvey, Gretel Pelto, James P. Smith, Bondan Sikoki, Cecep Sumantri, Wayan Suriastini

Chair: Paul Schultz; Discussant: John Mullahy Wed June 7, 2006 9:45-11:15 Room 332

Permalink

HIV Treatment Breakthroughs and Precaution Among the Uninfected

Presenter:

Darius Lakdawalla

Authors:

Darius Lakdawalla, Neeraj Sood

Chair: Arleen Leibowitz Wed June 7, 2006 9:45-11:15 Room 335

Permalink

Distance and Time Costs of HIV Testing

Presenter:

Arleen Leibowitz

Authors:

Arleen Leibowitz, Stephanie Taylor

Chair: Arleen Leibowitz Wed June 7, 2006 9:45-11:15 Room 335

Permalink

Estimating the Impact of Medical Innovation: A Case Study of HIV Antiretroviral Treatments

Presenter:

Mark Duggan

Authors:

Mark G. Duggan, William N. Evans

Chair: Arleen Leibowitz Wed June 7, 2006 9:45-11:15 Room 335

Permalink

The Effects of Free-Standing Ambulatory Surgery Centers on Hospital Surgery Volume

Presenter:

John Bian

Authors:

John Bian, Michael Morrisey

Chair: Dean Lillard; Discussant: James Marton Wed June 7, 2006 8:00-9:30 Room 121

Rationale: Hospitals traditionally compete among themselves for health services such as surgical procedures. The growing number of free-standing ambulatory surgery centers (ASCs) poses a potential threat to hospital surgery services. However, there has been little empirical evidence on the effect of ASCs on hospital surgery volume.

Objective: This study examined the effect of the growth in ASCs on community hospital outpatient and inpatient surgery volume.

Methods: Using secondary data of the 1992-2001 American Hospital Association Annual Survey files, the 2002 Medicare Online Survey Certification and Reporting System, an HMO penetration file, and the Area Resource Files, we constructed a balanced Metropolitan Statistical Area (MSA) panel dataset including 317 MSAs from 1992-2001. The unit of analysis was a MSA-year. Ordinary least squares regressions with MSA and year fixed effects were used to control for MSA-level heterogeneity and time trends. Three dependent variables were the log-transformed hospital outpatient, inpatient, and total (inpatient and outpatient combined) surgery volumes. The key explanatory variable was the number of ASCs per 100,000 people. Other covariates included health maintenance organization (HMO) penetration, hospital concentration measured by the Herfindahl-Hirschman Index, supplies of surgeons and physicians, and demographic and economic characteristics. The standard errors were adjusted via Huber standard errors correction.

Results: The panel data included a total of 3170 MSA-years. From 1992-2001, average outpatient surgery volume at the MSA-level increased by 23% from 58,783 to 72,111, average inpatient surgery volume decreased by 12% from 50,778 to 44,911, and average total surgery volume only increased by 7% from 109,561 to 117,022. During the same period, the number of ASCs increased by 143% from .66 to 1.69 per 100,000 people, HMO penetration nearly doubled, and hospital markets became more concentrated. In regression analysis, the number of ASCs per 100,000 people was inversely associated with outpatient and total surgery volumes (p<.01) but was not associated with inpatient surgery volume (p>.10). Other thing equal, an increase in 1 ASC per 100,000 people is associated with a decrease of 4.1% in outpatient surgery volume and a decrease of 1.8% in total surgery volume. In addition, our study showed that increased hospital concentration raised hospital outpatient and total surgery volumes. But we found no associations of HMO penetration and hospital surgery volume.

Conclusions: This was the first study using nationally representative MSA-level panel data from 1992-2001 to show that ASCs may lead to a decline in hospital total surgery volume, largely driven by decreased hospital outpatient surgery volume. Additional research needs to focus on the impact of ASCs on hospital provision of charitable care and quality of care.

Permalink

The Quality of Diabetes Care by Insurance Status in Community Health Centers

Presenter:

James Zhang

Authors:

James Zhang, Anne Kirchhoff, Jennifer Walk, Marshall Chin

Chair: Dean Lillard; Discussant: James Marton Wed June 7, 2006 8:00-9:30 Room 121

Rationale: Access to quality care is important to eliminate health disparities and increase the quality and years of healthy life for all persons in the United States. Community health centers (CHCs) provide critical primary health care to 12 million Americans with or without health insurance in medically underserved areas. Over forty percent of CHC patients are uninsured and over one-third are on Medicaid; therefore, understanding how insurance status relates to care at CHCs is important for improving medical care for vulnerable CHC patients.

Objective: The objective of this study is to compare the quality of care for diabetes patients by insurance status at 27 CHCs, including patients without health insurance and Medicaid/Medicare Dual Eligibles (DEs).

Methodology: We sampled 27 CHCs in 17 West Central and Midwest states in the year 2002. A total of 2,052 diabetes patients were enrolled in the study. We developed an algorithm to expand the insurance groupings commonly reported in the literature and categorized our diabetes patients into six mutually exclusive groups that included a Medicare/Medicaid dual eligible (DE) group, as well as groups for no insurance, Medicare without Medicaid, Medicaid without Medicare, private insurance, and other. We used a set of six quality of care indicators developed by National Committee for Quality Assurance (NCQA) to assess the quality of care. We applied multivariate regression analysis technique to analyze the association between the insurance coverage and quality of care, adjusting for age, gender, race, one dummy variable for urban location of services, seven dummy variables for medical comorbidity/complications, and CHC site fixed-effects. The quality of care in those CHCs were further compared to that in commercial managed care plans located in these states.

Results: This study reveals that in CHCs there are differences in quality of diabetes care by insurance status. Patients without insurance, and younger patients with Medicaid fared the worst in six comprehensive diabetes care quality indicators. In contrast, Medicaid/Medicare dual eligibles received better quality of care despite being more frail and vulnerable both medically and economically. Patients with private insurance did significantly better in four of the six quality indicators.

Conclusion: Our findings support the initiative to create and expand CHCs in communities to improve access to care for the poor and vulnerable, such as the Federal Health Center Growth Initiative supported by the current administration. Our study also argues for coupling expanding CHC service locations with increasing insurance coverage to achieve the best health outcomes for the hundreds of thousands indigent diabetes patients who rely on safety-net providers for their primary care.

Permalink

What is the Value of a Critical Access Hospital?

Presenter:

Paul McNamara

Authors:

Paul McNamara

Chair: Dean Lillard; Discussant: James Marton Wed June 7, 2006 8:00-9:30 Room 121

The Critical Access Hospital (CAH) Program, which provides cost-based reimbursement to isolated small rural hospitals, as opposed to the prospective reimbursement formula used for most other hospitals, provides an example of how equity concerns are implemented in the context of US health policy. However, no economic analysis exists that measures the value of providing access to rural hospital services delivered in a specific location. This paper measures the value of locally provided rural health services in the specific example of the Critical Access Hospital (CAH) program. Understanding this value is necessary for the evaluation of current federal rural health policy.

This paper proposes the travel-cost hospital choice method as the means of developing an estimate for the location-specific benefits provided by a small rural hospital participating in the CAH program. The travel cost or time cost method of measuring the value of access to health services builds upon Acton’s seminal work (1975), and has been broadly applied in the analysis of health care demand, with a notable example being the work of Gertler and van der Gaag (1990). However, the method and its extension to the valuation of services has not been applied often to rural health services in higher income countries, with two exceptions being the work by Clarke (1998) and McNamara (1999).

As an illustration of the general approach and to obtain an estimate of the value of a CAH in Illinois, this paper presents an analysis of the locational demand for hospital services by patients with pneumonia using Illinois 2001 hospitalization data. Using a 40% random sample of the pneumonia cases in the first quarter of the year leads to 4618 hospitalization cases. The per trip estimates range from $14.93 to $17.41 in the case of a small CAH in Benton, Illinois to between $43.44 to $46.91 on average for all the pneumonia cases of rural Illinoisans examined. Assuming the generalizability of these CV estimates, we find that a rural hospital provides significant benefits to the community simply as a function of its location and the community’s relative distance to alternative sources of care. Average estimates of the value of a rural hospital in Illinois (assuming 1200 inpatient admissions) ranged from $17,916 for a rural hospital in Benton, Illinois to an average value of $56,292 annually based on the entire data set of 4816 pneumonia hospitalizations. These estimates of value are much smaller than net financial cost (to Medicare compared to the CAH’s previous Medicare revenues) of the CAH program’s cost-based reimbursement for participating hospitals, which anecdotally is costing between $250,000 to over $1,000,000 per participating hospital. It should be noted that these estimates do not include some other important sources of value associated with a rural hospital: outpatient visits are not considered; the value of life saved through the presence of a nearby emergency room is not considered; and, the option value that rural residents experience because of the presence of a hospital in their community or area is not considered.

Permalink

Enrollment in Health Insurance Plans Fully Paid by Employers

Presenter:

Alice Zawacki

Authors:

Alice Zawacki, Amy Taylor

Chair: David Bradford; Discussant: David Bradford Wed June 7, 2006 8:00-9:30 Room 213

Rationale: In 2003, 44% of establishments in the U.S. offered at least one health insurance plan that required no contribution from the employee for single coverage. However, not all eligible employees enrolled in these plans. Some eligible employees might want family coverage, while only single coverage is fully paid, or perhaps these plans are unattractive.

Objective: The objective of this study is to examine health insurance enrollment in establishments that pay 100% for at least one plan and to study the plans that employees were enrolled in.

Methodology: We analyze data from the 1997-2003 MEPS-IC (Medical Expenditure Panel Survey - Insurance Component), which surveyed a nationally representative sample of establishments about health insurance plans offered to employees. Data includes premiums, contributions by employers and employees, and benefit characteristics. The MEPS-IC also provides information on employer characteristics (e.g., size, ownership, and industry) and workforce characteristics (e.g., percent female, unionization, and wages).

Bivariate analysis is used to look at the percent enrolled at both the establishment and plan level. We look at employers offering only one plan and more than one plan, further subsampled into those that paid 100% of the premium cost and those that did not.

To help explain why all eligible employees do not enroll in fully paid plans, we will compare enrollment in plans where single coverage is fully paid, but family coverage is not, and enrollment in plans with both single and family coverage fully paid. This will help identify whether eligible employees may not be enrolling in fully paid single coverage plans because they want family coverage.

To examine the impact of a plan’s attractiveness on enrollment in fully paid plans and those that are not, we compare their characteristics. We focus on attributes that might make a plan more attractive to some employees, such as provider choice, coverage for pre-existing conditions, and no gatekeeper. Multivariate analysis will also be done to explain the impact of multiple dimensions of attractiveness on enrollment and compare the plans paid 100% with those that are not. Here we will focus only on plans from establishments offering more than one plan, in order to control for workforce characteristics.

Results: In 2003, only 87% of eligible employees in establishments that offered only one plan, with fully paid single coverage, enrolled in the plan. At the same time, only 33% of eligible employees in establishments that offered more than one plan enrolled in plans that had fully paid single coverage. Plans that were paid 100 percent by employers generally had gatekeepers, did not cover pre-existing conditions or outpatient prescriptions, and had the highest out-of-pocket expense limits.

Conclusions: Preliminary results indicate that some eligible employees may not be enrolling in fully paid health plans because the plans are not attractive to them. Further analysis will be done to examine whether employees do not opt for plans with fully paid single coverage because they want family coverage.

Permalink

The Impact of Increased Tax Subsidies on the Insurance Coverage of Self-Employed Families: Evidence from the 1996-2003 Medical Expenditure Panel Survey

Presenter:

Thomas Selden

Authors:

Jessica Vistnes, Thomas Selden

Chair: David Bradford; Discussant: David Bradford Wed June 7, 2006 8:00-9:30 Room 213

Over the past two decades, tax policy has provided increasing subsidies for health insurance purchased by self-employed persons. The share of premiums paid by the self employed that are excludable from federal income taxation rose from a minimal share before 1986 to 30 percent by 1996 and 100 percent as of 2003, with many states mirroring or exceeding the pace of federal change. This paper examines the impact of increased tax subsidies on the insurance coverage of self-employed workers and their families using the 1996-2003 Medical Expenditure Panel Survey. Following Gruber and Poterba (1994), the analysis examines the impact of self-employed workers’ declining tax prices, using as a control group employed workers, a group for whom tax prices held relatively constant. The analysis yields an elasticity of private coverage for adults in self-employed families that exceeds -1 in magnitude when estimated using the most widespread measure of tax price. This elasticity estimate exceeds most estimates in the literature on employed workers. Using a more comprehensive measure of the relative price of insurance, the elasticity estimate for adults in self-employed families is even larger. These results suggest that self-employed workers do respond to tax subsidies by increasing private coverage. Increased tax subsidization of self employment coverage also appears to have reduced reliance on public coverage, although the magnitude of this “reverse crowd out” is small. Tax subsidies are found to have the expected effect on the coverage of children in self-employed families, although the impact on children’s coverage is only about half of that found for adults.

Permalink

The Effect of Mandatory Employer-Sponsored Health Insurance on the Use of Part-Time versus Full-Time Workers: The Case of Hawaii.

Presenter:

Gerard Russo

Authors:

Sang-Hyop Lee, Gerard Russo, Lawrence Nitz, Abdul Jabbar, Rui Wang, Thamana Lekprichakul

Chair: David Bradford; Discussant: David Bradford Wed June 7, 2006 8:00-9:30 Room 213

Authors: Sang-Hyop Lee (leesang@hawaii.edu), Gerard Russo (russo@hawaii.edu), Lawrence H. Nitz (lnitz@hawaii.edu), Abdul Jabbar (jabbar@hawaii.edu), Rui Wang (ruiw@hawaii.edu) and Thamana Lekprichakul (thamana@hawaii.edu), University of Hawaii.

Title: The Effect of Mandatory Employer-Sponsored Health Insurance on the Use of Part-Time versus Full-Time Workers: The Case of Hawaii.

Overview: The Hawaii Prepaid Health Care Act (PHCA) of 1974 is a unique law which requires private-sector firms to provide health insurance to their employees working at least 20 hours per week. It represents a natural experiment which is ideally suited for an investigation of the impact of mandatory ESI on labor force utilization with particular emphasis on part-time and full-time workers.

Hypotheses: The Hawaii 20-hour rule is thought to have two possible effects on the distribution of the workforce by hours worked. First, firms may seek to employ more part-time workers a legal avoidance of the mandated labor expense associated with ESI. Alternatively, we hypothesize an effect in the opposite direction. This arises from the recognition that mandated ESI represents a fixed-cost per employee per month. The lump-sum nature of health insurance premiums implies the incremental cost of utilizing a full-time worker more intensely is absent any additional ESI expense. This second hypothesis states that mandatory ESI will shift the distribution of equilibrium employment by hours worked towards full-time workers, implying total labor utilization will rise but total employment will fall. We conduct an empirical investigation of these hypotheses.

Data & Method: We produce direct and model-based estimates of the distribution of employees by hours-worked, using 12-years of the Current Population Survey (CPS), Basic Monthly Survey and Annual Social and Economic (ASEC) Supplement 1994-2005. These are estimates of employment patterns for working age adults for Hawaii, the U.S. as a whole and several comparative States including Nevada, Michigan, California and Florida controlling for worker characteristics and industrial structure. We treat the Hawaii distribution as the factual and the U.S., Michigan, Nevada, California and Florida distributions as the counter-factual.

Results: We find utilization of employees working between 20 and 35 hours per week, is reduced under Hawaii’s employer mandate. We also find that mandated ESI increases the proportion of part-time workers (i.e., less than 20-hours per week) and also increases the utilization of labor from full-time workers above 36-hours per week. Generally, the shifts in the distribution of labor force are modest but statistically significant.

Conclusion: Mandating ESI will increase the cost of labor for many firms who would not otherwise provide coverage and reduce cash wages as a proportion of total compensation. The equilibrium response is two-fold. First, more part-time employment will result as employers and employees legally avoid the mandate. Second, employers will utilize full time workers more intensely by increasing hours, thereby shifting the upper distribution of employees by hours worked further to the right. On net, the distribution of employment by hours worked will be “hollowed out” in the 20-35 hour range as these employees are the relatively most expensive to employ under a Hawaii 20-hour mandate.

Permalink

Does College Education Impact Obesity and Smoking? Evidence From the Pre-Lottery Vietnam Draft

Presenter:

Bo MacInnis

Authors:

Bo MacInnis

Chair: James Burgess; Discussant: James Burgess Wed June 7, 2006 8:00-9:30 Room 225

The drastic change in drafting rules from the pre-lottery Vietnam draft to the draft lottery creates a discontinuity in postsecondary educational attainment between males of cohorts 1946-1950 and their immediately adjacent cohorts. Unlike cohorts 1942-1944, males of cohorts 1946-1950, who were appropriately aged for both the draft and college, could easily obtain draft deferments by enrolling in college during the high-induction pre-lottery period. For males of cohorts 1951-1953, enrolling in college would no longer exempt them from the draft because of the elimination of college deferments by the draft lottery. Consequently, males’ college education rate rose noticeably for cohorts 1946-1950 and then fell abruptly for cohorts 1951-1953. We exploit this exogenous discontinuity gap in college education to infer a causal effect of education on health.

Using the National Health Interview Survey 1998-2003, we find the pre-lottery Vietnam draft caused approximately a 4-5% increase in college enrollment and 3-4% increase in college completion. We find strong evidence that college completion reduces the probability of obesity and increases the probability of smoking cessation. There is moderate evidence that college completion also reduces the probability of smoking initiation and Type 2 diabetes. The evidence of the impact of college enrollment or associate degree attainment on these health outcomes is moderate to weak.

The economic benefit of college completion, in terms of the reduced probability of smoking and obesity and consequent reduced medical expenditures and productivity losses that are attributable to smoking and obesity, amounts to more than $150,000 lifetime savings total per college graduate, with the lower bound of more than $75,000. The lower bound savings are more than three times the per-student amount of public financing of college education i.e. $22,234, and represent nearly 15% of the lifetime return to a college degree compared to a high school diploma in monetary earnings i.e. $500,000.

Permalink

The Effect of Education and Parental Education on Obesity

Presenter:

Mark Stehr

Authors:

Mark Stehr

Chair: James Burgess; Discussant: James Burgess Wed June 7, 2006 8:00-9:30 Room 225

Economists have expended a great deal of effort to determine the effect of education on wages and productivity. Recent research has broadened the scope of this investigation to include the non-pecuniary benefits that education may provide such as improvements in health. Lleras Muney (2002) finds that high school education decreases mortality, but is silent on the exact mechanisms through which education operates. DeWalque (2003) shows that college education has a causal role in lowering smoking rates, but more research is needed to understand the other channels through which education exerts its positive influence on health. At the same time, economists are actively investigating the relationship between markers of socioeconomic status, such as income and education, and child health (see Currie and Stabile, 2003).

This paper tests the hypothesis that more schooling at the college level leads to (a) lower levels of obesity and (b) lower levels of obesity among one’s children. Obesity is a particularly important health outcome because it is rapidly approaching smoking as a cause of premature morbidity and mortality. This hypothesis cannot be tested by examining a simple association between education and body mass index (BMI) because both of these outcomes may be influenced by unobservable characteristics of the individual. For example, the relative value individuals place on current and future consumption may vary. Those who place a high value on future consumption may make large investments in education and health while they are young that involve sacrifices in the form of foregone wages and leisure time. Those who place a high value on current consumption may not be willing to make these sacrifices when they are young, and as a consequence may enjoy lower earnings and health when they are older. Thus, to infer from the simple association between education and BMI that education reduces obesity risk is invalid.

Ideally, to isolate the effect of education on obesity, one would randomly assign individuals to different education levels and then follow the evolution of their BMI over time. Because this is clearly infeasible, I instead use a quasi-experimental design that attempts to mimic this random assignment. Use of this quasi-experiment requires that the experiment predict education, but have no direct effect on BMI. My quasi-experiment is the number of colleges and universities in an individual’s county of residence at age 17. Previous researchers have relied on this quasi-experiment to study the effect of education on wages (Card, 1995) and civic participation (Dee, 2004). To acquire data on schools, I use the Higher Education General Information Survey (HEGIS), which provides data on the number of 2-year and 4-year colleges in each county in the United States. Then, I match this measure of college availability by county with respondents from the NLSY79 and their children from the NLSY79 Child/Young Adult Survey. Preliminary OLS results indicate a strong negative association between education and obesity, but it is too early to report results from the quasi-experimental research design outlined above.

Permalink

How Does Parental Education Affect Child Health?

Presenter:

Kosali Simon

Authors:

Kosali Simon, Dean Lillard, Maki Ueyama

Chair: James Burgess; Discussant: James Burgess Wed June 7, 2006 8:00-9:30 Room 225

Estimating the causal effect of education on one’s own or one’s children’s health is complicated by the fact that unobserved variation across individuals could cause both health and education. Little prior research considers the effect of parent’s education on child health in a developed country context. A recent exception is Currie and Moretti (QJE, 2003) who show that exogenous increases in college education induced by college openings has a beneficial impact on an infant’s health. No study has examined how parental high school completion causally affects infant, child or adolescent health in the US. Our paper provides answers to this question, and investigates mechanisms that may be responsible for this effect. We exploit variation in state educational testing policies (graduation requirements and GED) that lead to exogenous differences in high school educational attainment. We conduct our study with two data sets; the National Longitudinal Survey of Youth, 1979 (NLSY79) cohort, and the Natality Detail Dat??????f health care services, or through health related behaviors.

Permalink

Intergenerational Obesity Transmission and Correlations of Human Capital Accumulation

Presenter:

Timothy Classen

Authors:

Timothy Classen

Chair: Vilma Carande-Kulis; Discussant: TBA Wed June 7, 2006 8:00-9:30 Room 226

The goal of this research is to provide an estimate of the intergenerational persistence of obesity and its influence on human capital accumulation. I measure the intergenerational correlation of weight status between women and their children when both are at similar stages of development. This study contributes to the literature on the role of health as a mechanism in the correlation of economic status between generations. Prior studies of obesity have found a strong relationship between weight status and economic outcomes. Thus, the transmission of obesity between generations may explain a portion of the intergenerational correlations of economic status that have previously been characterized. Using the National Longitudinal Survey of Youth 1979 (NLSY79) and the Children and Young Adults of the NLSY79, I compute the Body Mass Index (BMI) of women and their children when both generations are between the ages of 16 and 24. In the sample used, the measured intergenerational correlation of BMI is roughly 0.35. This result differs by the gender of the offspring with a BMI correlation between female children and their mothers of 0.38, compared to a significantly lower BMI correlation of 0.32 between mothers and their sons. Intragenerational correlations are slightly lower and are highest for same-gender siblings. Women who were overweight in early adulthood are found to have a lower likelihood of high school completion and produce offspring who are also less likely to complete high school.

Permalink

Food Insecurity, Food Storage, and Obesity

Presenter:

Sean Cash

Authors:

Sean Cash, David Zilberman

Chair: Vilma Carande-Kulis; Discussant: TBA Wed June 7, 2006 8:00-9:30 Room 226

Although individuals with poor food security might be expected to have reduced food intake, and therefore a lower likelihood of being overweight, some empirical evidence has indicated that overweight status is actually more prevalent among the food insecure (Townsend et al., 2001; Adams et al., 2003; Sarlio-Lähteenkorva and Lahelma, 2001; Alaimo et al., 2001). As obesity is associated with excessive energy intake, and hunger reflects an inadequate food supply, such observations would appear to be paradoxical (Dietz, 1995). We develop an economic model that shows that this apparently paradoxical result is consistent with rational behavior regarding food availability risk and the effectiveness of food storage options.

We construct a two-period model of utility maximization, in which periods may differ by the availability of food for harvest. The availability of food in the current period is deterministic, whereas the availability of food in the second period is stochastic. Utility in each period is determined by the contemporaneous consumption of food, health status, and time allocated to leisure. Health status is determined, in part, by the stock of internally stored energy. Individuals can either consume food, store food physically (externally) for a future period, or store energy internally (i.e., as body fat). Both forms of storage are subject to depreciation. Individuals seek to maximize utility by allocating time and first period consumption decisions, subject to both time and food availability constraints.

The model suggests that if physical storage is ineffective and the health effect in the second period dominates the consumption effect, then there will be extra consumption in the first period for storage of energy as body fat. The amount of internal storage increases as the variance of food productivity in the second period increases, which is consistent with the empirical observation of a positive relationship between food insecurity and the incidence of overweight. The model further indicates that higher climatic energy needs (e.g., colder climates) and lower efficiency of external food storage will also contribute to increased accumulation of body fat.

The theoretical model is then adapted for use in two simulations. The first simulation indicates that the model also predicts a higher probability of survival of individuals adapting an internal energy storage strategy in the face of increased food insecurity. This suggests that the model applies not just under an assumption of reasoned utility maximization, but also to explanations involving evolutionary behavior. In the second exercise, we adapt the model for use in an empirical simulation involving data from the U.S. Continuing Study of Food Intake in Individuals, showing that the model is consistent with previously noted results regarding the incidence of food insecurity and overweight in those data.

Permalink

Can you afford to exercise? Effects of time and money income on physical activity and the body mass

Presenter:

Nidhi Thakur

Authors:

Nidhi Thakur

Chair: Vilma Carande-Kulis; Discussant: Timothy Classen Wed June 7, 2006 8:00-9:30 Room 226

Rationale: Much has been told about how technological advancements have lowered the price of food and reduced physical demands of work, both of which have been attributed to increased body mass amongst Americans. However, while this explanation of obesity does explain long-term trends in body mass, it should not be forgotten that basically obesity is a micro-level phenomenon. At any point in time individuals take technological advancements as given and then they make behavioral choices which show up in heterogeneity in the weight outcomes. These individual choices are a function of various socio-economic processes which may interact in various ways.

Objectives: The objective of this paper is to analyse the determinants of physical exercise and its impact on body mass.

Methodology: We estimate a system of simultaneous equations where weight is determined by exercise levels besides various socio-economic factors and exercise level is in turn determined by weight and various socio-economic factors. We believe that exercise by its sheer time intensiveness has a large time cost, even when its direct monetary costs may be low, for example in a simple exercise of walking in the park. The time costs therefore add to the opportunity costs of undertaking physical exercise. To the extent that individuals are uniformly endowed with 24 hours in a day, clearly they will allocate their time such as to maximize their utility a crucial component of which is the money or budget constraints. Thus allocation of time to physically demanding activities will determine the body weight of an individual. However, the need to do exercise may itself be motivated by the body weight. We use the NLSY79 data for the years 1998-2002, the only years for which the NLSY explicitly asked questions on physical activity level. To our knowledge there is no other paper which has used this aspect of NLSY till now.

Results: Preliminary analysis suggests that lack of physical exercise can explain part of an increase in body-weight of an individual. Additionally we find that physical exercise, is most lacking in some income groups.

Conclusions: To the extent that individuals do not choose the technology levels in the economy, their choices are revealed through their use of their resources such as to maximize utility. Since overweight does have a disutility clearly individuals who still end up with an uncomfortable level of body weight must be constrained either by money income, or low discount rates, or time. We approached the problem of obesity from time perspective, and our analysis suggests that policy should intervene in the time use of individuals such as to gear them towards some regular basic level of physical exercise.

Permalink

Storm Clouds on the Horizon - Expected Adverse Selection in Medicare Prescription Drug Plans

Presenter:

Steven Pizer

Authors:

Steven Pizer, Austin Frakt, Roger Feldman

Chair: Carole Gresenz; Discussant: Bill Encinosa Wed June 7, 2006 8:00-9:30 Room 235

Objective: To estimate the costs and benefits from a Neonatal Intensive Care Unit (NICU) case management program. Background: Starting in 2005 ParadigmHealth (PH) began administering its NICU case management program for Blue Shield of California (BSC) members. During the 2 years prior there was no NICU case management program in place. This program assigns a NICU-trained RN case manager to each enrolled patient. The on-site RN case manager consults with ParadigmHealth’s neonatologists. The PH team works with the hospital’s neonatology staff to develop a treatment plan using evidence-based guidelines. In addition PH provides patient education and creates a post-discharge plan with the family. PH receives a regular data feed from the BSC inpatient authorization system. Patients are enrolled 2-3 days after admission to the NICU. This program is expected to reduce the initial length of stay as well as reduce readmissions. Methodology: This program was implemented system-wide in January 2005, with no control group. However BSC NICU data was available during a two-year period prior to program implementation (2003-2004). To control for possible differences in case-mix across time we stratified the cases into 21 groups. There were seven birth weight categories and three clinical groups (surgical, congenital abnormalities or respiratory distress, and other complications). 2015 NICU cases were included in the baseline analysis. Cases that were discharged within 3 days of admission and cases with missing or conflicting birth weight observations were excluded. There are also several clinical exclusions from the program. These conditions (such as transplant surgery and extracorporeal membrane oxygenation) are all rare within the BSC population. Average length of stay (ALOS) was calculated for each of the 21 weight/clinical groups. This length of stay included any readmissions or transfers. The analysis was repeated with data from 2005. We then calculated the difference in weighted ALOS between the 2003-2004 baseline and the 2005 treatment groups. Preliminary Results: The ALOS for the 2003-2004 baseline group was 24.3 days (N=2015). In the treatment group, 194 patients from 2005 1st quarter have complete follow-up. We found a reduction in ALOS of 3.5 days. These preliminary results are statistically significant with a p-value of 0.03. The program costs (vendor costs plus internal BSC costs) are substantially less then the cost savings from the reduction in NICU days. Readmission rates will be assessed when 2005 2nd quarter data is released. Limitations: Expensive NICU cases take longer to process then most hospital claims, with some cases not getting full adjudication for more the 180 days. Thus there may be further revisions to 2005 first quarter data. Data from the subsequent quarters will be available in late 2005 and early 2006.

Permalink

Potential for adverse selection for the new Medicare drug benefit

Presenter:

Patrick Bernet

Authors:

Patrick Bernet

Chair: Carole Gresenz; Discussant: Steve Zuckerman Wed June 7, 2006 8:00-9:30 Room 235

Rationale: Enrollment in the new drug benefit program is voluntary for most Medicare beneficiaries. Such voluntary enrollment is often associated with concern over adverse selection.

Objectives: This paper first explores the potential for adverse selection through an assessment of the likely proportion for which enrollment in the new Medicare drug plan will be mandatory. This is followed with a study of the insuring habits of older people who already use prescription drugs regularly.

Methodology: Medicare beneficiary data is employed to estimate the proportion of recipients for whom enrollment in the new drug plan is voluntary. This represents the size of the population for which adverse selection is a valid concern. Data from the Health and Retirement Survey (HRS) is then used to analyze the insurance purchasing habits of people who take prescription drugs regularly. While HRS does not have information on the new drug insurance product, it does have information on the other insurance decisions, including the level of Medicare supplemental insurance, life insurance and long-term care insurance. An analysis relates the propensity to insure with prescription drug use. This study draws on theories relating individual risk preference to medical history.

Results: Medicare recipients who use prescription drugs regularly are more likely to purchase higher levels of Medicare supplemental insurance and are more likely to own long-term care insurance policies.

Conclusions: The higher propensity of routine prescription drug users to insure in other areas is likely to carry forward under the new Medicare drug benefit. As such, voluntary Medicare drug plan enrollment will likely be skewed towards those who use more prescription drugs than average, bringing fears of adverse selection to fruition. As with other forms of health insurance, success for insures may be primarily a function of avoiding adverse selection.

Permalink

Income-Related Disparities in Kidney Transplant Graft Failures Are Eliminated by Medicare's Immunosuppression Coverage

Presenter:

Robert Woodward

Authors:

Robert Woodward, Ricardo Soares

Chair: Carole Gresenz; Discussant: Will Dow Wed June 7, 2006 8:00-9:30 Room 235

RATIONALE: The maintenance immunosuppressive (IS) regimens required following kidney transplantation cost approximately $1,000 per month, a financial burden that is especially difficult for lower income transplant recipients. The changing durations of Medicare’s coverage of IS medications for kidney transplant recipients have provided opportunities to estimate the impact of that coverage on income-related disparities in the survival of the kidney graft. Previously published analyses of Medicare’s 1994 extension from 1 to 3 years IS coverage demonstrated that the extra two years eliminated a previously significant income-related disparity in the 2nd and 3rd years post-transplant. In 2000, Medicare extend the IS coverage to the life of the graft for the elderly and disabled.

OBJECTIVE: This study seeks to estimate the proportion of the annual graft losses in the fourth and fifth year post-transplant that may be correlated with the loss of Medicare’s Immunosuppression coverage at the end of three years post-transplant.

METHODOLOGY: The most recent USRDS data allow up to a five year follow-up for patients transplanted between January of 1997 and December of 1999. Among this cohort, income-related differences in graft survival among individuals eligible for only three years of IS coverage were compared with income-related difference among individuals eligible for life-time IS coverage. Of the 13,491first kidney transplanted from cadaveric donors whose transplant was primarily financed by Medicare, 3,943 had grafts that survived uncensored for 3 years and 90 days or more and who used Medicare’s IS coverage for 3 years. Median income of the ZIP code of the patient’s residence from the 2000 Census was used as a proxy patient income. Kaplan-Meier plots and multivariate Cox Proportional Hazard models were used to identify significant differences income-related disparities between the graft survival of patients who then did use and did not use Medicare’s IS coverage in the first 3 months of the 4th year.

RESULTS: Kaplan-Meier model estimates indicated that: i) loss of Medicare’s Insurance Coverage increased graft failures by 56.3% (P<0.01) among low income individuals with incomes too high for Medicaid, and ii) 12.3% of all graft failures in years 4 and 5 after transplant may be attributed to the loss of Medicare’s insurance coverage.

CONCLUSION: Medicare’s extensions of IS coverage have consistently eliminated the previously existing income-related disparities in kidney graft survival. Unfortunately, individuals who received a transplanted kidney primarily financed by Medicare, who are not disabled, and who were under 62 at transplantation remain eligible for only 3 years of Medicare’s IS coverage.

Permalink

Cost-effectiveness of implementation research: the VA QUERI program

Presenter:

Mark Smith

Authors:

Mark Smith, Paul Barnett, Andrea Shane

Chair: David Meltzer; Discussant: David Meltzer Wed June 7, 2006 8:00-9:30 Room 309

Objectives: (1) to assess how traditional cost-effectiveness analysis (CEA) methods must be tailored and expanded for implementation research; (2) to introduce QUERI, the national VA program to support implementation of medical best practices.

Methods: We will describe traditional CEA methods and their use in current VA research. Next we will discuss how implementation research differs from traditional clinical trials, and how this impacts both the range and measurement of economic outcomes. We will present data on economic studies within the QUERI program and suggest ways in which future projects could be used to test new methods.

Results: Implementing an intervention raises economic issues beyond those encountered in standard CEA. Data are needed on the costs and impacts of both the underlying intervention and the program that implements it. The societal perspective of a standard CEA may be less influential than a managerial perspective that emphasizes a short time horizon, fixed staffing and facilities, and community standards of care. The comparator intervention (“usual care”) may depend on the level at which the intervention takes place, which in turn affects the range of study designs available.

We have two major lessons: (1) An implementation effort itself has costs and effectiveness separate from the intervention it is spreading. This has two implications. First, cost-effective interventions may not be cost-effective when implemented. Second, the best response to poor cost-effectiveness during implementation may be to redesign the implementation strategy rather than to alter or drop the underlying intervention. (2) Economic analyses of implementation research should reflect the needs of end users at a variety of levels. Policymakers at local, regional, and national levels are likely to view cost burdens differently and may have differing thresholds for cost-effectiveness.

VA presents many advantages for studying implementation methods. These include a uniform electronic patient record system nationwide, a comprehensive set of encounter-level data open to researchers, the ability to link VA and Medicare records, and administrative information on the 140+ local VA health systems and the regional networks they belong to. Although they belong to a single national system, there is still considerable diversity in medical practice and administration among VA facilities. Nationally, VA has committed to implementing best practices throughout its system and considers economic analysis an important element of the process.

Conclusions: Improved economic analyses can and should be implemented alongside implementation projects. Doing so is feasible in terms of both data collection and cost. Researchers should consider VA a laboratory for developing and testing new CEA methods alongside implementation projects.

Permalink

The Use of Time-Variant Attributes in Conjoint Analysis

Presenter:

Eric Nauenberg

Authors:

Eric Nauenberg, Richard Ito, Lusine Abrahamyan, Amir Azarpazhooh, Nancy Valencia Rojas

Chair: David Meltzer; Discussant: Scott Grosse Wed June 7, 2006 8:00-9:30 Room 309

Rationale: Conjoint analysis is a valuable tool in eliciting preferences for various treatment options and has helped to set priorities. The basic structure of such analyses involves making comparisons of alternatives against a status quo. The latter is one in which the values of attributes remain fixed while the attribute values for alternative treatments are varied to determine how preferences between options are affected. The requirement that the status quo be represented by a fixed set of attributes presents a problem in analyzing attributes that vary with time-so-called time-variant attributes or attributes that reflect first differences (e.g., X= X(t) - X(t+1)). By their nature, the baseline values (X(t)) of these attributes often vary alongside varying values for the first differences. With such attributes, the values associated with the status quo will also vary as the baseline values shift even though the first difference will always be zero reflecting no change. It appears, therefore, that such attributes can not be included in conjoint analyses.

Objective: This paper presents a method for coding time-variant attributes in conjoint analyses.

Methodology: This paper presents two treatment options-status quo and alternative—each with three attributes: 1. cost-effectiveness (CE), 2. budget impact (BI) in terms of dollars expended per HMO member per month (PMPM) and 3. change in newly budgeted dollars left unexpended after inclusion of alternative treatment in the formulary. The third attribute is a time-variant attribute. The following presents attribute levels for the two treatments.

Cost-Effectiveness, 3 levels (per QALY): 1. $50,000 vs. $50,000, 2. $50,000 vs. $100,000, 3. $50,000 vs. $180,000

Budget Impact, 2 levels (PMPM): 1. $0 vs. $200, 2. $0 vs. $400

Unexpended budget, 3 levels: 1. 5% (no change) vs. 5% to 4%, 2. 1% (no change) vs. 1% to 0%, 3. 1% (no change) vs. 1% to 0.5% over budget

Each of the time-invariant attributes could be coded as the value included in the above table. Regarding the time-variant attribute, the economist could use a set of dummy variables to represent the different levels and use “no change” as the reference group. Thus, “5% to 4%” could be represented by a dummy variable D1 and “1% to 0%”could be represented by a dummy variable D2 and “1% to 0.5%” over budget could be represented by a dummy variable D3. All three levels of this attribute for the status quo could be coded as “no change” or coded as a value of 0 for all three dummy variables. The value of the associated regression coefficients would capture the value that decision-makers place on both the magnitude of the change as well as differences at baseline.

Results: Probit regression is used to analyze the change in benefit (delta B) from adopting the alternative treatment.

Delta B = Beta1(CE) + Beta2( (BI) + Beta3( (D1) + Beta4( (D2) + Beta5( (D3)

The results of a November 2005 pilot analyzing preferences for alternative treatments for end-stage-kidney-disease will be presented (n=32).

Conclusion: This paper expands the scope of conjoint analysis by employing time-variant attributes.

Permalink

Validity and Reliability of Willingness-to-Pay Estimates: Evidence from Two Overlapping Discrete-Choice Experiments

Presenter:

Peter Zweifel

Authors:

Peter Zweifel, Harry Telser, Karolin Becker

Chair: David Meltzer; Discussant: William Cartwright Wed June 7, 2006 8:00-9:30 Room 309

Discrete-choice experiments, while becoming increasingly popular, have rarely been tested for validity and reliability. This contribution purports to provide some evidence of a rather uniqe type. Two surveys designed to measure willingess-to-accept (WTA) for reform options in Swiss health care and health insurance are used to provide independent information w th regard to two elements of reform. The issue to be addressed is whether WTA values converge although the three overlapping attributes (a more restrictive drug benefit, delayed access to medical innovation, and a change in the monthly insurance premium) are imbedded in widely differing choice sets. Experimentt A contains rather radical health system reform options. while experiment B concentrates on more familiar elements such as copayments and the benefit catalogue. EWile mean WTA values differ between experiments, the tend to vary in similar ways, suggesting at least theoretical validity and reliability.

Permalink

Information Technology Adoption and the Quality of Care in U.S. Hospitals

Presenter:

Timothy Simcoe

Authors:

Tim Simcoe, Dov Rothman

Chair: Stephen T. Parente; Discussant: Jonathan Ketcham Wed June 7, 2006 8:00-9:30 Room 313

There has been increasing interest over the past decade in using information technology (IT) to improve the quality of medical care provided by U.S. hospitals. For example, the Institutes of Medicine (IOM) frequently highlights the potential role of IT in a series of reports on quality improvement. Evidence for these claims frequently comes from event studies that focus on a particular technology and a narrow range of outcomes—such as prescription error rates following the adoption of Computerized Physician Order Entry (CPOE) systems. However, there is a broader literature on IT and productivity that generally finds substantial variation in the establishment-level benefits of IT adoption. This literature attributes variation in the productivity impact of IT to the presence of complementary business processes and organizational characteristics. This paper will examine the relationship between IT adoption and the quality of care at U.S. hospitals, with a focus on identifying complementarities between IT and particular business processes or local market characteristics. We plan to address the following questions: Do hospitals that adopt IT earlier or use it more intensively provide better quality care? If they do, which dimensions of quality are most influence by IT? Are there complementarities between different kinds of IT or between IT and other clinical technologies? And finally, are there complementarities between IT and other business practices or market characteristics? The primary methodological challenge for this line of research is the potential endogeneity of IT adoption. In particular, we might expect the hospitals adopting a particular technology to be precisely those facilities that will benefit most from it. To address this issue, we consider several approaches for isolating the causal impact of IT on the quality of care. One approach exploits small differences in the timing of adoption, by comparing quality outcomes at hospitals that have implemented IT to those who have contracted for the same technology but not yet implemented it. Another approach uses a statistical method called Inverse Probability of Treatment Weighted (IPTW) estimation, which generalizes propensity-score matching techniques that control for selection on observables to the case of time-varying treatments (i.e. adoption decisions).1 The data for this study come from several sources. Measures of hospital IT adoption come from The Dorenfest Complete IHDS+ Databases 1998-2002. This data contains demographic and information systems data for acute care facilities with 100 beds or more. We also plan to construct clinical technology adoption data using publicly available administrative discharge data from Florida (HCUP) and California (OSHPD). Finally, we measure quality using the Healthcare Cost and Utilization Project (HCUP) quality indicator set, and the Patient Safety Indicator Modules (PSI) available from the Agency for Healthcare Research and Quality (AHRQ).2…

Permalink

Network Effects and Diffusion of Health Information Technology

Presenter:

Michael Furukawa

Authors:

Michael F. Furukawa

Chair: Stephen T. Parente; Discussant: TBA Wed June 7, 2006 8:00-9:30 Room 313

Health information technology (HIT) is widely-regarded as a key strategic resource to increase efficiency and improve quality of care. Despite growing interest in HIT by managers and policymakers, few studies have examined the determinants of HIT adoption. In particular, little is known about the role of network effects in the diffusion of HIT across hospitals and affiliated organizations. This study seeks to address this gap in the literature by examining the effect of system membership on the adoption of HIT by hospitals, sub-acute, and ambulatory facilities. The primary data source is the 2004 HIMSS Analytics (HA) database, which contains detailed information on the adoption of HIT by 1,453 integrated health delivery systems. The sample includes 3,989 hospitals, 3,007 sub-acute, and 18,008 ambulatory care facilities. For hospitals and sub-acute facilities, the HIT variables are classified into 4 categories: financials and business office; medical records and administrative; management and human resources; and clinical and ancillary departments. Variables are also created for specific HIT, particularly advanced clinical applications such as computerized patient records and computerized physician order entry. The HA database is linked to two supplemental data sets: 1) 2002 AHA Annual Survey of Hospitals, which provides information on hospital (staffed beds, ownership type, teaching status, etc.) and system characteristics (centralization, physician arrangement, insurance product, etc.), and 2) 2004 Area Resource File, which provides information on area characteristics (rural, per capita income, etc.). I estimate multivariate regression models to test the effect of system characteristics on HIT adoption, controlling for facility and area characteristics. The first model uses logistic regression to estimate the factors correlated with the likelihood of adoption of specific HIT applications. The second model uses negative binomial regression to estimate the factors associated with the level of HIT adoption, as proxied by the count of applications adopted within each HIT category. Finally, I test the robustness of the results to the potential endogeneity of system membership using appropriate econometric methods. Preliminary results suggest that network effects are a strong determinant of HIT adoption. In particular, facilities in systems that are more centralized and offering an insurance product are much more likely to adopt advanced clinical IT as well as have a higher level of HIT adoption. The consolidation and integration of hospitals, physicians, and affiliated organizations into integrated health delivery systems has important implications for the adoption and diffusion of HIT. System membership may confer “network effects” because HIT adoption confers benefits to the adopter as well as other affiliated facilities. The study finds evidence that system governance and integration play a key role in HIT adoption. This implies that HIT adoption decisions are made at the system-level and that organizational structure plays a key role in the diffusion of HIT across facilities within a system.

Permalink

The financial and clinical value of hospital information technology: A study of complementarity between monitoring technologies and health system organization

Presenter:

Jeffrey McCullough

Authors:

Jeffrey S. McCullough

Chair: Stephen T. Parente; Discussant: Yunwei Gai Wed June 7, 2006 8:00-9:30 Room 313

Health information technology (IT) holds the promise of improved clinical outcomes at lower costs. The benefits of health information technology, however, may not be fully realized through the simple adoption of new technologies. Rather, health systems may need to simultaneously reorganize their operations to complement information technology. This study estimates the clinical and financial value of both monitoring technologies and integration as well as complementarities between these inputs; furthermore, this study allows for both adoption and integration decisions to be endogenously determined… Monitoring technologies present a particularly interesting case for understanding the relationship between technology and firm boundaries. Monitoring technologies might allow for more complete contracts, thus being complementary to less integration between physicians and hospitals—a decrease in firm boundaries. Conversely, monitoring might improve the efficiency of hierarchical management structures thus being complementary to greater physician-hospital integration—an increase in the boundaries of the firm. By estimating the complementarity between monitoring IT and integration, I test these competing hypotheses. This study utilizes data from the Dorenfest IHDS+ and 3000+ databases for the years 1994 through 2003. Together, these databases provide a nearly complete census of U.S. nongovernmental, acute-care hospitals with more than 100 beds for the years 1994 through 2003. These data provide detailed descriptions of the nature and timing of hospitals’ adoption of monitoring information system. Furthermore, the Dorenfest data have been combined with five additional sources of hospital and physician data for the years 1994 through 2003; specifically, American Hospital Association’s (AHA) database of hospital characteristics, the Medicare Cost Reports which provide hospital cost and case-mix-adjustment data, the HCUP data which provide further detail of hospital output and clinical outcomes, the Area Resource File, and Interstudy’s HMO and managed care data. Additional data regarding the hedonic price of hardware and technology worker wages have been obtained from the Bureau of Labor Statistics. This study will estimate models of hospital costs and quality as a function of technology adoption and organizational design. A system of simultaneous equations for hospital performance, vertical integration, and technology adoption will produce consistent estimates of complementarities while allowing for endogenous choices regarding both integration and adoption. Two sets of instruments will be utilized to identify the model: physician labor market conditions for the integration equation and hardware and IT labor costs for the technology adoption component. This approach is adapted from Athey and Stern’s (1998) framework for testing complementarity in organizational design.

Permalink

Does Information Matter? The Case of Nursing Home Report Cards

Presenter:

David Grabowski

Authors:

David C. Grabowski Jonathan Gruber, Robert J. Town

Chair: Edward Norton; Discussant: Jennifer Troyer Wed June 7, 2006 8:00-9:30 Room 325

Over the last decade, there has been significant growth in the use of public report cards as a mechanism to address information asymmetries on the part of health care consumers. However, the overall welfare implications of report cards are unclear. On the one hand, they may empower consumers to make more informed choices and increase quality competition among providers. However, they may also create disincentives to admit certain patient types (e.g., sicker patients), and they may also increase market power on the part of providers. In the nursing home sector, there have been recent Federal and State initiatives to publicize quality information to consumers. Most significantly, the Federal government introduced the Nursing Home Compare website with a detailed report card on every certified provider in the country. The early rollout of the Federal report cards in a collection of states on a pilot basis along with the various state report card efforts allow us to exploit within-state variation in these policies over time. Using a number of datasets, we assess the overall welfare implications of nursing home report cards. Specifically, we analyze the implications of the introduction of report cards on demand for care, quality, patient mix, private-pay prices, and competition.

Permalink

The Staffing-Outcomes Relationship in Nursing Homes

Presenter:

Sally Stearns

Authors:

R. Tamara Konetzka, Sally S. Stearns, Jeongyoung Park

Chair: Edward Norton; Discussant: John Nyman Wed June 7, 2006 8:00-9:30 Room 325

Permalink

The Cost-Quality Relationship: Evidence from Ontario Complex Continuing Care

Presenter:

Walter Wodchis

Authors:

Walter P. Wodchis, Gary F. Teare, Geoff M. Anderson

Chair: Edward Norton; Discussant: Melinda Beeuwkes Buntin Wed June 7, 2006 8:00-9:30 Room 325

The relationship between cost and quality in long-term care has not been carefully explored, yet policy is often made based on assumptions about this relationship. This research examines the relationship between cost performance (cost per weighted patient day) and four quality indicators in Ontario Complex Continuing Care hospital beds. We used five years of data from 99 Ontario facilities from the period 1996 through 2000. Facilities were stratified into high and low categories of cost and quality using quartiles of the indicator distributions. These distributions revealed a marked variability in both cost and quality. Cross-sectional and pooled time-series analyses were also conducted. Over time, cost and quality performance showed a high degree of serial correlation. Reduced costs were associated with lower pressure ulcer prevalence with lower incidence of bladder incontinence, while high costs were associated with high quality assessed as a lower pressure ulcer incidence rate. There was no significant relationship between costs and pain management, prevalence of incontinence, antipsychotic use or the use of physical restraints. The analyses to date indicate opportunities for facilities to improve both cost and quality performance in pressure ulcer management and incontinence care.

Permalink

The Incidence of the Healthcare Costs of Obesity

Presenter:

Jay Bhattacharya

Authors:

Jay Bhattacharya, M. Kate Bundorf

Chair: Katherine McDonald; Discussant: Jay Bhattacharya Wed June 7, 2006 8:00-9:30 Room 326

The incidence of obesity has increased dramatically in the U.S. Obese individuals tend to be sicker and spend more on health care, raising the question of who bears the incidence of obesity-related health care costs. This question is particularly interesting among those with group coverage through an employer given the lack of explicit risk adjustment of individual health insurance premiums in the group market. In this paper, we examine the incidence of the healthcare costs of obesity among full time workers. We find that the incremental healthcare costs associated with obesity are passed on to obese workers with employer-sponsored health insurance in the form of lower cash wages. Obese workers in firms without employer-sponsored insurance do not have a wage offset relative to their non-obese counterparts. Our estimate of the wage offset exceeds estimates of the expected incremental health care costs of these individuals for obese women, but not for men. We find that a substantial part of the lower wages among obese women attributed to labor market discrimination can be explained by the higher health insurance premiums required to cover them.

Permalink

JNC VI Had Modest Impact on Antihypertensive Prescribing Patterns in the US between 1993 and 2002

Presenter:

Jun Ma

Authors:

Jun Ma, Ky-Van Lee, Randall Stafford

Chair: Katherine McDonald; Discussant: Jun Ma Wed June 7, 2006 8:00-9:30 Room 326

Objective: Practice guidelines aim to guide physician practice according to the best available evidence. Data are mixed regarding the impact of practice guidelines on physician prescribing. We examined patterns of antihypertensive prescribing between 1993 and 2002 to assess the impact of JNC VI released in 1997. Methods: The 1993-2002 National Ambulatory Medical Care Survey (NAMCS) and National Hospital Ambulatory Medical Care Survey (NHAMCS) provided national estimates of clinician-reported antihypertensive prescribing during patient visits to private physician offices and hospital outpatient departments (OPDs). The main outcome measure was the use of antihypertensive drug classes as proportion of visits for adult patients having hypertension as primary diagnosis who were taking any antihypertensive medications (antihypertensive drug visits).
Results: The number of patient visits involving hypertension as the primary diagnosis ranged from 37,777 to 50,173 per annum, of which 62% to 78% received antihypertensive medications. ACE inhibitors (ACEIs) and calcium channel blockers (CCBs) were leading medication choices by physicians in private practice and OPDs for treating hypertension between 1993 and 1997. After the release of JNC VI, the increase in ACEIs seemed to slow down somewhat and the already present decline in CCBs continued. Also, their role as leading antihypertensive drug class was approached or replaced by thiazide diuretics. In 2002, thiazide diuretics accounted for 34% of antihypertensive drug visits in private physician offices and 40% in OPDs, as opposed to 24% in 1993. The use of other diuretics declined by nearly half from 27% of office-based antihypertensive drug visits to in 1993 to 14% in 2002; the degree of decline was smaller in OPDs from 20% to 16%. -blockers increased from a bit over one-fifth of office-based antihypertensive drug visits between 1993 and 1996 to 30% in 1997 but remained plateau thereafter. For visits in OPDs, however, -blockers increased from 19% in 1933 to 24% in 1997 and then to 35% in 2002. Conclusions: Our results suggest that practice guidelines can impact prescribing patterns, but the degree of impact seems modest. A range of other clinical and market factors may mitigate the impact of practice guidelines.

Permalink

The Effect of Rate Regulation on Access to Supplemental Health Insurance

Presenter:

Kate Bundorf

Authors:

M. Kate Bundorf, Kosali Simon

Chair: Katherine McDonald; Discussant: Kate Bundorf Wed June 7, 2006 8:00-9:30 Room 326

In this paper, we examine the implementation of rating restrictions in the market for privately purchased health insurance supplementing nearly universal, publicly financed Medicare coverage for the elderly in the U.S. Although Medicare is a federal program, private health insurance markets are generally regulated by the states, and during the 1990s, a subset of states adopted different types of rate regulation in this market. We exploit variation in the timing and design of these policies to identify the effects of rate regulation on access to supplemental health insurance. Our analysis is based on data from the 1992-1999 Medicare Current Beneficiary Survey (MCBS). We find that the strongest versions of these laws reduced access to supplemental coverage among low risks and increased access among high risks. We also find evidence of substitution among low risks toward managed care. The results suggest that rate regulation in this market had significant effects on coverage among the elderly.

Permalink

Dirichlet-Multinomial Regression

Presenter:

Paulo Guimaraes

Authors:

Paulo Guimaraes, Richard C. Lindrooth

Chair: Richard C. Lindrooth; Discussant: Will Manning Wed June 7, 2006 8:00-9:30 Room 335

Permalink

Weak instruments in nonlinear models with endogenous regressors

Presenter:

Partha Deb

Authors:

Partha Deb

Chair: Richard C. Lindrooth; Discussant: Will Manning Wed June 7, 2006 8:00-9:30 Room 335

Permalink

Structural Econometric Evaluation of Randomized Clinical Trials

Presenter:

Bart Hamilton

Authors:

Bart Hamilton

Chair: Richard C. Lindrooth; Discussant: Will Manning Wed June 7, 2006 8:00-9:30 Room 335

Permalink

Impact of Obesity on Labor Market Outcomes of the Elderly

Presenter:

Francesco Renna

Authors:

Francesco Renna, Nidhi Thakur

Chair: Don Kenkel; Discussant: TBA Tue June 6, 2006 15:30-17:00 Room 121

Rationale: With increased longevity, the population of the retired is increasing at a fast pace. It is thus important to understand the factors that might lead to an even earlier onset of retirement than the usual 65th year of life. Since health is often cited as one such factor, it becomes imperative to understand how different constituents of health affect the labor market decisions of the elderly. We believe that obesity can impact employment decisions directly by creating functional disabilities, and/or indirectly by aggravating or actually causing other health ailments which can in turn impact employment status. While the impact of obesity along these two lines can be seen in almost all ages, amongst the elderly obesity can act as a catalyst in the labor market exit decisions, since ‘retirement’ is a concept much in their horizon than in the horizon of the younger obese.

Objectives: The objective of this paper is to estimate the direct and indirect impact of obesity on the labor market decisions of the older population.

Methodology: BMI varies from 18.5 for normal individuals to more than 40 for extremely obese. We compute the BMI for a sample of older workers, and then look at how changes in BMI affect their labor market outcomes for the next 8 years, during which time they make one of three choices: they either take an early retirement, apply for a DI/SSI benefit or change the number of hours worked. Each of these decisions is mutually exclusive, allowing for a multinomial logit approach. We also understand that since obesity is to a large extent also a self-control issue, it is possible that decisions regarding the labor market, like number of hours supplied, retirement or applying for DI assistance might be affected by the same time preferences of the individual which affect consumption decisions and subsequently affect the body weight. Thus in addition to controlling for various socio-economic characteristics of the older worker we also allow for individual heterogeneity through fixed effects. To estimate the direct and indirect impact of obesity, we control for physical or mental illnesses and their interaction with BMI. The data comes from the HRS.

Results: Preliminary analysis suggests that higher BMI increases DI applications and the probability of working part time. The effect of BMI on employment is ambiguous: BMI may decrease employment because obesity is generally associated with an increase of chronic conditions. Still, many obese might have an incentive to not retire till 65, since they do not qualify for Medicare and they are more likely to need medical assistance. We also expect to find that disability associated with physical illness may be smaller than recent statistics suggest after accounting independently for the role of obesity.

Conclusions: The paper suggests that obesity plays an important role in the late work life of elderly workers because of its effects on physical mobility and chronic conditions that determine the probability a worker will apply for DI, may need medical assistance, and ultimately withdraw from the labor force.

Permalink

Is Being Overweight A Big Deal? Body Mass Index and Health Dynamics of Elderly Americans

Presenter:

Zhou Yang

Authors:

Zhou Yang

Chair: Don Kenkel; Discussant: TBA Tue June 6, 2006 15:30-17:00 Room 121

Abstract

There is much debate about whether being overweight is a serious concern of the health among American people recently. Up to the end of 1990s the general conclusion is the relationship between Body Mass Index (BMI) and mortality rate from all reasons is in a concave shape where the highest and lowest mortality rates are observed among people in the lowest and highest BMI level, and the optimal weight associated with lowest mortality rate was determined to be between 23.5 to 24.9 in men, and between 22.0 and 23.4 in women.

However, there has been much debate recently concerning the health risk related to obesity. The center of the debate is the publication on JAMA (Flegal et al.. ,2005 ) This study shows that the relative risk of death among obese people with BMI 30 or higher is significantly higher than the normal weight people (BMI 18.5 to 25), but the relative risk of death of overweight people with BMI between 25 and 30 is the same as the normal weight people. Other researchers questioned the results and methods of Flegal’s research, and specifically pointed out that the relationship between BMI and mortality is complicated with smoking, chronic diseases and natural aging process which involves loss of bone density and muscularity, and calls for longitudinal behavior research that could investigate the changes in weight, health care behaviors and health outcomes.

This study used a dynamic behavior model to explain that weight is a choice variable of each individual that relates to her health status, health care services consumption, personal habit, and life expectancy. At the same time, weight itself changes over time in a dynamic pattern.

A jointly estimated equation system is then developed based on the theoretical framework to estimate different aspects of dynamic changes of weight and health conditions. Longitudinal data of Medicare Current Beneficiary Survey from 1992 to 1998 was used in the empirical estimation. Discrete random error term was adopted in the maximum likelihood estimation to control for unobserved individual heterogeneity. Both a short-term simulation over the entire sample over 5 year, and a long-term simulation over 35 years on one cohort from 65 were conducted. The results showed that the econometric tools adopted in this study significantly controlled the possible estimation bias caused by unobserved individual heterogeneity. The highest mortality rate is observed among the obese and underweight cohort. The elderly who are overweight does experience higher mortality than the normal weight cohort. Survivors in all age of both the obese and overweight cohort have higher disability rate than the normal weight group. The average weight among the survivors in all age cohorts converge to BMI 24 over time. The obese cohort experienced the worse health outcomes, but highest health care expenditures in total, while the normal weight cohort have the best health and lowest health care expenditures. The optimal weight for the elderly is still suggested to be lower than, and close to 25 up to age 65 in this study.

Permalink

The effect of obesity on labor market outcomes

Presenter:

Euna Han

Authors:

Euna Han

Chair: Don Kenkel; Discussant: TBA Tue June 6, 2006 15:30-17:00 Room 121

Rationale: Previous research suggests that obesity has potentially important effects on labor market outcomes. Obese people may be discriminated against by consumers or employers due to their distaste for obese people. Employers also may not want to hire obese people due to higher expected healthcare costs if the employers provide health insurance to their employees. These may result in lower wages, low likelihood of being employed and the sorting of obese people into jobs where slimness is not rewarded.

Objective: The objective of this study is to understand the effect of obesity on wages. Although other studies have linked obesity to wages, the validity of their estimation results remains questionable due to potential weaknesses in the strategies employed to control for the endogeneity of obesity. I identified the effect of obesity on wages with exogenous state-level variation in multiple variables. The over-identification of obesity with exogenous instruments will provide valid parameter estimates if the identification is supported.

Methodology: This study employed an amplified dataset based on the National Longitudinal Survey of Youth 1979 (NLSY79). NLSY79 provides ongoing panel information with a nationally representative sample of 12,686 young men and women who were 14 to 22 years old when first surveyed in 1979. I have augmented the publicly available data by obtaining confidential geographic information for individuals.

Body-mass index (BMI) was used to measure the extent of obesity. Wages were assessed separately by gender as a function of BMI splines and interactions of BMI splines with two race dummies (non-Hispanic Black and Hispanic).

This study used two-stage estimation techniques to identify the effect of obesity on wages in conjunction with individual fixed effects model. I specified an overidentified first-stage equation using exogenous state-level variation to instrument individual obesity. Instruments for obesity included the following state-level variables: cigarette prices, per capita number of restaurants, per capita number of food stores, fast-food price, cost of alcoholic drinks (inclusive of beer, wine, liquor), and cost of food.

A Heckman selection model was used to control for the selection into the labor force with the following state-level identifying instruments: unemployment rate, number of business establishments, and number of Social Security Program beneficiaries.

Results: Specification tests support the exclusion of the instruments from the main equation and the strength of the instruments in the first-stage equation. Preliminary study results indicate that an increase in BMI after being overweight has a negative effect on wage earnings for both males and females, even after adjusting for selection into the labor force.

Conclusion: The results will support the understanding of the economic cost of obesity to an individual that arise from sources other than adverse health effects. This spillover effect will increase the total cost of obesity to both individuals and society as a whole. The negative effect of obesity on labor market outcomes could raise further attention to the epidemic of obesity.

Permalink

Prenatal Health Investments: What's the Child's Gender got to do with it?

Presenter:

Aparna Lhila

Authors:

Aparna Lhila, Kosali Simon

Chair: James Burgess; Discussant: TBA Tue June 6, 2006 15:30-17:00 Room 213

How individuals invest in their own health, and how parental resource allocation decisions translate into investments in child health, have long been a concern among social scientists. This paper studies one class of parental investment decisions that combines the two, i.e. prenatal health investments that impacts maternal health, but may ultimately be viewed as investments in child health. A large body of literature documents how son preference impacts health investments in developing countries, and recent evidence suggests that parents in the U.S. are guided by son preference in their marriage, fertility, divorce and custody decisions too. In this paper we examine whether gender preference among American parents affect their prenatal health investments, an outcome which to date has not been studied in the gender preference context in either the US or abroad. We pay particular attention to testing whether male-biased child health investments persist among first generation immigrant mothers who were born in countries with a history of son preference. Data from the 1989-2001 U.S. Natality Detail Files and the 1988 National Maternal and Infant Health Survey are used to compare the prenatal investment decisions of mothers who have an ultrasound and eventually have a girl versus a boy. Information on mothers who do not have an ultrasound is used to control for innate differences that may arise due to fetal gender. We find that knowing fetal gender is female is not systematically associated with any differences in prenatal health investments among the U.S. population. What is most striking is that Indian and Chinese immigrant mothers also do not exhibit son preference in any of their prenatal investment decisions.

Permalink

Providing a Healthier State to Life: The Impact of Conditional Cash Transfers on Infant Mortality

Presenter:

Tania Barham

Authors:

Tania Barham

Chair: James Burgess; Discussant: TBA Tue June 6, 2006 15:30-17:00 Room 213

Every year more than 10 million children die from preventable diseases such as malnutrition and intestinal infections in developing countries. The majority of these deaths take place during infancy. Consequently, finding effective policies to reduce mortality among infants is a key part of the development agenda. Conditional cash transfer programs are a popular type of social investment tool designed, amongst other goals, to improve the health of children, but which may also lead to important reductions in infant mortality. However, empirically establishing causality between the implementation of conditional cash transfers and infant mortality is difficult because the death of an infant is a relatively rare event. Even large household surveys commonly do not have a sufficient number of observations to examine infant mortality. In 1997, Mexico implemented one of the first, largest, and most innovative conditional income transfer programs, Progresa. Owing to its extensiveness, Progresa provides an opportunity to test the causality of conditional cash transfers on the infant mortality rate (IMR). In this paper, I use non-experimental methods tp examine if this new policy tool reduced the rural IMR in Mexico.

Previous research on Progresa has taken advantage of a randomized treatment and control evaluation database to investigate if the program improved various aspects of children’s health. This research has shown that the nutritional status of children improved and the number of days a mother reported her child ill decreased for treatment households as compared to those from similar families that did do not receive the transfer. These findings indicate that there are some important nutritional benefits of conditional cash transfers. This paper therefore focuses on infant mortality, which is a broader and more objective measure of children’s health.

Since the Progresa randomized treatment and control database is too small to accurately estimate the impact of the program on infant mortality, I construct municipal-level panel data from 1992 to 2001. I take advantage of the phasing-in of the program over time both between and within municipalities to identify the impact of the program. The econometric model employs municipality and time fixed effects, and includes variables associated with the program phase-in rule to control for program timing bias. The analysis also explicitly controls for changes in the supply of health care in rural areas. Additionally, the identification strategy takes advantage of the fact that Progresa was not provided in urban areas prior to 2000, and uses the urban IMR to test whether unobservable municipal time-variant variables are biasing the results. We find that Progresa led to an 11 percent decline in rural infant mortality among treated households. Given the relatively high incidence of infant mortality in rural areas, and that it fell by less than one percent a year over the five years before Progresa, this is an important decline. Reductions were even higher in communities where the population all spoke some Spanish and had better access to piped water.

Permalink

When Low Birth Weight Babies Grow Up: Can Parents Buffer Their Health Shock?

Presenter:

Shin-Yi Chou

Authors:

Ming-Jen Lin, Jin-Tan Liu, Shin-Yi Chou

Chair: James Burgess; Discussant: TBA Tue June 6, 2006 15:30-17:00 Room 213

The causes and consequences of adverse birth outcomes, particularly low birth weight (LBW), have long been a major concern for social scientists. This interest has largely tended to focus on the increased risk of LBW infants suffering later developmental difficulties, which can ultimately impose substantial costs upon society. The major advances in medical technologies over recent decades is a further factor for consideration, since such advances have also led to significant improvements in the survival rates of LBW infants in both developed and developing countries.

Some of the prior studies have identified various causal relationships between birth outcomes and family background factors, such as parental education, family income and maternal behavior. However, while there has been some examination of the short-term consequences of LBW, there remains a distinct gap in the literature with regard to the examination of the long-term consequences of LBW and the interactive effects with parental education.

Viewing birth weight as an ‘input’ into the production function (or the initial endowment of human capital), the prior studies have generally established a correlation between LBW and low test scores, violent crimes committed at the age of eighteen, poor adult health and lower wages. There is also some evidence to show the existence of a strong relationship between family background and children’s educational attainment. However, there has been precious little work undertaken with regard to observation of the interactive effects of parental background and the negative health shock on the long-term outcomes of children.

The prior literature has generally failed to establish the association between adverse birth outcomes and long-term developmental outcomes for two main reasons. First of all, there are relatively few datasets containing the necessary information on adverse birth outcomes in conjunction with long-term developmental outcomes on the same study sample, and secondly, it is even more difficult to obtain such a linkage at national level. The studies have therefore tended to rely heavily upon small and selective samples; clearly therefore, it may not be possible to generalize the findings to other contexts.

In this study, we set out to combine several unique datasets in order to examine the long-term outcomes of LBW and the effects of the interaction between such outcomes and parental education. Our main dataset is annual birth certificate records, with the birth information of the study sample then being matched to the College Entrance Examination and High School Entrance Examinations files to determine long-term educational achievements. Our main goals in using these unique datasets are to identify the long term effects of LBW and to investigate the interactive effects of LBW and parental education. Although our results show that LBW has significant negative long-term effects on educational attainment, they also suggest such negative effects can nevertheless be mitigated by parental education.

Permalink

A Review and Synthesis of the Cost and Benefits of Health Facilities Regulation in the U.S.

Presenter:

Christopher Conover

Authors:

Christopher Conover, Ilse Wiechers

Chair: Joel Hay; Discussant: TBA Tue June 6, 2006 15:30-17:00 Room 225

Rationale: No one previously has developed a comprehensive estimate of the costs and benefits of health facilities regulation.

Objectives: The central objective of this paper is to develop a comprehensive estimate of the costs and benefits of health facilities regulation derived by summing the results of fine-grained cost estimates in 18 different domains of health facilities regulation; a companion objective is to identify domains of health facilities regulation in which regulation currently appears not to be cost-effective.

Methodology: A formal literature search was conducted for each separate domain of health facilities regulation, including regulations focused on access to care (e.g., EMTALA), cost control (e.g., certificate of need) and quality (e.g., nursing home accreditation and licensure. For each domain, a synthetic estimate of regulatory costs is obtained using a standardized procedure that systematically assembles evidence from the literature regarding government regulatory costs (such as monitoring and enforcement costs) and compliance costs (including both industry costs and patient time losses), and indirect costs such as costs associated with health losses (morbidity and mortality losses, all monetized using a common value of statistical life year) or increases in uninsured risk (monetized using a standardized estimate of the external cost of being uninsured and the monetized value of the increased mortality risk faced by the uninsured). Corresponding monetized estimates of benefits were developed based on empirical evidence of benefits derived from the literature, including efficiency gains, quality improvements or reductions in uninsured risk. Lower and upper bound estimates were derived using minimum and maximum parameter estimates for the various components used to calculate regulatory costs.

Results: The preliminary base case result for the cost of health facilities regulation in the U.S. is $47.7 billion (2002 $); the minimum estimate is $23.7 billion and the upper bound estimate is $277.6 billion. [final estimates updated to 2004 $ to be available in early February]

Conclusions: The domains of health facilities regulation having the highest net cost include hospital accreditation/licensure, hospital uncompensated care pools and Clinical Laboratory Improvement Act (CLIA). These areas warrant closer examination to determine how regulatory objectives might be achieved more cost-effectively.

Permalink

Estimating the Cost of Capital for Pharmaceutical, Biotechnology, and Medical Device Firms

Presenter:

Scott Harrington

Authors:

Scott Harrington

Chair: Joel Hay; Discussant: TBA Tue June 6, 2006 15:30-17:00 Room 225

Author: Scott Harrington (harring@wharton.upenn.edu), Health Care Systems, Wharton School, University of Pennsylvania

Title: Estimating the Cost of Capital for Pharmaceutical, Biotechnology, and Medical Device Firms

Rationale: The financing, investment, and risk management decisions of pharmaceutical, biotechnology, and medical device firms are fundamentally important to the development and availability of innovative treatments to enhance health outcomes and the quality of life. A full understanding of these decisions is also important for the design and administration of government review and approval of new compounds and devices.

Objectives: This study provides new insight into health care research and development decisions by investigating factors that influence the cost of equity capital for publicly-traded pharmaceutical, biotechnology, and medical device firms, a subject that has received relatively little attention in the literature. The primary objective is to explore analytically the link between firms’ characteristics and risk factors that affect the cost of capital.

Methodology: According to conventional finance theory, the risk characteristics of firms and their associated effects on the cost of capital are of central importance in making optimal investment decisions. The study draws from corporate finance theory to develop an analytical model of how firm-specific characteristics and the “real option” features of pharmaceutical, biotechnology, and medical device firms’ investments affect their cost of capital and then investigates the empirical relation between those factors and capital costs. Two frameworks are employed for estimating firms’ cost of capital: (1) the traditional capital asset pricing model (CAPM), and (2) the empirically-driven three risk-factor model of Fama and French.

Results: The study provides evidence concerning three main questions:

1. How does the cost of capital vary across pharmaceutical, biotechnology, and medical device firms and within firms in each sector?
2. How do firm characteristics, including levels of R&D spending, product diversification, real option features, and primary care versus specialist focus affect firms’ cost of capital through their effects on risk factors that are priced in capital markets?
3. Can evidence on the relationship between firm characteristics and firm-level cost of capital estimates be used to develop reasonable estimates of the cost of capital for classes of products for use as inputs to both expected net present value analysis and real options valuation methods?

The results should be useful to corporate managers in developing efficient investment and financing strategies and relevant to public policy related to such investment.

Permalink

More Predictive Modeling of Total Healthcare Costs Using Pharmacy Claims Data: Adherence and Boosted Regression

Presenter:

M. Christopher Roebuck

Authors:

M. Christopher Roebuck

Chair: Joel Hay; Discussant: TBA Tue June 6, 2006 15:30-17:00 Room 225

OBJECTIVE A variety of disease state classification systems derived using claims data are used in risk stratification and predictive modeling applications. Algorithms based solely on pharmacy claims data have the advantages of timeliness, cleanliness, and availability, while still being robust to predicting prospective healthcare outcomes and costs relative to their integrated medical and pharmacy counterparts. Following on Powers et al. (2005), this study expands Pharmacy Health Dimensions (PHD) to include controls for patient adherence to drug therapy. The study also evaluates the use of boosted regression as an alternative to other econometric approaches for predicting commonly right-skewed and leptokurtotic healthcare cost data.

METHODS: Using 2003 and 2004 data from a large health plan (N=369,985), PHD is used to identify participants having any of five diseases: diabetes, congestive heart failure, asthma, hypercholesterolemia, and hypertension. A split-sample design is employed to train and validate the predictive models. For the base model, total healthcare costs in 2004 are estimated as a function of age, gender, 2003 pharmacy costs, and comorbidities using PHD. Subsequent models add two variables of patient adherence to drug therapy, a commonly used measure of compliance, Medication Possession Ratio (MPR), and a novel gauge of persistency, Maximum Gap in Therapy. It is hypothesized that MPR would be negatively related to future total healthcare costs, while Maximum Gap in Therapy would be positively associated with future total healthcare costs. Furthermore, the magnitude of these effects should vary substantially by disease state. Several econometric modeling techniques are explored included ordinary least squares (OLS), log-OLS, two-part modeling (with a generalized linear model second part), and boosted regression. Standard metrics of predictive model fit and accuracy are reported.

RESULTS: As expected, the inclusion of the two measures of adherence increased the proportion of explained variation in all five of the disease state models. Furthermore, MPR was significantly (p<0.10) and negatively related to future total healthcare costs in all models except for hypercholesterolemia and hypertension. Maximum Gap in Therapy was positively associated with future total healthcare costs in all model (p<0.10). As in the prior study (Powers et al., 2005), results from OLS regressions were comparable to those of log-OLS and two-part modeling approaches in terms of their validation R2, positive predictive value, and specificity. Boosted regressions, however, provided the most accurate predictive model of future total healthcare costs for several of the disease states.

CONCLUSION: Predictive modeling of future total healthcare costs using pharmacy claims only offers several advantages over using integrated medical and pharmacy data. The predictive power of basic models of demographic and disease classification can be enhanced by adding measures of patient compliance and persistency. The magnitude and significance of the relationship between adherence to drug therapy and future total healthcare costs varies substantially by health condition.

REFERENCE: Powers, C.A., C.M. Meyer, M.C. Roebuck, and B.Vaziri. 2005. Predictive Modeling of Total Healthcare Costs Using Pharmacy Claims Data: A Comparison of Alternative Econometric Cost Modeling Techniques. Medical Care 43(11): 1065-1072.

Permalink

Effect of proximity to death on participation in the long-term care market

Presenter:

France Priez

Authors:

France Priez, Sally Stearns

Chair: Willard Manning; Discussant: TBA Tue June 6, 2006 15:30-17:00 Room 226

Increasing longevity raises concerns about the growth in the per capita demand for health care services. However, the aging of the population may not increase nursing home and paid home care use to the extent that proximity to death rather than age determines use at the end of life. As the elderly get closer to death, their use of long-term care (LTC) changes, not because they become older, but because their health deteriorates irreversibly. This study examines the effect of proximity to death on the participation in the nursing home and paid in-home care markets and how it interplays with availability of informal care that is a substitute to market-based LTC.

Proximity to death is hypothesized to increase the use of LTC. But availability of spousal care and support from children is hypothesized to mitigate this effect. Linear probability models estimate the effect of being within two years of death on the probabilities of nursing home and paid home care use. Proximity to death is interacted with availability of spousal care (being married) and available support from children (co-residence with an adult child). Unobserved heterogeneity is addressed by using individual and time fixed effects. The potential simultaneity between proximity to death, residing with an adult child and LTC use is addressed by using instrumental variables. Instruments for co-residing with an adult child are a set of characteristics of the person’s children. No valid instruments were found for proximity to death. The analysis is conducted on the 1993-2002 Health and Retirement Study data, the target sample being the 70+ elderly. The final five-wave data set has 27,879 observations.

Results indicate that proximity to death significantly increases the likelihood of nursing home and paid home care use, while age 85+ has a positive but smaller effect. Furthermore, proximity to death has significantly different effects by marital status for both types of LTC, and by co-residence with children for nursing home use only. Among married elderly, being within two years of death significantly reduces nursing home use, from 1.0 to 0.2 percent. This effect is similar among elderly co-residing with children. In contrast, for non-married elderly, nursing home use increases significantly from 4.9 to 16.6 percent in the last two years of life. For paid home care, proximity to death has a significant effect only among non-married elderly, the likelihood of use going from 6.9 to 9.6 percent. Proximity to death does not significantly affect paid home care use among married elderly or elderly co-residing with an adult child.

Proximity to death affects participation in the LTC market, but it plays a significantly different role whether the elderly persons have spousal care or children support available to them. The aging of the population may not substantially increase LTC use, not only if proximity to death rather than age is one of the main determinants of use, but also if the proportion of married elderly continues to increase, as observed over the last decade.

Permalink

Spatial Analysis of Elderly Access to Primary Care Services: Informing the Debate on Physician Shortage in the U.S.

Presenter:

Lee Mobley

Authors:

Lee Mobley, Elisabeth Root, Luc Anselin, Nancy Gracia, Julia Koschinsky

Chair: Willard Manning; Discussant: TBA Tue June 6, 2006 15:30-17:00 Room 226

High rates of admissions for Ambulatory Care Sensitive Conditions (ACSCs) are signals of poor preventive care utilization. This paper examines the influence of geographic or market-level supply and demand factors on market-level rates of ACSC admissions among Original Medicare (FFS) beneficiaries in the latter nineties (1998-2000). This period follows implementation of the Balanced Budget Act of 1997, which reduced the level of prospective payments and introduced limits to home health care visits and other aspects of preventive care services for the elderly.

The conceptual model assumes that local area market conditions, serving as interventions along the pathways to healthcare utilization, can impact outcomes. Using natural markets defined by The Health Resources and Services Administration’s Primary Care Service Area (PCSA) Project, spatial regression is used to analyze admission rates for Ambulatory Care Sensitive Conditions (ACSCs) among all elderly FFS beneficiaries 1998-2000, controlling for disease severity using detailed information from the MEDPAR claims.

A spatial spillovers model is estimated to account for endogenous provider and beneficiary behavior across Primary Care Service Areas. Spatial multiplier effects are found to be quite large, suggesting that OLS estimates of marginal impacts from explanatory variables would be overstated by about 50 percent in magnitude of effect. GeoDa and R spatial analysis software, with additional code written in PYTHON, are used to estimate an instrumental variables version of the spatial lag model and conduct mis-specification and goodness-of-fit tests. R and Python are required to obtain robust estimates of parameters because the dependent variable does not meet the assumptions under the Maximum Likelihood estimation done in GeoDa (the dependent variable is highly skewed). Reported standard errors are robust to heteroskedasticity.

Our evidence suggests that elderly living in impoverished rural areas, or in sprawling suburban places, are about equally more likely to be admitted for ACSCs. Greater availability of physicians does not seem to matter, but greater prevalence of non-physician clinicians and international medical graduates, relative to traditional physicians, does seem to reduce ACSC admissions, especially in poor rural areas. The relative importance of these non-traditional physician groups in providing primary care in areas of greatest need can inform the ongoing debate regarding whether there is an impending shortage of physicians in the US.

Permalink

Spend It While You Can Still Enjoy It: Health, Longevity, Aging, and Consumption in the Life Cycle

Presenter:

Lauren Olsho

Authors:

Lauren Olsho

Chair: Willard Manning; Discussant: TBA Tue June 6, 2006 15:30-17:00 Room 226

Simple life-cycle models based on the permanent-income hypothesis predict that individuals will smooth consumption over their lifetimes. However, empirical studies consistently find that consumption levels fall steadily late in life. In this paper, I examine declining health status as one potential factor contributing to this pattern. I estimate a life-cycle model treating health as a form of human capital investment, and predict substantial decreases in consumption after age 50 that are consistent with empirical observations. When health and consumption are complements, decreasing health capital results in falling marginal utility of consumption with age. Thus rational agents will consume at higher levels when they are younger and healthier.

This effect is particularly pronounced among low-income groups, whose health deteriorates more rapidly than the health of their higher-income counterparts. A faster rate of health deterioration results in a steeper decline in marginal utility of consumption late in life among the poor. Additionally, larger investments in health increase longevity in high-income groups relative to low-income individuals, leading to lower rates of consumption and higher rates of savings for the rich across all years. I find that this disparity in longevity, paired with differences in the rate of health deterioration, can largely account for the divergence in life-cycle consumption patterns between the top and bottom income and asset quartiles of the United States population.

The theoretical framework is a dynamic version of Grossman’s seminal 1972 health capital model. While Grossman’s work is frequently cited as a motivation for empirical work in health economics, there have been few if any serious attempts to quantitatively solve the model and closely examine its key implications. In the version of the model presented here, health exogenously deteriorates at an increasing rate with age. To counteract this health deterioration, individuals can invest a portion of their assets into health production each period. If the total health stock falls below some minimal level, the individual dies with certainty. Health investment increases the marginal utility of consumption within each period, and extends the total length of life. Additionally, health is a productive good in the sense that it increases the amount of time an individual can spend in the labor market, thereby augmenting earnings potential. A bi-directional relationship between health and wealth thus emerges: higher wealth induces greater health investment and increased longevity, while better health increases labor market productivity. This feature of the model makes it particularly suited for examining health, wealth, and consumption inequalities.

I estimate the model using data from the Health and Retirement Survey panel, 1992-2002. I construct a health index based on participants’ answers to a series of questions regarding mental health and physical limitations, and find that deteriorating health can account for an annual decline in median consumption levels of about 2 percent after age 50, as compared to an actual drop of around 3 percent. Failing health accounts for a 3 percent annual decline in consumption for individuals with income and assets below the 25th percentile, in contrast to a decrease of only 1 percent per year among individuals with income and assets above the 75th percentile.

Permalink

Surplus Appropriation from R&D and Technology Assessment Procedures

Presenter:

Tomas Philipson

Authors:

Tomas Philipson

Chair: Catherine McLaughlin; Discussant: Daniel Eisenberg Tue June 6, 2006 15:30-17:00 Room 235

We discuss the difference between dynamic and static technology assessment criteria in healthcare. We argue that popular assessment criteria going under the rubric of “cost-effectiveness” often concern maximizing consumer surplus, which many times is consistent with maximizing static efficiency after an innovation has been developed. However, dynamic efficiency concerns aligning the social costs and benefits of R&D and is therefore determined by how much of the social surplus from the new technology is appropriated as producer surplus. We estimate that for the HIV/AIDS therapies that entered the market from the late 1980’s onwards, producers appropriated only 5% of the social surplus arising from these new technologies. We show how to translate standard findings of cost-effectiveness to estimates of innovator appropriation for standard studies of over 200 drugs, and find that these studies implicitly support a low degree of appropriation as well. Despite the high annual costs of drugs to patients, the low share of social surplus going to innovators raises concerns about advocating cost-effectiveness criteria that would further reduce appropriation by innovators, and hence further reduce dynamic efficiency.

Permalink

Is Managed Care Restraining the Adoption of Technologies by Hospitals?

Presenter:

Janice Seinfeld

Authors:

Nuria Mas, Janice Seinfeld

Chair: Catherine McLaughlin; Discussant: Loren Baker Tue June 6, 2006 15:30-17:00 Room 235

Authors: Núria Mas (IESE Business School, Spain) and Janice Seinfeld (Universidad del Pacífico, Peru)

Title: Is Managed Care Restraining the Adoption of Technologies by Hospitals?

Rationale: As health care costs keep increasing, cost-control mechanisms such as the ones introduced by managed care, have become more widespread and it is crucial to understand their implications for the health care market and, in particular, for the adoption of new technologies, which are one of the main determinants of health care cost growth.

Objectives: This paper examines the relationship between managed care activity and hospitals´ technology adoption. The existing studies on this subject have been limited to only a few technologies over short periods of time. This paper extends previous work in several important dimensions: first, we examine thirteen different technologies from 1982 to 1995. Second, we distinguish between these that diffused in the eighties, when managed care was rare in the US, from these that diffused in the nineties, when managed care was already a widespread phenomenon, since we expect managed care effect to be stronger for the second group of technologies. Finally, we rank technologies according to their cost-benefit ratio since, given that managed care gives strong incentives to minimize costs, we expect its effect to be stronger for these technologies with high cost-benefit ratios.

Methodology and Results: Managed care may affect hospital´s decisions to take on new technologies by imposing financial pressure on providers, changing the incentives associated with the utilization and purchase of new technology equipment. We use a hazard rate model to investigate whether high levels of managed care market share are associated with a decrease in medical technology adoption during the period 1982-1995. We analyze annual data on 5,390 US hospitals regarding the adoption of 13 different technologies from the American Hospital association (AHA). After adjusting for hospital characteristics, demographics, regulation and local market characteristics, we find that managed care has a negative effect on hospitals´technology acquisition for each of the thirteen technologies in our study, and this effect is stronger for technologies diffusing in the 1990s, when managed care sector was at its largest. We also take into account that cost-benefit analysis is one of the main dimensions considered by hospitals when deciding about the adoption of new technologies. We find that managed are enrollment has a significantly more negative effect on the adoption of technologies with higher cost-benefit ratios.

Conclusions: the negative effect of managed care on technology adoption may have important policy implications, such as its ability to slow down health care costs growth. Our results provide evidence that managed care contributes to costs savings in the health care market by limiting availability of technologies, especially of the more expensive ones.

Permalink

Does Liability for Medical Malpractice Drive Health Care Costs and Technology Adoption?

Presenter:

Seth Seabury

Authors:

Seth Seabury

Chair: Catherine McLaughlin; Discussant: Robert Town Tue June 6, 2006 15:30-17:00 Room 235

Liability for medical malpractice has been the focus of enormous controversy in recent years. Many have argued that it artificially inflates the price of medical services and limits access to health care. Our paper investigates the relationships between liability for medical malpractice, medical costs, and the adoption or utilization of new health care technologies. These relationships are complex, with causality running in different directions. Growth in litigiousness and liability could drive up the cost of receiving health care, as many charge. On the other hand, increases in the cost of medical care from other quarters could mechanically inflate the cost of recovering from medical errors. In this way, medical cost growth could be the cause of increased malpractice liability payments. Similarly, rapidly rising malpractice costs could stunt the adoption of new technology and make health care providers excessively risk-averse. On the other hand, malpractice risk may encourage the utilization of accepted technology and treatment methods; this is an important policy goal of the malpractice regime.

We employ several identification strategies to sort through these channels of causation and examine whether malpractice risk substantially contributes to health care costs. We first exploit county-level variability in the “generosity” of jury verdicts. Juries in certain parts of the country have exhibited a willingness to award larger damages for substantially similar transgressions. Jury generosity affects local malpractice claims levels, but plausibly has no direct effects on medical costs or technology adoption/utilization. We also exploit variability in non-malpractice measures of medical costs, such as practice expenses and the relative wage compensation paid to highly skilled workers. These arguably affect medical costs without having direct effects on malpractice payments. Finally, we exploit variation in malpractice premia and technology adoption across different specialties. These analyses will provide us with important evidence about how medical malpractice liability really impacts the functioning of the health care system.

Permalink

A Multimodal Model of Health Care Prices

Presenter:

Benjamin Craig

Authors:

Benjamin Craig, Partha Deb

Chair: Dave Vanness; Discussant: Christopher Hollenbeak Mon June 6, 2006 15:30-17:00 Room 235

In the demand for health care, price is a central determinant, yet the modeling of out-of-pocket prices is poorly understood. This paper studies the multimodality of out-of-pocket prices using a mixture of Gaussians that allows for the estimation of modal shifts and variation around the modes. We describe how prices may vary continuously due to market fluctuations, and may shift between modes because of contracts, search costs, bulk purchasing, and price discrimination. A better understanding of the endogeneity of out-of-pocket prices not only improves our ability to identify the demand for health care, these models better characterize the financial burden of poor health. To illustrate this point, we examine 1992-2002 data from the Medicare Current Beneficiary Survey, specifically the bimodality of medication prices among seniors.

Permalink

The Impact of Malpractice Liability Claims on Physician Outcomes and Practice Patterns

Presenter:

Gilbert Gimm

Authors:

Gilbert Gimm

Chair: Edward Norton; Discussant: Hua Wang Tue June 6, 2006 15:30-17:00 Room 309

ABSTRACT

Author: Gilbert W. Gimm, Ph.D. (ggimm@mathematica-mpr.com), Health Researcher, Mathematica Policy Research, Washington, DC.

Title: “The Impact of Malpractice Liability Claims on Physician Outcomes and Practice Patterns: Evidence from the State of Florida (1992-2000)”

Key Terms: malpractice claims, physician behavior, outcomes, quality

Objectives: Since the late 1960s, a dramatic rise in medical malpractice premiums has sparked both debate and concern among physicians and health policymakers. However, empirical evidence is limited to a few studies, which rely on market-level variation, so the impact of claims on the individual response of physicians is not well understood. This study examines whether malpractice claims have an impact on individual physician behavior, with respect to delivery volume, adverse outcome rates, and C-section rates. It also analyzes the factors that contribute to the likelihood of physician exit.

Methodology: This study uses a micro-level panel data set, which links inpatient deliveries in the state of Florida to malpractice claims from 1992-2000, with unique physician identifiers. The identification strategy for the impact on delivery volume is a fixed-effect OLS levels regression that controls for time-invariant, unobservable characteristics of individual physicians. Explanatory claim variables include whether the closing of a claim occurred in a specific year, as well as the award size and severity index measure associated with a malpractice claim. Other control variables include patient characteristics and medical risk factors. Finally, a binary logistic regression is used to estimate the likelihood of physician exit.

Results: Physicians perform 6-10 fewer inpatient deliveries per year in response to the closing of a malpractice claim, after controlling for fixed effects. However, malpractice claims do not appear to have a significant impact on the adverse outcome rate. Also, an incident occurrence is associated with a higher C-section rate among older physicians. Finally, physicians with a malpractice claim award of $500,000 or more have a significantly greater likelihood of exit from performing inpatient deliveries altogether. These empirical results suggest that malpractice claims do have a significant effect on individual physician practice patterns.

Disclosure Statement: This research study was funded by an AHRQ dissertation grant number 1-R36-HS014515-01.

Permalink

The Impact of Liability on the Physician Labor Market

Presenter:

Mark Showalter

Authors:

Eric Helland, Claremont McKenna, Mark Showalter

Chair: Edward Norton; Discussant: Jinny Kim Tue June 6, 2006 15:30-17:00 Room 309

Medical malpractice reform is a contentious issue. Despite the fact that states have implemented a variety of reforms over the past three decades, the implications of these reforms are still in dispute. Reductions in liability exposure have been shown to reduce malpractice premiums (Born and Viscusi, 1998), but the extent and magnitude of physicians’ direct behavioral responses are unclear. Some researchers have found evidence of a large physician response. Perhaps the best known empirical papers are by Kessler and McClellan (1996, 2000) which examine how health expenditures for Medicare recipients vary with changes in state liability reforms. They find that up to nine percent of expenditures on treatment for heart disease and heart attacks can be attributed to excessive care due to physicians practicing ‘defensive’ medicine, defined as care that would not be done in the absence of a liability risk. However, other researchers have tended to find smaller effects. For example, Dubay et al (1999, 2001) examine data on cesarean section procedures and find relatively small effects attributable to liability reform. Other work by Sloan et al. (1995) and the Congressional Budget Office (2003) also find little effect of liability reform on expenditures.

Our study reexamines the issue of liability reform and its implications for physician behavior. We develop a model in which physicians are unable to fully insure against liability risk, opening the possibility for ‘defensive’ medicine. We then use data on actual physician behavior from the Physician Practice Costs and Income Survey (PPCIS, 1983 and 1988). This data includes physician work hours, insurance level, patient mix, treatment practices, practice structure, among other variables, and we match this data with information on state-level changes in liability law. We investigate changes in physician behavior that correspond to changes in the legal climate. These two nationally representative surveys bracket a period of substantial state-level reform which provides the variation for our empirical strategy.

We use several measures of liability reform, but one measure unique to this paper is derived from malpractice litigation data for Florida from 1980 to 1985. This data allows us to estimate the impact of a damage cap on the distribution of awards by specialty and state. We estimate the impact of liability reform on hours worked per year, the hours spent working in emergency rooms, retirement, number of patients treated and the types of patients treated (i.e. uninsured, Medicaid or Medicare patients). We estimate the impact semi-parametrically using a full set of state-specialty controls. Our preliminary results suggest that doctors reduce their hours worked overall, reduce their hours worked in the emergency room and alter their patient mix in ways that are consistent with avoiding liability exposure. The results also suggest that physicians are unable to fully insure against liability risk through the purchase of malpractice insurance.

Permalink

Medical Malpractice: Examining its Effect on Hospital Efficiency

Presenter:

Shalini Bagga

Authors:

Shalini Bagga

Chair: Edward Norton; Discussant: Partha Deb Tue June 6, 2006 15:30-17:00 Room 309

Authors: Shalini Bagga (sbagga@tulane.edu); M. Mahmud Khan (khan@tulane.edu); Praveen Dhankhar (pdhankm@tulane.edu)

Title: Medical Malpractice: Examining its Effect on Hospital Efficiency

Rationale/Objective: There is a growing fear among the physicians that they will not be able to buy medical malpractice liability insurance due to the sharp increases in the malpractice premiums nationwide. A part of the increase in health care expenditures can be attributed to an increase in hospital expenditure. One way of decreasing expenditure on hospitals would be to increase the efficiency of hospitals. We attempt to look at one such dimension in this paper: how does malpractice pressure impact the efficiency of our health-care system, in particular, the efficiency of our hospitals? This side of the story has been neglected so far in the literature, and our paper attempts to fill this gap and introduce a new perspective.

Methodology: Data was combined from four sources: Nationwide Inpatient Sample, American Hospital Association’s Annual Survey, and Medicare Cost Reports, for the year 2001; National Practitioner Data Bank’s Public Use File for years 1998, 1999, and 2000. Malpractice variables were entered as a lag. Efficiency was estimated through the stochastic frontier analysis. We explicitly controlled for output heterogeneity and quality of the services provided: teaching status, location, ownership, region, insurance status, mortality. Two different types of malpractice variables were used: frequency of claims and severity of claims. Severity was in turn represented by: mean payments and median payments.

Results: Our results indicate that malpractice does affect the efficiency of hospitals. In fact, greater malpractice pressure, as given by the mean and median payments, increases technical efficiency of hospitals. The effect is larger for median payment. The sign for frequency of payments is also in the right direction, though insignificant. A point to be made: the existence of ‘corporate shield’ results in underreporting of cases in the NPDB. Therefore, the malpractice results that we present in this paper are going to represent the lower bounds in terms of their effects.

Discussion: The recent focus by the American Medical Association and physicians about the dramatic increases in medical malpractice insurance premiums, and their suggestion of a cap on non-economic damages, deserves a closer look. According to Baicker and Chandra (2004), increases in premiums are not affected by past or present malpractice payments, but may increase due to other unrelated factors. Chandra, Nundy, and Seabury (2005) find that the rising cost of medical services may explain the bulk of the growth of “compensatory awards”. They also find that the greatest ten percent of the malpractice payments have grown at a smaller pace than the average payment for the years 1991 and 2003. This means that the “medical malpractice crisis” is not necessarily fueled by the growth in malpractice payments. Furthermore, malpractice pressure actually forces our hospitals to be technically more efficient. This implies that existence of the medical malpractice system is beneficial, and its strength should not be diluted by either putting caps on non-economic damages or by decreasing the statute of limitations.

Permalink

The Effects of Specialty Hospitals on General Hospital Operating Margins, 1997-2003

Presenter:

John Schneider

Authors:

John Schneider, Robert Ohsfeldt, Michael Morrisey, Pengxiang Li, Bennet Zelner, Thomas Miller

Chair: Kathleen Carey; Discussant: Ginger Jin Tue June 6, 2006 15:30-17:00 Room 313

Authors: John E. Schneider (1,2) (john-schneider@uiowa.edu); Robert L. Ohsfeldt (3); Michael A. Morrisey (4); Pengxiang Li (1); Bennet A. Zelner (5); and Thomas R. Miller (1). [(1) Department of Health Management and Policy, University of Iowa; (2) Center for Research in the Implementation of Innovative Strategies in Practice (152) Iowa City VA Medical Center; (3) Department of Health Policy and Management, School of Rural Public Health, Texas A&M Health Science Center; (4) Department of Health Care Organization and Policy, and Lister Hill Center for Health Policy, University of Alabama Birmingham; (5) Haas School of Business, University of California Berkeley]

Title: The Effects of Specialty Hospitals on General Hospital Operating Margins, 1997-2003

Rationale: A recently expired moratorium on Medicare payments to new specialty hospitals was predicated in part on concern that the financial stability of general hospitals was being eroded by competition from specialty hospitals, thereby impairing their ability to cross-subsidize indigent care. If this were the case, general hospital operating margins in markets with specialty hospitals are expected to be lower, controlling for other factors affecting operating margins.

Objectives: Using a database of U.S. general hospitals and a sample of specialty hospitals from 1997 to 2003, we conduct econometric analyses of general hospital operating margins in markets with and without specialty hospitals.

Methodology: Data is from Medicare HCRIS Cost Reports, the American Hospital Association, a survey of specialty hospitals, and the Area Resource File. These sources were combined to form a panel data set of approximately 3000 hospitals over the seven year time period 1997 to 2003. Four different model specifications were compared: (1) exogenous entry with hospital random effects; (2) exogenous entry with hospital fixed effects; (3) endogenous entry with hospital fixed effects, where lagged mean county-level general hospital profit margin and certificate of need (CON) status serve as the instruments; and (4) endogenous entry with hospital fixed effects, using only CON status as an instrument.

Results: Counties with at least one specialty hospital consistently had higher mean operating margins than counties without specialty hospitals (p ≤ 0.05). In all four models, including the endogenous entry models, the presence of one or more specialty hospitals in the market was associated with higher general hospital profit margins (p ≤ 0.05).

Conclusions: Contrary to the conjecture that entry by specialty hospitals erodes the overall operating profits of general hospitals, general hospitals residing in markets with at least one specialty hospital have higher profit margins than those that do not compete with specialty hospitals.

Permalink

The Influence of Competitors' Performance on Hospital Efficiency

Presenter:

Vivian Valdmanis

Authors:

Gary Ferrier, Vivian Valdmanis

Chair: Kathleen Carey; Discussant: Katherine Ho Tue June 6, 2006 15:30-17:00 Room 313

Rationale: The focus to date of “peer effects” and “social interactions” has been on individuals. One important exception to this is the idea of “yardstick competition”-a regulatory scheme under which the performance of firms is compared in order to determine the appropriate level of prices and subsidies (see Schleifer, 1983).

In this paper we examine the role of “peer effects” on firm behavior. Specifically, we estimate the influence of a firm’s competitors’ efficiency levels on a firm’s own level of efficiency. In sports one often hears that a team “plays to the level of the competition;” we ask whether the same is true of hospitals. Given the evolving nature of competition in the hospital industry, this is an important question. .

Objectives: To determine the role of market interactions and peer effects in explaining variations in hospital efficiency.

Methodology: Hospital efficiency is measured by Farrell input-oriented technical efficiency scores whereby all inputs are proportionately scaled back until the observed levels of output are still just feasible. The input-orientation, rather than an output-orientation, is used because it is consistent with the goal of cost containment in the hospital industry. The efficiency scores are calculated using the linear programming approach of data envelopment analysis (DEA). The technical efficiency scores thus obtained were then used as the dependent variable in a regression model that seeks to explain variability in efficiency scores across hospitals. The focus of this second stage of the analysis is the effect of the average performance of peers on a hospital’s own performance. Thus the following regression was performed: (4) where Eff-i is the mean efficiency of hospital i’s peers and Xi includes a variety of other factor that might be expected to influence a hospital’s efficiency level. The other factors include the radius of a hospital’s market, the number of competitors within the radius, the HHI of each market, ownership status, network/alliance membership, teaching status, the share of Medicaid patients among all patients served by a hospital, and hospital size (proxied by number of beds).

Results: 38% of the variation in technical efficiency is explained by the model. The traditional measures of competition-the number of competitors and the Herfindahl-Hirschman Index-appear to have little effect on a hospital’s efficiency. Peer efficiency; however, does appear to influence how efficiently hospitals operate. A 10% increase in peers’ efficiency would result in a just over a 2% increase in a hospital’s own efficiency.

Other influences on efficiency include ownership form-both not-for-profit and public hospitals appear to be more efficient than their for-profit counterparts, teaching status-teaching hospitals are more efficient than non-teaching hospitals, and payer mix appears to matter as well-as the share of Medicaid patients increases, hospital efficiency declines. The regressions based on the radii from which 75% of a hospital’s patients are drawn. Qualitatively similar results were obtained when data based on the 90% radii were used instead.

Permalink

Single Specialty Hospitals and Competition in the Hospital Industry

Presenter:

Kathleen Carey

Authors:

Kathleen Carey, Gary Young, James Burgess

Chair: Kathleen Carey Tue June 6, 2006 15:30-17:00 Room 313

OBJECTIVES. Single specialty hospitals (SSHs) are a small but rapidly growing genre among U.S. hospitals. These providers, the majority of which are owned by physicians who refer patients to them, have raised considerable controversy over whether they promote economic efficiency though specialization or engender unfair competition. We undertake a broad based approach to investigating several benchmarks that jointly assess whether physician-owned SSHs are enhancing the competitive process in the U.S. hospital industry. These include cost efficiency, economies of scale and scope, and pricing for private payers.

DATA AND MEASURES. The primary databases are the Medicare Cost Reports, the American Hospital Association Annual Survey, and state discharge abstracts for the three states in which physician-owned SSHs are most heavily concentrated: Texas, California and Arizona. We examine three specialties: cardiac, orthopedic, and surgical hospitals for the years 1998 through 2004. For comparison, we choose competitor hospitals defined as those offering the same services and located in the same markets (identified as Hospital Referral Regions in the Dartmouth Atlas of Health Care). Key variables include measures of hospital total cost, discharges, length of stay, outpatient visits, case-mix, severity, quality, input price, disproportionate share of poor patients, teaching, ownership, and system affiliation.

ANALYSES. The basic analytic structure is a multiple output hospital total variable cost function estimated using stochastic frontier regression techniques for longitudinal data. This method will allow us to produce hospital specific inefficiency measures. We develop several groupings by which we compare inefficiency differences among SSH and competitor hospitals using parametric and non-parametric test of significance, and taking into account for-profit status, and system membership. We also estimate the cost functions for SSH and competitors separately in order to calculate measures of scale and scope economies for each hospital type. Finally, we construct average SSH vs. competitor hospital price indexes based on prevalent DRGs, standardized to control for patient characteristics, case-mix, and severity. We perform several analyses of mean price differences across various hospital groupings using parametric and nonparametric tests of significance.

SIGNIFICANCE. As of December 2003, Congress has declared a moratorium on physician referrals of Medicare and Medicaid patients to SSHs. The hospital industry is strongly advocating that the moratorium be made permanent. Yet in a March 2005 Report to Congress in support of the moratorium, MedPAC explicitly stated that it does not want to put an end to development of SSHs before gaining a fuller understanding of their efficiency benefits. Yet to date, no one has established the knowledge base that would support a laissez-faire policy. This research offers the first solid evidence on this matter by informing Congress as well as hospitals, judges, and policy makers on the economic logic of organization of hospital services around single specialties.

Permalink

Hospital Reimbursement and Treatment Intensity

Presenter:

Richard Lindrooth

Authors:

Richard C. Lindrooth, Gloria J. Bazzoli, Jan P. Clement, Mei Zhao

Chair: Gloria Bazzoli; Discussant: Jack Zwanziger Tue June 6, 2006 15:30-17:00 Room 325

Recent research suggests that a ‘new medical arms race’ may be beginning among hospitals due to the shift to less selective contracting and payment methods that do not motivate efficiency and also the emergence of new competitors (e.g., physician-owned specialty hospitals and diagnostic centers). In the context of the shift, we examine the extent to which a hospital’s service offerings explain its financial performance and test whether the importance of service mix relative to payer mix has changed over time. In doing so, we quantify the amount of variation in hospital financial performance that is due to service mix, payer mix, operational decisions, market characteristics, and ownership/mission. First, we regress hospital operating margins and cash flow ratios from the Medicare Cost Reports on the share of hospital inpatients in the 100 most common DRGs and MDC categories, payer mix, and a variety of ownership, market, and staffing variables. Next we calculate the partial (and adjusted) R-squared for each variable category. We then use inter-quantile regressions to assess which characteristics define poor versus strong financial performance. Finally, we control for the endogeneity of service offerings to assess the causal effect of service offerings on financial performance. Our sample includes all nonfederal, general short-term hospitals in operation between 1995 and 2002 in 11 states. Our service and payer mix variables are developed from inpatient admissions data reported in the HCUP-SID data for these states. The measures are combined with operating margins and cash flow data from the Medicare cost reports; hospital characteristics from the AHA Annual Survey; and market characteristics from the Area Resource file. The analysis controlling for endogeneity of service mix is based on a subset of 16 MSAs. Specifically, we estimate multinomial logits in the first stage for each MSA and include the residual from the multinomial logit in the second stage. Based on an initial analysis of 1995 to 2000 data, the OLS results and the quantile analysis for high performers indicate that hospital service mix became much more important in determining hospital financial performance beginning in 1997. Payer mix, in contrast, declined from being the most important determinant in 1995 to the third most important determinant in 2000. Looking specifically at hospitals with poorer than median performance, however, unprofitable payer mix remained the most important factor. The presence of uncompensated care and low reimbursement for Medicaid patients put hospitals with large shares of these patients at a significant competitive disadvantage. Certainly, a major factor that defines a strong performer is making the correct decisions about adding services that yield improvements in profitability. Strong performers also have more internal resources and access to capital to facilitate service expansions. Overall, this has led to an unequal distribution of financial performance among hospitals. This divergence in performance is likely to increase over time as long as the DRG weights of the Medicare Prospective Payment System make some services more profitable than others and as long as exogenous restraints on entry exist and are binding.

Permalink

The Effect of Hospital Safety Reports and a Tiered Hospital Network on Inpatient Referrals

Presenter:

Dennis Scanlon

Authors:

Dennis P. Scanlon, Jon B. Christianson, Eric W. Ford, Coleen Lucas

Chair: Gloria Bazzoli; Discussant: Jennifer Troyer Tue June 6, 2006 15:30-17:00 Room 325

Recent years have seen a movement towards both ‘consumer directed’ and ‘pay for performance’ programs in health care. Many of these programs utilize ‘tiered networks’ for hospital care, where consumer out-of-pocket co-payments vary based on the hospital chosen. While most of the early tiering efforts were based on hospitals’ charges only, there is an increasing movement towards placing hospitals into tiers based on efficiency and quality/safety indicators. However, little has been published about consumer response to tiered hospital benefits or the impact of tiered networks on hospital admissions and revenues. This paper examines a tiered hospital benefit in a commercially insured employed population. In conjunction with its major labor unions, the Boeing Company instituted the Hospital Safety Incentive (HSI) for union (i.e., hourly) employees enrolled in Boeing’s Traditional Medical Plan (TMP) in July 2004. The TMP is an ERISA self-funded health plan administered for Boeing by Regence Blue Shield of Washington. The HSI is unique because it gives patients a financial incentive to choose hospitals that meet the Leapfrog Group’s three patient safety leaps. While the TMP’s standard coverage for hospital care is 95% of allowed hospital charges (up to the annual out-of-pocket maximum), union beneficiaries enrolled in the TMP can achieve a benefit of 100% for hospital care if admitted to a hospital that meets the Leapfrog standards. Boeing’s actuaries have estimated the average value of the 5% payment to be approximately $450 per admission. We estimate the effect of the HSI on patients’ selection of hospital in two of Boeing’s major employment hubs (Seattle, WA and Wichita, KS). We utilize a pre-post study design and take advantage of the fact that the HSI did not apply to non-union (i.e., salaried) employees. We identify the effect by comparing the change in hospital admissions of hospitalized hourly and salaried beneficiaries after the HSI went into effect. To gauge awareness of the HSI, we also examine differences, pre-post, between hourly and salaried non-hospitalized beneficiaries. We identify enrolled beneficiaries from claims data, and while we examine changes in hospital market shares, our primary outcome variables come from answers collected during a 20 minute telephone survey. The telephone survey was necessary since patients are referred to hospitals by their physicians, and thus it is not clear if admissions decisions are made by patients, physicians, or jointly. Targeted respondents were randomly sampled from four groups in each period (union/non-union, hospitalized/non-hospitalized). The key outcomes include questions regarding the degree to which the patient was involved in the choice of hospital and awareness of the HSI. We completed approximately 1,200 interviews in each period and achieved a 60% survey participation rate…

Permalink

Effects of Financial Stress from Price Competition on Hospital Quality of Care

Presenter:

Kevin Volpp

Authors:

Kevin Volpp, R. Tamara Konetzka, Julie Sochalski, Jingsan Zhu

Chair: Gloria Bazzoli; Discussant: Leemore Dafny Tue June 6, 2006 15:30-17:00 Room 325

Hospital competition in the United States shifted from a quality/amenity basis to a price basis with the growth of managed care in the 1980s and 1990s. However, a significant managed care backlash starting in the mid-to-late 1990s may have blunted its effectiveness in reducing the rate of increase in hospital costs. This may have altered the balance between price and quality competition in hospitals over time. Preliminary analysis using hospital financial data from California showed that while from 1991-96 the mean annual rate of increase in hospital expenses was lower in markets with higher managed care penetration, these effects are reversed from 1997-2001. The difference in the rate at which expenses increased between high and low competition areas narrowed over these two time periods as well, suggesting that managed care no longer effectively facilitated price competition between hospitals during this later period. To assess the impacts on quality of the change in the effects of managed care penetration (MCP) in influencing hospital competition, we use annual hospital financial and patient-level discharge data from 1991 through 2001 linked with state death certificates from California to examine effects on 30-day mortality for a group of 4 common and severe conditions (AMI, stroke, GI bleed, hip fracture) as well as failure to rescue (death after post-operative complications) during two different time periods (1991-96, 1997-2001). Since managed care is expected to have a stronger effect in more competitive areas, our design incorporates interaction effects between market competitiveness and MCP. We control for baseline costs and quality in more competitive areas and adjust for differences in patient severity. We focus on testing whether the nature of competition between hospitals shifts from favoring price competition (1991-96) to greater emphasis on quality competition (1997-2001). We use long-difference regressions to examine the effect of changes in MCP on outcomes and the effects of changes in MCP in more and less concentrated hospital markets. Preliminary evidence suggests that increases in managed care penetration reduce mortality risk overall but to a lesser degree in more competitive markets. Hospital market concentration is also associated with better mortality outcomes. Next steps will include testing for differences in the relationship between managed care and hospital market concentration with outcomes in different time periods. In addition, we will explore effects by payer group and whether potentially more vulnerable populations, such as the uninsured, were adversely affected to a greater degree than the insured in terms of either the quality or quantity of services received. This work is important to developing a better understanding of how the nature of hospital competition has evolved in conjunction with significant changes in recent years in the effects of managed care. These results will contribute to debates on payment policy, cost containment, and quality improvement efforts.

Permalink

Understanding Local Variation in Physician Employment Arrangements: The Market for Hospitalist Services

Presenter:

Lorens Helmchen

Authors:

Lorens A. Helmchen, Guy David

Chair: Frank Sloan; Discussant: Alvin Headen Tue June 6, 2006 15:30-17:00 Room 326

Permalink

The Determinants of Public versus Private Provision of Emergency Medical Services

Presenter:

Guy David

Authors:

Guy David, Arthur Chiang

Chair: Frank Sloan; Discussant: Judith A. Shinogle Tue June 6, 2006 15:30-17:00 Room 326

The value that Americans place on Emergency Medical Services (EMS) has increased due in part to its improved capacity to respond to, stabilize and transport trauma patients, such as from myocardial infarction, car crashes or incidents of violence. The increase is due as well to the role these services play in saving lives and limiting casualties after large-scale catastrophic events, such as terrorist attacks or natural disasters. The viability and success of EMS hinges on infrastructure and flexibility. Municipalities delegate authority to emergency resources by awarding a contract to either governmental entities (e.g. fire districts) or to private ambulance service providers. In awarding such contracts, local governments consider not only costs of deploying existing municipal infrastructure, but also incentives for persistent implementation of novel technologies and fresh ideas. Economists’ views are controversial on the choices government should make between inhouse provision and contracting out. In order to analyze the organizational structure and the economics of emergency medical services, the paper advances a theoretical model drawing on spatial economics and scope of government ownership theory. To test our theoretical predictions, we study the empirical relationship between the organizational structure of EMS, local market and demographic characteristics, and system performance. To do so, we analyze a cross-section of the 200 largest US cities, combining data on EMS providers from the Journal of Emergency Medical Services (JEMS) with data from the National Fire Department Census, the Bureau of Labor Statistics (BLS), the FBI Uniform Crime Reporting (UCR), and the 1990 and 2000 US Census. Approximately 25% of Americans reside in the 200 largest cities. These cities provide a unique laboratory for studying the determinants of different EMS organizational forms, as only half rely solely on in-house provision of EMS, while the other rely either on private or a public-private mix.

Permalink

Pregnancy and the Labor Market over the Business Cycle

Presenter:

Melinda Pitts

Authors:

M. Melinda Pitts, Mary Beth Walker

Chair: Frank Sloan; Discussant: Rebecca Stein Tue June 6, 2006 15:30-17:00 Room 326

This study examines health outcomes resulting from participation in a prenatal social program established for Medicaid eligible women in the state of Georgia during the period 1994 through 2003, which included a period of rapid growth and a period of recession, as well as a period of rapid change in the welfare system. It is expected that the characteristics of the Medicaid recipients during this time period varied due to both welfare reform and the business cycle. Thus it is expected that there will be a differential impact of participation in health care programs designed for Medicaid beneficiaries. The impact of the change in characteristics is unknown. For example, one might expect that individuals moving on to Medicaid during the recession may not benefit from program participation as those already on Medicaid as these individuals have higher levels of income prerecession. Conversely, the benefit may be greater if the individuals moving on to Medicaid rolls were part of the uninsured working poor and thus come onto Medicaid with a lower stock of health. In this research, we examine the impact of the Perinatal Case Management program in Georgia, which is designed to assist pregnant women who are Medicaid eligible to gain access to medical, nutritional, social, and educational services. The goal of this research is to estimate the impact of program participation on prenatal care utilization, birth weight, and length of gestation, while considering the selectivity resulting from using the sample of live births to estimate these variables of interest and changes in welfare coverage and the business cycle. Estimation is carried out using both parametric and semi-parametric methods. The models are estimated using a rich data set covering women who gave birth in the years 1994-2003 in the state of Georgia. This data set includes information on the named parent(s)’s socioeconomic and demographic conditions and medical data on the pregnancy and related and unrelated comorbid conditions. The data is merged with information on employment histories from the state unemployment insurance data, as well as data on welfare program participation and neighborhood (tract level) characteristics. Preliminary results from one year of analysis (1997) indicate that the consumption of prenatal care is higher for participating women than non-participants. Furthermore, there is a positive direct impact on length of gestation and thus, an indirect benefit to birth weight.

Permalink

Employment of individuals with mental retardation in South Carolina, 1996-2003

Presenter:

Melayne Morgan McInnes

Authors:

Joshua R. Mann, Suzanne McDermott, Melayne M. McInnes, Huafeng Zhou

Chair: Frank Sloan; Discussant: Rebecca Stein Tue June 6, 2006 15:30-17:00 Room 326

Employment of adults with mental retardation is a central aspect of participation in society. It is important to identify factors that result in meaningful employment in this population, as knowledge of these factors can lead to policy and practice improvements. Each year since 1996, we have surveyed local South Carolina disability providers for information about each client’s occupational status, including an estimate of weekly hours worked and wages earned, in addition to employer identity. We also obtained descriptive information for individuals served by the South Carolina Department of Disabilities and Special Needs during the time period, including birth dates, IQ, and the presence or absence of a job coach. A weekly income of $50 was used as the cutoff defining employment. Approximately 3000 (18%) individuals who received some disability services earned $50 per week or more at some time between 1996 and 2003, and 244 worked for all 8 years. We will describe where individuals worked and what jobs they held. An analytic model will be presented showing the relationship between individual, community, and disability service provider characteristics and job acquisition, job tenure and job loss. Predictors of employment, job tenure, and wages will be examined, including job coach activities. Implications for policy and practice in supportive employment will be discussed.

Permalink

State Liquor Policies, Maternal Substance Use, and Child Outcomes

Presenter:

Tara Watson

Authors:

Angela Fertig, Tara Watson

Chair: Michael Grossman; Discussant: Ellen Meara Tue June 6, 2006 15:30-17:00 Room 332

State regulations on the consumption of alcohol by minors are widely credited with reducing teen drinking and alcohol-related traffic fatalities. Less often emphasized is the potential effect of these laws on pregnancy and drinking while pregnant, and subsequent outcomes for infants. Surprisingly little is known about whether, by reducing drinking by young women, these regulations also improve birth and infant outcomes. We focus on the changes in minimum legal drinking age (MLDA) laws that occurred in many states in the late 1970s and early 1980s. The principal objective of the study is to evaluate the consequences of drinking laws and maternal substance use on birth and infant outcomes. There are two channels by which substance use among teenagers affects the health of the next generation. First, by increasing risky sexual behavior, youth alcohol consumption could change the composition of births towards younger mothers and unintended pregnancies. Second, dependent of the compositional effect, drinking alcohol during pregnancy may directly cause poor health outcomes. Because alcohol and tobacco are often used jointly, minimum drinking age laws may affect maternal smoking as well. The specific aims of our project are: 1. To estimate the effects of MLDA laws and enforcement on substance use, with a particular emphasis on alcohol and tobacco use by young women and pregnant women. 2. To estimate the effects of MLDA laws and enforcement on sexual behavior, pregnancies, and births to young women. 3. To evaluate the impact of maternal alcohol and tobacco consumption on birth and early childhood outcomes, using changes in MLDA laws as a source of exogenous variation in alcohol and tobacco use. We use both the restricted version of the NLSY 79 (National Longitudinal Survey of Youth) and Vital Statistics microdata to address these questions. Preliminary results are as follows: 1. MLDA laws reduce alcohol and tobacco use for affected cohorts, and reduce drinking among pregnant young women (see Table 1). 2. The effect of MLDA laws on sexual activity and births is small, but may be larger for some sub-groups. 3. MLDA laws are associated with reduced incidence of low birthweight among infants born to affected cohorts (see Table 2).

Permalink

Does Greater Exposure to WIC Affect Maternal Behavior and Improve Infant Health? Evidence from the Pregnancy Nutrition Surveillance System

Presenter:

Cristina Yunzal

Authors:

Ted Joyce, Cristina Yunzal

Chair: Michael Grossman; Discussant: Robert Kaestner Tue June 6, 2006 15:30-17:00 Room 332

Benefits of the Supplemental Program for Women, Infants and Children (WIC) that may impact maternal behavior and birth outcomes include nutritional supplementation, nutritional counseling, health care referrals, and breastfeeding promotion. We propose to test whether earlier and sustained exposure to program benefits results in less maternal smoking, greater weight gain, more breastfeeding and therefore healthier pregnancies and infants. However, instead of comparing WIC to non-WIC participants, we compare women who enroll in WIC in the first trimester of pregnancy to women who either enroll later in pregnancy or in the postpartum period. In other words, everyone in our sample will have “selected” into WIC between conception and the first postpartum visit. We appreciate that many of the selection issues present in comparisons of WIC and non-WIC women may also be present in an analysis of WIC participants only. Therefore, we stratify the analysis by the parity and the timing of the first prenatal care visit in order to lessen selection within WIC participants. Specifically, we focus on women with no previous live births, all of whom begin prenatal care in the first trimester. Some are certified for WIC in the first trimester and some not until after birth. The key to our empirical strategy is data with a large number of WIC participants, detailed information on the timing of their WIC certification, and a rich set of maternal behaviors. The Pregnancy Nutritional Surveillance System (PNSS) in North Carolina and New Jersey meets these requirements.(http://www.cdc.gov/pednss/). The systems in North Carolina and New Jersey collect information on approximately 40,000 women and their infants each year. We will analyze data from both states from 1995-2003 and thus have upwards of 600,000 observations. The PNSS collects data on body mass index, hematocrit level, smoking cessation, and breastfeeding. The greater detail makes PNSS superior to birth certificates along numerous dimensions. However, like birth certificates the samples are large and thus we are able to stratify the analysis not only by the parity and the timing of prenatal care, but race and ethnicity as well. Preliminary results indicate that greater exposure to WIC is uncorrelated with smoking session and weight gain during pregnancy. Consistent with these findings, we find no difference in birth outcomes among women who enrolled in WIC in the first trimester, second or third trimester. These results are in agreement with recent work that has challenged a causal association between prenatal WIC participation and improved birth outcomes.

Permalink

Violence and Birth Outcomes

Presenter:

Anna Aizer

Authors:

Anna Aizer

Chair: Michael Grossman; Discussant: TBA Tue June 6, 2006 15:30-17:00 Room 332

Everyday roughly 14 thousand women in the US are battered and four are killed by their intimate partners, prompting former Surgeon General C. Everett Koop to label domestic violence “the single most important health issue in the US.” In this paper, we examine the impact of violence against women on birth outcomes. Estimates from the medical literature suggest that as many as one in five pregnant women is the victim of intimate partner violence. Trauma during pregnancy is associated with poor birth outcomes including low birth weight, preterm birth and neonatal mortality. Using a unique dataset that combines data on hospitalization for assaults with birth outcomes from vital statistics records for the state of California for 1990-2000, we estimate the impact of violence on birth outcomes. Because violence is highly correlated with poverty and other risk factors associated with poor birth outcomes, we instrument for violence using variation in the prosecution of domestic violence across counties and over time. Over this period, policies regarding the prosecution of domestic violence offenses changed dramatically and the probability that a man convicted of spouse abuse increased considerably. We use variation in these policies to instrument for the probability of assault. Based on the IV results, we find that reductions in violence are responsible for roughly three quarters of the improvements in birth outcomes witnessed over the last decade. This work sheds new light on the health production process as well as observed income gradients in health. Violence is considerably more prevalent among low income families. We find that violence against women is responsible for roughly 15 percent of observed income gradients in health.

Permalink

The Role of Patient vs. Provider-Level Factors in Treatment Patterns and Outcomes Under Managed Behavioral Health Care

Presenter:

Susan Ettner

Authors:

Susan Ettner

Chair: Susan H. Busch Tue June 6, 2006 15:30-17:00 Room 335

Permalink

The Effects of Expanded Mental Health Benefits on Treatment Costs

Presenter:

Anthony LoSasso

Authors:

Anthony T. LoSasso, Ithai Z. Lurie, Jhee Un Lee, Richard C. Lindrooth

Chair: Susan H. Busch Tue June 6, 2006 15:30-17:00 Room 335

Permalink

Estimating the Family Burden of Children with Mental Health Disorders

Presenter:

Susan Busch

Authors:

Colleen Barry, Susan Busch

Chair: Susan H. Busch Tue June 6, 2006 15:30-17:00 Room 335

Permalink

Human cost of disease in Italy

Presenter:

Francesco Mennini

Authors:

Francesco Mennini, Fabio Palazzo, G Stirparo, Susanna Conti, G Minelli, R Solimini, V Toccaceli

Chair: Melayne McInnes; Discussant: TBA Tue June 6, 2006 13:45-15:15 Room 121

Introduction: Mortality data represents essential elements for the quantification of health problems and, from an economic perspective, the most important element among human costs. They are recorded all over the country and coded following current and internationally approved criteria.

Death counts and related rates are among the simplest indicators to analyze mortality. They are a short and clear measure of a population’s mortality pattern, that may be used to establish and monitor health priorities or objectives. In particular, within the actual federal framework of the Italian health system, a mortality analysis at a regional level provides specific information which allows for detection of health priorities and objectives for each region.

Methodology: Official mortality data recorded and coded by ISTAT (Italian National Statistics Institute) in accordance with the IX Revision of the International Classification of Disease (ICD IX) were analyzed. The data refers to deaths occurred in Italy among the resident population during the year 2001 (most recent available year).

Taking mortality data, as a starting point, it is possible to calculate a series of indicators, such as: life expectancy, standardized mortality rates, Years of Potential Life Lost (YPLL) and rates of YPLL, which allow us to compare mortality through time and space.

Median ages at death were also calculated (in demography: “median life”) for both men and women and each cause of death. The median age is an indicator of longevity in a population and technically it is the age at which the reference population (i.e. regional population) halves. The causes of death taken into account are the most relevant in Italy: they contributed about 90% to the absolute number of deaths in our country in 2001.

Moreover, other causes of death were analyzed, according to their impact on general mortality and their relative weight within the above mentioned groups of causes. In particular, Acute Myocardial Infarction, Malignant Neoplasms such as Lung cancer, Breast cancer and Colon-rectum cancer, among Violent causes of death, Road accidents and Suicide.

Taking into account the federal debate and related problems, it is useful to underline that the spatial distribution of these indicators allows a comparison among regions in terms of years of potential life lost and age at death for each specific cause of death.

Results: From a federal perspective, YPLL and median ages at death result to be effective indicators of possible lack of the Regional Health Services. It is worth noticing the variability among Regions of the median age at death for Breast cancer; it is well known nowadays that mortality from this cause can be reduced thanks to early diagnosis and treatment, consequently regional values of the median ages lower than those at national level could suggest, apart from any possible epidemiological and clinical difference regarding women affected by this pathology, also the lack of effective screening programmes as well as treatments the Regional Health Services are in charge of. This analysis supports the financial and organizational considerations on different impacts of the Regional Health Services in Italy.

Permalink

Strategies to identify relevant behaviors and costs in family asthma management

Presenter:

Sheryl Magzamen

Authors:

Sheryl Magzamen, Sylvia Brandt, Ira Tager

Chair: Melayne McInnes; Discussant: TBA Tue June 6, 2006 13:45-15:15 Room 121

Despite promulgation of national and international guidelines on the effective management of pediatric asthma, asthma-related morbidity among urban children remains disproportionately high. While there is ample evidence of persistent health disparities by race/ethnicity and income, there is a lack of understanding of the causal mechanism(s)for these disparities. The lack of conception of the relevant causal pathway proves challenging for the implementation of interventions designed to mitigate the burden of asthma in urban communities. Further, without an appropriate causal model, measuring changes in health outcomes due to interventions, as well as determining the economic value of these changes, becomes problematic. These two methodological issues are central concerns in benefit-cost studies. Each of these issues are complicated by the correlation between poor health outcomes and socioeconomic characteristics. Models of the impact of health interventions on specific health endpoints are likely to be biased due to unobservable variables that determine health outcomes and are correlated with standard SES variables. The problem is especially likely to occur in cases of health programs that target high-risk populations such as urban children who suffer from asthma.

In this presentation, we outline how a methodology to value non-market goods, contingent valuation, can be used to collect data for both 1) modeling household behavior and asthma outcomes and 2) cost-benefit assessment of a health intervention program. We will provide empirical data obtained from a CDC asthma intervention, in Oakland, CA, targeting urban adolescents and their families. Our focus will be on the ecological and familial influences on health-related decisions using contingent valuation to elicit an individual’s willingness to pay (WTP) for a particular non-market good (i.e. a reduction in asthma symptoms) using a hypothetical scenario that characterizes a realistic market good that affords the non-market benefit of interest. The association of WTP with asthma severity and health beliefs points to factors to explicate family demand for asthma management programs. The degree and nature of family participation in asthma management may partially explain the observed disparities in asthma morbidity by socioeconomic and demographic characteristics. With respect to pediatric asthma, contingent valuation is useful framework to understand how families would assign a monetary value to reduce asthma severity and improve quality of life for their children. The correlation between change in behaviors and symptoms and WTP can identify sub-populations that may benefit the most from community-based asthma interventions.

In addition, utilizing this approach, assessment of non-market costs related to asthma can inform the design of interventions to mitigate the burden of asthma among urban populations. Although direct and indirect costs related to asthma care have been well established, the non-market costs of asthma provide a more accurate picture of the overall economic burden of asthma. The use of these non-market costs to measure asthma morbidity and gauge the effectiveness of programs to improve asthma management has profound implications for the delivery and the evaluation interventions designed to reduce asthma prevalence in urban communities.

Permalink

The Economic Burden of Experiencing a Major Complication during Percutaneous Coronary Intervention

Presenter:

Kirsten Long

Authors:

Kirsten Long, Erin McMurtry, Kent Bailey, James Naessens, Kurt Jacobson, Charanjit Rihal

Chair: Melayne McInnes; Discussant: TBA Tue June 6, 2006 13:45-15:15 Room 121

Objectives: Technological advances (including coronary stents and new anti-thrombotic regimens) have significantly reduced rates of ischemic complications and enabled percutaneous coronary intervention (PCI) to be applied to expanding indications. However, escalating costs are of concern to patients, providers and payers. This study assessed the incremental medical costs of treating major in-hospital procedural complications incurred by patients undergoing PCI.

Methods: We considered all patients undergoing elective, urgent, or emergent PCI at Mayo Clinic Rochester between 3/1/1998-3/31/2003 in analyses. Exclusions included elective, staged procedures during hospitalization (atypical events) and episodes for patients who denied research authorization. Clinical, angiographic, and outcome data were derived from the Mayo Clinic PCI Registry. In-hospital PCI complications included major adverse cardiac and cerebrovascular events (MACCE) (defined as death, myocardial infarction [MI], emergent coronary bypass surgery, repeat PCI, or stroke) and bleeding of clinical significance. Administrative data was used to estimate total costs (hospital and physician) in standardized, year 2004 constant-dollars. We used generalized linear modeling (inverse Gaussian with log link) to estimate the incremental costs associated with complications adjusting for demographic, clinical, angiographic, and procedural characteristics.

Results: 8,109 eligible PCIs episodes (7,027 treated patients) occurred during the study duration. In 1,071 (13.2%) of these episodes, patients experienced at least one of the selected complications during hospitalization. Both MACCE and bleed events occurred in 147 (13.7%) of these complicated procedural cases. Patients experiencing complications were older, more likely to present with emergent PCI, recent or prior MI, multi-vessel disease, B2/C type lesions, and comorbid conditions than patients who did not experience these events. Unadjusted total costs were, on average, $27,865 + $39,424 for patients who experienced any complication compared to $12,279 + $6,796 for those who were free of complications (p<0.001). A nearly 5-fold increase in inpatient costs was observed among patients with and without MACCE and major bleeding complications ($55,230 vs. $12,279, respectively; p<0.001). Adjusted mean total cost were $7,000 higher for patients experiencing complications compared with patients who were complication free (95% CI of cost difference: $5,854, $8,145). Incremental costs associated with only bleeding events, only MACCE, or for patients experiencing bleeding and MACCE events were $5,813, $5,151, and $15,699, respectively (p<0.001).

Conclusions: This observational study highlights the significant economic burden associated with in-hospital procedural complications. Bleeding complications alone contribute significantly to inpatient cost of care. Interventions to reduce the risk of adverse events likely enhance financial as well as clinical performance.

Permalink

The Impact of Antibiotic Resistant Infection Levels, Outbreaks, & Control Measures on Hospital Expenses-per-Admission

Presenter:

Gwendolyn Morrison

Authors:

Gwendolyn Morrison, Kim McCoy, Bradley Doebbeling

Chair: Michael Hagan; Discussant: Ed Miller Tue June 6, 2006 13:45-15:15 Room 213

Background: Antibiotic resistant infections are an increasing problem in hospitals in terms of the number of resistant organisms and their prevalence. Consequently, the costs of these infections, in terms of mortality, disability, and money, are also increasing. Antibiotic resistance (AR) increases the costs of care in various ways including increased length of stay, more admissions to ICU, and more intensive resource use by inpatients. Furthermore, these additional costs of care often go unreimbursed. Despite the implied incentive to adopt infection prevention guidelines to reduce these AR related costs, hospitals have implemented these guidelines to varying degrees.

Implementing clinical practice guidelines improves health care quality but, because providers incur the implementation costs while resulting savings accrue to payors, the third party payment system typically poses a barrier to guideline implementation. However, implementing infection prevention guidelines may generate provider savings by reducing hospital acquired, including AR, infections. If AR infection levels are to be contained or reduced, economic research must first focus on potential cost savings to the entity with the greatest influence over the adoption and implementation of preventive interventions-the hospital. In this study, we sought to assess the impact of AR levels, outbreaks, and control measures on hospital level expenses-per-admission.

Methods: We surveyed a nationally representative sample of 670 hospitals (stratified by bed size, teaching status, geographic region, and VA versus non-VA status) regarding levels and outbreaks of AR, their adherence to recommended AR prevention and control guidelines, and availability of information technology. Survey data were linked to the American Hospital Association Annual Survey to incorporate other hospital level financial and operational capacity characteristics. We use regression analysis to assess the impact of AR levels, outbreaks, and control measures on hospital expenses-per-admission.

Results: We estimate two cost models with expenses-per-admission as the dependent variable: one assesses the impact of AR levels and outbreaks; the other the impact of specific infection prevention measures. We find that hospital expenses-per-admission increase with levels of MRSA (the most common AR organism). This result was consistent both when pooling VA and non-VA hospitals, and analyzing non-VA hospitals separately. For non-VA hospitals, higher expenses-per-admission are also related to more recent outbreaks of another AR organism, K-ESBL. Cost models incorporating infection prevention measures found that, for non-VA hospitals, (1) using information technology to automate decisions to reduce errors (including, but not limited to, antibiotic prescribing) have lower expenses-per-admission, and (2) hospitals reporting active involvement of their infection control committee in AR prevention and control have higher expenses-per-admission. The former suggests a cost-saving means for hospitals equipped with the necessary IT, while the latter may just reflect the cost of improving quality.

Discussion: The successful implementation of infection prevention and control measures cannot occur without hospital administration support, and administrators will not support such measures until there is evidence that they are associated with cost savings. In presenting evidence that AR increases costs at the hospital level, the results of this study are a first step in providing the business case for improving infection prevention and control.

Permalink

Market Power in Dutch Hospital Sector: Bargaining between Hospitals and Insurers

Presenter:

Misja Mikkers

Authors:

Rein Halbersma, Misja Mikkers, Evgenia Motchenkova, Ingrid Seinen

Chair: Michael Hagan; Discussant: Herb Wong Tue June 6, 2006 13:45-15:15 Room 213

Rationale: Competition was introduced in parts of the hospital market in the Netherlands in 2005. According to a monitoring report of the Dutch Health Care Authority, the introduction of competition lead to an average increase of 5% of the estimated cost prices. However, prices differ amongst hospitals, diagnosis and regions. In this paper we investigate the consequences of market power and bargaining power of both hospitals and insures on prices after the institutional change. We have a unique dataset consisting of about 60.000 observations of both list prices per product and prices per product contracted between hospitals and insurers.

Objectives: In this paper we would like to show the effects of buyer and seller concentration on the price of hospital care. Next we investigate whether hospitals strategically interact in a transitional phase, just after the introduction of competition. Furthermore, we will try to identify possible unilateral market power of certain hospital types. Finally we estimate the relative bargaining power of hospitals and insurers in determining the negotiated prices.

Methodology: First we use a traditional Structural Conduct Performance model to estimate the effects of market structure on price mark-ups in the competitive segment of the hospital market. These models normally are based on the Cournot model in which it is assumed that there is strategic interaction between sellers.

Using the SCP model we will estimate to which extend hospitals are able to coordinate their actions in the period just after the introduction of competition. Within the SCP framework we identify possible unilateral market power by introducing dummy variables for different types of hospitals (e.g. academic hospitals).

We also model the interaction between a hospital and an insurer in the context of a bargaining model (similar to Brooks et al., 1997), by estimating the share a hospital receives (defined by the contracted price minus the cost price) from the total gain of contracting (defined as a list price for non insured consumers minus the cost price).

Conclusions: Preliminary estimation results show that hospitals are, just after the introduction of competition, unable to coordinate their actions. The concentration of hospitals does not have a significant effect on the observed price mark-ups. However, large insurers can negotiate significantly lower prices. Furthermore, academic hospitals are able to negotiate significantly higher prices for hospital services than general hospitals. Since patients are willing to travel larger distances to get a treatment from an academic hospital, we conclude that academic hospitals are able to exert unilateral market power. On average insurers have more bargaining power. We will also investigate the effect of the market structure on the division of bargaining power between hospitals and insurers.

Possible Extensions: We will try to include simultaneous testing for excessive pricing, predatory pricing, and countervailing power by insurers in the model. Furthermore, we might extend the paper with a spatial regression model, which would estimate the spatial interaction between neighboring hospitals.

Permalink

Hospital Financial Condition and Operational Decisions Related to Quality of Care

Presenter:

Gloria Bazzoli

Authors:

Gloria Bazzoli, Richard Lindrooth, Jan Clement, Hsueh-Fen Chen

Chair: Michael Hagan; Discussant: Michael Hagan Tue June 6, 2006 13:45-15:15 Room 213

Over the last decade, the US hospital industry has experienced intensifying financial pressures as labor and other costs grew rapidly relative to slower increases in hospital payments. In fact, the Medicare Prospective Payment Advisory Commission reported that nearly one-third of hospitals had negative total margins in 2003. These pressures have raised concern. Some institutions may close limiting access to care in some communities. However, closure may not be the biggest issue given existing research showing that hospitals frequently continue to operate despite dire financial circumstances. Instead, hospitals may be forestalling closure by reducing the quality of their services. Our research examines if this is the case, assessing the relationship between hospital financial condition and operational decisions related to quality of care.

Our theoretical model draws directly on economic theory of production. We model product quality as a derivative of the production process and hospital choice of inputs. A number of studies have examined the effect of financial pressures on hospital staffing decisions, but few have examined its effect on hospital processes and infrastructure supporting the care-giving process or on the upkeep of hospital facilities. We examine proxies for these latter dimensions, specifically hospital compliance with certain Joint Commission on the Accreditation of Healthcare Organization (JCAHO) performance areas and hospital investments in plant and equipment.

Although many JCAHO standards are not difficult to meet, a subset is viewed as being particularly challenging by the accreditation agency, with annual compliance rates as low as 50%. We focus on this subset, which includes such things as hospital procedures for ensuring medication or anesthesia safety. Our net hospital plant assets measure, on the other hand, relates to the quality of equipment and technology available for patient care. Low levels of net plant assets, for example, likely mean that antiquated facilities and technology are not being replaced.

For our empirical work, we examine a 6-year longitudinal database on 3,000 US general hospitals that were operational between 1995 and 2000. This database blends data from: the AHA Annual Survey; Medicare hospital cost reports; selected JCAHO performance areas; Area Resource File; and InterStudy. We estimate instrumental variables fixed and random effects models that control for unmeasured hospital and market characteristics. We use overidentification tests to assess the validity of our instruments, and Hausman specification tests to assess the adequacy of various models.

To date, our analysis has found that the quality of hospital infrastructure and management processes decline with reductions in cashflow and revenues per patient day. These negative effects are most apparent for our net plant assets measure and for selected JCAHO performance areas that relate to the infrastructure that supports initial patient assessment, ongoing staff competency assessment, standardization of patient data, and maintenance of infection control standards. Overall, our research is yielding new insights about the relationship between hospital financial condition and operational decisions related to patient care and outcomes, especially as they pertain to the quality of hospital infrastructure and processes to support patient care.

Permalink

Financial Incentives for Adherence to Diabetes Care Pathways: Discerning Pay-for-Performance Impact from Increasing Trends

Presenter:

James Burgess

Authors:

James Burgess, Gary Young, Dan Berlowitz, Bert White, Mark Meterko, Barbara Bokhour, Errol Baker, Karen Sautter, Howard Beckman, Robert Greene, Kathleen Curtin

Chair: Richard C. Lindrooth; Discussant: TBA Tue June 6, 2006 13:45-15:15 Room 225

Many organizations around the world have begun designing and implementing pay-for-performance or pay-for-quality programs for physicians as a way of attempting to induce practice behavior changes in physicians and improve patient outcomes. Quality improvement also can be occurring for other reasons even if physician payment schemes do not account for quality of service. Difference-in-difference approaches using state and national controls can be used to discern the intervention impact from the trend. We employ a three year pre-intervention and three year post-intervention design measuring an array of adherence scores on diabetes care pathways (Hemoglobin A1C, Microalbumin urinalysis, LDL cholesterol, and Retinal eye exam) that led to financial payouts based on rank ordering of provider outcomes in a large Independent Practice Association (IPA) to accomplish this test of pay-for-performance plans. We review the theory and practice of incentive experiments in this area. In the current sparse literature on physician pay-for-performance, e.g. Rosenthal et al. (2005), careful distinctions between the incentive designs and intended effects have not always been clear. Measures with a fixed threshold of performance have quite different incentive for behavior change than rank ordering methods that are dependent upon relative improvement across all subjects. Physicians already above a threshold do not have to improve further to gain the incentive, so the focus is not so much on gain or improving quality as on paying the highest performing physicians more. While physicians improving nominal scores as other physicians improve by more can get paid less under a rank ordering system, but the incentive for improvement is clearer across the range of all providers. In addition, the patient panels for a chronic disease like diabetes can be quite small for some physicians, which may offer a large reward to a small amount of effort. Results indicate that adherence scores are increasing over time for all measures of diabetes care; however, only some of the measures exhibit a change attributable to the pay-for-performance intervention. We track individual physicians over the six year period and test effects of the size of the exposed diabetes panel and changes in the size of the panel over the period as well as the effect of personal characteristics of the physicians on the results.

Permalink

Intended and Spillover Effects of Pay-for-Performance on Quality of Care

Presenter:

Meredith Rosenthal

Authors:

Meredith Rosenthal, Richard Frank, Zhonghe Li, Arnold Epstein

Chair: Richard C. Lindrooth; Discussant: TBA Tue June 6, 2006 13:45-15:15 Room 225

The number of health plans and purchasers in the United States that have adopted pay-for-performance mechanisms for quality improvement is growing rapidly. However, most of these programs are in the early stages of trial, evaluation, and adjustment. Although there is intense interest in and optimism about pay-for-performance programs among many policy makers and payers, there is little published research on pay-for-performance in health care. In fact, there are only a few studies demonstrating that pay-for-performance leads to improved quality of care.

We evaluated a natural experiment with pay-for-performance using administrative reports of physician group quality from a large health plan for an intervention group (California physician groups) and a contemporaneous comparison group (Pacific Northwest physician groups). Quality improvement reports were included from October 2001 through April 2004 issued to approximately 300 large physician organizations.

We examine the impact of the program on three targeted measures of clinical quality: cervical cancer screening, mammography, and hemoglobin A1c testing as well as a set of untargeted measures, including Chlamydia screening, appropriate asthma medication, and emergency room visits for asthma. Improvements in clinical quality scores were as follows: for cervical cancer screening, 5.3% for California vs 1.7% for Pacific Northwest; for mammography, 1.9% vs 0.2%; and for hemoglobin A1c, 2.1% vs 2.1%. Compared with physician groups in the Pacific Northwest, the California network demonstrated greater quality improvement after the pay-for-performance intervention only in cervical cancer screening (a 3.6% difference in improvement [P=.02]). In total, the plan awarded $3.4 million (27% of the amount set aside) in bonus payments between July 2003 and April 2004, the first year of the program. For all 3 measures, physician groups with baseline performance at or above the performance threshold for receipt of a bonus improved the least but garnered the largest share of the bonus payments. Patterns of improvement for untargeted measures were mixed, suggesting the existence of both positive and negative spillovers.

Permalink

The effect of financial incentives on the volume of diagnostic imaging ordered by physicians

Presenter:

Mythreyi Bhargavan

Authors:

Mythreyi Bhargavan, Cristian Meghea, Jonathan Sunshine

Chair: Richard C. Lindrooth; Discussant: TBA Tue June 6, 2006 13:45-15:15 Room 225

Rationale: Medical imaging is a large component of health care costs in the United States, with an estimated annual cost almost $100 billion, and has one of the fastest growth rates (approximately 10% per year) among all medical services. Imaging is usually provided through physician referral, and therefore, may be influenced by financial incentives faced by the referring physician.

In this study, we analyze the differences in imaging utilization between self-referring and radiologist-referring physicians, while controlling for patient characteristics and diagnoses. If a patient has an office visit and a subsequent related image, and the same physician bills for both events, the physician is identified as a self-referrer with a financial interest in the imaging procedure to which he or she refers the patient. If the image is performed by a radiologist, then the referring physician is flagged as having no financial interest in the referral. (Referrals to physicians other than radiologists are analyzed as a separate category because the incentives are unclear.)

Objective: To measure the effect of physician financial incentives, as captured by self- versus radiologist-referral, on the utilization of imaging studies, controlling for patient demographics and co-morbidities, geographic location, practice setting (physician office, outpatient hospital, or inpatient hospital), and other physician-related factors.

Methods: We use claims data from a large national employer plan for five years (1999-2003). We identify patients with certain conditions such as acute upper respiratory tract symptoms, headache, and knee pain. We construct episodes of care and for each episode, observe whether an image was ordered, the type of image, total imaging costs, and total patient care costs. Each medical condition is analyzed indepently. Outcomes of interest for each patient are (a) percent of episodes with imaging, (b) total costs of imaging, and (c) total medical costs, with costs measured in measured in global relative value units (RVUs) to control for price differences. Logit regression is used to analyze outcome (a) and log-linear regression for outcomes (b) and (c).

Preliminary Results: For cardiac imaging the number of images per patient is higher when the treating physician does his/her own imaging than when the patient is referred to a radiologist (2.8 ultrasounds and 2.9 nuclear medicine images per patient by self-referrers versus 2.6 and 2.7 respectively for radiologist-referrers). But the number of images per patient is even larger when patients are referred to an independent clinic (2.9 ultrasounds and 3 nuclear medicine procedures), and the financial motivations are unclear. Controlling for patient and neighborhood characteristics, and state physician supply, physicians who performed their own imaging for heart disease performed 6% more images than physicians who sent the patient to a different physician.

Conclusions: There is some evidence that imaging utilization is higher if a physician self-refers patients than if the patient is referred to a radiologist or a different physician. Analysis of costs of imaging may reveal sharper differences between self-referring and radiologist-referring physicians. The results of this study will assist policy makers and payers in designing effective incentive structures to ensure appropriate utilization.

Permalink

Does Participation in the Food Stamp Program Affect the Incidence of Obesity and Health Care Costs?

Presenter:

Yuriy Pylypchuk

Authors:

Yuriy Pylypchuk, Chad Meyerhoefer

Chair: Tom Selden; Discussant: TBA Tue June 6, 2006 13:45-15:15 Room 226

The Food Stamp Program (FSP) is a cornerstone of public assistance and poverty alleviation efforts in the United States, serving nearly 24 million individuals at a cost in excess of 25 billion dollars in 2004. Though the program plays an important role in alleviating hunger by supplementing food budgets and freeing income for non-food purchases, some have raised concerns that it may be contributing to the high incidence of overweight and obesity among the poor. Furthermore, many FSP participants are Medicaid recipients, so higher rates of obesity due to program participation may result in greater financial burden on public insurance programs if obesity-attributable medical expenditures increase. Several studies have shown that participation in the FSP leads to higher food expenditures and a greater marginal propensity to consume food out of program benefits than out of cash income (Fraker, 1990). However, higher expenditures on food do not necessarily translate into greater caloric intake. For example, a recent study by Mathematica Policy Research (2000) finds inconsistent and insignificant effects of program participation on the number of servings consumed, level of micronutrient intake, and indices of overall dietary quality for adults. Both Gibson (2002) and Chen et al. (2005) find that participation in the FSP is associated with higher rates of obesity for woman, but neither of these studies provides rigorous identification of the link between program participation and obesity.

We investigate the impact of participation in the FSP on adult obesity for the food stamp eligible population using the 2000 - 2002 Medical Expenditure Panel Survey (MEPS), a nationally representative survey of the U.S. civilian, non-institutionalized population. To identify the effect of program participation on overweight status we use instrumental variables methods that exploit exogenous variation in program exclusion criteria at the national level and differences in state level program rules and measures of access. We also make use of the longitudinal dimension of the MEPS to estimate random effects discrete factor models. These models are less sensitive to exclusion restrictions than single-equation instrumental variables methods and allow us to explicitly control for individual level time-invariant unobservables (Morz, 1999). We subsequently estimate the impact of FSP participation on medical expenditures, both directly, and indirectly through changes in overweight status. Obesity attributable increases in medical expenditures are then determined for the Medicaid eligible population as well as the sub-population of individuals suffering from certain chronic conditions, such as diabetes and hypertension. These estimates will help policy makers determine whether to investigate alternative benefit delivery mechanisms or provide supplemental nutrition education through the FSP.

Permalink

Labor Income Risk and Health Outcomes

Presenter:

Timothy Halliday

Authors:

Timothy Halliday

Chair: Tom Selden; Discussant: TBA Tue June 6, 2006 13:45-15:15 Room 226

This paper uses data from the Panel Study of Income Dynamics (PSID) to investigate the impact of earnings risk on health. We show that good health exhibits pro-cyclical variation in the PSID for working-aged men and women. To identify the impact of shocks to income and labor supply on health, we estimate a dynamic, linear model using moment restrictions that allow for health to impact labor supply. Doing this, we find that income shocks exert strong negative impacts on the health of working-aged men, but macroeconomic conditions do not impact health once we control for income and labor supply. For women, however, macroeconomic conditions still impact health even after controlling for income and labor supply. We provide some indirect evdence that suggests that this is reflecting the impact of shocks to the husband’s income on the health of the wife.

Permalink

You Can Never Be Too Rich or Too Thin: Causal Impact of Income on Obesity

Presenter:

Kosali Simon

Authors:

Kosali Simon, John Cawley, John Moran

Chair: Tom Selden; Discussant: TBA Tue June 6, 2006 13:45-15:15 Room 226

Over the past 25 years, many countries have experienced increases in obesity, which the World Health Organization attributes in part to increases in income. The link between income and obesity is also important for understanding racial differences in the prevalence of obesity in the U.S., and patterns of obesity across socioeconomic status among women in the U.S. A parallel literature examines whether low-income elderly in the U.S. have poor nutrition. All of these literatures relate to the impact of income on obesity, of which there is little evidence.

This paper estimates the causal impact of income on weight by exploiting a natural experiment: an error in the way Social Security benefits were calculated that led some retirees to receive windfall payments from the Social Security system. We exploit this natural experiment, known as the Social Security benefits notch, by estimating instrumental variables and reduced form models using data from the first wave of the Asset and Health Dynamics Among the Oldest Old (AHEAD). We find evidence that exogenous increases in Social Security income raise weight for relatively light low-income women. These changes in weight are economically meaningful; we estimate that the value of statistical lives saved because of changes in weight caused by the Social Security notch totals $7.4 billion.

Permalink

Web Analysis - Research on Members' Health Plan Website Use and Impact on Health Care Utilization

Presenter:

Mona Shah

Authors:

Charlotte Wu, Regina Levin, Ross Owen, Carole Bashaw, Mona Shah

Chair: Randall Ellis; Discussant: Jenny Kenney Tue June 6, 2006 13:45-15:15 Room 235

Rationale: A primary premise of consumer driven healthcare is that members will use reliable resources of information for healthcare decision making. Benefit, network and pricing information is specific to health plans and members must therefore rely upon their plan for this information. The health plan web site is a convenient way to make this information available to members. This study examines the prevalence of plan website use, what information members access, and the impact on health care utilization.

Objectives: To evaluate consumers’ pattern of use of a leading consumer-driven health plan member website, and to examine the relationship between web use and health outcomes and health care utilization behaviors.

Methodology: Our study population is all members enrolled in a leading consumer-driven health plan during the plan years of 2003 or 2004, which includes approximately 185,000 members from 70 employers. Using 2003 and 2004 clickstream data, we examined web site use overall and by the area accessed: “My Account”, “My Benefits”, “Doctors & Hospitals”, “Pharmacy”, “Healthcare Costs” and “Health Resources”. Family is used as the analysis unit since members of the same family will log onto website with subscriber’s identification. We categorized families into groups by number of times they logged on and compared these groups’ illness burden, their utilization behavior including office visits, ER visits, hospital admissions, and drug use. We also examined differences in when the Health Reimbursement Account was exhausted and the deductible was met. In addition, families were categorized by their aggregated illness burden into four groups and their web usage and healthcare utilization were examined by these different risk groups. Linear regression model was used to determine the effect of using any area on website on total healthcare cost, adjusted by their illness burden.

Results: More than half of members accessed the health plan website. In general, members who used web more have higher utilization of health care services. However, for members with high burden of illness, web-users have lower emergency room visits and hospital admissions than non-web-users, and their overall cost were not statistically significantly different from non-web-users. The frequency of access to some areas is strongly related to members’ healthcare cost after adjusting for their illness burden: cost increases with more access to “Doctors & Hospitals” information while decreases with use of “Healthcare Costs”.

Conclusions: Members are active in managing their healthcare utilization by accessing their health pan website. Members with higher utilization are more likely to use the health plan web site. When controlling for illness burden strata, in high health risk group the web-users’ costs were similar to non web-users’ with a reduction in use of acute health care services. This study suggests that there is a relationship between seeing health care costs and actual healthcare spending. Although a direct cause-effect relationship has not yet been established, we believe our findings provide a positive feedback on the benefit of using web-based consumer health information on cost savings.

Permalink

The Effects of Health Plan Performance Measurement and Reporting on Quality of Care for Medicare Beneficiaries

Presenter:

Kate Bundorf

Authors:

Kate Bundorf, Kavita Choudhry, Laurence Baker

Chair: Randall Ellis; Discussant: Margaret Byrne Tue June 6, 2006 13:45-15:15 Room 235

Background: Significant resources have devoted to the development and public dissemination of reports comparing the performance of health plans. The objective of these consumer report cards is to improve the quality of healthcare by providing consumers with the information necessary to incorporate quality into their decisions and plans with information on how their performance compares with others. While a number of studies have documented that these report cards have had relatively little effect on consumer choice of plans, relatively few studies have examined their impact on quality of care.

Objectives: The objective of this research is to examine the effect of health plan quality measurement and reporting on the use of measured services by Medicare beneficiaries. We examine how the implementation of mandatory health plan participation in standardized quality measurement and reporting affected the utilization of measured services among Medicare beneficiaries enrolled in these plans as well as those enrolled in traditional Medicare.

Methods: The primary data source for our analysis is the Medicare Current Beneficiary Survey, an annual survey of Medicare beneficiaries conducted by CMS. Using these data, we examine the utilization by Medicare beneficiaries of 4 services that correspond to performance indicators for Medicare Managed Care plans, including mammograms, flu shots, drugs for controlling high blood pressure, and beta blocker after heart attack. We also define a set of control indicator that includes services, such as pap smear, that are similar to, but not included in, the HEDIS performance indicator set. These models provide an empirical test of whether our findings for the performance indicators are due to report cards rather than other unobserved trends in service utilization. We examine the effect of performance measurement by comparing utilization of performance indicators among Medicare beneficiaries enrolled in managed care plans before and after the implementation of mandatory health plan participation, using the Medicare fee-for-service population as a control group. We use data on Managed Care market penetration to test for the impact of selection of beneficiaries across sectors and spillovers from the managed care to the FFS sector on our results.

Results: We find little evidence that the implementation of quality reporting increased rates of either flu shots or mammograms, both performance indicators. While rates of use of these services increased among Medicare Managed Care enrollees after the implementation of mandatory health plan quality reporting, rates of utilization of the performance indicators also increased among Medicare FFS enrollees who were not directly affected by the program.

Conclusions: Our findings to date indicate that the implementation of mandatory health plan quality reporting had little effect on the utilization of two performance indicators among Medicare beneficiaries enrolled in managed care plans. Our analysis will examine two additional performance indicators and also determine whether spillover or selection effects are driving these results.

Permalink

Patient information and pharmaceutical prices

Presenter:

Nicklas Rudholm

Authors:

Nicklas Rudholm

Chair: Randall Ellis; Discussant: Ting Liu Tue June 6, 2006 13:45-15:15 Room 235

Rationale: The most common motivation for regulating health care markets has been that the difference in information between the producer of health care and the patient is large, giving the producers of health care market power. In its most extreme form, this could lead to so called supplier induced demand, where producers of health care sell unnecessary treatments to uninformed patients. In other, less extreme cases, the information gap between producer and patient still gives the producer the possibility to charge prices above marginal cost. In this paper, the effects of increased consumer information about price differences in medically equivalent treatments (branded and generic drugs) on the market power of health care producers (pharmaceutical firms) will be examined.

Objective: The objective of this paper is to estimate the effect of increased consumer information about price differences between brand name and generic drugs on pharmaceutical prices.

Methodology: In October 2002, the Swedish pharmaceuticals market was reformed. The reforms required that pharmacists substitute the prescribed pharmaceutical product to the cheapest available generic product. The reforms also require that pharmacist inform the patient that there are substitute products available and that the products will be switched. However, they also have to inform the patient of the opportunity to buy the prescribed pharmaceutical product instead of the generic, paying the difference in price between the products themselves. This means that under the new regulations, patients have more information about the price difference between the prescribed (in most cases brand name) product and the cheapest available generic alternative. As such, the introduction of the substitution reform can be seen as a natural experiment where the information available for patients concerning price differences in medically equivalent treatments has been increased. Using a unique dataset covering 1.5 million pharmaceutical prescriptions made out in the county of Västerbotten, Sweden during the period January 2001 until June 2003, regression analysis is used to test how the increase in patient information has affected pharmaceutical prices.

Results: The results show that increased patient information about the price difference between available generic products and brand name drugs do lower the price of brand name pharmaceuticals.

Conclusions: The results from this paper indicate that increased patient information decreases the market power of pharmaceutical firms. This, in turn, means that it might be possible to reduce health care costs by increasing patient information about available treatments and the price differences between those treatments.

Permalink

Strategies and performances in the pharmaceutical Italian market

Presenter:

Lara Gitto

Authors:

Francesco Mennini, Lara Gitto, Marco Ratti

Chair: Richard Frank; Discussant: Chris Roebuck Tue June 6, 2006 13:45-15:15 Room 309

A structural analysis has put in evidence how variations in the economic results of pharmaceutical firms may be due both to exogenous and endogenous factors. The implementation of an innovating strategy based on the introduction of new products (due to a higher portion of investments in R&D activities), for example, might lead to higher profits. At the same time, the strategy selection process might depend on some structural “exogenous” factors, such as regulatory issues, concentration in the market, changes in the institutional framework, etc., likely to influence firms’ performances.

The aim of the present paper is to focus on the first of the two aspects above mentioned: here, a micro analysis, aimed at studying the behaviour of Italian pharmaceutical firms, that, across the period 1994-2004, achieved the 75% of total sales, is performed. We even analysed two sub-periods of time (1994-1999 and 1999-2004), characterised by a different regulatory interventions, to better understand the effects of regulation on pharmaceutical strategies.

At the same time, this paper aims at providing useful information to gain an insight into a few typical behaviors of businesses in the Italian pharmaceutical market and how business strategies changed in the last decade.

Variations in economic results might depend on the adoption of a given strategy (conservative vs. innovating), on the number of new products and/or best seller drugs marketed, on the circumstance that the firm belongs to an established group or experienced a merger in order to consolidate its position in the market, or on a combination of all these factors.

A limited impact of the latter would indicate how increasing profits are mainly due to institutional “exogenous” factors, likely to be investigated through a macro analysis.

Results of this paper are useful to understand trends in pharmaceutical industry in Italy and may be used in international comparisons, to analyse the impact of some “exogenous” factors that influence, for example, EU pharmaceutical market performance.

Permalink

The Market for Follow-On Biologics

Presenter:

David Ridley

Authors:

Henry Grabowski, David Ridley, Kevin Schulman

Chair: Richard Frank; Discussant: Francesco Mennini Tue June 6, 2006 13:45-15:15 Room 309

The Food and Drug Administration is examining whether biologics can and should be treated like pharmaceuticals with regard to generics. Follow-on biologics differ from generic pharmaceuticals in that equivalence of safety and efficacy is more challenging to demonstrate. Because of these differences, generic biologics are often referred to as “follow-on biologics.” The focus of the debate on follow-on biologics has been on legal and health concerns, but there are important economic questions. How will differences in development and manufacturing costs and associated regulations affect the market for follow-on biologics? Will follow-on biologics be as competitive and provide the substantial financial savings provided by generic pharmaceuticals?

Generic pharmaceuticals provide a substantial price discount over branded products. Nevertheless, it is not the mere presence of a generic product in the market but competition between multiple firms that results in aggressive price competition and discounting. To assess the potential economic advantages to consumers from follow-on biologic products we must assess the potential for firm entry into this new market and whether competition among manufacturers of follow-on biologics is likely to be as vigorous as that of manufacturers of generic pharmaceuticals.

We analyze market entry and prices for follow-on biologics in the United States using four analogs. First, we analyze the U.S. market for generic pharmaceuticals. We report estimates of the number of generic competitors as a function of expected returns. We then report estimates of relative generic prices as a function of the number of generic competitors. We use IMS Health’s Generic Spectra data. Second, we analyze the U.S. market for human growth hormone (HGH). In this market some patents have expired and there are multiple competitors, but none of the products entered through an abbreviated process bypassing clinical trials. We will use price data on the HGH market from Drug Topics Redbook. Third, while there are no generic biologics yet sold in the U.S., Australia became the first regulated market to approve a generic biologic. We will analyze price data for Omnitrope, a generic biologic human growth hormone approved by the Australian Therapeutic Goods Administration in October 2004. Fourth, we will report prices of biologics in emerging markets such as Brazil, China, India, and Mexico to provide insight on manufacturing costs and generic competition.

Given uncertainty about the market for follow-on biologics, it is challenging to accurately predict how the market will evolve. Nevertheless, these four analogs can provide a better understanding of the likely market for follow-on biologics. Policy makers should be cautious in projecting large financial benefits for follow-on biologics for consumers and payers based on the experiences of generic pharmaceuticals. They need to consider how follow-on biologics will differ in terms of economics as well as scientific and regulatory factors.

Permalink

An investigation of first-mover advantage in pharmaceutical advertising

Presenter:

Winghan Kwong

Authors:

Winghan Kwong

Chair: Richard Frank; Discussant: Richard Frank Tue June 6, 2006 13:45-15:15 Room 309

Rationale: Pharmaceuticals are experience goods because physicians often choose medications based on their clinical experience and they may be reluctant to try a new product due to uncertainty about its quality. Because of imperfect information, products that enter the market earlier may have a competitive advantage over new entrants.

Objective: The objective of this study was to determine if there is an asymmetry in the effectiveness of advertising between earlier and later entrants that would support the presence of a first-mover advantage in pharmaceutical advertising.

Study Design: Effectiveness of advertising was evaluated using estimates of advertising elasticity of demand (i.e., the percent change in a product’s sales per each percent change in its advertising expenditures). Advertising elasticity of demand of existing products and new entrants in the periods following new entry was estimated using time and disease fixed effects estimation. Because the effect of advertising on product sales is not immediate, one-period and two-period lag models were estimated. For instance, if a new product entered the market during period t, the advertising expenditures for existing products and the new product in period t+1 or t+2 were analyzed. The logarithm of a product’s sales was regressed on its own advertising expenditure, the total advertising expenditure of competing products in the same market, the number of competing products, number of years the product has been on market, a dummy variable indicating whether the product is a new entrant, and an interaction term between the dummy variable and the product’s own advertising expenditure.

Populations Studied: Advertising expenditure and product sales data were obtained from Scott-Levin Market Research Audit data. New product entry data were from United States Pharmacopeia Drug Information, Orange Book and the Scott-Levin Market Research Audit data. Data from January 1995 to December 2001 were used. The analysis was conducted on a quarterly basis. Products from eight therapy markets were examined: asthma, migraine, obesity, Parkinson’s disease, seizure, depression, lipid disorder, and gastric and duodenal ulcer.

Results: In both lag models, product sales significantly increased with a product’s own advertising expenditure, and significantly decreased with the total advertising expenditure of competing products and the number of competing products. For existing products, advertising elasticity of demand was estimated to be 0.068% in the one-period lag model and 0.062% in the two-period lag model. The elasticity estimates for new entrants was lower at 0.046% and 0.044%, respectively but not significantly different from existing products. Comparing results of the two lag models, advertising elasticity of demand depreciated at 8.7% per quarter.

Conclusion: The analysis did not find any significant first-mover advantage in the effectiveness of advertising between earlier and later entrants that would support the entry deterrence effect of advertising in the pharmaceutical market.

Permalink

Consequences of Serving in Vietnam for Health-Related Behaviors Later in Life

Presenter:

Daniel Eisenberg

Authors:

Daniel Eisenberg, Brian Rowe

Chair: Ching-to Albert Ma; Discussant: Christopher Carpenter Tue June 6, 2006 13:45-15:15 Room 313

Studies have shown that men who served in Vietnam were more likely to smoke, drink alcohol, and use other substances later in life than their peers who did not serve. It is not clear to what extent these differences are due to service-related effects or to differences in personal characteristics that cannot be controlled for in analyses (e.g. different preferences regarding health risks, or higher discount rates). We address this issue using an instrumental variable (IV) approach where the IV is based on draft numbers, which were randomly assigned in 1970, 1971, and 1972 to eligible men born between 1944 and 1953 according to their birth dates. This type of analysis has been previously conducted looking at outcomes such as mortality due to accidents and wages. These studies show that because draft numbers are based on random selection of birth dates, groups defined by draft numbers are unlikely to differ significantly by personal characteristics. Data are taken from two sources. First, draft numbers according to birth dates are published in tables by the Selective Service Commission. Second, individuallevel data on smoking habits and other characteristics are taken from all of the annual National Health Interview Survey (NHIS) data sets between 1980 and 2002. These data sets include approximately 3 million observations, of which approximately 200,000 are men who were potentially eligible for the Vietnam drafts (born between 1944 and 1953). We have secured access to restricted NHIS data (with birth dates) via the National Center for Health Statistics’ Research Data Center. Analysis of the data is ongoing. Our findings will improve the understanding of the health-related consequences of military service. This understanding should be helpful for optimizing care for current veterans and making informed decisions related to military service policies.

Permalink

Parity for Whom? Exemptions and the Extent of State Mental Health Parity Laws

Presenter:

Tom Buchmueller

Authors:

Tom Buchmueller, Mireille Jacobson, Philip Cooper, Samuel H. Zuvekas

Chair: Ching-to Albert Ma; Discussant: Christopher Carpenter Tue June 6, 2006 13:45-15:15 Room 313

The fact that private insurance coverage is typically more restrictive for mental health than for other types of medical care was the impetus for the 1996 Mental Health Parity Act (MHPA). While this Federal law was hailed by some as landmark legislation, it is fairly weak in terms of its effect on employers and the health benefits they offer. In the years since the MHPA was enacted, there has been considerable state-level legislative activity aimed at enhancing mental health benefits in private insurance policies, with many states passing laws that go far beyond the MHPA. However, the number of workers who are subject to these parity laws is limited by the fact that self-insured employer-sponsored health plans are not subject to state regulations and several state laws provide further exemptions for small firms. Using repeated cross-section data from a set of nationally representative employer surveys conducted between 1997 and 2003, we examine trends in the number of insured workers affected by different types of parity laws. The data are from the Medical Expenditure Panel Survey-Insurance Component (MEPS-IC). The MEPS-IC surveys provide detailed information on firm characteristics and health insurance offerings, which allow us to explicitly account for the effect of exemptions for different types of firms. The results show that the percentage of private sector workers potentially affected by strong state parity legislation increased steadily over this period. The growth in potential coverage varied regionally, and by firm size. However, the exemptions for small firms and those that self-insure substantially reduced the number of workers actually affected by these laws. Roughly half of workers potentially covered by state parity mandates are actually affected by the laws. The dilution of parity legislation by these exemptions is one possible reason why prior studies have found these laws to have little effect on access to care and utilization.

Permalink

Implications of Mental Illness and Nativity for Labor Market Participation and Health Insurance Outcomes

Presenter:

Victoria Ojeda

Authors:

Victoria D. Ojeda, Richard Frank, Thomas McGuire, Todd Gilmer

Chair: Ching-to Albert Ma; Discussant: Haiden Huskamp Tue June 6, 2006 13:45-15:15 Room 313

Some research indicates worse socioeconomic status and labor market outcomes for the mentally ill persons due to factors such as lower human capital, income, and fragmented labor market participation. Nevertheless, there is little research on the relationship between mental illness and nativity, vis-à-vis labor market and health insurance outcomes. The diversity, size, and permanence of the immigrant population in the U.S. call for an increased understanding of how mental illness may affect immigrant adults’ connections to the labor market as well their access to both public and private sources of health insurance coverage. We examine health insurance coverage since unmet need for mental health conditions is high in the U.S., and also related to financial factors. Overall, immigrants have high uninsured rates, a situation that poses challenges to receipt of timely and affordable health and mental health services. This research pools data from the nationally representative 2001 and 2002 National Survey on Drug Use and Health and focuses on non-elderly working adults ages 18-64. The unweighted sample includes 61,418 U.S.-born and 8,528 foreign-born adults, of whom 635 meet DSM-IV criteria for Serious Mental Illness. We examine sociodemographic and employment characteristics of adults stratifying by mental health status and nativity, and use both descriptive and multivariate logistic regression analyses. Preliminary findings from logistic regression models suggest that mentally ill immigrant adults exhibit similar total involvement in the labor market as a healthy, native-born control group, even after controlling for demographic, human capital, and time-cost factors. In contrast, native-born adults with serious mental illness are more likely to be out of the labor market than healthy natives. Differences in the unadjusted rates of disability between foreign and native-born adults with serious mental illness are also evident. Unadjusted rates of health insurance coverage demonstrate that foreign-born persons, both with and without serious mental illness, are highly vulnerable to being uninsured. The proportion of uninsured persons is similar in both immigrant groups. There is a slightly higher rate of access to public coverage, though, among mentally ill immigrants. Findings from this research especially have implications for policy discussions relating to the provision of public coverage and public policy formulation as these affect vulnerable populations, including immigrants and their families.

Permalink

The Changing Demand for Mental Health Treatment

Presenter:

Chad Meyerhoefer

Authors:

Chad D. Meyerhoefer, Samuel H. Zuvekas

Chair: Ching-to Albert Ma; Discussant: Todd Gilmer Tue June 6, 2006 13:45-15:15 Room 313

The availability of new pharmacological treatment alternatives and rapid rise of managed behavioral health care organizations during the 1990s have significantly influenced utilization patterns of mental health services in the United States. While inpatient costs were reduced through managed care and more effective outpatient treatment, the use of pharmacotherapy by specialty mental health providers and primary care physicians increased substantially. We seek to understand the underlying behavioral and economic dynamics driving changing utilization patterns, and in particular, the rising demand for pharmacotherapy and substitution between this treatment approach and behavioral therapy. Therefore, we derive a mental and non-mental health care demand model that incorporates the relevant costs influencing consumption decisions, including out-of-pocket payments (cost-sharing) for ambulatory services, out-of-pocket prescription drug costs, and insurance premiums. The model makes use of the expected end-of-year price concept (Ellis, 1986; Ellis and McGuire, 1986), to derive theoretically consistent measures of out-of-pocket price. Our model of the joint demand for pharmacotherapy, behavioral therapy, and non-mental health treatment is estimated using the 1996-2002 Medical Expenditure Panel Survey (MEPS), a nationally representative survey of the U.S. civilian, non-institutionalized population. We exploit the longitudinal dimension of the MEPS to control for the endogeneity of out-of-pocket prices and health insurance coverage using a correlated random effects specification (Chamberlain, 1982). This allows us to relax the untenable assumption of standard random effect models that out-of-pocket price and health insurance are uncorrelated with the individual specific random effects as well as account for time invariant measurement error processes. In addition, we use a Zero-Inflated Ordered Probit specification to model the skewed distribution of ambulatory visits and prescription drug fills. Elasticity estimates from the model suggest that the demand for ambulatory mental health treatment is now much less elastic than it was during the RAND Health Insurance Experiment. In fact, consumers are now less responsive to the price of ambulatory mental health treatment than non-mental health treatment. The elasticity of demand for mental health drugs, however, is found to be relatively large, and higher than the elasticity of demand for non-mental health drugs. In order to check the robustness of our findings, we re-estimate Horgan’s two-part model (1986) of the demand for specialty providers and find that the price elasticity of demand she estimated for 1977 has likewise decreased substantially. We also provide alternative instrumental variables estimates. The findings of our study suggest that moral hazard problems associated with the coverage of ambulatory mental health services may be less severe than previously thought, with the implication that coverage could be expanded without substantially increasing use of mental health services.

Permalink

Will Mandatory Nurse Staffing Ratios Lead to Better Patient Outcomes in Hospitals?

Presenter:

Jingsan Zhu

Authors:

Julie Sochalski, R. Tamara Konetzka, Jingsan Zhu, Kevin Volpp

Chair: Sally Stearns; Discussant: David Grabowski Tue June 6, 2006 13:45-15:15 Room 325

U.S. Congress require mandatory minimum nurse staffing ratios in hospitals, and are motivated by several prominent cross-sectional studies reporting significant associations between higher nurse staffing levels and better patient outcomes. Yet legislating arbitrary mandatory nurse staffing ratios for all hospitals on the basis of evidence from cross-sectional studies without knowing if and when changing the staffing levels to meet such ratios improves outcomes could result in an inefficient allocation of resources that may or may not improve quality. Using 1991-2003 panel data from California hospitals, which are broadly representative of hospitals nationally, this study pursues the following aims: (1) determining whether increasing nurse staffing improves patient outcomes, (2) characterizing the marginal benefit of increasing staffing, and (3) identifying subgroups of hospitals that benefit the most from increasing staffing levels. Staffing measures include patients-per-nurse ratios for RNs, LVNs, and nurse aides, as well as RN skill mix (i.e., the percent of nursing personnel that are RNs). Four inpatient quality indicators—30-day mortality among patients with a principal diagnosis of acute myocardial infarction, stroke, hip fracture, or gastro-intestinal hemorrhage—and two patient safety indicators—failure-to-rescue rate and rates of post-operative pulmonary embolism or deep vein thrombosis among general and orthopedic surgical patients—that are from AHRQ’s inventory of quality indicators will serve as outcome measures. We use multivariate regression analyses to determine whether changes in nurse staffing over this twelve year period contributed to changes in patient outcomes. We control for differences in the mix of patients across hospitals using validated and well-established severity-adjustment methods, and for secular changes over time that could influence overall staffing levels. We characterize the marginal benefit of staffing increases by examining whether the returns to quality diminish as staffing levels increase and whether there are thresholds in the relationship between staffing and outcomes. Stratified analyses of subgroups of hospitals are undertaken to identify hospitals demonstrating the greatest improvement in outcomes with changes in nurse staffing. We employ standard panel data techniques of fixed or random effects as appropriate and correct standard errors to account for clustering of patients within hospitals. This study extends prior work through its longitudinal design, its use of refined measures of nurse staffing uniquely available in California and nurse sensitive outcome measures from AHRQ’s quality indicators, and its thorough examination of the marginal benefit of increasing nurse staffing across all and subgroups of hospitals. The findings will help to explicate the quality and cost implications of increasing nurse staffing in hospitals and guide payers and policy makers on the development of strategies to achieve their quality goals.

Permalink

Nursing Staff Reductions, Workload Increases, and Adverse Events in Florida Hospitals 1992-2004: New Variable and Longitudinal Approaches

Presenter:

Lynn Unruh

Authors:

Lynn Unruh, Keon Lee, Ning Zhang

Chair: Sally Stearns; Discussant: Joanne Spetz Tue June 6, 2006 13:45-15:15 Room 325

Nurse staffing/outcomes studies have primarily used common staffing measures and crosssectional or repeated measures approaches. However, given new explanatory variables such as “nursing staff reductions” and “workload,” a latent response variable called “patient outcomes,” longitudinal data, and the use of growth curve modeling, additional information and more causal conclusions can be derived. This study examines the relationship between nursing staff reductions, workload increases, and patient adverse events in Florida hospitals, 1992-2004. Hypotheses are: 1) An increase in nursing workload during a one year period is associated with a proportional increase in adverse events in that same year; 2) A 5 percent or more reduction of the licensed nursing staff during year ti-1 is associated with higher rates of adverse events and worse patient outcomes in year ti-1 and year ti; 3) A 5 percent or more reduction of the licensed nursing staff during year ti-1 is associated with a higher nursing workload in year ti-1 and year ti; 4) A hospital characteristic such as for-profit status is associated with higher nursing workloads, and vice versa for a characteristic such as teaching status; and 5) Higher nursing workloads in year ti are associated with higher rates of adverse events and worse patient outcomes in year ti. Nursing staff reductions in year ti is a dichotomous variable defined as a drop in licensed nursing staff of 5 percent or more measured from the beginning of time ti-1 to the beginning of time ti. Nursing workload is the ratio of adjusted patient days of care to the numbers of RNs and LPNs taken separately and together (licensed nurses). Patient days of care are adjusted for outpatient care and for patient turnover, which affects the intensity of nursing care. Adverse events are hospital-level rates of nursing sensitive events such as urinary tract infections, atelectasis, pneumonia, decubitus ulcers, and failure to rescue. The patient outcomes variable is a latent measure derived from the adverse events rates in the measurement model. Other measures are patient case mix, and hospital characteristics such as ownership, size, location and teaching status. Staffing measures and hospital characteristics are from the American Hospital Association Annual Survey. Adverse events and case mix are extracted from patient discharge records obtained from the Agency for Health Care Administration in Florida. Multi-wave, multivariate, latent growth curve modeling is used to define the relationships between the trajectories of endogenous and exogenous variables over 13 waves of data, time invariant and time-varying covariates included. The time invariant variables are the hospital characteristics, while the time varying variables are the rest of the variables. Rates of adverse events are transformed into approximately normally distributed variables prior to introduction into the model. Maximum likelihood estimation methods are used.

Permalink

Nursing Home Staffing and Quality of Care

Presenter:

Jeongyoung Park

Authors:

Jeongyoung Park, Sally Stearns

Chair: Sally Stearns; Discussant: Sally C. Stearns Tue June 6, 2006 13:45-15:15 Room 325

Most studies have found that higher nursing home staffing leads to higher quality of care. The implications of previous findings for an association between staffing and quality of care, however, may not reflect the true relationship. Existing estimates of the magnitude of the effect may be biased because most of these analyses were based on limited model specifications and did not control for the potential endogeneity of staffing. This study attempts to remove bias in estimates of the relationship between staffing and quality of care by controlling for endogeneity in staffing choices made by the facilities using fixed effects (FE) or fixed effects with instrumental variables (FE-IV). The analyses are conducted using facility-level data from the Online Survey and Certification Reporting (OSCAR) system from 1998 to 2003. OSCAR data are linked to data on specific market conditions and state policies. Quality is measured by facility-level total survey deficiencies, the incidence of contractures and pressure sores. Staffing level is measured by hours per resident day. Staffing mix is measured by the proportion of RNs hours compared to total staff hours. State policies, market (county) level nurse supply and demand variables are chosen as instruments to predict the staffing changes over time. Instrumental variables are incorporated in the model in order to (a) identify how nursing homes respond to the changes in the exogenous state policy shocks (i.e., state minimum staffing standards, Medicaid payment rates, a wage pass-through legislation), the relative competitiveness of the market and local resource constraints, and (b) investigate how these changes interact with staffing to yield changes in quality of care. The results from the study will be useful for understanding the contributions of staffing level and mix to the quality of nursing home care. The analysis also has two other dimensions of assessment. First, the estimation will allow an assessment of the effect of policies including state minimum staffing standards and wage pass-through provisions. Second, the magnitude and direction of the effect of staffing on quality of care is hypothesized to differ, contingent upon variations in facility characteristics (e.g., payer mix, case mix, size) and market environments (e.g., market competition, excess demand). Structural differences in the relationship between staffing and quality of care for different types of facilities may suggest different policy implications.

Permalink

Disease Status and Health-Specific Moral Hazard Effects

Presenter:

Cagatay Koc

Authors:

Cagatay Koc

Chair: Richard Hirth; Discussant: Alan Monheit Tue June 6, 2006 13:45-15:15 Room 326

The effect of insurance on the demand for medical care depends on the marginal utility of health since medical care services are inputs to health production. Given that marginal utility of health is highly related to the consumer’s initial health, additional medical care used by the insured, i.e., the moral hazard, likely varies by health. This health-specific moral hazard may vary by dimension of health. A consumer may exhibit different actions in his/her decisions toward care associated with serious chronic conditions than toward care associated with less serious medical conditions. Consequently, the extent of the health-specific moral hazard may differ across medical care services associated with different disease states. Using a clinical classification code based on ICD-9-CM condition codes, this paper classifies medical care services by disease status in three groups: services associated with serious chronic conditions, services associated with acute conditions of any sort and minor chronic conditions, and services associated with no medical conditions. Using data from the Medical Expenditure Panel Survey and its Medical Conditions supplement, the paper examines the variation in the health-specific moral hazard across disease-specific medical care services. The paper illustrates that the welfare implications of health insurance depends on both the type of medical care service and the disease status associated with that service. The empirical analysis suggests that both efficient and inefficient moral hazard may exist depending on the disease status and type of care. These results may have implications for designing disease-specific insurance policies across medical care services.

Permalink

Fairness to All Parties: Resolving Utilization Review Disputes under ERISA with the Contingent Claims Contract Model

Presenter:

Dahlia Remler

Authors:

Dahlia K. Remler, Kamiar Khajavi

Chair: Richard Hirth; Discussant: Alan Monheit Tue June 6, 2006 13:45-15:15 Room 326

Insurance coverage decisions today are frequently clinical medical decisions made on an individual basis, as part of utilization review (UR). Coverage denials under UR have become the subject of legal disputes. The law surrounding these cases has proven to be confusing and unsatisfactory. The 2004 Supreme Court decision, Davila, has essentially left managed care organizations (MCOs) associated with ERISA-protected plans with no liability for the consequences of UR-based coverage denials. Such protection is alleged to bolster MCO automatic denial policies that do not consider the merits of individual cases. Despite widespread dissatisfaction with this state of affairs, both Congress and the courts are reluctant to expand MCO liability, apparently because they fear something like the medical malpractice situation will occur, undermining costcontainment strategies, raising health care expenditures, raising premiums, and increasing the number of uninsured. We review the history of court decisions on UR coverage denials, illustrating how all attempts to disentangle the medical and contractual/economic elements have proven impossible, resulting in the recent Davila decision. We build on the view that, for these UR disputes, contract law is preferable to tort law (Havighurst 1995, Danzon 1997). We illustrate how contract law with consequential damages could be applied to individual medical decisions using a contingent claims contract (CCC) model. We examine the ways in which real world health insurance contracts could become more like CCCs, by enhancing clinical specificity through a variety of means. We also analyze how transactions costs, selection, cognitive limitations, behavioral economics, and technological change limit the ability of real world health insurance to mimic CCCs. Thus, any real world contract will inevitably fall far short of such an ideal. Nonetheless, we illustrate the further value of the CCC model in providing courts with a conceptual test that can be used to settle UR-based coverage disputes. Specifically, in an economic framework, patient/plaintiffs are entitled to coverage for more expensive care if they have paid the higher actuarially fair premium needed to cover the more expensive care. The courts can ask if plaintiffs could reasonably think that they had purchased such coverage and if insurers reasonably thought that they had sold it. Such after-the-fact assessments of “meeting of minds” are common in contract law. We illustrate the forms of evidence that courts could use in coming to such decisions.

Permalink

The Effect of Private Health Insurance on Health Care Purchases and on Health in Brazil

Presenter:

John Nyman

Authors:

John A. Nyman, Nathan Barleen

Chair: Richard Hirth; Discussant: Alan Monheit Tue June 6, 2006 13:45-15:15 Room 326

The goal of health insurance programs in developing countries is often to allow citizens to gain access to additional health care, but according to conventional theory, this additional care–the moral hazard–is welfare decreasing. This paper uses data from the Living Standard Measurement Study (LSMS) Survey for Brazil, collected by the World Bank during 1996-7, to estimate the net welfare gain that is generated by the additional care that is in turn purchased because of having private health insurance. The paper uses logit regression analysis to show that those with private health insurance purchase more health care. Separate analyses are reported for those with an acute disease, a chronic disease, and for all respondents. Having established that those with insurance purchase more health care, a probit regression analysis is used to determine the effect of health insurance on health, as measured by changes in the probability of falling into the self-reported health status categories of excellent, very good, good, fair, and poor. We find that those with health insurance are more likely to categorize themselves as excellent, very good, or good, and less likely to categorize themselves as fair or poor. The welfare gain is determined by associating health-related quality of life scores with the various self-reported health states, and determining the increase in the average quality of life score that is generated by the change in the probabilities of a respondent classifying himself or herself in the various states that is caused by having private insurance. The change in health related quality of life score is then compared to the private insurance premium, a conservative, upper bound estimate of the cost of the moral hazard. We find that the cost of generating these additional quality-of-life gains appears to be consistent with incremental cost-utility ratios that would be deemed welfare increasing for new medical technologies. Sensitivity analysis is performed with regard to estimates of the cost of private insurance in Brazil, of various estimates of the healthrelated quality of life of self-reported health status, and of the value of a quality adjusted life year in Brazil, in order to determine the robustness of the results. The results generally suggest that private insurance in Brazil is welfare increasing.

Permalink

Why So Few Integrated Plans? Simulating the Impact of an Alternative Health Insurance Strategy

Presenter:

Katherine Ho

Authors:

Katherine Ho

Chair: David Dranove ; Discussant: Andrew Sfekas Tue June 6, 2006 13:45-15:15 Room 332

Permalink

The Welfare Consequences of Hospital Mergers

Presenter:

Robert Town

Authors:

Robert Town, Douglas Wholey, Roger Feldman, Lawton R. Burns

Chair: David Dranove ; Discussant: David Dranove Tue June 6, 2006 13:45-15:15 Room 332

Permalink

Patient Learning and Advertising in the diffusion of Cox-2 Inhibitors

Presenter:

Ginger Jin

Authors:

Pradeep Chintagunta, Renna Jiang, Ginger Jin

Chair: David Dranove ; Discussant: Leemore Dafny Tue June 6, 2006 13:45-15:15 Room 332

Permalink

Effects of Time Costs on the Quantity and Intensity of Physical Exercise

Presenter:

David Meltzer

Authors:

David Meltzer, Bapu Jena

Chair: David Meltzer Tue June 6, 2006 13:45-15:15 Room 335

The cost of time is the major determinant of the cost of exercise, but increasing intensity of exercise provides an approach to reduce the time costs of a given amount of exercise. This paper examines the choice of mode f exercise from this perspective of producing exercise in the context of high time costs. Our results suggest that increasing the intensity of exercise is an important response to higher time costs.

Permalink

The Effect of Retirement on Weight Gain

Presenter:

Darius Lakdawalla

Authors:

Darius Lakdawalla

Chair: David Meltzer Tue June 6, 2006 13:45-15:15 Room 335

Work and occupation are powerful predictors of weight, since individuals’ exercise patterns are heavily influenced by the kind of work they do. As a result, retirement can have very different effects on weight and health for different kinds of workers. We show that workers retiring from sedentary jobs end up losing weight, as they are no longer paid to be sedentary during the day. Conversely, workers retiring from strenuous jobs gain weight, since they are no longer paid to exercise. In spite of similar weight trajectories pre-retirement, retirees from strenuous jobs gain about 0.5 more BMI units than retirees from sedentary jobs.

Permalink

Title IX, School Sports Participation and Adolescent Health

Presenter:

Robert Kaestner

Authors:

Robert Kaestner

Chair: David Meltzer Tue June 6, 2006 13:45-15:15 Room 335

The primary goal of this research is to obtain estimates of the effects of the dramatic increase in girls’ participation in high school sports, as a result of Title IX legislation, on the physical activity, health and health behaviors of adolescent women. This large and unprecedented increase in girls’ sports participation provides a unique opportunity to study the effects of school-based interventions targeted at increasing physical activity of youth.

Permalink

The Impact of Diabetes and Diabetes Management on Labor Productivity: A Genetic IV Approach

Presenter:

Henry Brown

Authors:

Henry Brown

Chair: Michael Grossman; Discussant: Curtis Florence Tue June 6, 2006 10:45-12:15 Room 121

Authors: H. Shelton Brown (shelton.brown@utb.edu), School of Public Health, University of Texas, Jose A Pagan, University of Texas-Pan American, Craig Hanis, School of Public Health, University of Texas

Title: The Impact of Diabetes and Diabetes Management on Labor Productivity: A Genetic IV Approach

Rationale: Diabetes has been shown to have a detrimental impact on employment and labor market productivity, which results in lost work days and lower wages. However, unobservables are correlated with diabetes and labor productivity, leading to endogeneity problems in estimates. Brown, Pagan and Bastida have recently shown that ignoring endogeneity results in an overestimate (underestimate) of the negative impact of diabetes on female (male) employment (2005). With endogeneity addressed, there was no effect of diabetes on female working propensity in their study. It is possible that gender differences in diabetes management, where female diabetics adhere to management better than males, may account for the differential impact of diabetes on labor productivity.

Objective: To determine whether diabetes, when managed, is related to labor productivity by gender.

Methodology: To account for the endogeneity of diabetes, we use family history of diabetes as genetic instrumental variables in determining diabetes. In order to determine the level of self-management, diabetes is interacted with laboratory-measured blood-sugar levels. Note that in our data, diabetes is measured rather than self-reported. Wage equations and working propensity equations are estimated, by gender. The data are from a new random sample from a largely Mexican-American community in Texas on the border of Mexico.

Results: Self-management is an important, but heretofore, unobservable effect in labor supply. Our results are particularly relevant in the case of populations where genetic predisposition has an important role in the etiology of diabetes.

Conclusions: Our results shed light on the reasons for the differential impact of diabetes on male and female labor productivity. Further, our results inform policy-makers about how to allocate health care resources between prevention of diabetes and diabetes management for those already diagnosed with diabetes.

Disclosure information: Nil.

Permalink

Do Workplace Health Promotion Programs Improve Health-Related Behavior?

Presenter:

Curtis Florence

Authors:

Curtis Florence

Chair: Michael Grossman; Discussant: Henry Brown Tue June 6, 2006 10:45-12:15 Room 121

Rationale: Many employers offer workplace health promotion (WHP) programs, such as smoking cessation programs and exercise facilities. There at least two possible economic rationales for employers offering these programs. First, they may offer them in an attempt to improve the health and productivity of their workforce. Second, they may offer them as non-wage compensation to compete for workers. While these two rationales are not mutually exclusive, if the “employment benefit” rationale dominates the “health improvement” rationale, workplace health promotion programs may simply provide an outlet for healthier behavior to workers who would have participated in these activities anyway. For example, and employer who provides exercise facilities may attract workers who are likely to exercise, and offering the facility will not change the probability a worker exercises.

Objective: The objective of this paper is to estimate the effect of workplace health promotion programs on health related behavior, controlling for the potential endogeneity of the availability of the program.

Methodology: The 1998 National Health Interview Survey Prevention Module has extensive information on the availability of WHP programs and health related behavior for workers. It is possible to determine if a worker has access to a given WHP program and whether or not they participate in the given health related activity (either at work or outside of the workplace). I estimate bivariate probit models with two outcomes: whether or not the worker has access to a given type of WHP program, and whether or not they participate in the health related behavior. The WHP outcome is an endogenous variable in the health behavior equation. The model is identified by the number of employees at the workplace, which is assumed to affect the likelihood of the program being offered, but does not have an independent effect on health related behavior. I estimate the model for the following types of programs and health related behaviors: smoking cessation, exercise, screening for hypertension, hyperlipidemia and cancer, and nutrition/weight control.

Results: When ignoring the endogeneity of program availability, all of the programs except for smoking cessation are shown to be positively associated with increases in the given health-related behavior. However, when the endogeneity of program availability is controlled for, the only programs that show a significant and positive influence on health-related behaviors are the three types of screening programs. Exercise and nutrition/weight control WHP programs do not increase the rate of the targeted health-related behavior.

Conclusions: The continuing growth in health care expenditures has lead many employers to offer WHP programs. The results of this study show that these programs do not uniformly increase healthy behavior. Smoking cessation programs, exercise facilities and programs and nutrition/weight loss programs do not increase the likelihood that workers will participate in these health promoting activities-at least as these programs are currently constituted. The results could indicate a need for greater information and outreach to encourage employees to access the programs they are offered.

Permalink

Effect of Disability on the Employment of Other Household Members

Presenter:

Stephen Mennemeyer

Authors:

Stephen Mennemeyer

Chair: Michael Grossman; Discussant: Reagan Baughman Tue June 6, 2006 10:45-12:15 Room 121

Rationale: While much work has been done on the return-to-work of persons with disabilities, surprising little is know about how the employment of other members of a household is affected by the onset of a disability.

Objective: We examine how household labor decisions are affected by the onset of a disability and by the subsequent “aging” of the disability condition. The main interest here is the extent to which the disability encourages others to go to work to offset the income lost by the person with a disability or to stay home to help with caregiving.

Methodology: We use the Survey of Income and Program Participation 1996 to identify households where members have disabilities, either old (existing prior to the SIPP survey) or new (concurrent with the Survey). We estimate panel data models of employment for both the person with the disability and other household members that control for the type and age of the disability as well as various characteristics of the household and its members.

Results: Results are currently available for stroke with other conditions now under analysis. We find that the occurrence of a stroke tends to permanently reduce the labor supply of the victim. For the non-victims, there is an asymmetric effect: male household members tend to only mildly decrease their employment whereas females make a substantial reduction in employment that tends to persist.

Conclusions: Estimates of the lifetime cost of disabilities need to take account of the lost earnings of other household members

Permalink

Reducing Direct Care Turnover: Do Wage Subsidies Matter?

Presenter:

Reagan Baughman

Authors:

Reagan Baughman, Kristin Smith

Chair: Michael Grossman; Discussant: Stephen Mennemeyer Tue June 6, 2006 10:45-12:15 Room 121

Rationale: The aging of the baby boom cohort is likely to put a significant strain on the market for long term care services provided by direct care workers like nursing and home health aides. In fact, many states are already experiencing shortages and high turnover rates that are affecting the quality of patient care. In contrast to a relatively large literature devoted to the labor market for RNs, there is little in the way of systematic analysis of the labor market for direct care workers.

Objectives: The objectives of this paper are twofold. The first objective is to provide a descriptive analysis of the dynamics of direct care workforce behavior, including estimates of employment durations for direct care workers and comparison occupation groups. The second objective is to assess the effects of two types of wage subsidies on the length of time that direct care workers stay in a given job (or stay in the direct care workforce). One of wage subsidy policies that we consider, the Earned Income Tax Credit, represents an income rather than occupation-targeted subsidy. The other wage subsidy that we consider, “wage pass-through” provisions in state Medicaid programs, is a policy directed exclusively at direct care workers.

Methodology: We use data covering the years 1996 to 2000 from the 1996 Survey of Income and Program Participation to measure employment spells for the workers in the sample. This dataset provides information on 786 workers in direct care occupations and allows us to observe approximately 80 percent of direct care employment spells at the monthly level without censoring. We will use these spell measures to estimate discrete-time hazard models of the effects of both income-targeted and occupation-targeted subsidies on employment durations. The fact that the SIPP identifies state of residence for all but 5 states allows us to identify the effects of policy using state-level variation. By 2000 (the last year in our panel), 13 states had state-level EITCs, and many of these states had changed the real values of their credits during the preceding four years. The majority of the 26 states that currently have wage or benefit pass-throughs in their Medicaid programs either implemented or expanded programs during the late 1990s.

Results: Descriptively, we find that spells of employment for direct care workers are slightly lower than those for other workers. The median spell for direct care workers (all workers) is 17 months (19 months) and the mean spell is 40 months (51 months). Spells for direct care workers are longer than those for child care workers (median = 10 months; mean = 22 months) but much shorter than those for a group of better-educated health care workers, including R.N.s and therapists (median = 32 months; mean = 58 months). The second part of the analysis, focusing on the impact of wage subsidies, is currently in progress.

Permalink

Are Care Givers More Risk Averse Than Patients? Acceptable Tradeoffs between Efficacy and Adverse-Event Risks

Presenter:

Semra Ozdemir

Authors:

Semra Ozdemir, F. Reed Johnson, Carol Mansfield, Steven Hass, Jeff White

Chair: James Burgess; Discussant: TBA Tue June 6, 2006 10:45-12:15 Room 213

Understanding patients’ own risk perceptions and their willingness to accept risks in return for treatment benefits can help inform risk-management decision making. In the case of children, parent caregivers are responsible for treatment decisions and their risk tolerance may be quite different than adult patients’ risk tolerance.

This study compares utility-theoretic estimates of adult patients’ and parents’ willingness to accept adverse event risks in return for increased treatment efficacy for the treatment of an inflammatory bowel disease (IBD). There are a number of medications used to treat symptoms of IBD. However, treatments vary in efficacy and may be associated with risks of serious adverse events including death from tuberculosis, polyomavirus infection, and lymphoma. Treatment decisions thus often require weighing treatment efficacy against adverse-event risks.

Stated-choice (SC) or choice-format conjoint analysis is increasingly being used by health economists to quantify the relative importance of treatment processes and outcomes. SC methods postulate that the utility a person derives from a product or service can be expressed as a function of the positive and negative features of the product or service. The pattern of observed choices among a series of hypothetical treatment and outcome comparisons reveals the implicit relative importance of each attribute and marginal rates of substitution among attributes. The estimated marginal rates of substitution between efficacy and adverse-event risks facilitate calculating maximum acceptable risk (MAR) for various changes in health states.

This study used a pretested SC instrument to elicit patients’ and parents’ preferences for a range of treatment attributes, including various levels of both the benefits and risks of the treatments. The analyzed data were from 345 patients over the age of 18 and 150 parents of a child under the age of 18. Survey subjects were randomly sampled from a large internet panel.

We found that both patients and parents are willing to accept adverse-event risks in return for treatment efficacy. Risk tolerance increases with larger improvements in treatment efficacy and varies by the type of adverse-event risk. We find significant differences between MARs of patients and parents. Parents are more risk averse and have MARs about half as large as patients. Patients with more severe symptoms have higher MARs, but this is not the case for parents whose child has more severe symptoms.

These results confirm that caregivers and patients have different risk perceptions. Parents are more cautious about exposing their child to health risks, which is consistent with previous studies. Nevertheless, both parents and patients are willing to accept some risk in return for increased efficacy. These results may help inform clinical and regulatory decision making, as well as help identify more effective risk-management strategies.

Permalink

Psychosocial Consequences and Perceived Risk Following Screening for Lung Cancer

Presenter:

Margaret Byrne

Authors:

Margaret Byrne, Mark Roberts, Joel Weissfeld

Chair: James Burgess; Discussant: TBA Tue June 6, 2006 10:45-12:15 Room 213

Background: New technologies offer a promise of earlier detection and thus earlier treatment for lung cancer. Although there is still debate as to whether screening and early detection in fact decreases mortality from lung cancer, the incidence of individuals being screened for lung cancer has been increasing in recent years. However, 30-40% of individuals screened with scanning CT receive a finding of an “indeterminate non-calcified lung nodule”, with a recommendation for further periodic (annual) screening. The goal of this research is to measure the psychological and health care resource utilization consequences for individuals participating in a lung cancer screening program.

Methods: 400 participants were enrolled as they enrolled in a study exploring the efficacy of lung cancer screening. Participants completed surveys prior to screening and immediately following receipt of the screening results. Survey instruments included questionnaires on: health perceptions, cancer worry, social support, anxiety, and health care utilization. Demographic information and screening results were obtained from the original lung cancer screening study. Screening results were classified as 1) no diagnostic follow-up or physician referral recommended (n=200); 3) advise periodic follow-up CT for one or more indeterminate non-calcified lung nodule (n=133), and 4) strong physician referral for lung cancer suspicion (n=25). The results below consider the change in survey scores between the baseline and post-screening surveys.

Results: The overall score on health perceptions fell significantly for all participants, but most for those in categories 3 and 4. Similarly, the effects of worrying about cancer (e.g. having trouble sleeping) increased significantly for all categories, and the level of worry about having cancer increased for categories 3 and 4. There were no significant changes in any measures of social support. Individuals in categories 3 and 4 had significantly higher measures for state anxiety, and those in category 1 had neither a reduction in anxiety, nor in the level of worry that they might have or get cancer.

Individuals’ perceived risk that they had lung cancer currently (~20%) or would develop cancer over their life time (~30%) was similar for all categories at baseline. This is substantially higher than the true risk of approximately 1-3%. Following screening, perceived risk in category 1 individuals dropped slightly, and did not change for category 3 participants. Perceived risk in category 4 individuals increased significantly to 39% for current cancer and 49% for every developing lung cancer. Actual risk of cancer in category 4 individuals, which was relayed to them following screening, is approximately 15%, indicating that individuals do not update their believes efficiently.

Conclusions: Anxiety and worries that they might have cancer increased in those individuals who received an “indeterminate” finding from lung cancer screening. Individuals with a negative screen did not appear to have less anxiety or worries following screening. Individuals did not rationally update their perceived risk of cancer following screening and information dissemination. Recommendations for screening and screening decisions should be informed by these results.

Permalink

Quantifying Patients' Risk-Benefit Tradeoff Preferences: A Conceptual and Empirical Comparison of Methods

Presenter:

Reed Johnson

Authors:

Reed Johnson, George Van Houtven, Carol Mansfield, David W. Miller

Chair: James Burgess; Discussant: TBA Tue June 6, 2006 10:45-12:15 Room 213

OBJECTIVE: To compare conceptual models, empirical measurement, and results of alternative methods for measuring patients’ willingness to trade off risks of severe adverse events for specified health gains. METHODS: We define and compare the theoretical foundations of standard gamble (SG) and multi-attribute conjoint analysis (CA) methods. SG derives from von Neumann-Morgenstern expected-utility theory, while CA applies McFadden random-utility theory to hypothetical choices. We define theoretical conditions under which the two methods provide equivalent measures of health preferences, including linearity, separability, and risk-neutrality. We evaluate accepted empirical methods used in SG and CA studies and propose methods for incorporating risks as CA treatment attributes. We then compare empirical maximum acceptable-risk estimates from CA studies of multiple sclerosis and Crohn’s disease patients with and without restrictive SG assumptions, as well as with published SG estimates from other disease interventions. RESULTS: We find that SG can be used to estimate MARs for specific health outcomes only by imposing more restrictive assumptions on patient preferences than CA methods require. We show that CA methods can be used to test various theoretical restrictions imposed by the SG assumptions and find that risk neutrality and linearity are rejected statistically in most cases. By imposing SG assumptions on CA results and by comparing CA results to published SG estimates for chronic conditions, we find that SG assumptions increase MAR estimates by 20% to 150% relative to those obtained by CA methods. CONCLUSIONS: CA methods can be used to replicate SG tradeoff tasks and to test the restrictions required to interpret SG estimates as risk-preference measures. CA offers a more flexible and conceptually rigorous method than SG as conventionally applied for measuring treatment preferences and risk-benefit tradeoffs. Most importantly, multiattribute CA methods can more realistically simulate clinically relevant risk-benefit tradeoff ?_????

Permalink

Nurse staffing and skill mix and association with cost in the Veterans Health Administration

Presenter:

Anne Sales

Authors:

Anne Sales, Yu-Fang Li, Elliott Lowy, Chuan-Fen Liu

Chair: TBA; Discussant: TBA Tue June 6, 2006 10:45-12:15 Room 225

Rationale: Several large scale studies have found associations between nurse staffing and mortality outcomes for hospitalized patients, aggregated to the facility level. In this study, we provide the first large-scale analysis using nursing unit-level data to examine staffing levels, skill mix, and costs of care.

Objective: To examine the association between nurse staffing, skill mix, and costs of care at the unit and facility levels. We define skill mix as the mix between different types of nursing provider, with different levels of training and receiving different wage rates.

Methodology: Data came from several sources: DSS nursing labor input files (ALBCC); VA National Patient Care Databases for demographic, comorbidity, complication and utilization data on all patients admitted to VHA inpatient acute care between 2/03-6/03; Decision Support System (DSS) TRTIPD files, a DSS extract file linking inpatients to nursing units; and DSS cost extracts for patient level costs of care for each admission. We estimate costs of inpatient care for each admission by regressing on patient risk of complications (estimated in a prior analysis), and staffing hours by type of nurse provider (registered nurse- RN, licensed practical nurse- LPN, nurse aide- NA), divided into tertiles, and including interaction terms among staffing level tertiles of different types of nurse providers. We stratify the analysis by whether the first inpatient unit was intensive care or non-intensive acute care, and will conduct sensitivity analyses varying patient cohort definition, for example, by stratifying by whether or not the patient received any intensive care vs. those who have no intensive care. We will use GLM as our primary estimator, controlling for clustering at the unit and facility level, as well as multilevel modeling using Stata’s recently added xtmixed procedures with two levels of clustering, and controlling for market variation in wages using real dollars. We will accomplish this last piece by controlling for the Medicare/Medicaid wage index, which is specific to local markets in the United States.

Results: The analyses includes 126,382 patients from 463 nursing units in 119 VAMCs, admitted between February and June 2003. 184 units are intensive care, and 279 non-intensive acute care units. In prior analyses, we have found that skill mix and staffing have a non-linear, U-shaped association with mortality risk in both ICU and non-intensive care patients, and that multi-level modeling, adequate patient risk adjustment, and careful modeling of non-linear relationships are critical. One major finding in the analyses to date, which are ongoing, is that the curvilinear relationship in skill mix between different types of nurse provider has a significant impact on risk service utilization, including length of stay. As we complete our analyses, we anticipate that this difference will result in major differences in cost of care in nursing units using different skill mixes, controlling for quality of care. Differences in wage rates between RN and NA in the United States can be as high as five times greater wages for RNs than NAs. Analyses are ongoing, and we anticipate completion of these cost analyses by February 2006.

Permalink

The Effects of Competition on Community-Based Nursing Wages

Presenter:

Dara Zarnett

Authors:

Dara Zarnett, Audrey Laporte, Eric Nauenberg, Diane Doran, Peter Coyte

Chair: TBA; Discussant: TBA Tue June 6, 2006 10:45-12:15 Room 225

In 1997, Ontario’s Community Care Access Centres (CCACs) were given the authority to purchase home care services from for-profit and not-for-profit providers through a competitive bidding process. The purpose of this reform was to encourage competition between home care providers, with the twin objectives of lowering costs and increasing service quality. This paper investigates how competition amongst home care service providers in Ontario affected the gross nursing wages of private home care providers between 1995 and 2000. Regression analyses were performed to ascertain the relationship between measures of competition (Hirschman-Herfindahl Index (HHI), Participant Index (PI)), profit status of the provider agencies, and gross nursing wages for Registered Nurses (RN) and Registered Practical Nurses (RPN), while holding other factors constant. Using the HHI and PI as measurements of competition, RN gross wages significantly decreased by 11.7% and 0.9% respectively. The PI measured a significant decline in RPN gross wages of 1.5% in an increasingly competitive market. Furthermore, for-profits paid lower gross RN wages, by 6.8%, compared to not-for-provider agencies. It was also observed that there the distribution between for-profit and not-for-profit agencies has changed dramatically, and there have been large increases in the number and volume of for-profit service contracts. This implies that the for-profit agencies are winning more contracts and crowding out the not-for-profits. These findings correspond with previous literature which reports that competition can lead to lower staff wages and differential payments by provider types. These findings have the potential to lead to a destabilized labour market, which can negatively impact quality of care.

Permalink

Which Patients with Congestive Heart Failure (CHF) Benefit Most from Nurse-based Disease Management?

Presenter:

Paul Hebert

Authors:

Paul Hebert, Jane Sisk

Chair: TBA; Discussant: TBA Tue June 6, 2006 10:45-12:15 Room 225

Rationale: Several recently conducted randomized controlled trials (RCT) have demonstrated the benefits of a nurse-based disease management programs for patients with CHF. Most, although not all, trials have found that patients randomized to a nurse management program have fewer subsequent hospitalizations than patients in usual care. However, no studies have addressed whether the benefits of nurse management are limited to patients with certain clinical or socio-demographic characteristics. If this is true, administrators and policy makers may improve the cost effectiveness of nurse management by targeting patients most likely to benefit.

Objective: The objective of this paper is to reanalyze data from a successful RCT of nurse management for CHF to assess which patients benefited most from nurse-based disease management.

Methods: In the RCT, 406 patients with CHF from the four hospitals serving East and Central Harlem in New York City were randomized to nurse management (n=203) or usual care (n=203). Nurse managers gave patients information on CHF and tools designed to improve self-management, and worked with patients’ clinicians to optimize drug therapy. The outcome for this reanalysis was physical functioning as measured by the Short Form 12 (SF-12) questionnaire, which was measured at baseline and every 3 months for 1 year. We estimated the benefit of nurse management for each patient in the treatment group by estimating a random effects linear regression of SF-12 scores on time and patient characteristics at baseline for patients in the control group. We then used parameters from this model to estimate what functioning would have been for each patient in the nurse group if he/she had not received the intervention. The difference in actually and expected functioning for patients in the treatment group was then regressed on baseline patient characteristics to find characteristics of patients who benefited more than others.

Results: 406 patients provided 1693 follow-up observations. The patients were diverse in terms of age (59.4±13.7), race/ethnicity (46% black, 32% Hispanic, 15% White), education (46% < high school degree), and health literacy (28.9% inadequate). Nurse patients maintained better SF-12 functioning scores than control patients throughout the study (difference at 12 months +4.0, 95% C.I. (2.2, 5.9)). Preliminary results, suggest that Black patients (+4.1 ; p=0.021) and Hispanic (+5.9; p=0.005) benefited more than white Non-Hispanic patients. Patients age 40-59 faired better than patients age <40 (p<0.001). We found little evidence that patients with low health literacy skills, those who had been recently hospitalized, or those with more severe heart failure at baseline benefited more from nurse management.

Conclusion: Although these results are preliminary, they suggest that re-analyses of data from nurse management RCT may help to target patients with CHF who are most likely to benefit from nurse management programs.

Permalink

Middle School Behavioral Problems, High School Completion and Employment Outcomes in Early Adulthood

Presenter:

Mustafa Karakus

Authors:

Mustafa Karakus, David Salkever, Eric Slade, Nichalos Ialongo

Chair: Dan Polsky; Discussant: Daniel Polsky Tue June 6, 2006 10:45-12:15 Room 226

Rationale: High school completion is an important educational outcome in order to receive higher educational credentials and compete for better occupational prospects. ‘No Child Left Behind Act’ of 2002 aims to improve high school graduation by focusing on improvements of achievement scores such as gpa and standardized tests. However, simply focusing on these measures may not be sufficient to improve outcomes in high school graduation. Government data suggests that majority of adolescents who drop out of school present some level of behavioral problems. Thus, paying special attention to the needs of students with behavioral problems can increase the likelihood of education achievement and future occupational prospect.

Objectives: The objective of this study is to use data from the Fourth Follow-up to the National Education Longitudinal Study of 1988 (NELS 2000) to explore a causal relationship between middle school behavioral problems, high school completion, and subsequent labor market outcomes.

Methodology: We apply several statistical methods (multinomial logit, nested logit, and bivariate probit) that include a reduced form model as well as a structural model to test effects of middle school behavioral problems on high school graduation and future employment outcomes. We also test the possibility that both educational achievement and employment outcome may be caused by a common unobserved factor. Thus, our analysis includes school level instrumental variables and implements an instrumental variables methodology to control for possible endogeneity issues.

Conclusions: Preliminary results suggest that behavioral problems are important factors in explaining high school graduation and future employment prospects. Middle school behavioral problem indicators such as performing below ability, being passive or being disruptive in class and those students who are frequently absent are less likely to graduate from high school. Evidence on this relationship is stronger for male students. In addition, connection between high school graduation and employment outcome also differs by gender. Interestingly, unobserved characteristics that cause male students to drop out of school also produce more likelihood of employment. We do not observe the same result among female students. Policies that aim to improve educational achievements and future employment prospects should not be based on only observable characteristics such as gpa and other standardized test scores. We should pay special attention to the effects of behavioral problems and gender specific factors on educational and occupational success.

Permalink

The Impact of Illness on Family Labor Supply and Earnings

Presenter:

Virginia Wilcox-Gok

Authors:

Virginia Wilcox-Gok

Chair: Dan Polsky; Discussant: Kosali Simon Tue June 6, 2006 10:45-12:15 Room 226

As the population of the United States ages, Alzheimer’s disease has become, and will continue to be, a major public health concern. Because of the millions of individuals afflicted by AD, there are many more millions of family members indirectly burdened by the disease. The goal of this research is to estimate the indirect effects of illness on the labor market outcomes of family members.

Cost-of-illness studies examining labor market costs have generally focused on the primary caregiver and used small data sets that are not representative of the United States population. In this research, we estimate the labor market effects of illness for all family members using data from the Medical Expenditure Panel Survey (MEPS). MEPS is a large, nationally representative data set that contains information describing labor force outcomes, sociodemographics, and medical diagnoses and measure of severity of illness. By are able to link the records of persons living in the same household, thereby permitting examination of the effects of Alzheimer’s disease on the labor market outcomes of family members.

Our measures of labor market outcomes are labor force participation, employment, hours worked, and earnings. First we estimate a wage equation for all employed family members and predict a market wage for individuals who are not employed. We then estimate the effects of illness on labor force participation, employment, hours of work, and earnings using the predicted wage in a model allowing self-selection out of the labor force. This analytical model controls for the endogeneity between the market wage rate and the decision to be a fulltime caregiver.

Our preliminary findings indicate that the presence of a family member with Alzheimer’s disease increases the probability that a woman will drop out of the labor force. Further, among women who work, the presence of a family member with Alzheimer’s disease has a negative effect on weekly hours of work and annual earnings. In comparison, while men’s labor force participation is unaffected, we find that the average effect for men is to increase hours of work per week and increase annual earnings. This is the first empirical evidence that male family members respond to the family member’s illness by increasing hours of work. Our results indicate that average family earnings fall, reflecting the relatively larger magnitude of the effect on women’s earnings.

By examining the effects of the disease on all members of the family, rather than solely the primary caregiver, we provide a broader and more accurate picture of the family burden of illness. Our estimates demonstrate that there is substitution among family members to compensate for lost market earnings when one member reduces labor supply to provide caregiving. However, this compensation is insufficient to totally offset the loss in family earnings.

Permalink

The Effects of Mental Illness on Schooling

Presenter:

Noelle Molinari

Authors:

David Kalist, Noelle-Angelique Molinari

Chair: Dan Polsky; Discussant: Rosalie Pacula Tue June 6, 2006 10:45-12:15 Room 226

We examine how psychiatric disorders such as major depression and mild chronic depression (i.e., dysthymia) affect educational attainment, with the results presented separately for men and women. The Diagnostic Statistical Manual IV characterizes the psychiatric diagnoses. Using the first wave of the National Epidemiologic Survey on Alcohol and Related Conditions 2001-2002, we find that early onset of psychiatric disorders, especially before the age of 19, negatively affects level of schooling. Estimates from logit regressions indicate that for men early onset dysthymia reduces the probability of finishing high school by approximately 17 percent, whereas early onset major depression reduces the probability by 8 percent. For women the effect of early onset dysthymia is estimated less precisely; however, the negative effect on schooling of early onset major depression is similar to men. In addition to estimating standard logit models, we estimate censored ordered logit models, which take into account that some of the individuals in the sample are currently attending school. We further present results, from survival analysis regressions, showing how the onset of psychiatric disorders affects time to graduation.

This paper also adds to the economic literature on suicide by controlling for suicide attempts and thoughts of suicide. In general, we find that attempted suicide is associated with lower levels of schooling.

Permalink

Informal payments in developing countries

Presenter:

Ting Liu

Authors:

Ting Liu

Chair: Richard Scheffler; Discussant: Chris Garmon Tue June 6, 2006 10:45-12:15 Room 235

Rationale: Informal payments, which are payments to physicians in cash or in kind made outside official channels for services that are covered by the public health care system, are a widespread phenomenon in many developing countries. Various policies have been proposed by policy makers to deal with this problem. However, few studies investigate the rationale behind informal payments and the welfare implications of different policies.

Objectives: This paper examines the rationale for why patients offer informal payments and compares the effectiveness and welfare implication of various policies at alleviating this problem.

Methodology: We adopt a game theoretical approach to model the rise and spread of informal payments. There are two doctors, a skilled doctor and a naive doctor; two patients, a needy patient and a less needy patient. Each patient demands one unit of service while each doctor offers one unit service. Patients can distinguish doctors and prefer to be seen by the skilled doctor; however, the needy patient has a bigger incremental increase in utility if he is treated by the skilled doctor instead of the naive doctor. Government can not observe the difference in doctors, therefore set the same price for both types. Excess demand thus leads to competition between the two patients who simultaneously offer informal payments to the skilled doctor. Informal payments become sunk cost once patients pay. The skilled doctor treats the one who offers a higher informal payment. In case of a tie, he randomly picks one patient to treat. We derive the Nash Equilibrium in the game, which has implications for the size and the pattern of informal payments. Furthermore, we compare the welfare implication of three different regimes: informal payment banned, informal payment allowed, selling the right to choose doctors, i.e. the patient pays the highest price has the right to choose which doctor to see.

Results: We find that selling the right to choose doctors is always the best choice in three different regimes. When the difference in patients’ willingness to pay is small, banning informal payments is superior to allowing it; when the difference in patients’ willingness to pay is large, allowing informal payment is superior to banning it.

Conclusion: On one hand, in presence of asymmetric information between government and patients, allowing informal payments gives needy patient a bigger chance to be seen by the skilled doctor, therefore improves allocation efficiency; on the other hand, it may lead to wasteful competition between patients. The trade-off hinges on the heterogeneity in consumers’ willingness to pay for the difference services. When the heterogeneity is large, government should allow informal payments, otherwise it should ban informal payments

Permalink

International Comparison of Out-of-Pocket Costs and Medication Compliance

Presenter:

Richard Hirth

Authors:

Richard Hirth, John Piette, Scott Greer, Justin Albert, Eric Young

Chair: Richard Scheffler; Discussant: Jean Mitchell Tue June 6, 2006 10:45-12:15 Room 235

Background: Pharmaceutical spending has contributed disproportionately to increases in health care spending, causing drug costs to become a prominent health policy issue. Many studies have attempted to quantify cross-national drug price differences. This literature has focused on the full price of pharmaceuticals, regardless of who pays for the drug. However, across countries, and across patients within a country, the burden of these prices falls to differing extents on government health insurance systems, private insurers, and patients themselves. While variation in the full price of drugs across countries has policy implications for total health care expenditures, variation in OOP costs is more salient to clinical issues such as therapy adherence and patient outcomes. Little is known about differences across countries in the out-of-pocket (OOP) costs faced by patients.

Objective: The ability to study international variation in OOP costs and its effects on compliance has been hampered by the lack of comparable data from different countries. This paper analyzes a unique international survey that overcomes this limitation.

Data and Methods: The Dialysis Outcomes and Practice Patterns Study includes 10,398 randomly selected hemodialysis patients from 12 countries (US, Japan, Australia, New Zealand, Belgium, Canada, France, Germany, Italy, Spain, Sweden, and UK). Focusing on one illness and mode of treatment creates a more clinically homogeneous population across countries than would be the case with a more general survey. We compare rates across countries of patient-reported OOP costs for medication and non-purchase of medications due to cost. Logistic regression models identify characteristics associated with positive OOP costs and not purchasing medications due to cost.

Results: The proportion of patients paying OOP costs varies from 29% in France to 99% in Australia. The proportion of patients reporting non-purchase due to cost varies from 3.1% in Japan to 28.6% in the US. Japanese and Swedish patients were less likely to skip medications, and German patients were more likely to skip medications, than would be expected on the basis of costs, indicating that cultural factors or other aspects of the health care system influence cost-related compliance. The odds of paying positive OOP costs increased with income, education, private insurance, and membership in their country’s ethnic majority. Relative to Europe, the odds of facing OOP costs were higher in the US and Canada, and lower in Japan. The odds of cost-related non-purchase were higher for those who faced OOP costs, whose OOP costs exceeded their country’s average, had lower incomes, were younger, unemployed, and members of ethnic minorities. Relative to Europe, non-purchase was less likely in Japan and more likely in the US.

Conclusions: Substantial variation exists across countries in OOP costs and cost-related non-purchase of prescriptions. Few studies have considered the factors that modify patients’ adherence choices in response to medication cost pressures, leaving clinicians with few clues for how to support patients’ medication adherence given the costs that they will incur. The current study represents a significant advance by taking advantage of the variation in cost-sharing and other influences on patients’ medication use across 12 counties.

Permalink

The Impact of Cost-Sharing and Benefit Reductions in the Oregon Health Plan

Presenter:

Neal Wallace

Authors:

Neal Wallace, Kenneth McConnell, Charles Gallia

Chair: Richard Scheffler; Discussant: Tim Brown Tue June 6, 2006 10:45-12:15 Room 235

Rationale: In response to state budget shortfalls in 2003, the Oregon Health Plan imposed cost-sharing and eliminated some benefits for adult beneficiaries who were not part of the categorically eligible Medicaid population. Co-payments were imposed for inpatient care; emergency department use; hospital, clinic and individual practitioner ambulatory care; lab and x-ray services; and, prescription drugs. Coverage of specialty outpatient treatment for mental health and substance abuse, durable medical equipment and general medical supplies, eye care and dental services were eliminated. Limited experience and empirical evidence exists to assess the impact of these types of benefit changes on low-income individuals.

Objectives: This study estimates the impact of cost-sharing and benefit reductions on average monthly expenditures and rates of utilization per beneficiary, as well as average expenditures per service user, in total and by service type (inpatient, hospital outpatient, ambulatory professional, lab & radiology, emergency department and pharmaceuticals).

Methodology: The study uses a quasi-experimental design with a non-equivalent comparison group. Policy effects are identified as the difference in difference between non-categorically eligible adults affected by the policy change (known as “Standard” beneficiaries) and TANF eligible adults before and after policy implementation. FFS claims, MCO encounter data and monthly enrollment data for Oregon Health Plan beneficiaries were the primary data sources. Only services covered pre- and post- policy were analyzed. Expenditures per claim/encounter were calculated at the average “full” FFS payment rate during the study period (i.e. without reductions for co-pays, third party reimbursement or other adjustments). Individuals were included in the study sample if they were from 18-64 years old; had at least 6 months enrollment in each of two 12-month periods before and after the policy; and had at least 3 months enrollment in each of the two 6-month periods within the pre- and post-policy study periods. This yielded a study sample of 15,200 Standard and 7,540 TANF eligible subjects. Subjects were grouped into 94 primary care service areas (PCSAs). Average monthly expenditures, average expenditures per month with service use, and the average percentage of beneficiaries using service per month were calculated for the two eligibility groups, four 6-month study periods and 94 PCSAs yielding 752 aggregate observations. Fixed effects estimation is used to identify policy impacts. Expenditure changes are reported as percentage change from pre-policy levels.

Results: Average expenditures per beneficiary increased by 16% after cost-sharing and benefit reductions were imposed relative to the TANF control group. This was driven almost entirely by relative increases in inpatient care expenditures. Relative levels of expenditure were not different from the TANF controls for any of the other service categories. Use of ambulatory professional and lab/radiology services did decrease for the policy-affected group but expenditures per user rose at an equal rate.

Conclusions: The imposition of cost-sharing and benefit reductions for low-income adults in the Oregon Health Plan were found to raise expenditures per beneficiary and potentially shift treatment from ambulatory to inpatient settings. Expenditure increases were shared among consumers, providers and the state of Oregon due to the cost-sharing provisions.

Permalink

Before you advocate, do your sensitivity analysis: The evidence for reforming the Medicare physician payment system

Presenter:

Martey Dodoo

Authors:

Martey Dodoo, Robert Phillips

Chair: Timothy McBride; Discussant: TBA Tue June 6, 2006 10:45-12:15 Room 309

Rationale: Even though physician practice costs continue to grow, the formulae for updating the physician fees for Medicare services are projected to cut the fees by about 5% annually for several years beginning 2006. These projected cuts have raised concerns about appropriateness of the system for updating physician fees and physicians’ continued participation in Medicare. As they have done many times, physician groups decried the projected cuts and advocated for revisions in the formulae. Surprisingly, apart from a mandated GAO report in 2004, there have been no published quantitative studies assessing the sensitivity of the formulae to proposed revisions. This study contributes by filling that gap.

Objectives: To assess the sensitivity of the Medicare physician reimbursement rate formulae to the revisions proposed by various stakeholders.

Methods: We specified a model that simulated calculation of the Medicare physician fee conversion factor. We used the model to check sensitivity of the fee conversion factor to changes in five main parameters: Evaluation and Management (E&M) weights, GDP, SGR, MEI, and target expenditures over 10 years.

Results: Changes in all five parameters tested had no lasting net effects on the fee schedule. Five percent increases in MEI averted physician fee cuts only in the short-run.

Conclusion: This study shows that: (1) It would require a replacement of the entire reimbursement system to avoid physician fee cuts. (2) There is superior value in rigorous assessment of proposed policies and sensitivity analysis before undertaking advocacy.

Permalink

Physician Billing Behavior in Two State Programs

Presenter:

Eric Seiber

Authors:

Eric Seiber

Chair: Timothy McBride; Discussant: TBA Tue June 6, 2006 10:45-12:15 Room 309

Rationale: Physician billing is treated as exogenous in the academic literature, and has attracted very little econometric attention. A similar disinterest exists among Federal policy makers, with the Government Accountability Office frequently criticizing the reliance on generally weak state oversight and the lack of federal resources overseeing the $174 billion federal dollars contributed to the Medicaid program (2004). This weak oversight regime raises questions about whether a profit maximizing physician would accept their billing decisions as exogenously determined.

In state programs, physician prices are typically set by a fixed price schedule or through negotiations with the payer. Although price is fixed, physicians still have the power to choose the complexity level or billing code for the visit. If oversight is weak and probability of detection low, physicians can be expected to choose higher reimbursement codes or “upcode” on the margin.

Objectives: This study tests (1) whether physicians bill office visits at equal levels of complexity across state programs and (2) whether the billing behavior changes over time (a.k.a. code creep).

Methodology: The study uses 2001-2003 health care claims data (n=680,000) from the South Carolina Medicaid program and SC State Employee Health Plan to estimate a fixed effects ordered probit model of physician office visit billing where the provider assigns one of five complexity levels (billing codes) for the visit. An array of program dummies and physician-specific fixed effects and interaction terms control for physician practice and program wide differences, while an interaction term between the physician fixed effects and the Medicaid dummy test for differential billing between the two programs. Simulations demonstrate the magnitude of the effects on the two programs.

Results: Despite serving different demographics and the presence of 20% co-insurance in one program, physicians participating in both programs bill in a remarkably similar manner; the model found no significant difference between Medicaid and State Health Plan physician billing. Physicians participating in only one program or who bill under separate tax numbers deviated from this pattern and billed at 8% higher complexities. Some individual physicians demonstrated substantial differences between the two programs, with top 10% billing over 75% higher. Finally, all physicians demonstrate substantial annual “code creep”, with billing of higher complexity codes increasing 10% per year.

Conclusions: The results suggest that physicians do have pricing power, but the same results reject the hypothesis of differential billing. Physicians in this sample proved equally aggressive towards both state programs, increasing their diagnosis codes in every year of the sample. With similar billing patterns for both programs, the few outlying physicians billing higher complexities in one program compared to the other merit future attention, with their differential billing behavior meriting some degree of administrative review.

Permalink

Risk Adjusting Episodes of Care to Account for Illness Burden in Payment and Evaluation

Presenter:

Sharada Weir

Authors:

Sharada Weir, Rong Yi, Marilyn Kramer, Arlene Ash, Randall Ellis

Chair: Timothy McBride; Discussant: TBA Tue June 6, 2006 10:45-12:15 Room 309

There are many instances when it is convenient to use an ‘episode of care’ (e.g., a bout of pneumonia) as the unit of analysis rather than a patient-level measure (e.g., annual health care cost). The episode approach groups health care claims data into clinically homogenous periods of care typically based on diagnoses, procedures and dates. Episodes are widely used by health plans in profiling and in payment schemes to evaluate healthcare providers and are seen as constituting the critical analytic link between process and outcome measures, potentially providing an efficient tool for measuring quality of care. One criticism of the episode approach is that the classification system is too coarse and that some diagnosis data at the patient level are often ignored. Providers and analysts alike are concerned that episodes alone may not sufficiently account for the diversity of comorbidity among patients that may influence both total cost and therapeutic treatment choices within an episode. This paper explores this issue by evaluating the effect of risk-adjusting episodes of care with two years of data from MedStat’s MarketScan® database of large commercial employers. Episodes were created using Episode Treatment Group (ETG) software from Symmetry Health Data Systems Inc., and patient-level disease burden is computed using DxCG’s Hierarchical Condition Categories (HCCs) to predict concurrent year risk.

ETGs account for severity of illness and comorbid conditions by taking note of the presence of relevant surgical procedures and comorbid complicating conditions. For tractability reasons, included comorbidities are specific to an ETG. For instance, osteoporosis is a comorbidity when the diagnosis is arthritis but not a comorbidity for congestive heart failure. This approach is intuitively appealing and helps distinguish between more and less costly patients with a particular condition. However, ETGs stop short of fully risk adjusting for illness burden of the patient. This is illustrated by comparing predicted cost without risk-adjustment for paired ETGs, one with comorbidity and the other without. As expected, predicted cost for the ETG with comorbidity is higher that for the ETG without comorbidity. However, for many ETGs, the remaining variation in actual cost was found to be high and is further explained by the overall disease burden of the patient.

We compare the predictive power of unadjusted ETGs versus risk-adjusted ETGs by regressing actual episode payment separately as a function of: (1) unadjusted episode dummies (R2=0.29); and (2) risk-adjusted episode dummies (R2=0.34). However, part of the predictive power of the ETGs comes from grouping conditions with surgery separately from conditions without surgery, thereby classifying patients based on observed treatment rather than the disease condition alone. When ETGs are pooled so that only diagnoses are used and the regressions re-run, the gap widens (R2=0.31 for risk-adjusted ETGs vs. R2=0.24 for unadjusted ETGs). These findings support the case for risk adjusting episodes of care to promote fairness in comparing provider performance at the episode level and encourage efficient allocation of resources.

Permalink

Time allocation in primary care with competing demands

Presenter:

Ming Tai-Seale

Authors:

Ming Tai-Seale, Thomas McGuire

Chair: Michael Morrisey; Discussant: Jessica Vistnes Tue June 6, 2006 10:45-12:15 Room 313

Rationale: In analyses of physician behavior, how physicians allocate their time plays a central role. The time a physician spends during a visit is often put forward as the example of the “effort” the physician puts into a visit. Effort (time) is costly to the physician and the main input into the quality of a visit. In spite of its central importance, little is known about what determines how much time physicians spend with patients. While observable by the patient, time is regarded as not contractible. Empirical examination of how physicians spent clinic time is scarce.

Objectives: To develop a theoretical model of physician time allocation building on Becker’s theory of the allocation of time and to test the model, with videotaped behavior of physicians and patients. We study how time is allocated during a visit across the problems patients bring to the attention of their doctors.

Methodology: Direct observation of videotapes of 390 routine office visits that took place between 1998 and 2000 from three primary care practice sites in the Midwest and Southwestern regions of the U.S and surveys of participating patients and physicians inform the study. We break visits into “topics,” a natural unit of clinical decision making, and document how clinical time is spent. Using a mixed-level duration model, we analyze the effects of the nature of topics, the dynamics of time, and characteristics of patient, physician, and physician’s practice setting on how clinic time is spent.

Results. The average visit in our sample lasted 17.4 minutes and covered 6.5 topics. Patient and physician each spoke, on average, less than 1 minute per topic. Out of over 2,500 topics examined, more than 70% of the topics addressed biomedical issues. Psychosocial topics were 12%, and personal habits, 7%. Less than 4% of the topics involved mental health concerns. What happens in clinic is influenced by the contents of the discourse, physician practice setting characteristics, patient’s gender and race, and physician’s gender. When a topic was introduced during the visit and how much time has already been spent on it are also significant determinants of time allocation.

Conclusions. Physicians are likely to reduce time spent on a topic if it is raised later during the visit when the opportunity cost of physician time is higher. A large number of topics are discussed during a primary care visit, with little time spent on each topic. Efforts to improve the quality of primary care need to recognize the time pressure on both patients and physicians, and the time costs of improving information exchange.

Permalink

Physician Responses to Gainsharing

Presenter:

Jonathan Ketcham

Authors:

Jonathan Ketcham, Michael Furukawa

Chair: Michael Morrisey; Discussant: John Brooks Tue June 6, 2006 10:45-12:15 Room 313

Rationale: Financial incentives influence physician treatment decisions, yet regulation has limited the arrangements between hospitals and physicians. Hospital costs are the single largest component of health care spending, and clinical preference items (devices and drugs) account for up to 80 percent of the total cost spent on each patient, particularly for cardiology and orthopedics. The Office of Inspector General (OIG) has recently approved several gainsharing arrangements for cardiac care, where physicians receive equal shares of a hospital’s savings in a given service area. The aim of gainsharing is to control costs by promoting standardization of physicians’ treatment decisions. Although these incentives may influence behavior, a number of factors may limit their ability, including strong financial ties with device manufacturers, preference for non-taxable in-kind remuneration rather than cash, or free-riding. This paper provides empirical evidence about the impact of gainsharing from ongoing programs.

Objectives: We estimate the impact of hospital-physician gainsharing in the cardiac cath lab on hospital costs, quality of care, and access to technology. First, we decompose the sources of cost savings due to lower device utilization, lower price per item, and substitution of lower-priced items. Second, we assess the implications for patient care, including outcomes and complications, and the types of patients receiving treatment. Finally, we examine the impact on the availability of devices, adoption of new devices (e.g., drug eluting stents), and the extent of product standardization.

Methodology: The study uses data from Goodroe Healthcare Solutions from 2000-2005 for both gainsharing and non-gainsharing hospitals. These data include detailed information on patient demographics and risk-factors, physician identifiers, procedures and tests performed, which drugs and devices were used and what the hospital paid for them. Four hospitals have begun gainsharing for cardiac care since 2002. Difference-in-difference regression analyses are performed at the level of hospital, medical group, and physician. Although the difference-in-difference design eliminates a number of potential sources of bias, our estimates of gainsharing may be biased upward if, for example, physicians that participate are more responsive to financial incentives than average. We rely on propensity matching and instrumental variables to address such issues.

Results: We find evidence of significant cost savings due to gainsharing. One hospital reduced costs by $4.3 million and paid participating physicians $41,000. The majority of savings appears to be due to lower prices for a given device, and some due to fewer devices per patient. Treatment did not become more standardized, and neither patient outcomes nor risk factors changed.

Conclusions: Although preliminary results indicate large savings due to gainsharing, it has had little effect on standardization. It is unclear whether these savings are sustainable or whether they represent a onetime reduction in prices from manufacturers. The work will further consider which types of physicians appear most responsive to gainsharing. The findings of the study have important implications for a number of stakeholders, including supply chain managers and health systems who are considering the adoption of gainsharing, as well as government regulators and health services researchers.

Permalink

Geographic Variations in Medicare Physician Costs: The Role of Market Factors, Reimbursement, and the Practice Organization

Presenter:

James Reschovsky

Authors:

James Reschovsky

Chair: Michael Morrisey; Discussant: Michael Morrisey Tue June 6, 2006 10:45-12:15 Room 313

Authors. James Reschovsky (jreschovsky@hschange.org), Center for Studying Health System Change

Title: Geographic Variations in Medicare Physician Costs: The Role of Market Factors, Reimbursement, and the Practice Organization

Rationale: There are large geographic variations in Part B Medicare costs, even after controlling for differences in patient health status. There is a large literature, largely by John Wennberg and his Dartmouth colleagues, documenting these differences. However, their analyses of the underlying causes are limited because their data is largely at an ecological level. This analysis provides greater understanding of the factors underlying geographic cost variations using micro level data on physicians and their Medicare patients.

Objectives: The objective of this paper is to estimate a theoretically consistent model of physician provision of services to their Medicare patients that incorporates information regarding patient demand, the physician, his/her practice (including methods of physician compensation and revenue sources), Medicare payment and other market factors. Using this model, reasons underlying geographic cost variations will be explored.

Methodology: This uses a unique dataset containing information on a nationally representative sample of physicians from the 2000-01 Community Tracking Study Physician Survey merged with Medicare claims data on their patients. We examine how patient, physician, practice (size, type, how physicians are compensated, etc.), market, and Medicare programmatic factors contribute to the large geographic variations in Medicare physician costs. The CTS data are gathered in 60 representative local markets nationwide. Costs per beneficiary are specified as the product of the number of physicians per beneficiary in the market, the average quantity of services per beneficiary (in RVUs) among physicians treating Medicare patients, and the payment per RVU. Each is estimated in a theoretically consistent manner. The latter two components follow from previous work (Hadley and Reschovsky, 2005), in which physicians’ Medicare service volume was estimated using a specification that followed the theoretical work of McGuire and Pauly (1991). It recognized that the Medicare fee was potentially endogenous and that physicians might induce demand. In this paper, I decompose geographic variations in beneficiary costs into components that are attributable to patient, physician, practice, payment, and market factors.

Results: Preliminary results confirm that there are sizable variations in the average cost of treating Medicare patients across sites. Physician supply, the local mix of specialists and primary care physicians and demand factors all contribute to geographic variations, in part by influencing how intensively physicians treat their Medicare patients. Medicare’s fee system, in which physicians are, for the most part, paid a uniform rate according to the costs of providing services, also influences treatment patterns and serves to accentuate geographic cost differences.

Conclusions: Medicare’s payment system fails to incorporate both local demand side and supply side market differences that influence the quantity of services provided to beneficiaries. Hence, the uniform payment system encourages disparities in treatment patterns, and implicitly in the efficiency by which care is delivered across areas. Consideration should be given to setting localized sustainable growth rates or setting fees according to market characteristics in an effort to more closely resemble market prices.

Permalink

A Cost-Benefit Analysis of Female Primary Education as a Means of Reducing HIV/AIDS in Tanzania

Presenter:

Robert Brent

Authors:

Robert Brent

Chair: David Bishai; Discussant: TBA Tue June 6, 2006 10:45-12:15 Room 325

HIV-AIDS has affected Sub-Saharan Africa more than any other region, with the December 2004 estimate by UNAIDS being 26 million out of a world total of 39 million. One estimate of the consequences of this pandemic has it lowering life expectancies by a quarter, from 64 to 47 years. Much of the development progress in the region in the last half-century is in danger of being reversed. Intervention is clearly required. In order to establish priorities, cost-benefit analysis (CBA) is required as resources are very limited in Africa.

The issue as to which type of interventions to finance is especially important given the gender and age dimension of the Sub-Saharan African HIV-AIDS epidemic. The December 2003 UNAIDS update points out that African women are 1.2 times more likely to be infected with HIV than men. Among young people aged 15-24 (a good index of the number of new cases) women were 2.5 times more likely to be infected. Some of the gender differential is due to biological factors. But, some of the cause is due to the sociological power structure between the sexes in Sub-Saharan Africa. Interventions that seek to empower females, especially young females, would seem to be especially worth evaluating. Female education has been identified by the World Bank (2002) as one of most promising ways of combating HIV-AIDS because of the empowerment it provides.

In this article we carry out a CBA of female primary education in Tanzania based on panel data for 20 Mainland regions over the period 1994-2001. Schooling clearly targets the young. In Tanzania, primary schooling is the predominate form. 61% of females in 2000/1 had been through some primary education, while only 4% has been to secondary school, and 1% had a diploma or degree. The center-piece of our analysis involves establishing the effectiveness of female primary enrollments to reduce HIV-AIDS. This is the main contribution of the paper as there is an extensive literature which finds a positive relation between education and infection rates, contrary to most expectations. We treat this positive relation as the direct effect of schooling and regard it as coming from a static estimation framework. We then identify an indirect effect of education working through changes in income that has the opposite effect. Using a dynamic estimation approach developed by Arellano and Bond (1991) we are able to show that the indirect effect outweighs the direct effect, producing the net effect that education lowers infection rates.

The CBA that follows involves valuing the estimated number of HIV cases averted by their earnings, i.e., using the human capital approach, and comparing this with the costs for the enrollment numbers that generated the reduced number of infections. While the magnitude of the net-benefits depends crucially on the size of the discount rate used, our best estimate is that the benefits are between 1.3 to 2.9 times the costs. We highlight the role of external effects of female education as being the mechanism generating the favorable cost-benefit outcome.

Permalink

A Lower Price of Child Survival Lowers Fertility: Testing Quality Quantity Tradeoff In Africa

Presenter:

David Bishai

Authors:

Feng Zhao, David Bishai

Chair: David Bishai; Discussant: TBA Tue June 6, 2006 10:45-12:15 Room 325

The economist’s version of demographic transition theory stresses the role of price above the role of income in lowering first mortality and then fertility. Prior to the 19th century, households could not purchase significant improvements in child survival regardless of income. Technological developments that lowered the price of child survival are thought to have induced rapid substitution away from child quantity towards child quantity. According to this view: better child survival prices are the leading impetus for fertility reduction. The difficulty of measuring the price of child survival has impeded empirical tests of this central theory in economic demography.

We studied 15 Demographic and Health Surveys (DHS) from 9 sub-Saharan African countries. We theorized that there would be geographical variation in the ability of households to trade money for child survival. Using household level data stratified across 97 regions within the database we identified the regional price of child survival as the coefficient on the household asset score in a logistic regression on whether a child died before age 5. This produced a set of 97 regional coefficients (Pquality) which we interpreted as signifying geographical differences in the ability of households to trade wealth for child survival. In the second stage, we regressed the regional fertility rate against Pquality and aggregate indicators of economic development. In accordance with theory we found a positive elasticity of 1.71 (95% CI: 0.901 to 2.526) for the price of child quality on fertility and an elasticity of -0.193 (95% CI: -0.244 to -0.142) for the effects of regional wealth on regional fertility. A positive cross price elasticity between child quality and child quantity suggests that these are substitutes.

Our evidence is consistent with Becker’s theory of quality quantity tradeoff and emphasizes the central role of making child survival more affordable before expecting to see fertility declines in a population. When child quality (i.e. survival) was more expensive, households appeared to substitute towards child quantity. These results predict that if an uncontrolled epidemic (such as AIDS) makes child survival less affordable that fertility will rise even if educational attainments and household incomes remain constant or rise

Permalink

Is More Aggressive Treatment of Severe Brain Injuries in Children Worth It?

Presenter:

Mick Tilford

Authors:

Mick Tilford, Allen Goodman, P. David Adelson

Chair: David Bishai; Discussant: TBA Tue June 6, 2006 10:45-12:15 Room 325

Rationale: Management of traumatic brain injury varies according to treatments and in terms of aggressiveness. Variations in treatment aggressiveness are related to concerns over poor outcomes in survivors. Data describing health outcomes of children following traumatic brain injury that can be used in economic evaluations is scant.

Objective: To provide information on preference scores of children who survived a severe traumatic brain injury and use this information to calculate incremental cost-effectiveness ratios.

Methods: Information on life years gained and acute care costs from longitudinal hospital studies were combined with follow-up data on survivors. Follow-up data was based on telephone interviews with the primary caregiver of surviving children. In addition to demographic and insurance information, caregivers of survivors described use of services following discharge from the acute care hospital and reported on health-related quality of life using the Quality of Well-being (QWB) scale. Cost-effectiveness ratios were formed from the QWB scores and costs of acute and rehabilitative care.

Results: Over a 12-year period, approximately 6,500 children survived a traumatic brain that otherwise would have expired without more aggressive treatment. Follow-up data at 3 and 6-months following hospital discharge indicated a range in preference scores from 0.093 to 1.0 with a mean of 0.507 (0.201) at 3 months and 0.579 (0.227) at 6 months. Scores were related to probabilities of dying at the time of hospital admission. Children lacking health insurance had higher probabilities of dying and worse outcomes. Cost per quality adjusted life year was approximately $12,000 for the mean preference score estimates and $67,000 for the worst scores.

Conclusions: This study suggests that aggressive treatment in children with traumatic brain injury is warranted. Future research should further examine rehabilitation service use of uninsured children in relation to injury severity and outcomes.

Permalink

End-of-Life Transfers and the Decision to Care for a Parent

Presenter:

Meta Brown

Authors:

Meta Brown, Mark O. Wilhelm

Chair: Sally Stearns; Discussant: France Priez Tue June 6, 2006 10:45-12:15 Room 326

The largest and most common transfers from parents to children take the form of investments during childhood and early adulthood, while the largest and most common transfers from children to parents occur in parents’ old age. The life-cycle structure of intergenerational transfers poses a challenge for both theoretical and empirical analysis of transfer motives. Hypothesized transfer motives fall into three broad categories: social exchange, two-sided altruism, and role-modeling or demonstration. We use the recently available data from the HRS/AHEAD on both parents’ payment for their children’s college education and children’s supply of attention and informal care during their parents’ old age to investigate the predictions of these three theories. Within-family and within-gender variation in transfers is used to distinguish among them.

Permalink

Informal Care and Medicare Expenditures

Presenter:

Edward Norton

Authors:

Courtney Van Houtven, Edward Norton

Chair: Sally Stearns; Discussant: France Priez Tue June 6, 2006 10:45-12:15 Room 326

Permalink

Long-Term Care of the Disabled Elderly: Do Children Increase Caregiving by Spouses?

Presenter:

Liliana Pezzin

Authors:

Liliana Pezzin, Robert A. Pollak, Barbara S. Schone

Chair: Sally Stearns; Discussant: Hua Wang Tue June 6, 2006 10:45-12:15 Room 326

How do adult children affect the care that their elderly parents provide to each other? We develop two models in which the anticipated behavior of adult children provides an incentive for elderly parents to provide more care for their disabled spouses than they otherwise would. Our first model is based on a “demonstration effect” — adult children learn from a parent’s example that family caregiving is appropriate behavior. Our second model is based on a “punishment effect” — if the nondisabled spouse fails to provide care for the disabled spouse, then the children may respond by not providing future care for the nondisabled spouse when care becomes necessary. Both models assume that the nondisabled elderly parent recognizes that his or her caregiving behavior will affect the children’s willingness to provide care in the future. Moreover, the demonstration effect and the punishment effect increase the likelihood that nondisabled spouses will provide care for disabled spouses if they have children or, more precisely, if they have joint children. Joint children act as a commitment mechanism, increasing the probability that elderly spouses will provide care for one another. Stepchildren, depending upon the agree of parental attachment, provide weaker incentives than joint children for spousal care. Preliminary estimates find strong evidence that spouses provide more care when they have children with strong parent attachment.

Permalink

Certificate of Need Regulation in the Nursing Home Industry: Has it Outlived its Usefulness?

Presenter:

Barbara Caldwell

Authors:

Barbara Caldwell

Chair: Gabriel Picone; Discussant: Edward J. Schumacher Tue June 6, 2006 10:45-12:15 Room 332

In 1974 the United States Congress passed The National Health Planning and Resources Development Act (P.L. 93-641). The primary goals of this legislation were to (1) contain health care costs and (2) increase the accessibility and quality of health services. Certificate of need (CON) regulation is one attempt to constrain health care costs by limiting the supply of certain medical care facilities. In general, a CON program involves the regulation of the building, expansion, and modernization of health care facilities and capital equipment on the part of institutional health care providers, including hospitals, nursing homes, and home health agencies. With respect to the nursing home industry, prospective nursing home owners/operators are required to demonstrate that a “need” exists for more nursing home beds. Some States also imposed a construction moratorium that prevented any expansion of existing facilities or construction of new facilities regardless of whether or not a “need” existed. These CON/moratoria programs impose a supply side constraint that creates a potential barrier to entry and in the presence of excess demand may cause a nursing home bed shortage for those patients covered by Medicaid. In 1986 Congress allowed the Federal CON requirement to lapse due to its perceived anticompetitive and excessively regulatory nature. Today forty-one States and the District of Columbia continue to have a CON, a construction moratorium, or both for nursing home facilities. Yet maintaining these regulations comes at a cost. Since quality remains a controversial subject, access to care for Medicaid-eligible persons remains an issue in many States, and medical care expenses for nursing homes continue to climb, one might question the effectiveness of retaining these regulations. This paper investigates if any differences exist in the quality of nursing home care and the access to care for nursing home residents between those States that have continued with CON and/or construction moratoria and those States that have eliminated such policies. Using data for the years 1991 through 2003 for all Medicaid-certified nursing home facilities in the United States and employing the method of difference-indifferences we find that access to care for Medicaid residents is less in those States that still employ some type of regulation than those States that have eliminated the regulation. With respect to quality of care the results are mixed depending on the measure of quality that is employed. With the risk of becoming a nursing home patient at the age of 65 at 39 percent and at the age of 85 at 49 percent as well as the aging of the current population, the areas of access to care and quality of care remain important policy issues in the nursing home industry.

Permalink

Specialization and Sorting in the Obstetrics Market

Presenter:

Sean Nicholson

Authors:

Andrew Epstein, Jonathan Ketcham, Sean Nicholson

Chair: Robert Town; Discussant: TBA Tue June 6, 2006 10:45-12:15 Room 335

Permalink

Asymmetric Information in Physician Agency

Presenter:

Albert Ma

Authors:

Philippe Chone, Albert Ma

Chair: Robert Town; Discussant: TBA Tue June 6, 2006 10:45-12:15 Room 335

Permalink

Evaluating Estimators of Treatment Effects Using Detailed Measures of Illness Severity From Medicare Claims Data

Presenter:

Justin Trogdon

Authors:

Justin Trogdon, Ahmed Khwaja, Gabriel Picone, Martin Salm

Chair: Gabriel Picone; Discussant: Joe Terza Tue June 6, 2006 10:45-12:15 Room 332

An old and constant problem in health economics has been to estimate the effects of choices made by doctors or patients on health outcomes in the presence of unobserved heterogeneity. Economists have a broad range of estimators available, which fall into two categories. The first assumes that treatment is exogenous conditional on a set of covariates (selection on observables). Estimators based on this approach include ordinary least squares with a large number of controls, flexible regression methods, propensity score methods, and matching estimators. The second approach uses instrumental variables to identify the treatment effects of interest. However, each of these estimators makes different sets of assumptions and may even estimate different effects. In this study we compare these different estimators through their ability to estimate the effects of different treatments on survival outcomes for individuals with Acute Myocardial Infarction (AMI). The treatment effects studied are the effects of: (i) catheterization, (ii) admission to for-profit hospitals, and (iii) admission to low-volume hospitals. Our dependent variables are 30-day and one-year mortality following the AMI. We use data from the Cooperative Cardiovascular Project (CCP) merged with the American Hospital Association’s (AHA) annual survey of hospitals and the National Inpatient Survey (NIS). One significant limitation of the Medicare claims data used previously to study these treatment effects is that it does not contain complete measures of severity of illness. The strategy of this study is to use a richer data set to analyze how sensitive different estimators are to the addition of detailed severity of illness measures. Severity of illness is a primary factor driving selection into treatments and if omitted from the estimation leads to bias in the estimation of the treatment-outcome relationship. Lacking a combination of experimental and non-experimental data, as used in previous studies comparing treatment effect estimators, we use the availability of detailed severity of illness to compare the performance of different estimators. Our results show that for estimating average treatment effects, methods that rely on the selection on observables assumption produce almost identical results using similar sets of control variables and there are no advantages over using a standard least squares regression. Across methods, the addition of controls for severity of illness reduces the estimated average treatment effects. In fact, conditional on detailed patient data, the addition of hospital characteristics has little impact on the average treatment effects. Methods that rely on instruments tend to be very unstable across specifications with different sets of controls. We conclude that in estimating treatment effects there may be no alternative but to rely on good data.

Permalink

Compensating Wage Differentials and AIDS Risk

Presenter:

Jeff DeSimone

Authors:

Jeff DeSimone, Edward J. Schumacher

Chair: Gabriel Picone; Discussant: Sara Markowitz Tue June 6, 2006 10:45-12:15 Room 332

In the past two decades, AIDS has emerged as a top public health issue and had a profound effect on the health care sector; yet its impact on health care labor markets has not been studied. This paper examines how HIV infection risks affect the earnings of registered nurses (RNs) and other health care workers. From the standard theory of compensating wage differentials, we expect health care workers, particularly patient care providers, to require monetary compensation for the potential risk of contracting HIV and the burden of increased precautions required to avoid this risk, relative to similar jobs that do not involve HIV risk. Using annual CPS outgoing rotation group data from 1993– 2003, we estimate separate AIDS wage differentials for RNs, licensed practical nurses and nursing aides, and other non-physician health care workers. Our econometric specification allows both the wage effects of AIDS and wage trends to vary across sectors. We proxy for HIV infection risk proxy using state-specific AIDS prevalence rates, allocated to metropolitan areas using the fraction of state cumulative AIDS cases each year that occur in each metropolitan area, and include metropolitan area and year fixed effects in our regressions. Estimates show that a 10 percent rise in AIDS is associated with a wage increase, relative to workers outside the health care sector, of 1.3 percent for RNs and 0.9 percent for non-nursing health practitioners. Additional results suggest that these effects do not spuriously represent the impact of working in a hospital, are not driven by metropolitan areas with particularly high AIDS prevalence rates, and are unrelated to the introduction of antiretroviral drugs during the sample period. These findings imply that policies that reduce the prevalence of HIV and AIDS will reduce the wage necessary to attract individuals into jobs involving patient care; in particular, a 10 percent AIDS rate decline would save over $1 billion per year in RN wage payments.

Permalink

Testing the latest empirical approach to hospital mergers and market power evaluation

Presenter:

Yunwei Gai

Authors:

Yunwei Gai, Gary Fournier

Chair: Gabriel Picone; Discussant: Gabriel Picone Tue June 6, 2006 10:45-12:15 Room 332

Mergers of hospitals in the same locality raise interest because they restrict the choices that patients have when hospital care is consumed, and they give hospitals market power in contracting with the local insurers and managed care organizations. For at least the past twenty years, little progress has been made in quantifying the market structure and market power of hospitals for lack of convincing methodology. The inability to demonstrate the exercise of market power in local hospital markets has significant policy implications: over the past 12 years or more, the federal government has lost every major antitrust case involving mergers of local hospitals, often for lack of a convincing market delineation method. In effect, existing techniques assume patients have full insurance and have the mobility and savings to search out the better hospitals outside the general vicinity of their residential location, while in truth the ‘silent majority’ of patients have incomplete insurance arrangements from managed care organizations that restrict their choice set. This paper uses Florida hospital discharge and financial data and a new econometric approach (Capps et al. 2003) to study hospital market power and to show how it affects pricing and the efficiency of hospital care. The econometric model is flexible enough to detect groups of patients who must rely on local hospitals. The Capps et al. study demonstrated, with some success, its application on San Diego, Ca. data. But at this early stage the reliability of the empirical methodology remains unverified and needs to be subjected to further scrutiny. The current paper will examine several questions about the implementation of these methods on hospital mergers in Florida in the 1990s. Is this a truly feasible method that can be readily applied to many other situations in the market? Does the model produce stable and reliable estimates of the effects, particularly given the acute demands of merger litigation? How does it change our thinking about the extent of the problem of hospital market power in local markets and its impact on consumers?

Permalink

Using Quality Reports to Improve Health Care Markets

Presenter:

Jacob Glazer

Authors:

Jacob Glazer, Thomas McGuire

Chair: Robert Town; Discussant: TBA Tue June 6, 2006 10:45-12:15 Room 335

Permalink

Demand for Health under ex ante Moral Hazard

Presenter:

Rui Wang

Authors:

Rui Wang, Gerard Russo

Chair: James Burgess; Discussant: TBA Tue June 6, 2006 8:00-9:30 Room 121

Authors: Rui Wang (ruiw@hawaii.edu) and Gerard Russo (russo@hawaii.edu), Department of Economics, University of Hawaii at Manoa

Title: Demand for Health under ex ante Moral Hazard

Rationale: Although the United States ranks highest in health care spending per capita among the industrialized societies and it has the most advanced medical technology available to public, health status of Americans consistently ranks poorly relative to that of residents of other industrialized nations. One possible cause of this puzzle, which has been largely neglected by researchers, is insufficient preventive care and increased population risk factors as a result. Unlike social insurers in most other industrialized countries, private insurers in the United States do not have long term incentives to subsidize preventive healthcare since most insurance policies cover the insured for only a relatively short period of time. A lack of preventive care could explain the coexistence of high medical expenditure and poor population health outcomes.

Objectives: The objective of this paper is to develop a dynamic, stochastic framework of demand for health that examines individual lifecycle behavior of health investment and addresses the issue of ex ante moral hazard under insurance incentives.

Methodology: Following the line of literature beginning with Michael Grossman’s famous human capital model, this paper develops a demand for health framework that employs a random process and optimal control. It explicitly recognizes illness as random utility loss which could be compensated with curative medical care. Illness occurs over time according to a non-homogeneous Poisson process, whose intensity is inversely related to health stock. A representative individual’s lifecycle behavior of consumption and investment in health capital is then examined. With the introduction of health insurance, individual behavior under ex ante moral hazard is examined and conditions for optimal social insurance under ex ante moral hazard are derived.

Results: Under incentives of health insurance, individuals tend to invest sub-optimally in their health capital, which in turn leads to higher morbidity and increased curative health spending. Such distorted incentives under ex ante moral hazard are mitigated when subsidy to preventive care is provided. Under optimality the rate of such subsidy should be equal to the rate of reduction in utility loss by curative care paid for by insurance.

Conclusions: Investment in health during early ages is efficient because it serves the individual for a relatively long period of time under relatively low depreciation rates. Society should subsidize medical indigent families with young kids out of efficiency considerations besides preference for health equity. Private insurers do not have strong enough incentives to subsidize preventive healthcare. Lack of preventive care leads to higher morbidity and increased health spending, which may partially explain the American puzzle of high health spending coupled with relatively poor health outcomes.

Disclosure information: Rui Wang is a Ph.D. candidate and Dr. Gerard Russo is an associate professor in the Department of Economics, University of Hawaii at Manoa. This research is part of Rui’s dissertation project.

Permalink

Separating selection and moral hazard effects in Swiss health insurance - a nonparametric bounds analysis

Presenter:

Robert Leu

Authors:

Robert Leu, Michael Gerfin

Chair: James Burgess; Discussant: TBA Tue June 6, 2006 8:00-9:30 Room 121

There is an ongoing debate in health economics whether cost sharing in health insurance changes the behaviour of patients. The well documented negative correlation between the degree of cost sharing and health care costs can be attributed to changes in behaviour (reduction of moral hazard) or selection (more healthy people select higher cost sharing). Previous empirical evidence for Switzerland, which was obtained based on strong identification assumptions using non-experimental data, is mixed. In this paper we employ nonparametric bounding techniques to minimise statistical assumptions. This methodology yields the minimum and maximum value a treatment effect can take given the assumptions made. If nothing can be assumed about the data generating process we obtain the so-called no assumption (or worst case) bounds. These bounds can be tightened if we are willing to impose additional assumptions.

Treatment is ordered in our application. We consider three possible treatment states: low deductible, medium deductible and high deductible. These deductible levels are chosen by the individuals at the start of their health insurance contract. The outcome variable is the probability of going to the doctor. Our empirical analysis shows that we need some assumptions to tighten the bounds; the no-assumption bounds are too wide to be useful. We consider two further assumptions: treatment response is mean independent of an instrument, and treatment response is monotone (i.e. we assume the sign of the treatment effect to be known). The mean independence assumption is also weakened by assuming that mean response varies weakly monotonically with the instrument. Under the first two assumptions we estimate bounds for the treatment effect of the high deductible compared to the low deductible that are below zero. Hence given these two assumptions we conclude that there is a negative treatment effect (i.e. there is a change in behaviour) of at least -0.08. Given that the observed difference is -0.24 at least one third of the difference in the probabilities of going to a doctor can be attributed to a reduced moral hazard effect. Weakening the mean independence assumption reduces the treatment effect to -.04, but it remains different from zero. This finding is in contrast to previous empirical analyses based on the same data.

Permalink

Cost-Effectiveness Analysis under Uncertainty: Investigating the Role of the Cost-Effectiveness Frontier (CEF)

Presenter:

Gary Zarkin

Authors:

Gary Zarkin, Jeremy Bray, Debanjali Mitra, Mohan Bala, David Couper

Chair: James Burgess; Discussant: TBA Tue June 6, 2006 8:00-9:30 Room 121

Rationale: Several methods of handling uncertainty of cost-effectiveness ratios exist. A limitation of these methods is that they do not adequately address the uncertainty associated with interventions being dominated. If an intervention is dominated in either a strong (greater cost and lower effectiveness) or weak sense (greater cost-effectiveness ratio), it is no longer included in the cost-effectiveness frontier and it drops out of further cost-effectiveness analysis. Existing methods of dealing with uncertainty in CE ratios do not adequately incorporate variability in the CEF.

Objectives: The objective of this methodological paper is to introduce uncertainty of the CEF into cost-effectiveness analysis. The paper lays out the methodology that researchers should follow to address uncertainty in the CEF. We demonstrate how this uncertainty changes the typical cost-effectiveness analysis.

Methodology: We assume that the cost and effectiveness realizations for each intervention arm are drawn from a bivariate normal (BVN) distribution. We draw a cost and effectiveness pair from each BVN distribution and calculate the implied CEF. We draw N=1000 draws for each intervention arm and calculate the implied CEF for each draw. Based on these outcomes, we calculate the probability associated with each realization of the CEF.

We propose to use cost and effectiveness data from the COMBINE trial to estimate the parameters for the BVN distribution. These parameters include the mean cost, mean effectiveness (measured by days heavy drinking), the standard deviation of cost and effectiveness, and the correlation of cost and effectiveness for each of the 9 arms of the COMBINE trial. An advantage of using the COMBINE data is that it provides an anchor to actual clinical trial, which increases the relevance of the analyses.

Permalink

Healthcare Markets, the Safety Net, and Access to Care Among the Uninsured

Presenter:

Carole Gresenz

Authors:

Carole Gresenz, Jeannette Rogowski, Jose Escarce

Chair: Joel Hay; Discussant: TBA Tue June 6, 2006 8:00-9:30 Room 213

Research Objective : We use nationally representative data including observations on more than 8,000 uninsured adults to provide new insight into the relationship between access to care and not only the structure and capacity of the safety net but also the healthcare market more generally.

Study Design: We use Medical Expenditure Panel Survey (MEPS) data linked with data from multiple secondary sources. To describe the healthcare safety net and market structure in each individual’s location, we derive variables from the American Hospital Association Annual Survey of Hospitals, Area Resource File, the InterStudy Regional Market Analysis database, the Bureau of Primary Healthcare (BPHC) Physicians Uniform Data System, the Current Population Survey, Census of Governments, and the Census Bureau’s Annual Survey of State and Local Government Finances. A major innovation of this research is the calculation of detailed measures of safety net availability: We calculate the distance between each uninsured individual and the nearest BPHC provider, public hospital and emergency room. We analyze outpatient care utilization (office-based physician and non-physician visits, emergency department visits) and medical expenditures. We simulate utilization for a number of health care market and safety net values using standardized predictions.

Population Studied: Uninsured adults in the U.S.

Findings: Distances between the rural uninsured and safety net providers such as hospital emergency rooms, public hospitals, migrant health centers, public housing primary care programs, and community health centers are significantly associated with utilization of a variety of healthcare services. In urban areas, we find that the percentage of individuals in the area who are uninsured and the pervasiveness and competitiveness of managed care have a significant relationship with healthcare utilization.

Conclusions: Absent the universal provision of health insurance, policy approaches to alleviating the barriers to access facing the uninsured include incremental efforts to increase the affordability and availability of public or private health insurance as well as measures to increase the accessibility of healthcare for the remaining uninsured. Our findings shed light on areas of focus for the latter class of measures. Facilitating transport to safety net providers and increasing the number of such providers are likely to improve access to care among the rural uninsured. ur findings for urban areas suggest that particular attention be paid to the uninsured living in areas where many of those insured are covered by managed care, and especially so where little competition among managed care organizations exists. Ironically, the “backlash” against managed care may result in improved access to care for some uninsured, although the salutary effects would be offset to the extent that the backlash also results in increasing healthcare costs, greater numbers of uninsured, and more competition for healthcare resources.

Permalink

Managed Care and the Safety Net. More Pain for the Uninsured?

Presenter:

Nuria Mas

Authors:

Nuria Mas

Chair: Joel Hay; Discussant: TBA Tue June 6, 2006 8:00-9:30 Room 213

Author: Núria Mas, IESE Business School

Title: Managed Care and the Safety Net. More Pain for the Uninsured?

Rationale: Health care insurance in the US has changed enormously in the last 20 years, since managed care radically altered incentives to providers. The impact of the managed care boom has been subject to close scrutiny. However, most of the literature has concentrated either in the role of managed care in bringing efficiency gains to the health care market or on studying its effect on utilization and quality of care for its enrollees. Research analyzing the impact of managed care on the uninsured and on safety net hospitals has been minimal.

Objectives: This paper contributes to fill this gap and analyses the impact of managed care on access to care and quality of care for the uninsured.

Methodology: The United States has always relied on an institutional safety net to meet the basic health care needs of the uninsured. Traditionally, hospitals in the US had been able to finance charity care through a complex system of cross-subsidies where privately insured patients were charged higher prices. By imposing stricter financial restrictions to hospitals and doctors, managed care reduced prices charged to their insured patients, making it very difficult for hospitals to obtain excess funds for their uncompensated care. In this paper we test the hypothesis that this increased financial pressure has worsened both access to care and quality of care for the uninsured, by affecting the safety net hospitals. We expand Frank and Salkever’s (1991) model to analyze hospitals’ decisions to provide charity care and use a probit model to test the results empirically. We use hospital data from the American Hospital Association and patient data from the Office of Statewide Healthcare Planning and Development (OSHPD). Managed care data is obtained from the Area Resource File.

Results: First, using US data, our results show that manage care has increased safety net hospital closures and it has encouraged the termination of those services (ER, obstetrics, etc) most frequently used by the uninsured, hence negatively affecting their access to care. Second, our results confirm that uninsured patients shift toward government hospitals. The results also seem to point toward the possibility of a trade-off between the number of charity care patients and the quality of care provided. Finally, we directly analyze the impact of managed care on health outcomes for the uninsured (as measured by the probability of dying after a heart attack). Our results indicate that managed care penetration has a negative impact on the health of charity care patients and also of those that go to government hospitals.

Conclusions: the results of this paper confirm that the impact of managed care goes beyond its effect on its enrollees and on efficiency. They also have important policy implications since the gap between socioeconomic groups could widen if the uninsured not only see their access to care reduced, but also go to hospitals which quality is declining.

Permalink

Partially Identifying Treatment Effects in the Presence of Unobserved Treatments: Covering the Uninsured

Presenter:

Steven Hill

Authors:

Brent Kreider, Steven Hill

Chair: Joel Hay; Discussant: TBA Tue June 6, 2006 8:00-9:30 Room 213

Policymakers have long been interested in identifying the number of people lacking health insurance, the consequences of uninsurance for access to health care, and the potential cost of covering the uninsured. Household surveys provide the primary source of information about the uninsured, but many validation studies find some respondents misreport insurance status. Reporting errors may lead to inaccurate estimates of the number of uninsured and bias estimates of the effects of insurance on access, use, and costs. Using data from the 1996 Medical Expenditure Panel Survey (MEPS), we investigate what can be learned in the presence of arbitrary health insurance reporting errors about (a) the gap between the insured and uninsured in the use of health services and (b) the impact of universal insurance coverage on the use of services. We exploit information from insurance cards, policy booklets, and follow-back interviews with employers and insurance companies to construct validation data for a nonrandom portion of the sample. There are 18,851 nonelderly in the 1996 MEPS, and about two-thirds of them have evidence from at least one source validating their reported insurance status. Extending the theoretical literature on nonparametric bounds and treatment effects, we formally characterize the identification problem and assess the identifying power of a variety of verification, monotonicity, and independence assumptions. Molinari’s (2002) “missing treatments” bounds are tightened when there is information about the potential degree of reporting error within the subpopulation without validation data. For some results, we estimate rates of false negatives and false positives on a small subsample whose employers responded to the follow-back interviews and use them to extrapolate potential rates of reporting errors to the third of the sample that does not have validation data. Under a set of relatively strong nonparametric assumptions, we preliminarily estimate that coverage the uninsured would increase the proportion of the population using health care in a month by no more than 2.5 percentage points for adults (an 11% increase) and by no more than 0.8 percentage points for children (a 4.8% increase).

Permalink

Perspective in Drug Abuse Treatment Economic Evaluations

Presenter:

William Cartwright

Authors:

William Cartwright, Paul Solano

Chair: Willard Manning; Discussant: TBA Tue June 6, 2006 8:00-9:30 Room 225

Rationale: Economic evaluation of drug abuse treatment must consider the perspective of the study and the standing of those potentially affected by the intervention. Economic evaluations are often presented form the taxpayer’s point of view without sufficient analysis of the implications of this adoption.

Objectives: This paper defines the issues related to perspective and standing in economic evaluations. In particular, the problems related to evaluating drug abuse policies and interventions are discussed. Adjustments are recommended to the cost-benefit estimates in the usual cost saving approach. In addition, the willingness-to-pay approach is recommended to adjust questionnaires and target populations.

Methodology: The recommendations are developed in light of public finance issues surrounding the raising of revenue through taxation to fund government expenditures for drug abuse treatment. Both the cost saving approach and the willingness-to-pay approach of cost-benefit analysis are utilized for valuation.

Results: The recommended adjustments for the typical cost-benefit analysis involve adjusting benefits with the marginal cost of financing the government program. The financing mix in a federal system alters the marginal cost of financing and thus requires a further analysis of the specific state and local tax structure in line with the Federal tax structure. The actual standing of the drug abuse treatment client is a standing issue for which no definitive recommendations can be made. With regards to theft losses, treating these as transfers is not recommended because of the lack of standing that the criminal would have. However, many policy options certainly require standing be given to drug users and those in treatment. The willingness-to-pay approach must be designed with sensitivity to the issues of perspective and standing.

Conclusions: Some fairly straightforward adjustments can be made to standard cost-benefit analysis. Issues of standing require the analyst to establish the correct target population for the study, weighting of that population, and sensitivity to the various population groups affected by a particular intervention or policy.

Permalink

Longevity Bias in Cost-Effectiveness Analysis

Presenter:

Liqun Liu

Authors:

Liqun Liu, Andrew Rettenmaier, Thomas Saving

Chair: Willard Manning; Discussant: TBA Tue June 6, 2006 8:00-9:30 Room 225

The stated goal of medical CEA is to maximize health outcomes for a given amount of resources allocated to health care (Weinstein and Stanson 1977). The rationale for this goal seems to be that, for the overall optimality of health care resource allocation (which includes not only the allocation of a given health care budget but also the determination of the size of the budget), maximum health benefits must be achieved under any chosen health care budget. This characterization of the goal of medical CEA is so widely accepted that in many unresolved debates regarding the theoretical foundation and appropriate execution of CEA, both sides of the debate claim to abide by this goal.

Using a utility specification consistent with the QALY (quality-adjusted life-year) measure of health outcomes, we study both the “fixed budget problem” and the “optimal budget size problem” in medical resource allocation. The fixed budget problem is about how a given health budget should be allocated between improving longevity and improving health-related quality of life. The logically subsequent optimal budget size problem is about whether or not further increase in the size of health budget is welfare enhancing.

We compare two approaches to the fixed budget problem: the health gain approach that maximizes the QALY measure for a given budget and the total gain approach that maximizes the total utility for a given budget. The health gain approach implements the stated goal of CEA but it is the total gain approach that is consistent with welfare maximization. We show that the health gain approach or CEA results in a longevity bias: the CEA-based division of a given total medical expenditure between extending life and improving health gives the former a larger share than is called for by welfare maximization.

After discussing the fixed budget problem, we then study the optimal budget size problem and relate it to the issue of indirect costs of medical interventions (e.g., Garber and Phelps 1997 and Meltzer 1997). Under either approach (the incorrect health gain approach or the correct total gain approach) to the fixed budget problem, the fixed budget phase of resource allocation generates a QALY function with the size of the health budget as its argument. When considering whether to expand the health budget, one must compare the QALY gains with the consumption losses from such an expansion. We will show that regardless of the approach adopted in the fixed budget phase, the criterion for evaluating an expansion in the health budget is characterized by a cost-effectiveness ratio that includes, in its numerator, not only the direct health care costs, but also the indirect consumption costs and earnings from longer life. This means that appropriately accounting for indirect costs in the evaluation of new interventions (increases in the medical budget) alone cannot correct the longevity bias resulting from the maximization of QALYs under a given medical budget.

Permalink

Towards a Consistent Payer Perspective: A Microeconomic Approach to Economic Evaluation

Presenter:

Scott Grosse

Authors:

Scott Grosse

Chair: Willard Manning; Discussant: TBA Tue June 6, 2006 8:00-9:30 Room 225

Rationale: Conventional cost-effectiveness analysis (CEA) and cost-benefit analysis (CBA) are regarded by many decision makers as largely irrelevant. One reason is that most studies do not model the costs and benefits unique to different economic agents. Second, the outcomes that are modeled do not necessarily match decision makers’ objective function. Third, the comparators may not represent relevant alternatives. Even CEAs that purport to represent the payer perspective rarely do so in a way that is consistent with realistic models of decision making.

Objectives: The objective of this paper is to outline a framework for conducting economic evaluations from the perspective of payers and purchasers, including individual consumers, health plans, integrated health care delivery systems, and employer and government purchasers of health care.

Methodology: This paper reviews analytic methods used to evaluate health care interventions, including CEA, CBA, cost-minimization analysis (CMA), budget impact analysis (BIA), and return on investment (ROI) analysis. These methods are compared in terms of which costs are included, how costs are measured, analytic horizons, use of discounting, and the distribution of costs and benefits. The types of outcomes considered and the valuations placed on each outcome vary by method and intended audience. A hypothetical screening test scenario is modeled to compare the usefulness of the different approaches for distinct types of payers.

Results: A hypothetical screening test, which together with standard treatments is assumed to reduce mortality from X disease at an average delay of 5 years, provides net economic benefits from the societal perspective and is classified as cost-effective using conventional CEA methods. Using BIA and ROI methods, the financial impact is not favorable to private payers. Using CEA and CBA methods adapted to the perspective of different payers, including adjustment for differential discount rates and attrition or turnover among the covered population, the results are intermediate between societal CEA and ROI findings.

Conclusions: Each type of method has strengths and limitations. Economic evaluations should model the distribution of costs and benefits among different types of payers and assess the economic attractiveness of the intervention from each group’s perspective. This can make economic evaluations more relevant to health insurers, employers, and other payers or purchasers of health care and increase their use in resource allocation decisions.

Permalink

The Health Care Crisis for People with Disabilities Who Do Not Receive Social Insurance Benefits

Presenter:

Tricia Johnson

Authors:

Tricia Johnson, William Johnson

Chair: Catherine McLaughlin; Discussant: Dan Polsky Tue June 6, 2006 8:00-9:30 Room 226

Rationale: The increasing number of people with disabilities under age 65 will profoundly affect public policy towards disability by increasing the administrative burden on programs such as Social Security Disability Insurance (SSDI) and Medicare. The aging of the baby boom population will substantially increase the number of people with disabilities who do not qualify for disability benefits. Failure to address the health care needs for persons with disabilities prior to retirement is likely to increase long terms costs to Medicare by increasing the severity of illness once these persons reach retirement age.

Objectives: This paper estimates the number of people with disabilities under age 65 who do not receive disability benefits and projects the number without benefits through 2025 to quantify the uncompensated health care burden of disability. A number of simulations are conducted to understand how changes in the underlying characteristics of the population will affect these projections.

Methodology: This study first estimates the number of people with disabilities in the population using a logistic regression model. Second, the proportion of people with disabilities who do not receive SSDI is estimated from this model. Third, health care expenditures for people with disabilities who do not receive benefits are estimated using a two-part generalized linear model. Finally, a simulation model is developed to explore the effects of different assumptions about the demographic characteristics and health conditions of the population on the prevalence of disability and health care expenditures. The 1996 panel of the Survey of Income and Program Participation is used to estimate the number of people with disabilities who do not receive benefits. The 1997 to 2000 Cost and Use Files of the Medicare Current Beneficiary Survey are used to estimate health care expenditures for people with disabilities who do not receive benefits, simulating costs as if they received benefits.

Results: Nearly 23% (2.8 million [M]) of people with disabilities did not receive SSDI benefits and had no recent work experience in 2000, and 67% (8.9M) of those with disabilities did not receive SSDI benefits overall. These estimates will increase to 3.7M and 11.4M respectively by 2025. Health care expenditures for people with disabilities without recent work experience and no SSDI benefits were estimated at $20.7 billion (B) in 2000 and will increase to $29.3B in 2025.

Conclusions: People with disabilities who do not receive SSDI benefits represent the hidden population of Americans with disabilities. The reported prevalence of disability is sharply reduced if only Social Security statistics are used. Results emphasize that the problem of uninsured losses among Americans with disabilities who do not qualify for social insurance benefits will be one of the most important, unsolved dilemmas for society created by the baby boom generation.

Permalink

Do transportation brokerage services decrease expenditures and improve health outcomes of Medicaid children?

Presenter:

Jinkyung Kim

Authors:

Jinkyung Kim, Edward Norton, Sally Stearns

Chair: Catherine McLaughlin; Discussant: Jenny Kenney Tue June 6, 2006 8:00-9:30 Room 226

Research Objective: To help poor or disabled patients with transportation needs, Medicaid pays for non-emergency medical transportation (NEMT). Tight state budgets and rising Medicaid costs have prompted states to find efficient way of providing the mandatory NEMT services. Twenty-one states now contract with private brokerage companies, which coordinate and finance all service delivery under capitated payments. This study analyzes the effect of transportation brokerage services on Medicaid beneficiaries’ access to care.

Study Design: The study uses retrospective data analysis with a quasi-experimental design. Georgia implemented statewide transportation brokerage services in October 1997. We specified a 21-month pre- and a 21-month post-period using four calendar years of Georgia Medicaid Eligibility and Claims Data, 1996-1999. We use two-part models to predict the effect of transportation brokerage services on monthly Medicaid expenditures, controlling for personal characteristics and month. The dependent variables of interest are total, inpatient, and outpatient Medicaid expenditures. Total expenditures are the sum of all types of expenditures except transportation expenditures.

Population Studied: The study sample is a 20% random sample of children with asthma in Georgia. We identified children with asthma using ICD-9 diagnosis codes and national drug codes. The final study sample had 938,141 observations on 42,157 children at the person-month level. Inpatient, outpatient, and emergency room expenditures are analyzed.

Principal Findings: The shift to transportation brokerage services decreased the probability of any inpatient expenditures by 1.9 percentage points (from a mean of 2 percent) and the probability of outpatient expenditures by 7.4 percentage points (mean of 28 percent). The monthly expenditures per person per month also decreased by $77.28 for inpatient and by $34.35 for outpatient services. Separate analysis for children age 1-2 and age 3-18 revealed that the bulk of savings are from younger children. Incremental effects of per person per month expenditures are statistically significant using bootstrapped standard errors. The probability of any emergency room use decreased by 1.1 percentage points (mean of 6.7 percent), but the decreased emergency room expenditures were not statistically significant. The probability of any emergency room use due to asthma attack is decreased by .2 percentage points (mean of .8 percent).

Conclusions: Transportation brokerage services are associated with an overall decrease in inpatient and outpatient Medicaid expenditures and the probability of using Medicaid services. A larger decrease in inpatient expenditures is offset by a relatively smaller decrease in outpatient expenditures due to greater number of outpatient services users than inpatient services users. Decreased Medicaid expenditures are cost effective if beneficiaries’ health is not compromised. The decreased probability of emergency room visits due to asthma attack provides some evidence that health outcomes for Medicaid children with asthma improved with transportation brokerage services.

Permalink

Safety Net Health Care as Low-Quality Universal Coverage: Can it be Optimal?

Presenter:

John Goddeeris

Authors:

John Goddeeris

Chair: Catherine McLaughlin; Discussant: Stephen Zuckerman Tue June 6, 2006 8:00-9:30 Room 226

Despite decades of national debate about creating a system of universal health coverage, a substantial segment of the US population remains without formal health insurance. The persistence of this situation suggests that in some sense universal coverage is too costly to be provided in political equilibrium. Yet the uninsured have some access to health care beyond what they can pay for, through social safety net institutions, largely publicly funded (Hadley and Holahan, Health Affairs, 2003). The health care safety net can in fact be viewed as providing some level of health insurance for all, albeit at a low quality level, arguably lower than many of the uninsured would choose for themselves if the safety net did not exist.

This paper explores, using theory and numerical simulations, the conditions under which a low quality safety net can be part of an optimal income redistribution scheme. Safety net health care is modeled as a publicly-provided good which consumers may accept or opt out of, but may not supplement (that is, those who acquire their own insurance do not have access to the safety net). In a model of optimal nonlinear income taxation, with two types of individuals distinguished by ability, public provision of a private good without supplementation can only enhance social welfare when demand for the good has certain characteristics that do not seem to apply to health insurance. However, if available policy tools are more limited (for example, if only an optimal linear income tax is available), public provision of a good at a level that those of low ability will accept but those of high ability will not can be social-welfare enhancing under broader conditions. Nonetheless, depending on various parameters, it may not be possible to improve upon what can be accomplished with tax and cash transfer policy alone, or optimal public provision may be at a higher quality than consumers of low ability would choose for themselves. I explore the sensitivity of (a) whether such public provision of a safety net can be welfare-enhancing and (b) the optimal quality of the safety net, to several factors, including: the distribution of abilities in the population, the income elasticity of demand for health insurance and its substitutability with leisure, the wage elasticity of labor supply, and the degree of aversion to inequality in the social welfare function. The implications for optimal public coverage of an inability to commit not to provide some minimum quality safety net (because of a Samaritan’s dilemma (Coate AER, 1995)) are also explored. I comment on the relevance of the findings to the current US health care financing system.

Permalink

Do Risk-takers Floss?: Experiments and Health

Presenter:

Julian Jamison

Authors:

Julian Jamison

Chair: David Bradford; Discussant: TBA Tue June 6, 2006 8:00-9:30 Room 235

Experimental economics methodologies provide a valuable and broad set of new tools with which to approach many issues in population health. The underlying idea is to create controlled decision problems (either strategic games or single-person decisions) in a laboratory or field environment, and to pay subjects in cash in order to induce truthful (i.e. incentive compatible) responses. These problems can be used to measure a wide variety of attributes, including: risk and time preferences, fairness/generosity, trust, cooperation/coordination, theory of mind, strategic reasoning, reciprocity, and so on.

There are two main ways to incorporate such methods into health research. The first is to ask subjects about their health behaviors (smoking, exercise, flossing, etc.) and to correlate these answers with their responses in the [context-free] decision problems described above. This has been done so far with a sample of approximately 150 Berkeley MPH students, after which we created an index of health behaviors (with higher scores corresponding to more health-protective behaviors). Interestingly, we find that this index is not correlated with time preference (financial patience), but it is correlated with risk-avoidance (in a financial gamble over losses). This suggests that policy interventions based solely on ‘investment in the future’ are unlikely to be effective. We will be extending this work in the coming months to subject populations in Africa and elsewhere in order to check the robustness of the results in other relevant settings.

The other application of these methods is to utilize them as measures of cognitive response following various stimuli. Specifically, ongoing projects are looking at the effects of alcohol (in a population of young social drinkers) and of sleep deprivation (in a population of emergency medicine residents). In both studies, we will be able to use the economic decision problems in order to assess the cognitive behavioral changes in a novel manner (when compared with the standard psychological tools). Finally, it is possible to use these games as a more fine-tuned instrument to analyze the exact pathways affected by physiological reactivity (e.g. cortisol levels) for determination of health behaviors. The policy implications in all of these cases include both intervention design and more effective targeting of vulnerable populations.

Permalink

The Effects of Family Health on Health Insurance Status in Fragile Families

Presenter:

Hope Corman

Authors:

Hope Corman, Kelly Noonan, Nancy Reichman, Anne Carroll

Chair: David Bradford; Discussant: TBA Tue June 6, 2006 8:00-9:30 Room 235

Children’s characteristics may affect both a father’s willingness to both maintain a close relationship with the mother and to invest in the child. A number of studies found that married couples are more likely to divorce when their child has a serious health problem. Reichman, Corman and Noonan (2004) found even stronger effects among a sample of mostly unwed parents. That study found that one-year-old children with serious health problems are less likely than their healthy peers to live with their fathers and that their parents’ relationship commitment is more likely to have declined since they were born. This may translate to fewer paternal investments in the child.

The purpose of this study is to systematically explore the determinants of unmarried fathers’ financial and non-financial investments in their three- year-old children, focusing on the role of the child’s health status (i.e., whether they are very low birth weight or have serious health deficits) and other characteristics (gender, multiple birth, birth order), but also considering the nature and trajectory of the parents’ relationship as well as detailed individual characteristics of both parents.

We use data from the Fragile Families and Child Wellbeing (FFCWB) study, a national longitudinal birth cohort study of 5,000 sets of new parents with children born between 1998 and 2000. By design, three-quarters of the births were to unmarried parents. Interviews were conducted with both parents directly following the birth, and again one and three years later. We hypothesize that unmarried fathers will provide less support when they have children of lower quality (i.e., are unhealthy).

The FFCWB data set is particularly suitable for this analysis because it includes detailed information on unmarried fathers even if they never lived in the child’s household-including how much time they spend in the child’s household, how much time they spend with the child, and the extent of support they provide in the form of money (both formal and informal child support) and goods. In addition, the survey captures the fluidity of relationships between unmarried parents; at each wave, parents were asked about their cohabitation status, changes in relationship status, and other aspects of their relationship. Finally, we have collected detailed information from the hospital medical records of the mothers and children, enabling us to better characterize serious child health problems than with the surveys alone, and have merged information on the availability of prenatal health inputs at the neighborhood level.

We consider three potential sources of paternal support during the child’s first three years of life: money, goods, and time. We estimate the effect of the child’s health status and other characteristics on paternal support, controlling for maternal and paternal characteristics including race/ethnicity, education, age, work history, health status, children with other partners, and wantedness of the child, all of which may be associated with both the child’s health and the father’s ability or willingness to provide support.

Permalink

Do Your Neighbors Influence Your Health?

Presenter:

Heather Bednarek

Authors:

Heather Bednarek, Rowena Pecchenino, Sally Stearns

Chair: David Bradford; Discussant: TBA Tue June 6, 2006 8:00-9:30 Room 235

Recent trends in the U.S. suggest that the health of Americans is on the decline as measured by the increased percentage of obese and overweight individuals and their decreased physical activity. Several explanations for these trends have been put forth in both the theoretical and empirical literature. Hypotheses include technological change that makes production of food cheaper and time spent at work more sedentary, the rise in the number of restaurants, addiction to calories, reduction in smoking, changes in rates of time preference, urban sprawl and the increased reliance on automobiles. The role of social norms on health has also been discussed as a possible factor affecting weight, mostly in a theoretical context.

Social norms matter when it comes to relative comparisons of income, savings, consumption, or happiness as suggested by several studies. However, less has been done to examine such relative comparisons of healthiness. We examine whether, and in what direction, social norms influence one’s relative healthiness. For example, observing a larger number of overweight or obese individuals may provide comfort or acceptance of that level of health for an individual. Alternatively, it may be provide the impetus for an individual to maintain a healthier weight.

Using the National Health Interview Survey (NHIS), we empirically examine the relationship between healthiness and social norms. Health is measured in terms of weight (body mass index), self-reported health, and time spent at leisure healthy activities. Measures of health social norms are constructed from the health information of others in the local geographic region around a given individual. The standard complement of socio-demographic and economic controls for individual characteristics is also included.

If public health officials and policymakers believe there is a role for public policy in improving the health via decreased obesity and increased exercise of individuals, then it is critical to know how social norms influence the weight and physical activity of an individual. Preliminary results suggest that, indeed, social norms matter. If one is surrounded by others with high BMIs, then that increases one’s own weight. Similar results are found with regard to physical activity; the more (less) you see your neighbors engage in exercise, the more (less) you exercise. It appears that when individuals make comparisons of outwardly visible indicators of health, “peer effects” act as a negative force on health whereby an individual feels comfort that he is one among many overweight individuals. Thus, for any policy to be effective at improving health, it must target not just the individual but the entire community.

Permalink

Overweight in Adolescents and Implications for Health Expenditures

Presenter:

Alan Monheit

Authors:

Alan Monheit, Jessica Vistnes, Jeannette Rogowski

Chair: Albert Okunade; Discussant: TBA Tue June 6, 2006 8:00-9:30 Room 309

Rationale: Adolescents who are overweight (> 95th percentile of the age and gender-adjusted body mass index [BMI] distribution) or at risk of overweight (85th to 95th percentiles) are likely to become overweight adults and be at risk for health conditions associated with high adult health expenditures. However, little is known as to whether being overweight leads to higher health expenditures during adolescence compared to those of normal weight and thus, whether the health expenditures of overweight adults are also present during adolescence. Moreover, little is known regarding the relative contribution of economic factors, parental and family characteristics, and neighborhood attributes in explaining variation in adolescent bodyweight.

Objectives: This paper has two objectives. First, we examine the contribution of economic factors, parental and family attributes, and neighborhood characteristics to adolescent bodyweight, the likelihood of being overweight, or being at risk for overweight. Controlling for parent’s BMI, and thus for adolescent susceptibility to being overweight, we consider whether adolescent weight status is responsive to the price and availability of fast food, the labor supply, educational levels, and health habits of parents, and neighborhood characteristics. Next, we consider the relationship between adolescent bodyweight and health expenditures, controlling for health insurance status, parental educational levels, and household attributes. We consider the sensitivity of our findings to possible endogeneity arising from unobserved attributes of adolescent health status that may be correlated with bodyweight and expenditures, and from measurement error in reported bodyweight.

Methodology: We use data from the Medical Expenditure Panel Survey (MEPS) merged with geographic data on the price/availability of fast food and neighborhood attributes to model variation in bodyweight status among adolescents aged 12 to 19. Our models include a crucial and often omitted variable: parental bodyweight which can reflect both genetic predisposition for adolescent overweight as well as attitudes toward good nutrition. These models also include a comprehensive set of variables capturing economic, household, and neighborhood characteristics. We estimate models of health expenditures to consider whether overweight in adolescents results in enhanced levels of medical spending, and use these results to predict expenditures for adolescents by bodyweight status. We examine the sensitivity of our expenditure estimates to alternative models and to the use of instrumental variable estimates of adolescent bodyweight using parents’ bodyweight and prices of fast and convenience foods as instruments.

Results: Preliminary econometric results restricted to parental and household characteristics reveal that adolescents who are overweight/at risk for being overweight have increased health expenditures compared to those of normal weight in both single-parent and two-parent families. Adolescent overweight is strongly correlated with parental bodyweight and to other factors such as parental educational levels, smoking activity, and high family incomes.

Conclusions: Our findings to date indicate that adolescents who are overweight/at risk of being overweight incur higher health expenditures than those of normal weight. Thus, it appears that high health expenditures associated with being an overweight adult may also be present during adolescence.

Permalink

Participation in Food Assistance, Maternal Employment, and Child Obesity

Presenter:

Haiyong Liu

Authors:

Haiyong Liu

Chair: Albert Okunade; Discussant: TBA Tue June 6, 2006 8:00-9:30 Room 309

Rationale: The effect of Food Stamp Program (FSP) participation on mitigating food insecurity and childhood obesity is likely to be confounded with maternal employment decisions. For instance, supplemental food subsidies may free up mothers’ time spent in home production, namely food preparation, and enable them to work away from home. It is important to account for the potential income effect and substitution effect when assessing the effects of the FSP on weight health outcomes because both income and maternal time are crucial inputs for child health.

Objectives: The objective of this study is to evaluate the impacts of poverty, FSP, and maternal employment on the risk of obesity in early childhood.

Methodology: This study investigates the interactions between poverty, FSP participation, and maternal employment, as well as their impacts on the risk of obesity among children who are raised in single mother families. The main data source in this study is from the matched mother/child data from the National Longitudinal Survey of Youth (NLSY79). Utilizing the full information maximum likelihood (FIML) method, the empirical model estimates the joint decisions of FSP enrollment and maternal employment along with child health (weight) production function simultaneously to account for the potential endogeneity of these health inputs. The identification strategy used in this paper exploits the time series interactions of all current and lagged exogenous variables and a set of instruments that capture the variations of welfare regulations across states and over time.

Results: The results suggest that FSP participation mitigates the risk of childhood obesity among the poor while maternal employment is positively linked to incidents of obesity over the whole socio-economic spectrum. Both findings are statistically significant. In addition, the racial gap of the weight health problem is widening over years, even after accounting for family income and maternal employment. Finally, estimation methods without adequate control for endogeneity of FSP participation and maternal employment are likely to yield biased results.

Conclusions: The childhood health disparities among different socio-economic stratums are widening, especially for racial minorities. The food assistance programs modestly mitigate the risk of obesity among impoverished children while the risk could be aggravated because of mother’s excessive time spent in the labor market. These findings imply that when single mothers transitioning from welfare to the labor market, policy considerations should be given to their family, especially on relaxed eligibility rules for food assistance.

Permalink

The Connection between Maternal Employment and Childhood Obesity: Inspecting the Mechanisms

Presenter:

Rusty Tchernis

Authors:

Angela Fertig, Gerhard Glomm, Rusty Tchernis

Chair: Albert Okunade; Discussant: TBA Tue June 6, 2006 8:00-9:30 Room 309

This paper investigates the channels through which maternal employment affects childhood obesity. We use time diaries and interview responses from the Child Development Supplement of the Panel Study of Income Dynamics which combine vital information on children’s time allocation and mother’s labor force participation. Our empirical strategy involves estimating the effect of children’s activities and meal routines on children’s body mass index (BMI), estimating the effect of maternal employment on these activities and routines and then combining these two estimates. We find that the effect of activities on BMI and the effect of maternal employment on activities vary greatly by the mother’s educational status. In particular, when mothers are highly educated, mother’s employment significantly increases time spent watching TV, which in turn, significantly increases a child’s BMI, while when mothers have little education, mother’s employment significantly increases time spent in school, which in turn, significantly decreases a child’s BMI. However, both groups share the same nutritional effect - higher levels of employment reduce the number of meals consumed per day by children, which increases their BMI.

Permalink

Demand Response of Mental Health Services to Cost Sharing Under Managed Care

Presenter:

Chunling Lu

Authors:

Chunling Lu, Thomas McGuire, Richard Frank

Chair: Glenn Blomquist; Discussant: Mark Dickie Tue June 6, 2006 8:00-9:30 Room 313

Rationale: Demand-side cost sharing on mental health services is usually higher than that on general health services, based on the efficiency rationale that because the demand for mental health care is more responsive to insurance coverage than for general health care services, the welfare loss associated with insurance is higher. Previous studies of price elasticity of demand for mental health services under traditional health care delivery systems, which were mainly done in 80s, provide the supporting evidence by demonstrating a large, statistically significant effect of demand-side cost sharing on the utilization level of mental health services. The new feature of mental health care delivery since 1980s - managed care - was not considered in these studies. The role of demand-side cost sharing in controlling expenditures is different between traditional indemnity plans and managed care plans. Under an indemnity plan, demand-side cost sharing is the only plan policy affecting utilization, whereas in managed care plans, both demand-side cost sharing and managed care mechanisms play roles in determining individuals’ expenditures.

Objective: Demand response under managed care could well be less than under traditional plans, and such a finding would have important implications for designing insurance coverage for mental health care. Although some studies of demand response in managed care have been conducted, none have made an explicit comparison of demand response in traditional and managed care plans. In this study, we use 1996 Medical Expenditure Panel Survey data to examine the demand response of mental health services to demand-side cost sharing under managed health care.

Methodology: The 1996 data are used because this is the only year in which sufficient detail is available on coverage and forms of insurance in order to make the desired comparison. To address the selection problem, we focus on employees (and their dependents) who are privately insured and who have no choice of health plan, a strategy for minimizing the adverse selection problem at least with respect to choice of insurance plans. Couples with more than one insurance plan are also excluded from the analysis. We use logit models to analyze the effect of prices on the probability of any ambulatory mental health uses. We compare the estimated demand response to demand-side cost sharing between managed care plans and non-managed care plans by examining how demand prices affect the likelihood of seeking mental health services.

Results: In the ranges observed, deductibles have no impact on the likelihood of utilization or the level of spending. The coinsurance rate has significantly negative effects on the likelihood of seeking mental health services under conventional plans.

Conclusion: The effect of coinsurance rate on likelihood of seeking mental health services under managed care plan is smaller than that under conventional plans but not significant, implies that demand response to a coinsurance rate under managed care plans is less than that under conventional plans.

Permalink

Financing and Use of Public Mental Health Services and Costs Among the Seriously Mentally Ill

Presenter:

Todd Gilmer

Authors:

Todd Gilmer

Chair: Glenn Blomquist; Discussant: Mark Dickie Tue June 6, 2006 8:00-9:30 Room 313

Background: Research has shown that how mental health services are financed affects the types and amounts of services provided. Most of these studies have considered differences between fee-for-service (FFS) providers and capitated provider groups. To our knowledge, no published study has examined the characteristics of contractual relationships typical of county-based public mental health systems in California.

Objectives: To analyze the relationship between financing and service delivery in a large community public mental health system among three provider types: county providers, county-contracted providers, and FFS providers.

Research Design: We examined outpatient services, hospitalizations and use of emergency services, and costs for persons treated for schizophrenia, bipolar disorder, or major depression in San Diego County in fiscal years 2002-2003 and 2003-2004 (N=20,562). We assigned clients to one of the three provider types based on where they received the majority of their medication management services. Costs were determined by detailed examination of contracts and cost reports, and were linked to individual level encounter data. Multivariate models were used to adjust for a number of socio-demographic and clinical characteristics expected to affect service use.

Results: Outpatient service use was greatest, and total costs lowest, among clients of county-contracted providers. Clients of FFS providers were more likely than clients of county or county-contracted providers to be admitted to an acute psychiatric facility.

Conclusions: Mental health services use and costs in this population are related to provider type.

Permalink

The Effect of Health Literacy on Mental Health Service Outcomes in Adult Medicaid Patients

Presenter:

Richard Smith

Authors:

Richard Smith, Christine Davis, Mary Armstrong, Louis de la Parte

Chair: Glenn Blomquist; Discussant: Mark Dickie Tue June 6, 2006 8:00-9:30 Room 313

Rationale: There is growing consensus about the importance of involving health care consumers in medical decisions, as manifested in the rise of consumer-driven insurance plans. The recently proposed reforms for Florida Medicaid represent the introduction of a more consumer-driven model to populations with low economic and social resources. Consumer involvement is a challenge for those who cannot effectively understand health information. Poor health literacy has been found to result in poor health outcomes and excess costs in the tens of billions annually. Little is known, though, about the impact of health literacy on health outcomes within more disadvantaged populations.

Objective: The objective of this study is to develop an empirical model to examine the relationship of patient mental health literacy to utilization, cost, health service quality, and quality of life among current and former Medicaid recipients of mental health services in Florida.

Data: The data for this analysis comes from two sources: 1) the 2002-03 Florida Medicaid claims data, as furnished by the Florida Mental Health Institute (FMHI), and 2) the 2003 Florida Health Services Survey (FHSS), also from FMHI. These cross-sectional data are linked, consisting of all Medicaid-eligible, non-elderly adult consumers, age 20 to 65, who responded to the 2003 FHSS.

Methodology: We specify a three-equation model of the relationship between an individual consumer and his or her mental health care provider, in which outcomes are utilization, quality of care, and quality of life (cost is an outcome derived from utilization). Following earlier theoretical approaches to the provider-patient relationship, and which have recently been extended by one of the authors, we model the relationship as a two-stage process, in which the provider determines quality in the first stage, while the patient, after observing the level of quality, determines quantity (i.e. utilization) in the second stage. In addition, the model tests for consumer “empowerment” (i.e. patient control over care and having it meet needs) as a mediating determining factor between literacy and mental health care outcomes.

We employ two methods for estimating the model. Under the assumption that consumer literacy and empowerment are unrelated, independent determinants in each equation, we use the method of ordinary least squares (OLS) to estimate the model. However, under the assumption that health literacy is mediated in its effect on utilization, quality of care, and quality of life by consumer empowerment, we use the method of two-stage least squares (2SLS) to estimate the model. With 2SLS, the measure of health literacy serves as the identifying instrument for consumer empowerment. We run a standard Hausman specification test to determine which estimating approach (OLS or 2SLS) produces consistent results.

Results: Results are pending and expected to be available by December, 2005. We will relate these results to earlier work by the presenter on the relationship between patient health knowledge and outcomes in the area of general health.

Conclusions: The results will offer initial insights on applying more consumer-driven plans to public health insurance programs intended for low-income and other disadvantaged populations.

Permalink

Bodyweight and Academic Performance in High School

Presenter:

Edward Schumacher

Authors:

Edward J. Schumacher, Jeff DeSimone

Chair: Glenn Blomquist; Discussant: Alison Evans Cuellar Tue June 6, 2006 8:00-9:30 Room 325

Although childhood obesity has become a major public health issue in the U.S., little evidence exists regarding the impact of being overweight on academic performance. This paper accordingly examines the relationship between bodyweight and grades, using data on high school students in the 2001 and 2003 Youth Risk Behavior Surveys. We isolate a causal effect of bodyweight by specifying an instrumental variable model in which indicators of self-assessed weight relative to the ideal serve as instruments for actual weight. To control for likely sources of deviations between the instruments and actual bodyweight, we also include proxies for time preference and self-esteem. Empirically, our identification strategy works quite well for boys: extremely large partial correlations between the instruments and observed weight produce standard errors sufficiently small to allow for precise inferences, and little heterogeneity with residuals from the academic performance equations is evident. Estimates indicate that additional pounds, being overweight and obesity each reduce grades by a significant amount that exceeds the corresponding OLS effect. Results are similar for girls, though the validity of our procedure is suspect in some specifications.

Permalink

The Relationship between Education and Health Behaviors: Is it Causal?

Presenter:

Adele Kirk

Authors:

Adele Kirk

Chair: Glenn Blomquist; Discussant: Judith Shinogle Tue June 6, 2006 8:00-9:30 Room 325

An extensive body of literature documents a relationship between formal education and health that is strong, broad, and persistent. When presented with such a robust association, it is natural to make the leap, implicitly if not explicitly, to a presumption of causality. However, some have questioned the causal link on both conceptual and empirical grounds, arguing that the apparent relationship between education and health might in fact be due in some part to third factors, such as time preferences or self-efficacy, common to both educational attainment and health, but generally omitted from empirical models. The omitted-variables problem is exacerbated by the nature of most health-specific surveys. Such surveys, while rich in health data, generally provide sparse socioeconomic information about respondents, and in particular, do not provide much, if any, information about the respondent’s family background and socioeconomic circumstances in youth, when educational intentions and possible determinants of adult health behaviors, such as time preferences, are formed. This paper uses a relatively data-rich longitudinal dataset (NLSY79) and instrumental variables methods to investigate the nature of the observed relationship between educational attainment and health behaviors in midlife (ages 35-40), including smoking, heavy drinking, exercise, recent check-up, and weight control. I first estimate a series of models and compare the estimated effects of education on behaviors when other key variables, such as family background measures, a measure of ability (AFQT), and a measure of locus of control, are omitted and then included. I then estimate instrumental variables (IV) models for each dependent variable, using college proximity, area unemployment rates at the time of schooling, and availability of household reading materials in youth as instruments for educational attainment. Preliminary analyses indicate modest but persistently significant effects for education that are generally robust to the inclusion of covariates in OLS and logit/probit models. IV models of drinking and BMI yield estimates that are comparable in magnitude to OLS/probit models but with considerably larger standard errors. But for models of exercise frequency, recent checkup, and smoking, the IV estimates are larger in magnitude than the naïve estimates, and remain significant. Tests of exogeneity indicate that education is exogenous in models of drinking, BMI, and smoking, but endogenous in models of exercise and recent checkup. Overidentification tests indicate that all instruments are excludable except for unemployment in the case of drinking and household reading materials in the case of smoking. Because theory would suggest that naïve estimates are biased upwards, further analyses will explore why some IV estimates are larger in magnitude, not smaller as expected.

Permalink

High School Substance Use and Young Adult Labor Market Outcomes

Presenter:

Pinka Chatterji

Authors:

Pinka Chatterji, Jeff DeSimone

Chair: Glenn Blomquist; Discussant: Sandra Decker Tue June 6, 2006 8:00-9:30 Room 325

This paper uses data from the National Education Longitudinal Study (NELS) to estimate the association between high school alcohol and marijuana use and labor market outcomes in young adulthood. Our empirical strategy is twofold. First, we estimate standard OLS models that incrementally add variables that may confound an observed association between substance use and labor market outcomes. These variables draw on the rich, longitudinal data available in NELS and include personal characteristics, family background, and achievement test scores. Next, we instrument for alcohol use using parental beliefs regarding adolescent substance use (measured when the respondent was in high school) as identifying variables – these factors would be expected to be highly correlated with high school substance use, but not directly related to labor market outcomes of the respondent, after controlling for indirect pathways. Overidentification tests support our identification strategy. Our results offer evidence that marijuana use detracts from adult labor market outcomes but findings for alcohol use are less consistent.

Permalink

Overweight Status and Suicidal Behaviors among Adolescents: Correlation or Causation?

Presenter:

Dhaval Dave

Authors:

Dhaval Dave, Inas Rashad

Chair: Glenn Blomquist; Discussant: Zhun Cao Tue June 6, 2006 8:00-9:30 Room 325

Suicide is the third leading cause of death for adolescents. The suicide rate for individuals 15-19 years of age in 2002, while having recently declined, still remains twice that in 1960. At the same time, the prevalence of being overweight has also steadily risen among adolescents. In 2002, 16.1 percent of adolescents ages 12-19 were overweight, compared to 4.6 percent in 1968. This study utilizes data from the Youth Risk Behavioral Surveillance System (1991-2003) and the National Comorbidity Survey (1991) to explore the causal relationship between overweight (and the perception of overweight) on suicide ideation and attempts. Studies have shown a high degree of correlation between overweight, depressive disorders, and suicidal behaviors. Overweight or the perception of overweight may have a causal impact on depressive disorders, may lead to or reinforce other problem behaviors, and in turn may trigger suicidal thoughts and attempts. On the other hand, adolescents may not be randomly selected into overweight and suicidal behaviors, in which case the unobserved heterogeneity may be driving the association. Propensity score matching and constrained bivariate probit methods are used to assess the degree to which the observed correlation between overweight status and suicidal behaviors is driven by systematic selection rather than true causality. If overweight status is found causally to raise the propensity of engaging in suicidal ideation and attempts, then risk of suicide should be added to the economic costs of obesity. Isolating causal risk factors for suicidal behaviors may also guide targeted interventions for identifying and assisting atrisk youths.

Permalink

A Healthy Economy Can Break Your Heart

Presenter:

Christopher Ruhm

Authors:

Christopher J. Ruhm

Chair: Thomas C. Buchmueller; Discussant: John Cawley Tue June 6, 2006 8:00-9:30 Room 326

This paper uses panel data econometric methods on data from thods the 20 largest states over the period from 1979 to 1998 to investigate how the risk of death from acute myocardial infarction (AMI) varies with macroeconomic conditions after controlling for demographic factors, fixed state characteristics, general time effects and state-specific time trends. The results indicate that a one percentage point reduction in unemployment is associated with a 1.3 percent increase in the AMI mortality rate. This effect corresponds to 2,500 additional deaths. The increase in relative risk is similar for males and females and is larger for 20 to 44 year olds than for older adults. Nevertheless, because the absolute AMI rate is higher for senior citizens, they account for a greater number of deaths. This suggests the importance of factors like air pollution and traffic congestion that increase with economic activity, are linked to coronary heart disease and may have particularly strong effects on vulnerable segments of the population, such as the frail elderly. AMI mortality risk quickly rises when the economy strengthens and then increases further if the favorable economic conditions persist. This is consistent with strong effects of other short-term factors on heart attack risk and with health being a durable capital stock that is affected by flows of lifestyle behaviors and environmental conditions whose affects accumulate over time.

Permalink

Local Labor Market Conditions and Health: Is there a Connection and for Whom?

Presenter:

Philip DeCicca

Authors:

Philip DeCicca, Kerwin Charles

Chair: Thomas C. Buchmueller; Discussant: John Cawley Tue June 6, 2006 8:00-9:30 Room 326

Our paper examines the relationship between local labor market conditions and several measures of health and health behaviors for a sample of individuals living in the fifty-eight largest metropolitan areas in the United States. Since the effect of labor market conditions on health may depend on the extent to which one’s employment status is impacted by them, we split our sample into groups whose employment prospects are potentially more and less likely to be affected by such fluctuations. For example, we allow the impact of local labor market conditions to vary with individuals’ ex-ante employment probability. We also allow it to vary by race and education level since previous research suggests the labor market outcomes of non-white and less educated individuals are relatively more affected by economic fluctuations. In preliminary models, we find evidence of procyclical relationships for body weight and psychological well-being for men with low employment probabilities, but no systematic relationships for a variety of health behaviors including cigarette smoking, heavy alcohol consumption, and various levels of physical exercise for these same men. Separate estimates suggest worsening labor market conditions lead to weight gains and reduced psychological well-being among African American men and lower psychological well-being among less educated males. Among these findings, those related to psychological well-being are most pronounced.

Permalink

Macroeconomic Conditions, Mortality and Health: Evidence from France

Presenter:

Michel Grignon

Authors:

Michel Grignon, Thomas Buchmueller, Marc Perronnin, Florence Jusot

Chair: Thomas C. Buchmueller; Discussant: Sara Markowitz Tue June 6, 2006 8:00-9:30 Room 326

In contrast to the conventional belief that health deteriorates during economic downturns, a growing body of research suggests that mortality and morbidity decline when the economy temporarily weakens. Most studies in this area use data from the U.S. As a result, it is not clear whether this relationship holds in other countries or is an artifact of the unique features of the U.S. health care system and social safety net. This paper extends the literature by using aggregate and individual-level data from France to investigate the relationship between macroeconomic conditions and health outcomes. The aggregate analysis is based on data on mortality and unemployment rates from each of France’s 95 départementes for the period 1982 to 2004. In addition to examining the relationship between overall mortality and unemployment, we conduct separate analyses by source of mortality. The individual-level data come from the Enquête sur la Santé et la Protection Sociale (ESPS), a survey of the French population that has been conducted biennially since 1988. The ESPS provides detailed information on health status, health behaviors and the utilization of medical care. We relate several health outcomes to local economic conditions, controlling for individual demographic and socioeconomic characteristics, general time trends and area fixed effects.

Permalink

Keeping Up with the Jones and Staying Ahead of the Smiths: Evidence from Suicide Data

Presenter:

Mary Daly

Authors:

Mary C. Daly, Daniel Wilson

Chair: Thomas C. Buchmueller; Discussant: Sara Markowitz Tue June 6, 2006 8:00-9:30 Room 326

A growing theoretical literature posits that individuals’ happiness is a function of both absolute income and relative income. Empirical work on this issue is much more limited, generally relying on surveys of happiness or experimental economics. While such empirical studies point to a role for relative income in the utility function, the extent of inferences based on subjective data or the artificial (and nonrandom) environments of experiments is quite limited. Considering suicide as a revealed preference, we show theoretically how suicide data may be used to draw inferences on utility. We build a simple theoretical model of utility which depends on own income and income relative to a reference group. Suicide is treated as a choice variable reflecting one’s assessment of current and expected future utility. Within this model are nested a number of hypotheses that have been suggested in the theoretical literature on interdependent preferences and relative status. We take this model to a dataset containing suicide rates by county, gender, age, and race cells along with measures of absolute and relative income (by county) and numerous control variables thought to be related to suicide risk. We find strong support the notion that individuals look to others when evaluating their own utility and, in fact, care about the status of those above as well as those below them. Suicide risk for those near the middle of the income distribution rises as inequality in the upper tail (90/50 income ratio) rises. Consistent with a basic keeping-up with-the-Jones model, the suicide risk for the low income population rises with inequality in the lower tail (50/10 ratio). Importantly, we also find that individuals care about the income of those below them: suicide risk for the middle income population falls as inequality in the lower half of the income distribution (50/10) rises, while suicide risk for the high income population falls as inequality in the upper tail (90/50) rises. Overall, these results point to an important role for relative income in models of individual utility.

Permalink

The Association between Tobacco Marketing Practices and Youth Smoking Attitudes, Beliefs and Behavior

Presenter:

Frank Chaloupka

Authors:

Frank J. Chaloupka, Sandy Slater

Chair: Michael Grossman; Discussant: Amy Wolaver Tue June 6, 2006 8:00-9:30 Room 332

Existing evidence suggests tobacco remains the least regulated consumer product in the U.S. and childhood experimentation with cigarettes and progression to regular use remains a public health concern. Therefore, the purpose of the present study was to examine the association of tobacco point-of-purchase (POP) marketing practices on youth smoking behavior using a nationally representative sample of adolescents. The study used cross-sectional data collected during a five-year period (1999 through 2003) from a nationally representative sample of 8th, 10th and 12th grade students and the communities in which they reside. For the five years of data, there are a total of 109,308 students and 966 community areas. Students were surveyed about their smoking attitudes, beliefs and behavior and tobacco marketing retail information on cigarette placement, advertising, promotions, and prices, as well as tobacco store density, and industry sponsored and health-related tobacco control signage was collected from the communities. Because students were clustered within communities, a generalized estimating equations (GEE) approach was used for empirical analyses with weighted data. All analyses controlled for demographic and socioeconomic characteristics. Lower cigarette prices, less self-service placement, increased levels of tobacco store density, and advertising were found to be associated with more positive attitudes and beliefs about smoking. In addition, lower cigarette prices, self-service placement, higher levels of advertising, and promotions were found to be associated with increased youth smoking behavior. Results vary by attitudinal and behavioral outcome, but, overall, higher cigarette prices had the strongest impact on increasing disapproval and perceived harm, and decreasing prevalence and consumption. The impact of price on youth smoking behavior is conservative when compared to existing literature. However, analyses indicate by accounting for other important dimensions of tobacco marketing practices that previous research did not, may have resulted in the effects of price on youth smoking behavior being overstated. Results also suggest the tobacco retail environment works both directly and indirectly through attitudes and beliefs on behavior. Analyses indicate the tobacco retail environment does influence youth smoking attitudes, beliefs, and behavior, and additional research is need to help translate these findings into successful policies and tobacco control programs.

Permalink

Access to Fast Food and Food Prices: The Relationship with Fruit and Vegetable Consumption and Overweight Status among Adolescents

Presenter:

Lisa Powell

Authors:

Lisa Powell, M. Christopher Auld, Frank J. Chaloupka, Patrick M. O’Malley, Lloyd D. Johnston

Chair: Michael Grossman; Discussant: Kerry Anne McGeary Tue June 6, 2006 8:00-9:30 Room 332

This study examines the extent to which food prices and restaurant outlet density are associated with adolescent fruit and vegetable consumption, body mass index (BMI) and the probability of overweight. We use repeated cross-sections of individual-level data on adolescents from the Monitoring the Future Surveys from 1997-2003 combined by geocode identifiers with fast food and fruit and vegetable prices obtained from the American Chamber of Commerce Researchers Association and fast food and full-service restaurant outlet density measures obtained from Dun & Bradstreet. The results suggest that the price of a fast food meal is an important determinant of adolescents’ body weight and eating habits: a 10% increase in the price of a fast food meal leads to a 3.0% increase in the probability of frequent fruit and vegetable consumption, a 0.3% decrease in BMI, and a 5.5% decrease in prevalence of overweight. The price of fruits and vegetables and restaurant outlet density are less important determinants, although these variables typically have the expected sign and are often statistically associated with our outcome measures. Despite these findings, changes in all observed economic and socio-demographic characteristics together only explain roughly one-quarter of the change in mean BMI and one-fifth of the change in overweight over the 1997-2003 sampling period.

Permalink

The Effects of Point-of-Purchase Alcohol Marketing and Promotion on Youth Drinking Behaviors

Presenter:

Lan Liang

Authors:

Lan Liang, Frank J. Chaloupka

Chair: Michael Grossman; Discussant: Shin-Yi Chou Tue June 6, 2006 8:00-9:30 Room 332

Point-of-Purchase (POP) alcohol marketing is widespread in stores frequented by U.S. teenagers and young adults. The alcohol industry has credited substantial increases in alcoholic beverage sales to these on-site marketing campaigns. Little is known, however, about the extent to which POP alcohol marketing contributes to the underage drinking and binge drinking problems. To fill in this gap in the literature, we collected POP alcohol marketing information for a sample of stores in the catchments areas of the schools that participated in the Monitoring the Future Surveys in each of the five years from 1999 to 2003. We then merged this information to the Monitoring the Future Surveys to analyze the effects of these POP marketing practices on the prevalence and frequency of youth drinking and binge drinking. The POP information collected included the placement and intensity of exterior and interior alcohol advertisement, sales and promotion of alcohol, alcohol-branded functional objects, placement of beers, and anti-alcohol signage. Analysis based on the merged data provides evidence on which POP marketing and counter marketing device is an important determinant of youth drinking and binge drinking. This research will help design a more nuanced guideline on POP marketing to curb sales practices that might cause increased underage risky drinking and its related negative outcomes.

Permalink

Competition and Charity Care

Presenter:

Chris Garmon

Authors:

Chris Garmon

Chair: Deborah Haas-Wilson; Discussant: Gloria Bazzoli Tue June 6, 2006 8:00-9:30 Room 335

Permalink

Buyer/Seller Bargaining in Option Demand Networks

Presenter:

David Dranove

Authors:

David Dranove, Mark Sattherwaite

Chair: Deborah Haas-Wilson; Discussant: James Langenfeld Tue June 6, 2006 8:00-9:30 Room 335

Permalink

Hospital Mergers and the Consequences for Underserved Populations

Presenter:

Robert Town

Authors:

Robert Town

Chair: Deborah Haas-Wilson; Discussant: Ted Frech Tue June 6, 2006 8:00-9:30 Room 335

Permalink

Evaluating Appropriateness of Antibiotic Use for Children with Otitis Media in Rhode Island

Presenter:

Sylvia Kuo

Authors:

Sylvia Kuo, Susan Miller, James Burrill

Chair: Melayne McInnes; Discussant: TBA Mon June 5, 2006 17:15-18:45 Room 121

Background: The rapid growth in antibiotic use in the 1980s spurred concerns about the spread of antibiotic resistance. Otitis media (OM) is one of the leading indicators for outpatient antibiotic use despite the fact that it is clinically self-resolving. Additionally, most children have at least one OM episode by age 7. These factors led to evidence-based guidelines for the treatment of OM which recommend that antibiotic therapy (i.e. high-dose amoxicillin) should be the default treatment only for children under age 6 months, with use in more limited circumstances in older children. Existing studies have been restricted to either national office-based surveys or state Medicaid populations; little is known of antibiotic practices within a market despite the local nature of practice patterns and disease spread. Objectives: To assess whether antibiotic use among children with otitis media insured by Blue Cross & Blue Shield of Rhode Island (BCBSRI), the dominant insurer in the state of Rhode Island, is consistent with established evidence-based guidelines, and whether this use varies by visit or demographic characteristics.

Methodology: The study uses a cohort of children age 13 or under as of January 1, 2004 who had a diagnosis of otitis media at any time during 2004, and were continuously enrolled with BCBSRI with both medical and drug coverage. The visit-based analysis uses logistic regressions on the probability of receipt of any antibiotic, and the probability that the child receive amoxicillin given that an antibiotic was received; covariates included visit location (hospital inpatient, hospital outpatient, physician office or urgent care), insurance coverage type (PPO, HMO and Medicaid managed care), age (<6 months, 6 months - 2 years, over 2 years), and an indicator for whether the visit occurred on the weekend.

Data: Data come from BCBSRI administrative encounter and pharmacy claims in 2004. Antibiotic prescriptions filled within 2 days of an otitis-related encounter were matched to encounter data. The sample included 12,436 children, who were associated with 26,090 otitis-related encounters and 9,318 matched antibiotic prescriptions.

Results: Preliminary results show that the use of antibiotics by age is consistent with evidence-based practices; children under age 6 months are more likely to receive an antibiotic, with most children receiving the recommended antibiotic of amoxicillin. Interestingly, no differences were found by insurance coverage type. However, the likelihood of receiving an antibiotic are much higher for care provided at urgent care centers (versus physician offices or the hospital emergency room) and for visits occurring on the weekend (versus weekday).

Conclusions: Antibiotics may be used less judiciously for children with otitis media when care is sought outside of their usual source of care. Sites such as urgent care centers may have practice patterns that deviate from recommended pediatric practices because of less familiarity with both the patients themselves, and pediatric patients in general. At the same time, urgent care centers may play a role in reducing costly emergency room visits for acute visits.

Permalink

Financial Incentives and Physician Prescribing Behavior

Presenter:

Sarah Neyaz

Authors:

Sarah Neyaz, Kosali Simon, Adam Mathios

Chair: Melayne McInnes; Discussant: TBA Mon June 5, 2006 17:15-18:45 Room 121

As patents on prescription drugs expire and the threat of generics increases, pharmaceutical companies may move a drug into the over-the-counter (OTC) market to take advantage of brand recognition and maximize profits. Little research to date addressed how this change in prescription status affects the use of this drug or other drugs in its class. This purpose of this paper is to examine whether physicians change their prescribing behavior when a drug switches from prescription to OTC status, and how this varies by the insurance status of the patient and the drug class being considered.

Under a standard principal agent model, the physician’s choice of a drug for a patient’s ailment should not be influenced by its OTC vs Rx status, even though a physician authorization is required only for prescription drugs. However, a switch in regulatory status may entail a change in the cost to the patient depending on their insurance coverage. For example, an OTC version of a drug may cost more if the patient has drug insurance which would pay only for Rx medications. Physicians may also face different financial incentives in the decision to prescribe an OTC vs Rx drug, for example, if a capitated managed care contract implies that Rx drugs count negatively in the revenue equation but OTC drugs do not. Physicians may also face incentives in the decision to prescribe OTC drugs could change the probability of the patient continuing to come in for office visits. We hypothesize that under fee-for-service reimbursement, the physician may have greater incentives to recommend prescription drugs (vs OTC) relative to capitated managed care arrangements. When taking an OTC medication, patients do not have to visit their physicians repeatedly since they can purchase these drugs on their own, without physician authorization.

We propose to use micro data from the National Ambulatory Medical Survey, a nationally representative survey of patients who visit physicians, to investigate the relationship between financial incentives and prescribing patterns.

Permalink

Variations in the Prescribing Patterns of Statins for Persons Diagnosed with Dyslipidemia and CHD

Presenter:

Charlie Link

Authors:

Simon Condliffe, Charles Link

Chair: Melayne McInnes; Discussant: TBA Mon June 5, 2006 17:15-18:45 Room 121

Rationale: Events arising from coronary heart disease (CHD) are the most frequent cause of death in the US. While the onset of this condition occurs naturally with age, the identification of high cholesterol as a risk factor for CHD has prompted implementation of prevention programs, with drug therapies playing a large role. The introduction of statins in the early 1990s has provided a new and more effective treatment for lowering levels of low-density lipoproteins. Statins are promising in terms of prevention of CHD. As studies have shown for many drugs, large disparities exist across insurance types, regions of the country, and various demographic characteristics in terms of whether a patient is receiving the line of treatment recommended by clinical guidelines. That this may be the case for statins is disheartening since the technology not only exists to combat CHD through both primary and secondary prevention but also that statins are an effective treatment for all segments of the population.

Objectives: The objective of the paper is to estimate, using multinomial logit models, the prescribing patterns for statins among two groups of people, those diagnosed with CHD and dyslipidemia respectively (separate equations for each group for each of the 3 data sets). The question: Do the above-mentioned disparities exist with statins? Three databases provide several thousand observations with a host of socioeconomic and demographic information along with diagnosis information including secondary diagnoses. The first two data sets are obtained from office-based physician files (the National Ambulatory Medical Care Surveys (NAMCS)) and hospital settings (the National Hospital Ambulatory Medical Care Surveys (NHAMCS)) and provide 11 years of data, 1992 - 2002. We are in the process of running the analyses for 1996 through 2002 using the Medical Expenditure Panel Survey (MEPS). In addition to being a panel, an advantage of MEPS is that it has much richer socioeconomic data about each respondent.

Results based on NAMCS and NHAMCS: In separate samples of patients with dyslipidemia and with CHD, variations were identified across minorities, particularly for Blacks and, more drastically, regions of the US. While at times directions and magnitudes of these variations differ across data sets and medical conditions, it is evident that variations do exist. Hispanics are less likely to receive statins by 12 to 15 percent, and blacks with CHD at physicians’ offices are 7 percent less likely to receive statins. Patients may be up to 17 percent less likely to receive statins if living outside the Northeast. Medicaid patients are less likely to receive statins. Finally, as time passes patients are more likely to receive a statin, with a yearly increase of 1-5 percent.

We are currently conducting the analyses for 1996 through 2002 for the MEPS data and these will be incorporated into the paper.

Permalink

Hospital Competition, Managed Care and Mortality After Hospitalization for Medical Conditions: Evidence From Three States

Presenter:

Jeannette Rogowski

Authors:

Jeannette Rogowski, Jose Escarce, Arvind Jain

Chair: Ryan Mutter; Discussant: Lan Liang Mon June 5, 2006 17:15-18:45 Room 213

Objective: This study assessed the effect of hospital competition and HMO penetration on mortality after hospitalization for six medical conditions in California, New York and Wisconsin.

Data: Linked hospital discharge and vital statistics data were used to study adults hospitalized for myocardial infarction, hip fracture, stroke, gastrointestinal hemorrhage, congestive heart failure or diabetes. Methods: Logistic regression models were estimated with death within 30 days of admission as the dependent variable and hospital competition, HMO penetration, and hospital and patient characteristics as explanatory variables. Standard errors were corrected for clustering of admissions within hospitals using a Huber-White sandwich estimator.

Results: Higher hospital competition was associated with lower mortality in California and New York, but not Wisconsin. Higher HMO penetration was associated with lower mortality in California, but higher mortality in New York. Conclusion: In the context of the study states’ history with managed care, these findings suggest that hospitals in highly competitive markets compete on quality even in the absence of mature managed care markets. The findings also underscore the need to consider geographic effects in studies of market structure and hospital quality.

Permalink

The Impact of TennCare on Hospital Efficiency

Presenter:

Jennifer Troyer

Authors:

Jennifer Troyer, Cyril Chang

Chair: Ryan Mutter; Discussant: Michael Rosko Mon June 5, 2006 17:15-18:45 Room 213

On January 1, 1994, Tennessee implemented an innovative new health care reform plan called TennCare. The objective of TennCare was to extend health coverage to both individuals in the Medicaid population and people deemed uninsured or uninsurable. A decade later, TennCare still covers the Medicaid-eligible, individuals who are uninsurable as determined by an insurance company, and children whose individual family incomes are below 200% of the federal poverty level. TennCare services have been and continue to be offered through managed care organizations.

Several key facts point to the impact of TennCare on Tennessee hospitals. Between 1993 and 2000, occupancy rates in Tennessee hospitals fell from 47.2 percent to 42.2 percent, and the average length of stay fell from 6.1 days to 5.0 days (Tennessee Department of Health, 2002). In addition, there was a reduction in care disbursed to Medicaid patients in safety net hospitals and an increase in care disbursed to TennCare patients in non-safety net hospitals (Conover and Davies, 2000). Finally, a study sponsored by the Tennessee Hospital Association found that TennCare provided $0.58 for every dollar of inpatient care provided by hospitals (Meyer and Blumenthal, 1996.) Given the dual financial pressures caused by TennCare and the Balanced Budget Act of 1997, questions remain regarding how hospitals in Tennessee have responded to the reductions in reimbursement.

The objective of this study is to consider the effect of TennCare on the efficiency of hospitals operating in Tennessee. Toward that end, we will use information for hospitals operating in Tennessee from 1990-2001. The data allow us to examine efficiency prior to TennCare’s implementation (1990-1993), during the initial years of TennCare (1994-1996), and in subsequent years (1997-2001). Using the complete panel, we will use a multiple-output stochastic frontier approach to simultaneously examine the level of technical efficiency and the effect of key variables on technical efficiency in Tennessee hospitals. This method is appealing as it allows us to consider determinants of cost efficiency, where two groups of explanatory variables are used: 1) traditional measures of outputs, input prices, teaching status, etc. and 2) explanatory variables thought to influence efficiency, including binary variables for the pre-TennCare and post-TennCare periods described above, the proportion of TennCare patient days in the hospital, ownership type, etc. Our examination will allow us to determine the extent to which TennCare has affected hospital efficiency. In addition, we will explore the relationship between hospital inefficiency and characteristics of Tennessee hospitals.

References

Conover, Christopher J., and Hester H. Davies. The Role of TennCare in Health Policy for Low-Income People in Tennessee. Occasional Paper Number 33, The Urban Institute, February 2000.

Meyer, Gregg S., and David Blumenthal. “TennCare and Academic Medical Centers: The Lessons from Tennessee.” Journal of the American Medical Association 276 (9, September 4): 672-676.

Tennessee Department of Health, Health Statistics and Research. Joint Annual Report of Hospitals. September 2002.

Permalink

Insurance Matters: Prevalence and Costs of Potentially Preventable Hospitalizations in Tennessee by Insurance Type

Presenter:

Cyril Chang

Authors:

Cyril Chang

Chair: Ryan Mutter; Discussant: Ryan Mutter Mon June 5, 2006 17:15-18:45 Room 213

Purpose: This paper analyzes Tennessee hospital discharge records and uses insurance type as a factor to predict the likelihood of hospitalization for an ambulatory care-sensitive (ACS) condition.

Background: ACS conditions are those for which timely and effective outpatient care can potentially prevent the need for hospitalization. High incidence of preventable hospitalizations such as asthma, diabetes, hypertension and congestive heart failure may be indicative of underlying problems with access to primary care or deficiencies in outpatient care management. Analysis of preventable hospitalizations can thus provide information for health system evaluation. TennCare is Tennessee’s experimental Medicaid managed care program that began in 1994. An overarching goal of this reform model was to deliver prevention-centered longitudinal care to reduce inappropriate use of inpatient care. The resulting savings, it was theorized, could then be used to expand insurance coverage. However, the expected outcome of better use of primary care of this $8.4 billion program that covers 1 in every 4 Tennesseans has never been investigated.

Hypotheses: We will test two hypotheses. The first is that patients insured by TennCare are more likely to be admitted for ACS conditions than patients insured by other insurance types after controlling for race, age, gender, and co-morbidities. The second hypothesis is that the likelihood of ACS hospitalization varies among the different TennCare Managed Care Organizations (MCOs), with patients insured by nonprofit, academic-medical-center-affiliated MCOs having a lower likelihood of ACS hospitalization than those insured by other MCOs.

Data Sources: The primary source is the Hospital Inpatient Discharge Data System (HDDS) maintained by the Tennessee Department of Health for 1997-2002. HDDS receives information from UB-92 forms on all inpatient discharges in Tennessee. Each form contains information on patient diagnoses, procedures performed, charges, and selected patient demographics. We will also use county-level Area Resource File data to describe the environments in which patients live.

Methodology: We will use the AHRQ definition of Potentially Avoidable Hospitalizations which include those for 16 specific ACS conditions. There will be two dependent variables for each of the two hypotheses: a dummy variable with 1 signifying an ACS admission and 0 otherwise; the total charges of the admission. The key independent variable for the logistic regressions is a categorical variable representing insurance type (TennCare, Medicare, Private insurance, Self Pay, etc.). Control variables include patient demographics such as age, gender, race, patient risk strata that consist of principal and co-morbidity diagnoses, and county characteristics such as population density, percent poverty, supply of primary care physicians per 100,000 population. In the second regression analysis with the hospital charges as the dependent variable, a dummy variable representing whether the admission is for an ACS condition or not will be added.

Potential Contribution: Little is known about TennCare’s effectiveness in the provision and deployment of primary care. If significance differences in the prevalence and expenditures of primary care-sensitive hospitalizations are found between TennCare and other insurance types after controlling for confounding factors, the results will shed new light on the performance of TennCare as a model of public-sector Medicaid reform.

Permalink

An Alternative Approach to Estimating Medical Costs: The Case of Bariatric Surgery

Presenter:

Ritesh Banerjee

Authors:

Ritesh Banerjee, Nilay Shah

Chair: F. Reed Johnson; Discussant: TBA Mon June 5, 2006 17:15-18:45 Room 225

Rationale: Traditional cost accounting data are not always a good measure of resource costs for economic evaluations in health care because the fixed component of total costs is large.

Background: Price is likely to be a major factor determining the demand for discretionary medical procedures that require high out-of-pocket payments. This context allows us to evaluate firm behavior using standard economic theory.

Objectives: We estimate a cost function for Bariatric surgery without using traditional cost data.

Methodology: Our approach provides an alternative method for deriving costs when accounting data may be an unreliable estimate of true opportunity costs. We assume the market for Bariatric surgery is imperfectly competitive and that medical institutions are profit maximizers. Rosse (1970) shows how to empirically fit a cost function under these assumptions using data on price, quantity and institutional features. We estimate the cost function for Bariatric surgery using data from the National Inpatient Sample for 2003. We derive our measure of price from the reported expected payer source.

Results: Preliminary analysis reveals a mildly positively sloped marginal cost curve. This is robust to a variety of linear specifications. We find an inverse marginal cost elasticity of 0.14.

Conclusions: Rising marginal costs may be evidence of capacity constraints. However, a lack of data precludes us from controlling for other potentially important determinants of demand such as advertising, reputation effects and principal-agent issues. These may modify the estimated relationship. Typically, U-shaped marginal costs are considered evidence of large fixed costs. These are likely to be observed in time-series data for an individual firm or in a panel of firms. They may also be observed in a context where firms have different levels of “experience” in production. Our on-going analysis will use panel data from 1997-2003 to further study these factors. We believe the approach we take in this study provides a useful alternative to more commonly used techniques that may not be as appropriate in the health care industry.

Permalink

Assessment of Costs of Obtaining Improved Diabetes Outcomes by Efficient and Inefficient Patients

Presenter:

Arthur Williams

Authors:

Arthur Williams, Matthew Johnson, Sandra Bryant, Steven Smith, Teresa Christianson, Susan Bjornsen

Chair: F. Reed Johnson; Discussant: TBA Mon June 5, 2006 17:15-18:45 Room 225

Rationale: Treatment and management of chronic illness is an increasing concern as both costs of medical care rise and the number of persons with chronic illnesses grows. A better understanding is needed of the effectiveness and costs of interventions designed to improve patient control over chronic conditions.

Objectives: This study focuses on a cohort of patients (N=558) with type 2 diabetes in 1997-2004. This cohort received an intervention, UNITED Planned Care (UPC), that appeared to equalize treatment outcomes between groups of patients who were highly efficient (N=95) in managing their care and an inefficient group (N=95) who were poor managers of care in 1997 and 2004. The objective of this study is to measure the provider costs of equalizing the performance of the two groups on three outcomes: HbgA1c, LDL, and SBP.

Methodology: Provider costs of care to adult type 2 diabetes patients (N=190) enrolled in UPC will be measured for 1997-2004. Costs of the highly efficient and inefficient groups of patients will be compared. Data to measure utilization and costs are in the UPC database and the Olmsted County Healthcare Expenditure and Utilization Database (OCHEUD).

The provider costs to be measured are a) the costs of services and resources from UPC utilized by patients and b) the total costs of all health care and medical services utilized by patients, including hospitalization, over the seven years. Differences in costs between the two groups will be obtained and reported in both nominal and 2004 dollars.

Data Envelopment Analysis (DEA) has been used to assign efficiency classifications to UPC patients. These results have been reported at the Karolinska Institute and other venues. 95 patients on the efficiency frontier were considered very efficient managers of care, while 95 patients farthest from the frontier were considered inefficient managers of care. Input variables in the analysis included BMI, frequency of self glucose monitoring, the use of diabetes medications, statins, cardiovascular medications. Output variables included: HgbA1c, SBP, LDL, cardiovascular events; myocardial infarction, stroke, vascular procedure.

Results: Highly efficient patients had uniformly superior diabetes outcomes in 1997 compared to inefficient patients. The OR (95%CI) were 0.52 (0.42, 0.63) HgbA1c, 0.98 (0.97, 0.99) LDL, and 0.94 (0.92, 0.96) SBP. In 2004, statistical differences in diabetes outcomes between the groups had disappeared: 0.91 (0.76, 1.09) HgbA1c, 0.99 (0.98, 1.00) LDL, and 0.98 (0.97, 1.00) SBP. Patients in the low efficiency DEA group, however, remained in that group.

Conclusions: The apparent lack of efficiency improvements while outcome differences decreased suggests that the better performance of inefficient patients on the outcome measures may be due to increased inputs. We hypothesize that our study will show that a) improvements in outcomes were obtained at modest UPC cost differences but b) total health care costs were substantially higher in the inefficient group while narrowing over the seven years.

Permalink

Comparisons of Utilization and Costs among Women Veterans Obtaining Primary Care in Community Clinics and Veterans Affairs (VA) Medical Centers

Presenter:

Chuan-Fen Liu

Authors:

Chuan-Fen Liu, Elizabeth Yano, Scott Ransom, Matthew Maciejewski

Chair: F. Reed Johnson; Discussant: TBA Mon June 5, 2006 17:15-18:45 Room 225

Rationale: Women are one of the fastest growing segments of VA users, projected to be 10% of the total population of veteran users by 2010. About 11% of 1.6 million women veterans use the VA for some or all of their health care. The lack of awareness of the historic role women have played in the military and their numerical minority have created ethical and logistical problems in trying to create delivery systems that assure their equitable access to high quality comprehensive health services. The VA health care system has developed a system of community-based outpatient clinics (CBOCs) to complement VA Medical Centers (VAMCs) to optimize veterans’ access to primary care services.

Objectives: This study compares health care costs and utilization for women veterans seeking primary care at VAMCs and CBOCs.

Methods: Utilization and expenditure data were obtained from VA DSS Inpatient and Outpatient National Extracts (2000-01) among CBOCs (n=108; 800 women) and their affiliated VAMCs (n=78; 1,044 women) if in operation in 1999, were serving 200+ veterans, and could be tracked independently in DSS. We estimated utilization using negative binomial models and costs using generalized linear models, controlling for patient characteristics, and adjusting for sampling-weights and intracluster-correlation.

Results: Women using CBOCs were older (53 vs 48), had lower service-related disability (15% vs 21%), and were more likely to be a new patient (23% vs 18%) than women using VAMCs (p<0.0001). CBOC patients were less likely to use outpatient services, including specialty care, laboratory, radiology, and other services than VAMC patients (p<0.0001). There was no significant difference in the number of primary care visits between VAMC and CBOC patients; however, VAMC patients had significantly higher primary care costs than CBOC patients due to higher costs-per-primary care visit ($180 vs $118). CBOC patients had significantly lower total outpatient costs (-832, p<0.0001) and total costs (-$894, p<0.0001) than VAMC patients.

Conclusions: CBOC women receive comparable primary care, but fewer specialty and ancillary services than those at VAMCs. CBOC patients appear to be healthier, while CBOCs also offer fewer required services than their VAMC counterparts. Access to evidence-based clinical services is critical to the health of female veterans. VA care should differ only for patient-related factors, not in facility-specific service availability, provider knowledge or expertise, or logistic challenges for veterans. Given the rapidly expanding female patient population, further research is required to assure that veteran women are offered appropriate access to high quality and gender-sensitive care at both CBOCs and VAMCs.

Permalink

The Effect of Maternal Education on Fertility and Infant Health: Evidence From School Entry Policies

Presenter:

Heather Royer

Authors:

Heather Royer, Justin McCrary

Chair: Katherine Swartz; Discussant: Dahlia Remler Mon June 5, 2006 17:15-18:45 Room 226

This paper uses age-at-school-entry policies to identify the effect of maternal education on fertility and infant health. In particular, we exploit the fact that the year in which a child starts school is a discontinuous function of exact date of birth. For example, in California and Texas, our two study states, children must be 5 years old on December 1st (California) or September 1st (Texas) in the year in which they begin kindergarten. As a consequence of these policies, children born within one or two days of one another enter school at different ages and have different levels of education throughout school enrollment. Because individuals born near in time are likely similar, any differences in education at motherhood for individuals born near these entry dates are arguably exogenous. The crux of our identification strategy is to compare fertility and infant health outcomes for mothers born just prior to, and just subsequent to, the school entry date, and to relate the magnitude of these differences to the maternal education discontinuity.

Our analysis points to five key conclusions:

1. School entry policies have a significant impact on schooling at motherhood: one-fourth of young Texas mothers born after the school entry date have a year less education than they otherwise would, had they been born before the entry date. For California, our estimate is one-seventh of a year. These education discontinuities are estimated precisely, with a t-ratio over 20 for Texas and over 8 for California.
2. Maternal education does not significantly impact fertility: women born just before and after the school entry date are equally likely to become mothers, and these women give birth at similar ages.
3. Maternal education does not significantly impact infant health: women born just before and after the entry date give birth to children of similar health endowment, as proxied by birth weight, prematurity, and rate of infant mortality.
4. Maternal education does improve mating market outcomes: women born just after the entry date have younger and lesser educated mates than women born just before.
5. The research design is valid: women born just before and after the entry date have similar characteristics at their own birth, and their grandparents have similar attributes.

Given the fourth effect on mate quality, our infant health results are particularly surprising, as the combined effect of the direct impact of mother’s schooling and the indirect impact of her mate’s quality is zero in terms of infant health. Assortative mating effects would have been expected to lead to important infant health differences, even if maternal education were unimportant in isolation.

Our findings suggest that school entry policies manipulate primarily the education of young women at risk of dropping out of school. For these women, education may play a limited role in fertility and child investment.

Permalink

Do Abortion Restrictions Affect Child Fatal Injury? A Longitudinal Analysis

Presenter:

Bisakha Sen

Authors:

Bisakha Sen, Martha Wingate

Chair: Katherine Swartz; Discussant: Martha Bailey Mon June 5, 2006 17:15-18:45 Room 226

Rationale: The 1973 Supreme Court decision on Roe v. Wade gave American women the constitutional right to terminate a pregnancy via abortion. Subsequently, many states have placed restrictions on abortion access, like denying public funding for abortion, requiring parental consent or notification for minors, and mandating waiting periods. Previous research has indicated that legalizing abortion reduced the prevalence of children living in poverty, in single-parent households, infant mortality, and child abuse. Analogously, it is probable that abortion restrictions would increase these and related phenomena.

Objectives: The objective of this study is to explore the association of state abortion restrictions to two child outcomes - fatal injuries from violence/homicide and fatal injuries from unintentional causes/accidents. The underlying logic is that abortion restrictions are likely to increase the number of ‘unwanted’ births - leading to child physical abuse and neglect, and also increase the number of births to women with inadequate resources and abilities to protect their children from harm.

Methodology: The study utilizes National Center for Health Statistics (NCHS) data over 1981-2001 on deaths to children aged <1 years to 5 years, where the ICD code shows cause of death to be violence or unintentional injury. The data includes information on states of death and birth. This is linked to state-level cross-sectional time-series data on the following abortion restrictions: enforced parental consent laws and parental notification laws, absence of public funding for abortion (barring cases of rape and incest), and mandatory waiting periods. Additional controls include the number of bordering states with no parental involvement law in effect, abortion-provider availability per 1000 women, state poverty rate, percentage of population in rural areas, per capita alcohol consumption, maximum monthly AFDC payments for family of 3, and existence of ‘family cap’ laws. The main hypothesis is that existence of abortion restrictions in the state and year of birth will be associated with subsequent increases in child fatal injury due to violence and due to unintentional causes. The study employs count data models with state and year fixed effects to control for unobserved heterogeneity, and further uses fatal injury rates among young adults from the same causes to control for time-variant state level unobservables. The analyses are done separately for white and black children.

Preliminary Results: Preliminary results using NCHS fatal injury data aggregated over the 0-4 year age-group (available from WISQAR, Centers for Disease Control & Prevention) show that parental consent laws are associated with increases in violence-resultant fatal injury rates for white children, less so for black children; no public funding and mandatory waiting periods are associated with increases in unintentional fatal injury rates for black children. The results are robust to a variety of specification tests. This study is particularly relevant at this time, since recent changes in the U.S. Supreme court have raised questions about the future of abortion rights. It is important from a public policy perspective to understand the effects of existing abortion restrictions so as to gauge what the effects of future restrictions might be.

Permalink

Demand for the Influenza Vaccine among the Elderly

Presenter:

Padmaja Ayyagari

Authors:

Padmaja Ayyagari

Chair: Katherine Swartz; Discussant: Steven Pizer Mon June 5, 2006 17:15-18:45 Room 226

This study examines the determinants of influenza vaccine demand among the elderly U.S. population and it’s effect on the cost and usage of care. The analysis is done using data from the Medicare Current Beneficiary Survey (MCBS). In contrast to other studies that use a non-representative sample and are cross-sectional in nature, MCBS data is a panel and a representative sample of the Medicare population. Influenza or complications arising from it are responsible for a large number of hospitalizations and deaths among the elderly and as such, vaccinations have large potential benefits. I examine the following determinants of demand: the price of the vaccine to the individual which depends on insurance coverage, demographics such as age, race, gender, education and income, family composition such as whether the person lives alone or not, health conditions including chronic ones (e.g, asthma) that may be complicated by influenza. I also examine the effect of an individual’s expectation about the risk of infection. These individual expectations depend on the prevalence of the disease as well as on the individual’s own past experience with the flu and the vaccine on demand. Heterogeneity in expectations could give rise to differences in demand among individuals who are similar in other aspects. This project also studies the effect of getting a flu shot on hospitalizations and on payments by Medicare. I use the fact that Medicare Part B started covering flu shots in 1993 to identify the price elasticity of demand. This is likely to induce demand among part B beneficiaries but not among those with only part A coverage. While, the induced demand would increase payments by Medicare, vaccination itself might reduce hospitalizations and other medical care use. Vaccinations also affect mortality and in turn the cost of care conditional on survival. Thus, the exact impact on Medicare costs and medical care use is not clear without an empirical analysis.

Permalink

Determinants of growth in prescription drug expenditures

Presenter:

Nilay Shah

Authors:

Nilay Shah, John Mullahy, Maureen Smith, David Vanness

Chair: Randall Ellis; Discussant: Allen Goodman Mon June 5, 2006 17:15-18:45 Room 235

Rationale: Over the last decade there has been a large growth in prescription drug expenditures in the United States. Spending on drugs more than tripled between 1993 and 2003, increasing from $51 billion to $179 billion (in 2003$). There is an increasing interest in understanding the factors affecting the growth in prescription drug expenditures as their impact on public and private healthcare budgets continues to increase.

Objectives: This study seeks to evaluate the role of changing population characteristics on the growth in prescription drug expenditures between 1997/1998 and 2001/2002.

Methodology: Using data from the Medical Expenditure Panel Survey (MEPS), we employ non-linear extensions of the Oaxaca-Blinder regression-based decomposition method to attribute the sources of growth in prescription drug expenditures. These methods decompose the total observed growth in expenditures into growth attributable to changes in population characteristics (covariates) and growth attributable to the changes in expenditure structure associated with population characteristics (parameter estimates). We implement the decomposition analysis using generalized linear models. Analyses are weighted reflecting both the sample design of the MEPS to represent the non-institutionalized, civilian, U.S. population and survey non-response.

Results: Total annual prescription drug expenditures almost doubled between 1997/1998 and 2001/2002, increasing from $73 billion to $143 billion. Average per capita expenditures grew by more than 70 percent, increasing from $292 to $501. Annual per capita prescription fills increased by 2 prescriptions per person per year. All major therapeutic classes and drug classes saw a significant growth in total annual prescription drug expenditures between 1997/1998 and 2001/2002. More than 40 percent of the growth in expenditures between these time periods can be attributed to changing population characteristics. Much of this growth can be attributed to the increasing prevalence of chronic conditions and “aging” of the population. Specifically, increase in prevalence for hyperlipidemia, hypertension, allergic rhinitis, psychiatric conditions (such as psychoses and schizophrenia) and gastroesophageal reflux disease explained 23 percent of the growth in prescription drug expenditures.

Conclusions: Our analyses found a significant growth in prescription drug expenditures in the U.S. between 1997/1998 and 2001/2002. We also observed significant growth in expenditures for all demographic characteristics, as well as, all major therapeutic classes for prescription drugs. These analyses revealed that a significant portion of the growth in prescription drug expenditures may be attributed to the changing prevalence of chronic disease in the U.S. population. These findings have important policy implications for future expenditures, especially with the projected demographic changes in the U.S. population.

Permalink

Costs and Outcomes of Chiropractic Treatment for Low Back Pain: Evidence from the US and Australia

Presenter:

Allan Brown

Authors:

Allan Brown, Doug Angus, Stella Chen, Zhiliu Tang, Sarah Milne, Juergen Pfaff, Huimin Li, Shaila Mensinkai

Chair: Randall Ellis; Discussant: Karen Eggleston Mon June 5, 2006 17:15-18:45 Room 235

Issue: Low back pain (LBP) is a common and costly medical problem in industrialized countries. It is experienced by 70% to 80% of adults at some time during their lives. There is uncertainty about the clinical and cost effectiveness of chiropractic care for LBP relative to standard medical treatment or physical therapy.

Objective: To systematically review the evidence on clinical effectiveness, comparative cost, and cost-effectiveness of chiropractic for LBP.

Methodology: The comparators for chiropractic treatment of LBP were conventional medicine and physical therapy. A search strategy was developed for existing systematic reviews of clinical effectiveness, supplemented by a literature search of new RCTs since the end of 2002. Non-RCTs were also covered. A search strategy was also developed for the review of economic evidence. Qualitative analysis was performed on the included effectiveness studies and economic studies. A cost comparison analysis was done using results of the economic review. Quality of the included studies was assessed using the Oxman Guyatt Scale, Jadad Scale, Newcastle Ottawa Quality Assessment Scale, and a quality check for cost studies.

Results: For the effectiveness review, our selection criteria were met by eighteen systematic reviews, and four trials published after the reviews were completed. Ten studies were identified in the economic review. Four were cost comparisons and six were cost consequence studies. Nine were from the US and one was from Australia.

Conclusions: Chiropractic for LBP is similar in effectiveness to standard medical care and physical therapy. The higher quality reviews did not find significant differences in effectiveness. There is no clear cost advantage for any of the three methods studied. In terms of improving lost time from work, chiropractic care was similar to physical therapy; and as effective as or better than standard medical care.

Permalink

Is the price of cancer treatment falling?

Presenter:

Alexandra Constant

Authors:

Alexandra Constant, Marie-Chantal Benda, Ruolz Ariste, Charles Mallory

Chair: Randall Ellis; Discussant: Tzu-chun Kuo Mon June 5, 2006 17:15-18:45 Room 235

Rationale: Cancer care represents one of the most costly sectors of the medical care. Rapidly rising costs form medical care pose a crisis of “sustainability” of publicly funded health care systems. With rapidly evolving medical technology, it is unclear whether price increases or greater utilization is responsible for more of the increase in health care system costs.

Objectives: The goal of the study is to assess the impact of changing technology for cancer treatments on health care costs to distinguish price changes from quality/quantity changes.

Methodology: We estimate two types of price indexes: a service price index (SPI) that reflects the price of particular treatments over time, and a cost-of-living index (COLI) which measures an outcome-adjusted cost of treating a specific health problem and reflects price variations by discounting any technological/quality effects. The two indexes are estimated for the period 1995/96 and 2001/02 by combining data on cancer patients admitted to hospitals participating in the Ontario Case Costing Initiative (OCCI) with death data from the Canadian Cancer Registry (CCR).

Results: Hospital cost per patient increased slightly in Ontario for lung, breast, prostate and colorectal cancer. The SPI rose at an average annual rate of 2.0%. However, taking into account health outcomes from better health care technology, it is clear that the price of cancer treatment is falling rapidly. The COLI fell by 9.2% annually. Although cancer prevalence is increasing, more effective treatments are helping to hold costs down.

Conclusions: Recognition of the cost drivers in the total medical cost is important, as policy implications would vary. Given that cost increases are attributable to the wider use of more effective treatments there is a net social benefit associated with the expenditure increase. This study supports the growing international evidence that costs of care for some diseases are falling.

Permalink

Accounting for changes in morbidity among the non-elderly in the U.S.

Presenter:

Kumiko Imai

Authors:

Kumiko Imai, Edward Gregg, Yiling Cheng, Ping Zhang

Chair: Jeffrey Pyne; Discussant: David Cutler Mon June 5, 2006 17:15-18:45 Room 309

Background and objective: Evidence suggests that disability among the elderly in the U.S. has declined by 1-2 percent per year during the past several decades, but it is unclear whether disability has also decreased among the non-elderly population. The purpose of this paper is two-fold: (1) to estimate trends in morbidity among the non-elderly using more recent national surveillance data; (2) to determine if changes in disease severity or disease prevalence are responsible for these trends.

Methodology: We used data from the National Health Interview Survey (NHIS), a yearly nationally representative probability sample of the non-institutionalized U.S. population. We estimated trends in disability among individuals aged 18-64 using the 1997-2004 NHIS, and conducted decomposition analysis to determine the degree to which these changes are influenced by changes in prevalence of chronic diseases (including cardiovascular disease, diabetes, and hypertension) and changes in disease severity. Disability was defined as having personal-care or routine-needs limitations. We also looked at the proportion of individuals whose self-reported health is poor or fair. The analysis used a complex sample survey method with controls for age, sex, race/ethnicity, employment, and education.

Results: The data show that disability among the non-elderly population increased over 20 percent between 1997 and 2004 (2.2% in 1997; 2.4% in 2004). The proportion of individuals whose self-reported health status was fair or poor also rose over 14 percent during the same period (9.1% in 1997; 10.4% in 2004). It is estimated that the increased prevalence of cardiovascular disease, hypertension and diabetes among those aged 45-64 accounts for about 15 percent of the increase in disability and 20 percent of the lower self-reported status. There is little evidence of increased severity among individuals with these conditions.

Conclusions: Our study found that, after accounting for demographic and socioeconomic changes, disability among the non-elderly has increased in the past several years. The proportion of those who report poor or fair health status has also risen. The results also show that, in contrast to what has been suggested for the elderly, increased chronic disease prevalence accounts almost all of worsening disability and self-reported health status. The findings highlight the importance of primary prevention of chronic diseases in reducing morbidity among the middle-aged population.

Permalink

Exercise? May be another time: The Cost of Sedentary Life Styles

Presenter:

Nazmi Sari

Authors:

Nazmi Sari

Chair: Jeffrey Pyne; Discussant: Jeffrey Pyne Mon June 5, 2006 17:15-18:45 Room 309

Physical inactivity, one of the major health issues in the modern world, is estimated to cause 1.9 million premature deaths worldwide annually. Globally, it is identified as a cause of 10 to 16 percent of breast cancer, colon cancers, and diabetes, and about 22 percent of ischemic heart disease. Therefore, physical inactivity is a major contributor to the rising healthcare costs. Only in the USA estimated healthcare cost due to physical inactivity was around $75 billion in 2000. In Canada, physical inactivity accounts for about 6 percent of total healthcare costs. Although the level of inactivity showed a decrease from 62 percent in 1994 to 56 percent in 2001, two-thirds of Canadians aged 25 to 55 are not physically active enough to meet the guidelines for sufficient physical activity set out in the Canada’s Physical Activity Guide. Therefore, physical inactivity continues to be a burden on collectively funded health insurance system of Canada.

Physical activity improves health status and has the potential to reduce disability days and utilization of expensive healthcare services. Increase in healthcare costs can be controlled with health promotion efforts aimed at changing life styles of individuals. Such health promotion efforts have the potential of increasing life satisfaction and at the same time saving money for the society.

This area of research has been neglected by economists, and the estimates provided above are based on the cost of illness studies. These studies do not consider impacts of choices made now on future resources; therefore estimates are not calculated by taking lifetime costs and benefits into account. The aim of this paper is to estimate the healthcare costs and productivity loss due to sedentary life styles, and to calculate the lifetime cost savings if inactive people would instead be active throughout their life. We, therefore, estimate implied benefits from a health promotion program aimed to change life styles of inactive people.

The paper uses the Canadian Community Health Survey 2.1, which includes population level information on health determinants, health status and healthcare utilization. The sample size is approximately 130,000 Canadians and provides information at the level of provincial health regions. The dependent variables are non-negative count of health service utilizations and disability days. The sample mean and standard deviation are quite different, suggesting that the over-dispersion can cause a downward bias in the standard errors resulting from Poisson estimation. At the same time, the dependent variables take zero for a significant proportion of the sample. As a solution, we use zero inflated Poisson and/or Negative Binomial models.

Our results show that physically inactive people use more healthcare services compared to the active group. The range varies from 18 to 78 percent depending on the type of care. And the same pattern is true for disability days, around 43 percent more for inactive group. After obtaining estimates from the regression analysis, we estimate implied savings under the assumption that inactive people will become active throughout their life.

Permalink

Adult Body Mass Index as a Determinant of All-Cause Mortality in the United States: 2000 - 2003

Presenter:

Albert Okunade

Authors:

Albert Okunade, Rose Rubin

Chair: Jeffrey Pyne; Discussant: Nazmi Sari Mon June 5, 2006 17:15-18:45 Room 309

Medical care and the social science disciplines, together with global bodies (e.g., The UN Food and Agriculture Organization, and The World Health Organization) concerned with human health conditions, agree that the rising obesity epidemic (the US, worldwide) and related ill-health outcomes (e.g., discounted quality of life, atrophied labor productivity, preventable mortality and morbidity) are increasingly problematic. The resource costs of unhealthy Body Mass Index (BMI) increasingly individuals and the society (US DHHS, Healthy People 2010. Washington, DC: USGPO, November 2000). The BMI, a number adjusting adult body weight (in lbs. or kgs.) for height (in ft. or m.), classifies individuals as ‘not overweight/obese’ (BMI< 24.9), ‘overweight’ (25 30). The wider human health and medical literature increasingly links specific-cause and all-cause mortality to unhealthy BMI values. During 1990 and 2000, poor diet and physical inactivity ranked second, next to tobacco, as the second leading actual causes of death in the US (National Center for Chronic Disease Prevention and Health Promotion, 2000). Heart disease, cancer, and stroke, the three leading causes of US deaths, are positively correlated with the conditions of being overweight and obese. Obesity has been linked to almost all causes of death and morbidity, including chronic respiratory diseases and diabetes (respectively, also the 4th and 6th leading cause of deaths in the US in 2000).

Therefore, the goal of this study is to construct an econometric model linking all-cause mortality to the status of ‘overweight’ and ‘obese’ separately, as measured by BMI. Data are taken from the Behavioral Risk Factor Surveillance System, for the four years 2000, 2001, 2002, and 2003 (for which BMI data are consistently reported). Data of the 50 US states and the District of Columbia are modeled. Separate regression models linking all-cause mortality to ‘not overweight/ obese’, ‘overweight’, and ‘obese’ adult BMI, controlling for a set of independent other effects (e.g., per capita income, race/ethnic classification, number of toxic waste sites on location, others). Annual and panel regression model estimation first tested for, confirmed, and adjusted the model for the skewed data distribution of all-cause mortality per 100,000 people, and then modeled it as a Box-Cox power-transformed response to symmetry. Separate regression models of the BMI effects of mortality are reported for each year, and for the panel of years. Specific intervention policies are linked with the effect of BMI on mortality to assess their effectiveness. This paper concludes on the effectiveness of the various health promotion and disease prevention programs in specific US states, and reflects on the differences in health and economic consequences of unhealthy BMI for the ‘overweight’ and ‘obese’ in the US, from 2000 to 2003.

Permalink

Does Practice make Perfect? Evidence from Cardiac Surgery

Presenter:

Subramaniam Ramanarayanan

Authors:

Subramaniam Ramanarayanan

Chair: Didem Bernard; Discussant: Bill Encinosa Mon June 5, 2006 17:15-18:45 Room 313

Background: A number of studies in the health economics literature document the presence of a correlation between provider (hospital and individual physician) volume and patient outcomes for a variety of procedures. However, few studies attempt to translate this correlation into a well established causal relationship between procedure volume and outcome. It is important to unambiguously determine causality, as the two hypotheses have contrasting implications for policy.

Contributions: This study makes a contribution to the literature by using a novel instrumental variables technique to rule out alternate explanations (the “selective referral” hypothesis) and arrive at a consistent estimate of the impact of individual learning-by-doing on quality. The use of individual (surgeon) level data also allows me to test whether skills acquired by surgeons are equally effective across hospitals.

Data: The data for this study come from the Hospital Inpatient Data Files provided by the Florida Agency for Health Care Administration (AHCA) for the years 1994-2003. I restrict my analysis to Coronary Artery Bypass Grafts (CABG) performed in this period.

Methodology: The volume term in a regression of outcomes on provider volume is endogenous if one takes into account the fact that physicians of higher quality (i.e. with better outcomes) attract more patients. An ideal instrument should help explain variation in physician procedure volume, but have no causal relationship with physician quality (i.e. outcomes of a physician’s patients). This paper uses physician exit as an exogenous identifier. The estimation strategy works as follows: if a hospital has a set of practicing surgeons, and one of them exits for exogenous reasons (say retirement) at time t, this leads to an exogenous increase in volume for the remaining physicians at time t, as the volume of the exiting physician at time t-1 gets redistributed among the remaining physicians at time t. Assuming this shock to physician volume is not correlated with unobservable determinants of physician quality, I can use it to estimate the impact of provider volume on patient outcomes.

Findings: First-stage regression results indicate that the instrument is strongly correlated with the endogenous predictor variable. The second-stage results reveal that a one-unit increase in physician procedure volume leads to a statistically significant decline in mortality of .03 percentage points, which translates to nearly a 1% drop in mortality. This is indicative of a strong learning-by-doing effect. Also, learning is not firm-specific: a one-unit increase in physician volume at any hospital has equal impact on mortality at all hospitals where the physician operates.

Conclusions: The proposed instrument helps to unambiguously establish the presence of a learning-by-doing effect for individual surgeons in CABG procedures. The findings are in favor of federal and state regulations that support consolidation of providers.

Permalink

The Migration of Highly-Skilled Workers: The Case of Physicians

Presenter:

Pierre Leger

Authors:

Pierre Leger, Benoit Dostie

Chair: Didem Bernard; Discussant: Jim Walker Mon June 5, 2006 17:15-18:45 Room 313

The international migration of medical professionals has garnered much media attention of late in many countries. Although many reports have been alarming, a casual look at the data suggests that the brain drain may not be an important issue, as only a small proportion of physicians migrate internationally each year. Governments have, nevertheless, been preoccupied by issues of migration. This is because, from a public-policy perspective, it is not only the number of individuals who migrate that is of interest, but also who in particular migrates (and, in the case of intra-country migration, where they migrate to). If those who migrate are the relatively high-skilled or productive, then simply looking at net migration to determine the impact of migration may be misleading.

The goals of this paper are threefold. First, we wish to contribute to the general literature on the fundamental drivers of international and interprovincial mobility of highly-skilled workers. Second, we wish to uncover the link between observable and unobservable skills and the propensity to migrate. Finally, we wish to contribute to the recent empirical literature on the Borjas (1987) hypothesis, which posits that individuals sort according to skill levels and regional returns to these skills, by considering not only observable skills (as in the previous literature) but also unobservable skills (which has not, to our knowledge, been considered before).

In order to do so, we focus on the migration patterns of Canadian physicians (using a 10 year panel). We choose this particular group for several reasons. First, physicians (especially specialists) are often singled out, in Canada, as a highly-educated group who frequently experience both international and interprovincial migration. Furthermore, focusing on physicians allows us to study the migration decisions of a relatively homogenous set of workers in terms of formal education. Finally, Canadian physicians are generally paid on a fee-for-service (i.e., a fee-per-consultation) basis. Thus, wage rates (or fees) can be considered exogenous to both observable and unobservable individual characteristics as they reflect exclusively the physician’s specialty and province of practice. Consequently, we can use the physician’s total billings as a measure of his or her output and offer a precise definition of a highly-skilled physician: a physician who has greater billings (i.e., has a greater output of health-care services) than would otherwise be predicted by their observable characteristics and the exogenous wage-rate they face.

We first estimate a model which allow us to uncover whether or not individuals who earn more than would otherwise be predicted by their observable characteristics (i.e., are highly-skilled according to our definition) are also those who are more likely to migrate. We then turn to the impact of destination-specific characteristics (particularly earning differentials) on the choice of physician location. Finally, we examine whether differences in the returns to observed human capital and unobserved skills across different province are an important factor in physician mobility.

Permalink

Is Entry Efficient? Lessons from Cardiac Surgery

Presenter:

Jonathan Kolstad

Authors:

David Cutler, Robert Huckman, Jonathan Kolstad

Chair: Didem Bernard; Discussant: TBA Mon June 5, 2006 17:15-18:45 Room 313

By regulating entry into services with substantial fixed costs, certificate of need (CON) regulation has been viewed as a means of combating fears of a “medical arms race”, in which hospital markets would experience excess capacity and increased costs due to non-price competition between providers. Further, a common view in the health economics literature has been that there is a positive relationship between a surgeon’s volume (in aggregate or at a specific hospital) and his or her outcomes. If this is the case, entry limitations may improve quality to the extent they increase the volume per surgeon at a specific hospital.

We develop a model suggesting a different view of the role of entry on firm level competition. Instead of identical surgeons, our model is based on a distribution of surgeon quality both generally and at a firm-specific level. In this setting, hospital competition on the basis of both quality and quantity theoretically could be welfare enhancing without increasing quantity supplied. This would occur if free entry changes the contracting between hospitals and surgeons possessing heterogeneous levels of skill so as to create a more efficient distribution of surgeon skill across hospitals.

We provide empirical tests of the model using data from Pennsylvania for coronary artery bypass grafts (CABG) procedures performed between 1993 and 2003, both before and after the lifting of the statewide CON law in 1996. One critique of CON repeal is that it might encourage more surgeons to split their time across multiple hospitals, thereby reducing the benefits of hospital-specific volume-outcome effects. We find that, during the CON period in Pennsylvania, splitting appears to reduce surgeon performance. In the post-CON period, however, volume-outcome effects remain but are no longer firm specific, suggesting that splitting time between hospitals was not detrimental to outcomes. Specifically, the repeal of CON encouraged hospitals to compete more aggressively for higher quality surgeons; these high-quality surgeons accordingly split their time across institutions without reductions in performance. We are also able to test whether entry induced demand and find that entry was not associated with increases in CABG volume at a market level. In total, we find that the repeal of CON in Pennsylvania improved welfare without increasing the quantity of surgeries performed.

Permalink

A New Explanation for Differential Trends in Infant Mortality by Race

Presenter:

Authors:

Ellen Meara, David M. Cutler, Seth Richards

Chair: David Meltzer; Discussant: Anna Aizer Mon June 5, 2006 17:15-18:45 Room 325

Racial disparities in the rate of infant mortality have been documented throughout the 20th century in the United States. Despite progress in reducing black white differentials in the late 1960s, the gap between black and white infant mortality has again increased in the past two decades. The ratio of black to white infant mortality is higher now than at any point in the past century. We consider explanations for this trend. We estimate that two thirds of this adverse shift in relative black infant mortality relates to greater improvement in survival for whites of any birth weight. We further argue that the faster improvement for white infants stems from biased technological innovation. Because research and investment tends to favor the leading causes of death overall, medical advances that improve infant mortality automatically favor infants from the majority group. In support of our hypothesis, we document that between 1983 and 1998 1) mortality declines are greatest for causes of infant mortality that were most prevalent in 1983 2) mortality declines increase with the level of research devoted to a given cause of death; and 3) both initial mortality rates and measures of technological innovation are positively related to the increase in relative black mortality during this period. Our theory also draws support from case studies of particular causes such as respiratory distress syndrome. Our findings suggest that even in the absence of disparities in access to medical care or discrimination in the medical system, racial disparities in infant mortality can increase as medical technology improves.

Permalink

Causes and Consequences of the Increase in Treated Disease Prevalence

Presenter:

David Howard

Authors:

David H. Howard, Susan Busch

Chair: David Meltzer; Discussant: Christopher J. Ruhm Mon June 5, 2006 17:15-18:45 Room 325

The proportion of the population treated for major medical conditions, including heart disease, diabetes, cancer, and mental illness, increased rapidly during the 1990s. We document these trends using data from the Medical Expenditure Panel Survey and Medicare claims data and show that increases in treated prevalence are responsible for a large share of the growth in health care spending. We then review evidence for and against the following explanations for increases in treated prevalence: changes in risk factors, availability of new medical technology, changes in the market structure of treatment provision (i.e. approval of new on-patent drug treatments), and the rise of a “culture of early diagnosis” in medicine. We find that obesity explains a substantial share of treated prevalence increases (34% for diabetes, 71% for high blood pressure, 10% for back problem, etc.), but large residuals remain for most conditions. We conclude that improvements in technology combined with an increasing emphasis in medicine on early detection are responsible for most of the increases. Understanding the impact of technology on incentives to diagnose new cases is important for explaining the mechanism by which medical technology increases health care costs.

Permalink

Chronic Disease and Trends in Severe Disability in Working Age Populations

Presenter:

Jay Bhattacharya

Authors:

Jay Bhattacharya, Kavita Choudhry, Darius Lakdawalla

Chair: David Meltzer; Discussant: Zhou Yang Mon June 5, 2006 17:15-18:45 Room 325

Recent work has shown that rates of severe disability, measured by the inability to perform basic activities of daily living, have been rising in working age populations. We examine the extent to which chronic disease trends can explain these disability trends. Our primary findings are that for 30 to 45 year-old populations between 1984 and 1996: (1) disability prevalence fell dramatically among the non-chronically ill; (2) rising obesity prevalence explains about 40% of the rise in disability attributable to trends chronic illness; and (3) rising disability prevalence among the chronically ill explains about 60% of the rise in disability attributable to trends in chronic illness.

Permalink

Hospital Ownership and Quality: A Quantitative Research Review

Presenter:

Yu-Chu Shen

Authors:

Yu-Chu Shen, Karen Eggleston, Jia Chan, Joseph Lau, Christopher Schmid

Chair: Kevin Volpp; Discussant: David Becker Mon June 5, 2006 17:15-18:45 Room 326

Permalink

A Competition Index for Differentiated Products Oligopoly with an Application to Hospitals

Presenter:

Yaa Akosa Antwi

Authors:

Yaa Akosa Antwi, Martin Gaynor, William B Vogt

Chair: Kevin Volpp; Discussant: Frank Sloan Mon June 5, 2006 17:15-18:45 Room 326

We develop a competition index for differentiated products oligopoly and apply it to assessing the impact of concentration on price in hospital markets in California. Our index, which we term LOCI, is bounded between zero and one and increases with the competitiveness of a market. We use 1992-1995 hospital data from California to estimate the impact of concentration on price using our new concentration measure. We find that on average, hospital prices decrease significantly as markets are more competitive. A hypothetical merger that decreases the number of firms from 3 to 2 leads to a price increase of $800, or 16%. The estimated average demand elasticity for hospital services is -3.55. Government hospitals face a less elastic demand of -3.14 than all ownership types. Demand elasticity for for-profit and not-for-profit hospitals is -4.5 and -3.25 respectively.

Permalink

Regulatory Exploitation and the Market for Corporate Control

Presenter:

Leemore Dafny

Authors:

Leemore Dafny, David Dranove

Chair: Kevin Volpp; Discussant: Robert Town Mon June 5, 2006 17:15-18:45 Room 326

Permalink

Medicare and Disparities in Mental Health Treatment and Outcomes

Presenter:

Sandra Decker

Authors:

Sandra L. Decker, Judith Shinogle, Donald Cherry

Chair: Sherry Glied; Discussant: Carole Gresenz Mon June 5, 2006 17:15-18:45 Room 332

Mental health disorders may affect up to half of all Americans, and can result in substantial lost productivity and premature death. Although most mental disorders are treatable using medication and other therapies, most sufferers do not get help. One reason may be due to incomplete health insurance coverage. In particular, approximately 13 percent of Americans lack any kind of health insurance coverage before the age of 65, a fraction that is even higher for minorities and individuals living in areas with low socioeconomic status. Turning age 65 and becoming Medicare eligible results in a significant improvement in mental health insurance coverage for Americans who are uninsured before the age of 65. (For most individuals age 65 and over, Medicare Part A covers psychiatric hospitalizations, while Medicare Part B covers 80% of outpatient visits for psychiatric medication management and 50% of approved charges for psychotherapy.) The effect of this improvement on the use of mental health services and on mental health will be tested using data from the National Ambulatory Medicare Care Survey (NAMCS), 1989, 1990, 1995-1996, and 1999-2003. The analysis estimates differences in mental health treatment by race and socioeconomic status of resident zip code both before and after age 65. Dependent variables will include the probability of a mental health visit among psychiatrists and primary care practitioners, as well as differences in the prescribing of psychotropic drugs including antidepressants, antipsychotics, and central nervous system (CNS) stimulants. Disparities in mental health treatment before and after age 65 also are examined using data on emergency room visits from the National Hospital Ambulatory Medical Care Survey (NHAMCS), psychiatric hospitalizations from the National Hospital Discharge survey, and suicide rates from vital statistics.

Permalink

Return to Work or Receipt of SSDI: Competing Risks Model for those with Mental Health Disability

Presenter:

Judith Shinogle

Authors:

Judith Shinogle, David S. Salkever

Chair: Sherry Glied; Discussant: Pinka Chatterji Mon June 5, 2006 17:15-18:45 Room 332

Previous research examining employees with long term disability found that health benefita may affect the probability of returning to work as well as duration of claims. This previous research failed to examine the disabled worker applying and receiving Social Security Disability Insurance (SSDI). We fill this void by examining how the relationship between health benefits, other fringe benefits, disability management practices and employer characteristics affect the employee on long term disability insurance’s return to work compared to receiving SSDI. Disability claims data, employer benefit data and surveyed information on disability management from 116 employers insured for long term disability from one large insurer. Data was analyzed using bivariate probit and competing risk models. For the competing risk models, unordered failure events of different types were assumed. The two types of failures in the models were 1) Return to work or 2) Receive Social Security. These models allowed the baseline hazard to vary by failure type but constrain the coefficients to be the same across each failure type. These models assumed that each failure event can only occur once per subject. Next, we allowed the coefficients to vary by each failure type and incorporated unobserved heterogeneity of the individual through the use of frailty models. Preliminary results found that disability insurance characteristics such as the ratio of benefits to wages increase duration on private disability insurance as well as increase time to receipt of SSDI, while certain management principles lengthened claim duration. Availability of short term disability (formal or informal) increases duration on private disability insurance while supervisor involved in short term disability and ability to change jobs decreases duration. Health plan characteristics, such as having a mental health carve out or fee for service, decrease the likelihood of returning to work as well as receipt of SSDI, while having a high deductible for mental health insurance increased duration on private disability insurance. Employer disability management practices that decrease the barriers to work lowered the time the employee was on long term disability insurance. On the other hand restrictive health benefits, such as high deductibles could increase this duration and certain health plans such as mental health carve outs may decrease duration on private disability insurance and shorten time to receipt of SSDI. Policy makers and benefits providers should be cognizant of the interaction between health and disability benefits.

Permalink

How Do Youth with Mental Health Disorders Fare in the Juvenile Justice System?

Presenter:

Alison Evans Cuellar

Authors:

Alison Evans Cuellar, Pinka Chatterji

Chair: Sherry Glied; Discussant: Sherry Glied Mon June 5, 2006 17:15-18:45 Room 332

Several studies have documented that youth with emotional and substance use disorders are overrepresented in the juvenile justice system. Whether youth with mental health problems receive harsher treatment in the juvenile justice system compared to similar youth without mental health problems has been the subject of public policy concern, but there have been few systematic, empirical tests of this claim. Using data from a nationally representative sample of adolescents, this study assesses whether youth who use substances or have emotional disorders are sanctioned more heavily than other youth by the juvenile justice system, controlling for the youths’ crime and criminal history. This study uses data from the National Longitudinal Study of Adolescent Health (AddHealth) and offers several distinct advantages over existing research: First, it relies on data from a large, national sample of youth increasing the generalizability of the findings. Second, youth in the sample are first interviewed in the community rather than in institutions increasing the interpretability of the results and allowing us to address the question of bias in the juvenile justice system. Finally, AddHealth data allow for the inclusion of important control variables, such as delinquency history, emotional and substance abuse disorders, as well as demographic, education, and community variables. We apply conceptual and empirical methods that have been used to analyze disparities in health care. The performance and suitability of propensity score and rank and replace methods are assessed in this context. We also demonstrate the importance of the correlation between poor mental health and educational outcomes. Preliminary results of this study find evidence of bias in the justice system, particularly against youth with substance abuse and ADHD disorders.

Permalink

The Impact of DCA on the Use and Effectiveness of Statin Drugs

Presenter:

David Bradford

Authors:

W. David Bradford, Andrew N. Kliet, Paul J. Nietert, Steven Ornstein

Chair: Ernst R. Berndt; Discussant: Mon June 5, 2006 17:15-18:45 Room 335

In August of 1997 the Food and Drug Administration (FDA) relaxed the rules governing television direct to consumer advertising (DCA) of prescription pharmaceutical products. Little is currently known about the effects of DCA for the efficient allocation of prescription drugs. In this study, we will examine how DCA affects physician prescribing patterns, courses of care, and health outcomes for patients suffering from high cholesterol (hyperlipidemia). Currently, Congress and the Food and Drug Administration are investigating whether the 1997 regulatory relaxation should be reversed. In the face of this pressure, the pharmaceutical industry has voluntarily agreed to abstain from advertising a new drug within the first six months of its release. In addition, the recent expansion of Medicare to include a prescription drug benefit has enhanced the potential of DCA to impact Federal budgets. Thus, this research, by investigating whether DCA does or does not have a pernicious effect on the efficiency of health care delivery in primary care settings, will be of high policy relevance. The primary data for this study are taken from a geographically diverse national research network of primary care practices. As of 2004, 81 practices in 32 states, with 348 physicians, are or have been network members. Currently, the entire research network database has clinical chart and demographic information on 604,111 patients, including 3.8 million prescription records. We extract a sub-set of 51,853 patients who have a diagnosis of hyperlipidemia (for years 1998 – 2004). Advertising data is collected at the level metropolitan areas, for the top 75 media markets. Brand specific advertising data is collected at the monthly-level for: local television, network television, national magazine and newspaper and local magazine and newspaper.
The first part of the paper seeks to determine the degree to which DCA affects physician patient populations. We estimate the effect of aggregate and brand-specific advertisement on the number of hyperlipidemia-related visits at the physician practice level. Estimating the effect of DCA on patient populations is a necessary first step in controlling sample selection in the patient level analysis. We then estimate the impact of DCA on a disaggregated (patient level) population. DCA may have contradictory effects on patients: it may improve efficiency of care by encouraging patients to accept and adhere to effective treatments; on the other hand, DCA may reduce the efficiency of care by attracting patients who have only marginally severe conditions, and whose health state does not therefore have much room for improvement. To assess this, we estimate the effect of DCA on the joint the likelihood that the patient is prescribed one of the study drugs (Zocor, Pracachol or Lipitor) and the likelihood that their cholesterol levels are brought to within guidelines.

Permalink

Do Smokers Respond to Smoking Cessation Product Advertising?

Presenter:

Alan Mathios

Authors:

Rosemary Avery, Donald Kenkel, Dean R. Lillard, Alan Mathios

Chair: Ernst R. Berndt; Discussant: Mon June 5, 2006 17:15-18:45 Room 335

The welfare implications of DTC advertising of pharmaceutical products has been a hotly debated policy issue. In some markets however, the incentives of the pharmaceutical companies are well aligned with public health goals. The market for smoking cessation products is one such example. Because of their proven effectiveness, these products could be the key to meeting public health goals to reduce smoking. In other work, we demonstrate that when these products are regulated as prescription drugs, companies advertise them less. In this paper we examine whether consumers are more or less likely to try to quit smoking when they are exposed to more smoking cessation product advertising. Unlike previous research that uses data on whether individuals recall having seen advertisements, we link data on magazines people read with counts of advertisements that appeared in those magazines. Because this more objective measure relies only on accurate reports of magazine reading habits and not on accurate recall of advertising exposure, we avoid many of the measurement and identification issues associated with most research on the effects of advertising.

Permalink

Direct-to-Consumer Advertising, Media Publicity and Utilization of Prescription Drugs

Presenter:

Marta Wosinska

Authors:

David Bradford, Marta Wosinska

Chair: Ernst R. Berndt; Discussant: Mon June 5, 2006 17:15-18:45 Room 335

Ever since pharmaceutical companies began to communicate directly with consumers, direct-to-consumer advertising (DTCA) has been under great scrutiny from policymakers, consumer groups and health plans. A small but growing literature has found that DTCA tends to increase category demand for pharmaceuticals. At the same time, another important source of information for consumers and physicians about prescription drugs - news and popular media stories (media coverage) - has not been explored.
However, studying the effects of media coverage on the utilization of advertised prescription drugs is important for at least four reasons. First, if media coverage is correlated with DTCA expenditures, then current estimates of DTCA effects may overstate the ability of DTCA to affect demand. Second, since media coverage is not subject to regulatory oversight (and may be perceived by consumers as more objective than DTCA), the relative effectiveness of DTCA and media coverage is also important. Third, from a marketing perspective it is also possible to assign a monetary value to the firm from media coverage by assessing what advertising spending would be necessary to get the same boost in utilization that is observed from, say, one positive front page story in a national magazine on a drug. Thus the policy debate surrounding DTCA will be incomplete without an understanding of what the interactions are between media coverage and DTCA. To address these issues, we combine visit, prescription, health outcome, and demographic data on nearly 70,0000 patients (from more than 65 different primary care physician practices) in over 60 markets from 1998 through 2002 with DTCA expenditures and media coverage in those markets. We consider two product categories: Cox-2 inhibitors (for approximately 18,000 patients with osteoarthritis) and statins (for approximately 55,000 patients with hypercholesterolemia). We model changes to the flow of patients with the relevant disease categories (osteoarthritis and hypercholesterolemia) into the practices, and changes in physician prescribing patterns, as a function of class- and brand-specific DTCA, relevant media coverage, and other characteristics of the patients and clinicians.

Permalink

Health Insurance Demand Responses from New Price Structures Offered by Consumer Directed Health Plans

Presenter:

Stephen Parente

Authors:

Stephen Parente, Roger Feldman, Jean Abraham, Jon Christianson

Chair: Willard Manning; Discussant: TBA Mon June 5, 2006 15:30-17:00 Room 121

Rationale: Consumer directed health plans (CDHPs) are attracting attention from consumers, employers, and policy-makers. CDHPs are high-deductible health insurance plans coupled with a tax-advantaged account that can be used to pay for eligible medical expenses. CDHP enrollment is currently estimated at approximately three to five million covered lives.

Objectives: The purpose of this paper is to examine the effect of ‘price’ on health plan choice. More specifically, we will determine whether the health care spending account in the new CDHP designs provides an incentive for consumers to choose these plans over conventional health insurance plans; and we will estimate the extent to which the ‘donut hole’ (i.e. the gap between the account and point where insurance coverage begins) is a disincentive for health plan choice. These two ‘prices’ are largely absent from the design of traditional health insurance plans, which feature relatively low deductibles and do not have health spending accounts. The research questions of this examination are:

• How does the introduction of Consumer Directed Health Plans (CDHPs) into mainstream health insurance affect plan choice?

• Specifically, what is the impact of the account, donut hole and deductible on the own-price elasticity of CDHPs?

Methodology: This paper builds upon the plan choice estimation analysis from an earlier CDHP investigation. In that analysis, a logistic regression conditional upon the number of plan choices offered to a given employee was used. Our primary source of data for this analysis is the employee health plan choices from four large employers participating in a Robert Wood Johnson Foundation (RWJF) funded study. The data from employers represented approximately 250,000 covered lives. Each of these employers offers a CDHP along with traditional managed care plans.

Results & Implications: Early results show a positive impact of the health care spending account from which medical expenses are debited and negative responses to the new CDHP deductible developed as the difference between a traditional high deductible health plan and the consumer’s spending account. With respect to their respective elasticities, the CDHP deductible, known as the ‘donut hole,’ has a far more elastic price response than the health spending account.

We examined the price elasticities in our health plan choice model, including employee premium, health account, donut hole and coinsurance. The largest of the elasticities is for the tax-adjusted employee premium. The least elastic response is for the employee health account. An interesting finding is the greater elasticity of coinsurance compared with the donut hole, or the difference between the deductible and the health account. This result suggests that consumers have greater sensitivity to variations in coinsurance than to changes in the donut hole structure. This finding may challenge some detractors of CDHP plans who suggest consumers will not embrace a plan design with obvious increased cost sharing in the form of a large potential deductible, compared with coinsurance, a much more conventional method of cost-sharing.

Next Steps include adding chronic illness into our plan choice and cross price elasticity estimate using a nested logit model of plan choice.

Permalink

Pent-Up Demand and the Discovery of New Health Conditions after Medicare Enrollment

Presenter:

Jody Schimmel

Authors:

Jody Schimmel

Chair: Willard Manning; Discussant: TBA Mon June 5, 2006 15:30-17:00 Room 121

Recent evidence using hospital discharge data has shown increases in many forms of medical care utilization after age 65. Other work using longitudinal data has shown that those who are uninsured prior to Medicare enrollment experience larger increases in the utilization of certain types of preventive and curative care once on Medicare than those who were continuously insured. Increases in utilization upon obtaining insurance coverage may occur because the previously uninsured have known health conditions that have not received proper treatment or because the increased use of medical care leads to the new diagnosis of conditions. The “discovery effect” refers to the latter, the knowledge one gains about individual health status upon obtaining access to medical care. This effect predicts that increases in utilization after Medicare will lead to more new diagnoses among those who lacked insurance coverage before age 65 than those who had coverage.

Longitudinal data from the 1992-2004 waves of the Health and Retirement Study (HRS) are used to follow a panel of individuals as they progress from ages 56-61 to Medicare enrollment at age 65 and then for several years after age 65. At each interview, HRS respondents are asked whether a doctor has ever told them that they have: a heart condition, a lung condition, high blood pressure, diabetes, cancer, stroke, or arthritis. Because the discovery effect occurs as a result of increased medical care utilization, it is expected that the incidence of new diagnoses will be higher in the years immediately following Medicare enrollment. In order to test the discovery effect hypothesis, Cox models with time-varying covariates are used to assess the differential rate of diagnosis after age 65 by insurance status prior to Medicare.

Results in this paper indicate that the increased utilization experienced after age 65 by those who were uninsured prior to Medicare leads to an elevated hazard of diagnosis (relative to the insured) for virtually every chronic condition considered, for both men and women. The magnitudes of these effects are clinically meaningful; differential increases after Medicare are between 20% and 400% larger among the previously uninsured compared to the insured. These effects are particularly strong for men with heart conditions and for women with lung conditions or cancer. Estimates from Cox models also show that the previously uninsured are less likely to be diagnosed compared to those who were insured in the years prior to Medicare.

The additional diagnoses of chronic conditions among the previously uninsured after age 65 confirms that the uninsured are sicker than the insured, and that lack of access to medical care while uninsured results in untimely diagnoses. Though the uninsured are sicker, they have lower levels of chronic conditions prior to Medicare enrollment due to this lack of diagnosis. Thus, judging the relative health of the uninsured by considering only the presence of chronic conditions in a cross-section is a poor way to assess relative health status and will make the uninsured look healthier than they actually are.

Permalink

The Impact of an Increase in KCHIP Premiums on Insurance Coverage of Kentucky Children

Presenter:

James Marton

Authors:

James Marton

Chair: Willard Manning; Discussant: TBA Mon June 5, 2006 15:30-17:00 Room 121

The state of Kentucky has been very successful in implementing a health insurance program for low income, uninsured children under the guidelines established by the State Children’s Health Insurance Program (SCHIP) which was created by the Balanced Budget Act of 1997 under Title XXI of the Social Security Act.

Since its inception, the Kentucky Children’s Health Insurance Program (KCHIP) has not charged any premium for this insurance coverage. For many reasons it has become more difficult for Kentucky to finance the KCHIP program in its current form. The recent recession has caused state tax revenue to fall nationwide. At the same time, there has been acceleration in the growth of KCHIP spending. Nationwide growth in CHIP spending can be attributed to causes such as enrollment increases, increases in the cost of prescription drugs, and provider payment increases. An additional constraint on the state’s ability to maintain its KCHIP program is that Congress reauthorized federal funding at a lower level starting in July 2003. Finally, as other states build their CHIP programs, the federal money available for redistribution will decrease. For these reasons, the state of Kentucky began in December 2003 charging a $20 monthly premium for families with children covered by KCHIP with income levels between 151% and 200% of the federal poverty level (FPL). This service category is known as KCHIP III.

The purpose of this paper is to examine the impact of the introduction of this premium on enrollment durations in KCHIP III. Do children in KCHIP III exit the program more rapidly after the premium was introduced? A Cox proportional hazards model of KCHIP III duration will be estimated with the policy change modeled as a time varying covariate which is equal to 0 prior to the December 2003 and equal 1 afterwards. A similar model can be estimated for KCHIP II (coverage for children with family income between 100% and 150% FPL) enrollment. Because children in KCHIP II do not have to pay any premiums, they are a naturally provided control group.

In order to isolate the effect of non-payment of the premium as a potential exit route from KCHIP III, movements to and between other state level insurance programs, such as KCHIP II and Medicaid, are controlled for. Exits for other reasons, such as aging out of the program or moving out of the state are also controlled for.

Preliminary results suggest that children were 4 times as likely as in a typical month to exit KCHIP III if they remain enrolled until the premium is introduced. The average exit probability in a typical month in the sample is 9%. The preliminary results also show large effects of yearly recertification on exiting. Non-white children have a greater chance of exiting when compared to white children and children covered under managed care KCHIP III have a greater chance of exiting when compared to those covered under fee for service KCHIP III.

Permalink

Selection and the Effect of Prenatal Smoking

Presenter:

Angela Fertig

Authors:

Angela Fertig

Chair: Phil Cook; Discussant: TBA Mon June 5, 2006 15:30-17:00 Room 213

Author. Angela Fertig (afertig@uga.edu). College of Public Health and Carl Vinson Institute of Government, University of Georgia.

Title: Selection and the Effect of Prenatal Smoking

Rationale: A large number of studies find that smoking during pregnancy is associated with poor infant health outcomes. However, there is an on-going debate about the extent to which this association is causal. It is difficult to determine whether these poor health outcomes are the result of prenatal smoking or are also attributable to characteristics of the mother which are correlated with prenatal smoking.

Objectives: The objective of this paper is to examine the importance of selection on the effect of prenatal smoking by comparing the effect sizes across groups whose selection varies.

Methodology: I use three British birth cohorts where the mothers’ knowledge about the harms of prenatal smoking varied substantially. Specifically, little was known about the harms of smoking prior to the 1964 U.S. Surgeon General’s Report, nor the particular detrimental relationship between prenatal smoking and birth outcomes until the 1969 Surgeon General’s Report. Thus, mothers who smoked during pregnancies that occurred prior to these reports were a less select group than mothers who smoked during pregnancies following them. I compare the effect sizes across three British birth cohorts - those children born in 1958, 1970, and 2000 - and expect that the effect is larger for the latter two cohorts than for the earliest cohort if selection plays an important part in the measured effect.

Results: I find that the effect of prenatal smoking on low birth weight for gestational age among children born in 2000 is twice that of children born in 1958, implying that selection plays an important part in the association between smoking and birth outcomes.

Conclusions: If adverse selection is responsible for the increase in effect size, the implication is that policy interventions to discourage smoking among pregnant women may not prevent as many unhealthy births as usually reported. At the same time, the evidence presented here is further confirmation that prenatal smoking has a causal effect on birth weight conditional on gestation since I find that there is a significant effect in the absence of adverse selection.

Permalink

The Effect of Taxes and Bans on Passive Smoking

Presenter:

Francesca Cornaglia

Authors:

Francesca Cornaglia, Jerome Adda

Chair: Phil Cook; Discussant: TBA Mon June 5, 2006 15:30-17:00 Room 213

This paper evaluates the effect of excise taxes and bans on smoking in public places on the exposure to tobacco smoke of non-smokers. We use a novel way of quantifying passive smoking. We use data on cotinine concentration- a metabolite of nicotine- measured in a large population of non-smokers over time. Exploiting state and time variation across US states, we reach two important conclusions. First, excise taxes have a significant effect on passive smoking. Second, smoking bans have on average no effects on non smokers. The reason is that smokers replace smoking at banned locations with smoking in places, where non smokers are in turn more likely exposed. In particular, while bans in public transportation or in schools decrease the exposure of non smokers, bans in recreational public places can in fact perversely increase their exposure by displacing smokers to private places where they contaminate non smokers. Bans affect different social groups differently: we find that smoking bans increase the exposure of poorer individuals, while it decreases the exposure of richer individuals, leading to widening health disparities.

Permalink

The Effects of Workplace Smoking Bans on Exposure to Smoke

Presenter:

Christopher Carpenter

Authors:

Christopher Carpenter, Marianne Bitler, Madeline Zavodny, Agnes Scott

Chair: Phil Cook; Discussant: TBA Mon June 5, 2006 15:30-17:00 Room 213

Rationale: In recent years, a number of states and localities have implemented bans on smoking in private workplaces. While these bans may have the effect of reducing own smoking, much of the motivation for adopting these bans was to reduce exposure to environmental tobacco smoke (ETS). Little research has comprehensively examined this issue while also paying attention to own smoking and the effect of other policies such as cigarette taxes.

Objectives: The objective of this paper is to examine the effect of state and local smoking bans in the workplace on exposure to environmental tobacco smoke, taking account of own smoking and other policies.

Methodology: This paper will use difference-in-differences methods to examine own smoking and exposure to environmental tobacco smoke before and after imposition of state and local bans on smoking in the workplace. We will study bans in California (and possibly other states) that staggered implementation in different types of workplaces, such as restaurants and stand-alone bars. We will use data from the California Tobacco Use Surveys, the Tobacco Use supplements to the Current Population Survey, and the California Department of Health Bar Establishment Surveys. We will combine these data with contextual information on smoking bans, cigarette taxes, and other public policies toward tobacco. Our empirical models will include controls for demographic characteristics and economic conditions as well as location and time.

Results: Preliminary results indicate that statewide bans on smoking in private worksites are associated with significant increases in the likelihood that affected workers report a complete ban on smoking at their worksite. For smoking bans that cover bars and restaurants, these increases appear particularly strong for specific groups of affected workers, such as bartenders and waiters/waitresses. When completed, this project will identify the degree to which these increases in reported ban presence have reduced own smoking and exposure to ETS. Preliminary results also suggest a sizable reduction in own smoking among bartenders and waitstaff after the California bans were implemented.

Conclusions: State and local bans on smoking in private worksites are becoming increasingly popular. While reducing exposure to environmental tobacco smoke is a main motivation for adoption of these bans, little research has carefully examined the effects. Our preliminary results from California suggest that these bans have been effective at increasing the likelihood that affected workers report comprehensive smoking restrictions at their place of work. The final goal of this research-understanding the degree to which these bans have ultimately been effective at reducing ETS exposure and own smoking-is important for a comprehensive evaluation of these policies.

Permalink

The Effects of Eliminating Behavioral Health Benefits in a Medicaid Population

Presenter:

Kenneth McConnell

Authors:

Kenneth McConnell, Neal Wallace

Chair: Jose Escarce; Discussant: Ana Balsa Mon June 5, 2006 15:30-17:00 Room 225

Rationale: In response to state budget shortfalls in 2003, the Oregon Health Plan made several changes to its benefit package offered to the enrollees who were not part of the “Categorically Eligible” Medicaid population. One important change was the elimination of coverage for outpatient substance abuse treatment, including methadone treatment for heroin addiction. As other states consider cuts in their coverage of substance abuse treatment, the Oregon experience can provide evidence about the expected impact on costs and utilization.

Objectives: The objective of this paper is to compare the effect of an elimination of the substance abuse benefit on methadone users’ utilization of medical services and expenditures for those services.

Methodology: We use data on claims from members of the Oregon Health Plan who were subject to the cuts in their benefit package. We selected individuals who were enrolled in the plan for at least six months before and after the 2003 benefit changes. 381 methadone users and 15,713 non-users were identified through the claims data. Following Hirano, Imbens, and Ridder, we used matching with an estimated propensity score to evaluate the difference-in-differences in outcomes between methadone users and non-users. We estimated the change in office visits to physicians, emergency department visits, and hospital admissions, and expenditures for these services.

Results: After the elimination of substance abuse treatment benefits, visits and expenditures for physician services and emergency services showed significant increases after the substance abuse benefit was eliminated. Inpatient admissions increased, but at a rate that was less than the control group.

Conclusions: Our results suggest that the elimination of outpatient substance abuse benefits increased expenses associated with emergency departments and office visits. Elimination of substance abuse programs may not be a effective method of reducing expenditures for Medicaid programs.

Permalink

Cost-effectiveness of paying drug users to abstain

Presenter:

Todd Olmstead

Authors:

Jody Sindelar, Todd Olmstead, Nancy Petry

Chair: Jose Escarce; Discussant: Anthony Lo Sasso Mon June 5, 2006 15:30-17:00 Room 225

We analyze the cost-effectiveness of a novel approach to treating drug abuse that pays drug users not to use drugs. In this approach, drug users in treatment programs are tested and paid small incentives when they abstain from using specific drugs. The standard approach of such a Contingency Management (CM) approach is to pay fixed increments on an escalating basis. The CM under study is a lower cost prize-based version of in which instead of earning fixed awards, the successful patient draws chips from an urn. The chips indicate whether or not a prize has been won and the exact prize. The prize value and probability of winning each type of prize can be varied so that the expected value of payments can be varied. With increased abstinence, the patient gets to take more draws and the number of draws increases at an escalating rate to encourage longer term abstinence.

The effectiveness data were obtained from the Motivational Incentives-Drug Free Clinics study conducted in the context of the National Drug Abuse Treatment Clinical Trials Network. Patient outcome measures include total number of stimulant-negative urine samples provided and longest duration of confirmed stimulant abstinence. The cost data were obtained via surveys sent to the eight participating psychosocial drug abuse treatment clinics. Patients (N = 412) were randomly assigned within each clinic to usual care (UC) or usual care plus contingency management (CM). Clinic costs included: counseling, patient drug testing, and administration of the incentives component (value of prizes, administrative time to draw prizes, and administrative time to inventory, shop, and stock prizes).

Results are presented for the incremental cost-effectiveness ratios (ICERs) of CM relative to UC for all sites combined, as well as for each site separately. A special emphasis of this cost-effectiveness analysis was on the multi-site nature of the trial and the interpretation of the site differences. The study determined that the ICER for an additional stimulant-negative urine sample was $131 per additional negative urine sample for all sites combined, and ranged by site from $74 to $252 across the sites. The ICERs for lengthening the longest duration of confirmed stimulant abstinence by one week were also varied across the sites. In addition, scatterplots of the bootstrapped replicates and the corresponding acceptability curves were presented to illustrate the uncertainty inherent in the incremental cost-effectiveness ratios (ICERs) and to provide policy relevant interpretation of the ICERs. The overall and site-by-site scatterplots of the ICERs and the acceptability curves are provided. We discuss external data that can be brought to bear regarding the policy interpretation of the ICERs, e.g. what is the value gained in terms of crimes and spread of disease averted by lengthening longest duration of abstinence and a greater number of negative drug urine tests? We also discuss a variety of policy, implementation and equity issues regarding the wider spread adoption of this novel treatment.

Permalink

The Impact of Adolescent Perceptions of the Future on Substance Abuse

Presenter:

Elizabeth Vigdor

Authors:

Elizabeth Vigdor

Chair: Jose Escarce; Discussant: David Bishai Mon June 5, 2006 15:30-17:00 Room 225

Background: Understanding how perceptions of uncertainty and risk affect behavior is important for the economic analysis of substance abuse and for designing and targeting prevention efforts. This is particularly relevant for adolescents, who may be more likely to deviate from the classical rational choice model of economics by misperceiving risk and future uncertainty.

Objective: The primary hypothesis is that some adolescents have mistaken perceptions about future uncertainty that affect the likelihood that they use illicit drugs. If an adolescent is overly pessimistic, that is, her belief that an event will transpire is much greater than the actuarial likelihood of the event occurring, her expected future utility will be lower . Therefore we would expect her to be more likely to engage in activities with potential adverse consequences for the future, such as illicit drug use.

Methods: We used the 1997 National Longitudinal Survey of Youth (NLSY-97) to examine marijuana utilization as a function of perceptions of future risk, controlling for demographic characteristics and other social and environmental factors. The sample was restricted to adolescents who did not use marijuana at Wave 1. Outcome variables were initiation of marijuana use by Wave 4 and frequency of use at Wave 4. Perceptions of future risk were measured using self-reported chances of dying within the next year and before age 20. The actuarial probability of dying within this time frame was calculated from Social Security life tables.

Results: Approximately 30 percent of wave 1 respondents reported a probability of dying within 1 year of 0, which is the correct response in expectation. Adolescents who overestimated their risk of dying in 1 year were significantly more likely to initiate marijuana use by wave 4 than similar adolescents who assessed the risk of death more accurately (26.8% vs. 21.8% for death within 1 year), and significantly more likely to smoke marijuana on a regular basis at Wave 4 (8.8% vs. 7.9%).

Adolescents with a self reported risk of dying within one year of 1-10 percent, 11-20 percent, and 21-30 percent are significantly more likely to initiate marijuana use by Wave 4 than those who report a risk of 0, but there is no significant relationship between higher self-reported probabilities of dying and marijuana initiation. For regular marijuana use, only the teenagers who report a 1-year mortality risk of 1-10 percent have significantly higher rates of use than similar teens who report no risk of dying in the next year. Results are similar but less significant when the probability of dying by age 20 is used instead.

Conclusions: Adolescents’ perceptions of mortality risk are correlated with future use of illicit drugs, which has implications for targeting prevention efforts at teenagers most at risk for substance abuse. Further research is needed to explore whether the lack of correlation between very high reported probabilities of death and substance abuse is a result of the high responses being protest or careless responses rather than miscalculations of risk.

Permalink

Designing Effective Expansions of Insurance Coverage: Learning from New York's Health Reform Efforts

Presenter:

Sharon Long

Authors:

Sharon Long, John Graves, Stephen Zuckerman

Chair: Ted Frech; Discussant: Chris Garmon Mon June 5, 2006 15:30-17:00 Room 226

Lack of health insurance is a persistent problem in the U.S. health care system. Although there is increasing political consensus as to the need for new policies to increase insurance coverage, there is little consensus as to the form such actions should take: either further expansions of public programs or incentives to increase coverage in the private sector. Federal and State efforts to design effective strategies for expanding insurance coverage should build on the current efforts underway in several innovative states. One such state is New York: New York’s Health Care Reform Act of 2000 included a number of coverage expansions, including Family Health Plus for parents of Medicaid-eligible children and subsidies to make health insurance more affordable and available to small employers and their employees, sole proprietors, and working individuals whose employers do not provide health coverage. This paper assesses the effects of New York’s expansion on eligible parents with respect to public coverage, uninsurance, the displacement of private coverage, and health care access and use. A key challenge to evaluating state health reform initiatives is often a lack of state-specific data. As part of this study, we explore the feasibility of using the data from the National Health Interview Survey (NHIS) to evaluate New York’s expansion efforts. The NHIS samples from all states and the District of Columbia and, although not designed to provide state-level estimates for all states, provides state-level estimates for the largest states, including New York. For this study, we use data from the 1998-2004 rounds of the NHIS. As a check on our analysis, we compare the estimates of the effects of the New York’s expansion effort on insurance coverage from the NHIS to those obtained using comparable years of the Current Population Survey.

Permalink

How Much Choice? Nonlinear Relationships between the Number of Health Plan Options and the Choices of Medicare Beneficiaries

Presenter:

Brian Elbel

Authors:

Brian Elbel, Mark Schlesinger

Chair: Ted Frech; Discussant: Roger Feldman Mon June 5, 2006 15:30-17:00 Room 226

Background: Expanding the choices available to Medicare beneficiaries is a central aspect of recent Medicare reform. Economic theory predicts that as the number of health plan options available increases, beneficiaries will be more likely to find a plan that matches their preferences. Improved matching should lead to a) an increase in the take-up of Medicare Advantage (formerly, Medicare+Choice) plans as an alternative to fee-for-service Medicare, and b) an increase in the probability that individuals in poorly performing plans will exit when dissatisfied.
However, research on bounded rationality indicates that these assumptions might be wrong. This literature demonstrates that large numbers of options can be associated with choice deferral—foregoing options that would have been chosen if presented in a smaller choice set. However, most of this work (with one exception) has examined only low-cost, everyday goods. Whether the same is true for health plans is not clear. Objective: To determine how the health plan enrollment and disenrollment decisions of Medicare beneficiaries are influenced by the size of the beneficiary’s choice set. Data: We examine data from a separate survey of Medicare beneficiaries fielded in 2000.
Methods: We examine both the probability of enrollment into a Medicare health plan (when beneficiaries first become eligible for Medicare) and the probability of switching health plans. We also examine two additional outcomes to confirm our interpretation of the number of plans variable: Did those who did not enroll consider doing so, and How long did the beneficiary take considering their options. The key independent variable is the number of Medicare Advantage plans available in their county of residence. We control for other individual and market characteristics known to be related to enrollment and disenrollment and allow for a non-linear relationship between choice set size and our dependent variables. Results: We find that increasing the size of the choice set from 1-2 to 3-4 options does seem to both increase the probability of enrollment in a Medicare Health Plan, as well as switching plans. However, after 3-4 plan options, there is no subsequent increase in either enrollment or disenrollment, and some evidence of a decrease in both. These results seem robust across various model specifications and sub-samples. Conclusions: While some choice is beneficial in inducing health plan enrollment and switching, the relationship between number of options and enrollment and switching is non-linear. As health policy moves towards increasing the number of choices that consumes are asked to make, the behavioral economics and bounded rationality literature could have important contributions.

Permalink

Valuing Variety: How Much Do Workers Value Having Choices Among Health Insurance Plans?

Presenter:

Jean Abraham

Authors:

Jean Abraham

Chair: Ted Frech; Discussant: Ted Frech Mon June 5, 2006 15:30-17:00 Room 226

Authors: Jean Abraham, University of Minnesota (abrah042@umn.edu) and Anne Beeson Royalty, Indiana University Purdue University Indianapolis (royalty@iupui.edu)

Title: Valuing Variety: How Much Do Workers Value Having Choices Among Health Insurance Plans?

Rationale: In response to rising health care costs, insurers are introducing new products and employers are making changes to benefit designs, both in terms of the number and types of plans they offer to workers. Employers face tradeoffs when offering plan choice including fixed administrative costs per plan, higher premiums given lower per plan enrollment, adverse selection, and spiraling premiums. In contrast, there may be benefits associated with plan choice, particularly if the workforce exhibits heterogeneous preferences.

Objective: The primary objective of this study is to develop a model derived from the welfare economics literature to empirically estimate the monetary value that workers place on having variety with respect to the number and types of health plans offered to them through their place of employment.

Methodology: Obtaining estimates of the gains or losses associated with changes in the number or types of plans offered to workers is a two-step process. The first step is to empirically estimate multinomial logit and nested multinomial logit models of a worker’s choice of health plan, specifying the determinants of demand to include the employee’s premium contribution, the coinsurance rate, deductible, and covered benefits (e.g., prescription drug coverage). Second, we use integration methods to obtain the surplus function for our probabilistic health plan demand functions. Using the surplus function and specific values of the explanatory variables, we then estimate the gain or loss realized when the set of offered health plans changes. Calculations are performed for several simulated changes in plan offerings. We use data on 1,530 workers from the 1999 and 2001 linked Household Component-Insurance Component MEPS database. The Household Component (HC) is a random sample of the civilian, non-institutionalized U.S. population containing data on demographics, employment status, and health insurance. For individuals with an offer of employer coverage, surveyors asked for employer contact information and these employers were surveyed about their health plans offerings. The results comprise the MEPS Insurance Component (IC) database which contains information on premiums, employee contributions, plan type (e.g., HMOs, PPOs, and FFS plans), coinsurance, deductibles, and covered benefits for up to four comprehensive health plans offered to employees of private establishments and all plans offered by public employers.

Results: Preliminary estimates of the compensating variation for changes in the number and types of plans range from $165 to $1992.

Conclusions: As policymakers assess the effects of the managed care revolution and the consumer-driven health care movement, it would clearly be useful to know the value that workers put on having a variety of health plans from which to choose, as well as the extent of the losses incurred as they lose options. Employers would benefit from knowing how valuable a variety of health plan choices really is to workers, particularly since they face tradeoffs when designing their health insurance benefits.

Permalink

Modeling Efficiency at the Process Level

Presenter:

Robert Lee

Authors:

Robert Lee, Roma Taunton, Byron Gajewski, Marjorie Bott

Chair: Dave Vanness; Discussant: Brian Denton Mon June 5, 2006 15:30-17:00 Room 235

The health economics literature includes a significant number of efficiency studies, yet most have not fully engaged Newhouse’s (1994) concerns that standard data sources may omit important inputs and that output measures may ignore important sources of heterogeneity. A third concern is that Data Envelopment Analysis and Stochastic Frontier Analysis often disagree about efficiency rankings (Chirikos and Sear, 2000), suggesting that the results are sensitive to untested modeling assumptions. In looking at the performance of a sample of nursing homes, this study seeks to address all three issues.

This paper examines a single process: how nursing homes complete a standardized, mandatory assessment of residents. The paper is based on a study that also gathered detailed data about variations in the quality of these assessments and detailed data about differences in the characteristics of residents. The product should be homogeneous, and we have data to test this assumption. The paper analyzes primary data collected from a stratified random sample of 107 nursing homes during the last two years. The first step in collecting the underlying data was the construction and validation of a process map for each facility. As a result, the steps in the process for each nursing home and the resources used at each step were identified and verified. It is unlikely that data on any significant inputs into the production process are missing. Moreover, these data allow us to identify at which steps production is inefficient, not just which resources appear to be overused.

These data were augmented with data from Medicaid cost reports, the Minimum Data Set, and the Online Survey Certification and Reporting System. As a result, we have the data needed to conduct a variety of case mix-adjusted and quality-adjusted Data Envelopment and Stochastic Frontier Analyses.

At this point we have just finished data cleaning and have done initial DEA analyses. These first analyses find that some facilities are quite inefficient. Our analyses will proceed on three tracks. One will be to examine the correlations of efficiency scores as we change assumptions about input aggregation and returns to scale. Because we have data on input prices, the next track will repeat this process with DEA models that identify technical and allocative inefficiencies. We will again examine the impact of doing so on the correlations of efficiency scores. Finally, we will estimate alternative stochastic frontier cost functions that replicate the DEA assumptions about input aggregation and returns to scale. Again we will examine the correlations of the efficiency scores with the other models. For each of these three models we will estimate auxiliary regressions that examine whether ownership, chain membership, the structure of the care planning process, size, physical structure, market structure, or manager training affect efficiency scores.

Inefficient production is one of the hallmarks of the American health care system, yet economics has yet to offer profound insights into its sources. This unique data set should help us understand why alternative models paint different pictures of efficiency.

Permalink

Projecting the Health and Economic Impact of a Quadrivalent HPV Vaccine using a Multi-type Transmission Dynamic Model

Presenter:

Elamin Elbasha

Authors:

Elamin Elbasha, Erik Dasbach, Ralph Insinga

Chair: Dave Vanness; Discussant: David Vanness Mon June 5, 2006 15:30-17:00 Room 235

Background: The health and economic burden of HPV disease is significant. A safe, efficacious, prophylactic quadrivalent HPV (6/11/16/18) vaccine that reduces the overall incidence of cervical cancer, cervical intraepithelial neoplasias (CIN), genital warts, and HPV infection may soon be available. Objectives: To assess the cost-effectiveness of alternative HPV vaccination strategies (i.e., routine vaccination of 12-year old females versus 12-year old females and males) in settings with established cervical cancer screening programs in the U.S.

Methods: A non-linear, deterministic, age-structured, mathematical model of the transmission dynamics of HPV and cancer development in a U.S. population stratified by sex and sexual activity was developed and integrated with an economic model that accounts for the effectiveness and economic consequences of vaccination strategies. Inputs for the model were obtained from public data sources, published literature, and clinical trials. We assumed a vaccine uptake of 70%. We varied duration of protection from 10 years to lifetime and the cost of the vaccination series from $300-$500.

Results: Assuming lifetime duration of protection, vaccination combined with screening reduced the prevalence of high risk HPV by 60% and the incidence of genital warts, CIN, and cervical cancer by 76%, 43%, and 44% respectively after 100 years. The incremental cost-effectiveness ratios ranged from cost-saving to ~$124,000 per quality adjusted life year gained. The following parameters exerted the greatest influence on the results: vaccine characteristics of degree and duration of protection, preference weights, and cost of vaccination. Conclusion: Vaccination with a prophylactic quadrivalent HPV vaccine can reduce the incidence of cervical cancer, CIN, and genital warts and provide survival benefits and quality of life improvements at a cost within the range accepted as cost-effective for a reasonably wide range of model input values.

Permalink

Sicker and Poorer? The Role of Health Status for Welfare Leavers

Presenter:

Bianca Frogner

Authors:

Robert Moffitt, Bianca Frogner

Chair: Curtis Florence; Discussant: Curtis Florence Mon June 5, 2006 15:30-17:00 Room 309

Rationale: Temporary Assistance for Needy Families (TANF) emerged from major welfare reform at the federal and state level in the late 1990’s. Welfare reform imposed strict time limits, work requirements, and eligibility criteria, resulting in a large drop in the welfare rolls. Studies to date have shown that a majority of former welfare recipients, or welfare leavers, did relatively well in terms of employment and income, but a subpopulation of extremely poor welfare recipients were adversely affected. In addition, many welfare leavers were dropped from Medicaid even when still eligible for coverage. An issue that has not been well addressed is the distribution of health status of welfare leavers and whether it is bimodal. In other words, welfare recipients with relatively good health status are more likely to leave welfare, but there may be a subpopulation with very poor health that also left.

Objectives: To calculate the distribution of health status among welfare leavers and its change over time, with particular attention to possible bimodality; to calculate the likelihood that a person returns to welfare versus stays off welfare in the long run as related to where they are in the health status distribution; and to calculate health insurance coverage rates for welfare leavers of different health status levels.

Data: This study will use longitudinal household survey data from the Welfare, Children, and Families Study, commonly called the Three-City Study. The data collection process involved three rounds of in-person interviews ending in 1999, 2001 and 2005, allowing researchers to study the evolving effects of welfare reform. A random sample of approximately 2,400 households with children was drawn from racially and ethnically diverse low- and middle-income neighborhoods in three cities: Boston, Chicago, and San Antonio. The study interviewed both welfare and non-welfare recipients from the same neighborhood. Among all welfare reform data sets in the country, the study is unique in sample size and in length: the sample size is larger than the low income, welfare eligible portions of the Panel Study of Income Dynamics (PSID), and is longer than the largest national panel, the Survey of Income and Program Participation.

Methodology: We will analyze survey participants who have completed all three waves. We will use the many health measures in the data set to construct an index of health status with factor analysis. We will then see how an individual’s location in this distribution affects the likelihood of leaving welfare, returning to welfare, and staying off welfare. We will examine income, employment, and health insurance coverage conditional on each of these states. Hazard models with unobserved heterogeneity will be used to examine exit from and re-entry to welfare. Income, employment, and health insurance coverage will be estimated conditional on these transitions to allow for selection bias. One’s location in the health status distribution will enter all hazards and regression equations.

Results: We have not completed our analysis and do not have results at this time. We expect to have results by the time of the conference.

Permalink

Trends in the Health of the Poor and Near Poor: Have the Poor and Near Poor Been Catching Up to the Non Poor in the Last 25 Years?

Presenter:

Elise Gould

Authors:

Elise Gould, Timothy Smeeding, Barbara Wolfe

Chair: Curtis Florence; Discussant: David Becker Mon June 5, 2006 15:30-17:00 Room 309

Objective: We investigate how health status and the relationship between health status and income have changed over time.

Methodology: First, we document patterns of health status among children, prime-age adults, and adults over 65 using the National Health Interview Survey for selected years between 1978 and 2003. We measure income using three categories of poverty: poor for below 100% of poverty, near poor for between 100 and 200% of poverty, and non poor for above 200% of poverty. At the same time, we utilize three measures of health: self-reported fair or poor health, limitations in work, school, or home activities due to health problems, and a health-related quality of life index equal to one if individuals report both limitations and in fair or poor health. We track changes in health status by poverty status for all three age groups over the 25 year period. Next, we perform multivariate regression of the likelihood of having poor health for all three measures. Besides our investigation into the relationship between income and health, our probit model includes control variables for age, sex, race, ethnicity, education, and region.

Results: As expected, we find that poor health status and poverty are closely linked. For every age group and every heath indicator, the health of the poor is worse than that of the near poor or non poor. Adults show some slight improvement in health over time, using both measures. However, the gains appear greatest among the non poor, suggesting an increase in health disparities as income disparities increase. Health status as measured by health limitations has declined for all children from 1978 to 2003, though the greatest rate of increase was among the near poor children. In the probit estimates, we find poverty to have strong correlation with poorer health status and clear statistical evidence of the increasing association between income and health for nearly all age groups and all three measures of health.

Conclusions: The evidence suggests that poverty and near poverty play an increasingly important role in determining health status. The increase in income inequality when combined with the increase in reported disparities in limitations for children may have long run negative consequences for future earnings inequality and more fundamental measures of well-being. In conjunction with programs specifically directed to provide health insurance to the poor and near poor, public policies targeted at reducing poverty could have long term consequences on health.

Implications: We make a case that ill health is a separate and very important aspect of poverty that needs to be continuously monitored and reported in the same way that the number of uninsured are monitored and reported and the number of poor themselves are monitored and reported. Only when health status, SES, and access to care can be easily and systematically compared will we be in a position to say whether the poor and near poor are receiving adequate health care or are approaching the non poor in terms of access and health status.

Permalink

Choosing Health versus Wealth: A Laboratory Survey

Presenter:

Robert Graboyes

Authors:

Robert Graboyes

Chair: Curtis Florence; Discussant: Reed Johnson Mon June 5, 2006 15:30-17:00 Room 309

Rationale: We wish to develop a technique for revealing the monetary value that individuals place on stochastic health. Our methodology seeks a way to circumvent the tradeoff between moral hazard and risk-bearing in health care markets. Aside from this principal question, we wish to develop a prototype for a field experiment and suggest a possible structure for health insurance contracts.

Objectives: We wish to answer the following questions: How much are individuals willing to accept in lieu of medical treatments with a low probability of success? Is that willingness systematically related to the probability? Do they place higher value on loved ones than on themselves or on total strangers? Does fear of regret play a role in such choices?

Methodology: We use a contingent valuation method to estimate individuals’ willingness to accept cash in lieu of a treatment that has a low probability of success. We offer healthy individuals a choice of Policy T (full insurance coverage of an expensive treatment) or Policy C (a large, unfettered lump-sum cash indemnity upon diagnosis). For simplicity, we concern ourselves only with a single hypothetical medical contingency. With Policy C, consumers bear the marginal cost of a high-cost treatment without suffering the risk of catastrophic financial loss. In our survey, the disease and policies were abstract and hypothetical and the sample of subjects relatively small.

Results: Subjects choose the treatment policy more often when medicine is more efficacious. They choose the treatment policy most often for loved ones, less often for themselves, and least of all for anonymous strangers. Strikingly, subjects choose the treatment policy more often when ex post regret over the choice is likeliest, even when no differences exist in treatment efficacy. The responses were strongly consistent internally and comported well with economic logic. Despite the small sample, the results were highly significant. While healthy, the subjects were capable of signaling the marginal value they placed on a small probability of life. They were able to do so consistently over long lists of random scenarios.

Conclusions: The results suggest a low-cost methodology for estimating health-wealth tradeoffs and a potential means of inducing consumers to self-limit use of high-cost, low-benefit treatments. The results present several puzzles related to the valuation that individuals place on their own health versus the health of loved ones or strangers.

Permalink

Health Risky Behaviors and Social Capital: Four Approaches to the Issue of Causality

Presenter:

Sherman Folland

Authors:

Sherman Folland

Chair: Richard Scheffler; Discussant: Chris Seplaki Mon June 5, 2006 15:30-17:00 Room 313

Author: Sherman Folland (folland@oakland.edu), Department of Economics, Oakland University

Title: Health Risky Behaviors and Social Capital: Four Approaches to the Issue of Causality

Rationale: Social capital elements such as spouse, children or community are well known to be statistically associated with better health indicators. The thorny question, however, is whether these elements improve health or whether they are correlated with health for other, perhaps spurious, reasons. We need ideas and research to disentangle this issue.

Objectives: The objective of this paper is to describe and test four approaches to identifying whether the observed positive correlation of health with social capital indicators is causal.

Methodology: The four proposed tests are as follows: 1) instrument the community social capital indicator, following Putnam; 2) test the associations of community level social capital with individual smoking and excessive drinking behaviors; 3) apply the DeLeire/Levy data on risk valuing behaviors of married versus unmarried men and women with or without children to individual longitudinal data including the risky behaviors of smoking and excessive drinking; and 4) produce a natural experiment in which movers to other MSAs with different community social capital levels are compared to stayers regarding their smoking and drinking behavior. These tests are made possible through a unique database that combines the NLSY79 Geocode data with a large marketing database used by Putnam and combining these in turn with area health resource and MSA community characteristics data. The attempt is made throughout to spot potential endogeneities and address them as best possible, likewise to address and discuss the identification problems of social capital research described by Durlauf.

Results: Preliminary results show that the instrumental variables approach to the Putnam community social capital index produces similarly strong and positive associations with health behaviors as does the raw index. The Putnam community level social capital measures are associated with reduced smoking for older (30+ year olds) but not for younger ages. As of this writing, the DeLeire/Levy application and the natural experiment described above are still in process, the combined database for this, however, is completed.

Conclusions: In the spirit of Karl Popper, causality is never proved, we can only apply more and better challenges to the hypothesis. Of the tests I have completed both in this research and prior work, the hypothesis tends to survive reasonable tests. Given the importance that would probably be attached to the “discovery” that social capital is a substantial influence on one’s health, this research should probably conclude that the ‘social capital improves health’ hypothesis has earned the right to further study.

Permalink

Social Capital Externalities and Individual Mortality Risks in Swedish Municipalities

Presenter:

Kamrul Islam

Authors:

Kamrul Islam, Ulf-G. Gerdtham, Bo Gullberg, Martin Lindström, Juan Merlo

Chair: Richard Scheffler; Discussant: Chris Seplaki Mon June 5, 2006 15:30-17:00 Room 313

Rationale: Social interaction in various forms and degrees is common to all economic agents and externalities are a fundamental aspect of a modern interdependent society. As a societal characteristic the effects of social capital have been hypothesized to influence society in a multitude of ways, including health. Evidence has been found that the community social capital is important to individual health. If such a societal externality exists then this may have important implications for health policy.

Objectives: To analyze community social capital externalities on individual’s all-cause and cause-specific mortality risk in Swedish municipalities.

Methodology: Envisaged from conventional economic premises, this paper proposes a mechanism where social capital can be thought as aggregate factor of health production in a community which generates production externalities for the individuals live in the same community and explores the relationship within established paradigm of the Grossman model (Grossman, 1972). The study uses pooled data from Statistic Sweden’s Survey of Living Conditions (the ULF survey) from the annual interviews conducted in 1980-1997 for all the subjects aged 20-84 years who were followed up for 4-21 years. Two municipality social capital indicators- registered election-participation rates and registered crime rates are used to measure area level social capital. We estimate social capital impact of the mortality risk, controlling for the individual’s initial health status and a number of exogenous individual characteristics using extended Cox model (with time-varying covariates). A number of sensitivity analyses are performed with respect to different specifications. We also include additional macro variables such as municipality level income, urbanization, income inequality (estimated by the Gini coefficient) and aggregate level of education as potentially can confound the association between municipality social capital and mortality risk.

Results: Our results suggest that both municipality social capital indicators are significantly influence individual’s risk form all-cause mortality for males but not for females. In an analysis of cause-specific mortality, results indicate that higher rate of election-participation significantly decrease the risk of mortality from cancer and favorable effects are also noticed (but not significant at 5% level) on cardiovascular mortality or deaths due to ‘suicide’ or ‘other external causes’ (motor vehicle accidents, accidents and homicides). Analogous results are also observed for the crime rates indicator. The results are evident and robust to the alternative specifications tried in the sensitivity analysis for individual’s age 65 years or more.

Conclusions: We conclude that municipality social capital has spillover effects on mortality risk over and above individual personal characteristics for elderly population.

Permalink

The Role of Community Social Capital in Reducing the Prevalence of Serious Mental Illness

Presenter:

Richard Scheffler

Authors:

Richard Scheffler, Timothy Brown

Chair: Richard Scheffler; Discussant: Chris Seplaki Mon June 5, 2006 15:30-17:00 Room 313

Authors: Richard M. Scheffler (rscheff@berkeley.edu) and Timothy T. Brown (tbpetris@berkeley.edu, Nicholas C. Petris Center on Health Care Markets and Consumer Welfare, School of Public Health, University of California at Berkeley.

Title: The Role of Community Social Capital in Reducing the Prevalence of Serious Mental Illness.

Rationale. Community social capital has been theorized to positively affect individual mental health through its effects on information flows and social norms regarding social support. However, three recent reviews of this topic have found that the research literature to date suffers from methodological weaknesses.

Objectives: The objective of this paper is to estimate the association of community social capital on the likelihood of an individual suffering from serious mental illness.

Methodology: We estimate a mental health production function focusing on the association of community-level social capital (CSC) and serious mental illness. We use a new validated measure of community-level social capital, the Petris Social Capital Index, which is based on an actual ecological measure of CSC rather than on an aggregation of individual survey responses. We measure CSC at the level of the Metropolitan Statistical Area (MSA) in order to capture the effects of CSC in both the areas of an individual’s residence and employment. Our measure of serious mental illness is a reliable and valid indicator of serious mental illness: the Kessler K6. We link our measure of CSC to individual-level data on 48,222 adults in 48 MSAs for the years 1999-2001 and control for individual-level demographics, individual-level social capital, and MSA-level fixed effects. Lagged CSC is used rather than contemporaneous CSC in order to avoid endogeneity. We test contemporaneous individual-level social capital (ISC) for endogeneity, using instrumental variables and include tests that show the strength and exogeneity of our instruments. All estimates are conducted using linear probability models and sensitivity tests are conducted estimating the same specifications using probit models.

Results: We find that the level of CSC is inversely related to the probability of an individual having serious mental illness. However, CSC is subject to strong diminishing returns and only has an effect for the bottom quartile of individuals who live in those MSAs with the lowest levels of CSC. The effect was also only found when including MSA-level fixed effects, indicating the presence of strong omitted variable bias when MSA-level fixed effects are not included.

Conclusions: Community social capital is potentially an important environmental mediator of serious mental illness.

Disclosure information: This research was supported by the Center for Health Research at the University of California at Berkeley and the Centers for Disease Control and Prevention, Atlanta, GA through Cooperative Agreement Number U48/cCU 909706. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention. The authors are employees of the University of California at Berkeley and have no conflict of interest with regard to this project.

Permalink

Cost-Effectiveness of Delivering Adult Influenza Vaccination in Non-Traditional Settings

Presenter:

Megan O'Brien

Authors:

Lisa Prosser, Megan O’Brien, Noelle-Angelique Molinari, Katherine Hohman, Kristin Nichol, Pascale Wortley, Mark Messonnier, Tracy Lieu

Chair: Noelle-Angelique M. Molinari; Discussant: Bo-Hyun Cho Mon June 5, 2006 15:30-17:00 Room 325

Despite the morbidity and mortality associated with influenza illness, vaccination coverage rates remain well below Healthy People 2010 goals for groups that are recommended to receive influenza vaccination. Non-traditional settings may represent an opportunity for boosting influenza immunization coverage rates for recommended adults. The objective of this project was to evaluate the cost-effectiveness of delivering inactivated influenza vaccinations to adults in non-traditional settings. A decision tree was developed to predict the costs and health effects of inactivated influenza vaccination in various settings compared with no vaccination. Non-traditional settings included in the analysis were (1) mass vaccination clinics that can be conducted at a variety of locations such as employer sites, retail stores, or pharmacies and (2) pharmacies that use pharmacists to deliver vaccinations. The traditional setting was defined as the physician office with a scheduled appointment for vaccination. Adults were stratified by age and risk status for influenza-related complications. Probabilities and costs (direct and opportunity) for uncomplicated influenza illness, outpatient visits, hospitalizations, deaths, vaccination and vaccine adverse events were based on primary and secondary data. Costs of vaccination in non-traditional settings were collected via telephone surveys with representatives of organizations that conduct mass vaccination clinics and pharmacies. Primary outcomes reported were $/influenza episode averted, $/hospitalization averted, and $/death averted. For healthy 18-49 year olds, cost per influenza episode averted for vaccination in the pharmacy setting was $140 (95% probability interval: cost saving - $1200), $260 (cost saving-$1800) in the mass vaccination setting, and $1150 ($390-$6200) in the traditional setting. Vaccination was more likely to be cost-saving as age increased. Cost-effectiveness ratios were more favorable for high-risk as compared with healthy adults in the same age group. Results were sensitive to assumptions regarding probability of influenza illness, costs of vaccination (including patient time costs), and vaccine effectiveness. Non-traditional settings may represent a more convenient setting for vaccination at comparable or better cost-effectiveness than the traditional physician office setting.

Permalink

The impact of universal influenza recommendation at the clinic level

Presenter:

Michael Washington

Authors:

Michael L. Washington, Bo-Hyun Cho, Maggie Coleman

Chair: Noelle-Angelique M. Molinari; Discussant: Ismael Ortega-Sanchez Mon June 5, 2006 15:30-17:00 Room 325

While the US Advisory Committee on Immunization Practices considers expanding influenza recommendations to children up to age 18, little discussion has occurred on the logistics and economic costs to physician’s offices if such a recommendation is made. Clinics would have to determine ways to accommodate the influx of clients seeking care during the influenza season while addressing the spike in demand for influenza vaccine. Will providers need additional rooms, resources, and time? How much while this extra effort cost? One way to answer such a question is with computer simulation. The author will present a clinic model that will describe the behavior of the clinic before and after the peak influenza season and measure the logistical and economic impact.

Permalink

Economics of Increasing Immunization against Influenza in the US: Assessing Strategies toward Universal Vaccination

Presenter:

Ismael Ortega-Sanchez

Authors:

Ismael R. Ortega-Sanchez, Noelle-Angelique Molinari

Chair: Noelle-Angelique M. Molinari; Discussant: Megan O'Brien Mon June 5, 2006 15:30-17:00 Room 325

Rates of influenza illness severity and mortality have remained significantly high despite vaccination efforts: about 36,000 deaths and 200,000 hospitalizations are annually attributable to influenza in the US. More effective and extensive vaccination strategies leading to a universal annual vaccination against influenza have been under consideration. Yet, no systematic and comprehensive analyses of their prospective cost and benefits are available. In this study a probabilistic model is built for a US age- and influenza-disease-risk representative population sample to analyze the health and economic impact of strategies increasing actual vaccination coverage rates and effectiveness. Data on disease burden and resource utilization attributable to outpatient care, hospitalizations and deaths in annual epidemics is used in the model, as are estimates of incidence of co morbid conditions and productivity impact of temporal disability or premature death. Scenarios with increasing vaccination are also built and simulated using Monte Carlo algorithms. Among them projected vaccine effectiveness and hypothetical herd immunity impacts in the unvaccinated across age and risk groups are considered. After including vaccination costs and performing simulation analyses this study provides estimated ranges for health outcomes and resource utilization, estimates the potential net health impact and cost value of possible vaccination strategies, assesses the dollar value of prospective vaccination strategies, and evaluates the effects of using different criteria to expand vaccination policies across age and risk groups (e.g., based on risk of death, indirect protection or net economic returns). Sensitivity analyses are also conducted and projections to the US population calculated.

Permalink

Cost function Estimation of Mass Influenza Vaccination Clinics - Theoretical Frameworks and Challenges

Presenter:

Bo-Hyun Cho

Authors:

Bo-Hyun Cho, Mark L. Messonnier, Michael L. Washington

Chair: Noelle-Angelique M. Molinari; Discussant: Michael Washington Mon June 5, 2006 15:30-17:00 Room 325

Abstract: From the perspective of vaccination providers, cost function estimation can provide useful information of the cost impact of output changes as well as input prices. Yet, a conventional cost function does not consider undesirable outputs such as long patient waiting times. In this study, the theoretical background and statistical properties of two empirical models using scaling and translating methods will be addressed and compared to those of the conventional cost function. The immediate advantage of such an approach is that both models preserve the conventional properties of the cost function such as homogeneity, convexity in output and input prices while including the impact of an undesirable output. Simulation results will illustrate the advantages and challenges of the proposed models.

Permalink

Adjusting the Value of a Statistical Life for Age and Cohort Effects

Presenter:

Joseph Aldy

Authors:

Joseph E. Aldy, W. Kip Viscusi

Chair: Glenn Blomquist; Discussant: Don Kenkel Mon June 5, 2006 15:30-17:00 Room 326

To resolve the theoretical ambiguity in the effect of age on the value of statistical life (VSL) this article uses a novel, age-dependent fatal risk measure to estimate age-specific hedonic wage regressions. VSL exhibits an inverted-U shaped relationship with age. In the year 2000 crosssection, workers’ VSL rises from $3.2 million (ages 18-24), to $9.9 million (35-44), and declines to $3.8 million (55-62). Controlling for birth-year cohort effects in a minimum distance estimator yields a peak VSL of $7.8 million at age 46, and flattens the VSL-age relationship. The value of statistical life-year also follows an inverted-U shape with age.

Permalink

Altruism and Environmental Risks to Health of Parents and their Children

Presenter:

Mark Dickie

Authors:

Mark Dickie, Shelby Gerking

Chair: TBA; Discussant: Allen Goodman Mon June 5, 2006 15:30-17:00 Room 326

Special protection of young children from environmental hazards has become a worldwide priority of government policies to improve human health. The fundamental tension between parental altruism and self-interest looms as the crucial behavioral factor determining the effectiveness of these policies. This paper tests altruism in a unique field experiment by estimating marginal rates of substitution between skin cancer risks faced by 488 parents and their children between the ages of 3 and 12 years. Tests are guided by a reformulation of a standard model of altruistic family behavior that incorporates household production of latent health risk. The model demonstrates that while the marginal rate of substitution between parent consumption and child consumption of market goods is equal to unity, the marginal rate of substitution between risks faced by the parent and child is equal to the ratio of marginal risk reduction costs. Econometric estimates presented support the altruism hypothesis.

Permalink

Willingness to Pay for Improved Health: A Comparison of Stated and Revealed Preferences Models

Presenter:

Sylvia Brandt

Authors:

Sylvia Brandt, Michael Hanemann

Chair: TBA; Discussant: Allen Goodman Mon June 5, 2006 15:30-17:00 Room 326

The valuation of asthma morbidity is significant for environmental and health policies. The total direct and indirect cost of asthma in the U.S. was estimated to be $6.2 billion in 1990 and $12.7 billion in 1998. Both the intangible psychosocial costs of asthma and disutility of risk avoiding/mitigating behavior are omitted from these estimates of the economic burden of asthma. Cost of illness studies therefore should be taken as a lower bound of the true cost of asthma. In this paper we discuss two approaches to estimating the willingness to pay (WTP) for reduced asthma morbidity, contingent valuation and health production function. The study population includes 250 children ages 5-11 with clinically diagnosed asthma, residing in a section of Fresno County, California. Asthma symptoms, including coughing, wheezing and/or shortness of breath, ranged from mild and intermittent to severe and persistent in this group. Detailed health measures (including atopy and pulmonary function), utilization of health services, levels of antigens in the households and exposures to criteria air pollutants were collected as part of a five-year epidemiological study. We administered two economic surveys to measure 1) households’ perceptions of risks to an asthmatic child, 2) averting and/or mitigating actions taken, and 3) households’ stated willingness-to-pay for a reduction in their children’s asthma morbidity. In the health production model the health outcome is a function of exposure to asthma triggers, mitigating and averting behavior and household’s perceived risks. We find that variation in WTP is explained by attitudes towards asthma specific health investments including concerns of associated risks and perceived effectiveness. Furthermore, the observed racial differences in willingness to pay are strongly correlated with the elements of the Health Beliefs Model and measures of self-efficacy. The survey data indicate that households select from a small number of large, discrete health investments and that most risk reducing behavior are daily behavioral modifications with no relevant market prices. We argue that the discrete nature of health investments and socio-cultural patterns of health care utilization make the revealed preference approach inadequate for the case of asthma. As an alternative we present a contingent valuation scenario that was specifically developed to minimize systematic variation in preferences for characteristics related to the scenario rather than the reduction in asthma morbidity. For this purpose, guided by extensive testing in focus groups, we selected a scenario based on a hypothetical asthma monitor that provides to the wearer an indicator of current asthma status. We will present the results of a dichotomous choice experiment using this hypothetical asthma monitor. We then summarize how the difference between households’ stated and revealed WTP vary by socio-demographic variables and asthma severity.

Permalink

Willingness to Pay for the Detection and Treatment of Vulnerable Plaque Related to Heart Attacks

Presenter:

Patricia Ryan

Authors:

Patricia L. Ryan, Glenn C. Blomquist

Chair: TBA; Discussant: Don Kenkel Mon June 5, 2006 15:30-17:00 Room 326

Recent medical studies have led cardiologists to revise theories regarding the cause of heart attacks. Rather than a gradual clogging of the arteries, eruption of a “vulnerable plaque” is thought to be the cause of approximately 75% of all heart attacks. As a result, traditional risk factors are no longer sufficient indicators of who is at risk for a heart attack. Our research investigates willingness to pay (WTP) for a new, hypothetical detection (screening) and treatment method for vulnerable plaque. We develop a survey instrument that exploits the visual and interactive aspects of the Internet. We provide information about screening and treatment and get individual’s perceptions of the effectiveness and risks associated with the screening and treatment. Using a web-based survey, which follows a contingent valuation format, we use an iterative bidding process to elicit the respondent’s WTP for either the screening or treatment of vulnerable plaque. The iterative bidding allows us to obtain a maximum willingness to pay value from each respondent. Internet, on-line surveys are often prone to coverage bias; however, Knowledge Networks has established a panel of randomly selected households, which allows researchers to achieve a nationally representative sample using online surveys. Our survey on screening (a simple blood test) resulted in a sample of 268 adults based on sampling representative of the general population. Our survey on treatment (a more invasive heart catheterization procedure) was administered only to those who had doctor-diagnosed heart problems and were therefore more familiar with these types of medical decisions and resulted in a sample of 295 adults. We find an average WTP for screening of approximately $90 and estimates for treatment that are approximately an order of magnitude greater. Our experience suggests that our approach can provide useful information for decision making dealing with screening and treatment related to heart attack.

Permalink

Go Out or Stay In? The Effects of Zero Tolerance Laws on Alcohol Use and Drinking and Driving Patterns among College Students

Presenter:

Lan Liang

Authors:

Lan Liang, Jidong Huang

Chair: Jody L. Sindelar; Discussant: Daniel Polsky Mon June 5, 2006 15:30-17:00 Room 332

Zero tolerance laws make it illegal per se for anyone under age 21 to drive with any measurable trace of blood alcohol. While a link has been established between zero tolerance laws and lower motor vehicle fatalities, research has not produced strong evidence on how zero tolerance laws influence individual alcohol use and drinking and driving behaviors. Using a unique data set and a difference-in-difference research design, we are able to analyze a number of pathways through which zero tolerance laws can work among an important underage population, college students. We find that zero tolerance laws change both drinking patterns and drinking and driving behaviors. In particular, we find that zero tolerance laws reduce drinking frequency and the probability of drinking and driving. More importantly, we are able to infer from the data whether drinking occurs at home or away from home. Our results indicate that zero tolerance laws reduce drinking intensity when drinking occurs away from home and reduce the probability of drinking and driving for those who report drinking away from home.

Permalink

The Impact of Marijuana Prices on Emergency Department Admissions

Presenter:

Rosalie Pacula

Authors:

Rosalie Pacula, Khoa Truong

Chair: Jody L. Sindelar; Discussant: Edward Norton Mon June 5, 2006 15:30-17:00 Room 332

There is a common belief among the general public that the health consequences of marijuana use are relatively minor even though the rates of dependence have grown significantly over the past decade. In this paper we examine the association between marijuana use and the utilization of emergency department (ED) services using quarterly data from the 1994-2002 Drug Abuse Warning Network (DAWN). Reduced form models evaluate the association between local marijuana prices and the number of marijuana-involved ED episodes. Additional controls are included to account for high rates of alcohol and cocaine use. We find that even when we control for alcohol and cocaine prices, counties with lower marijuana prices have statistically higher marijuana-involved ED mentions. This relationship holds for marijuana-only mentions as well as encounters involving marijuana and other drugs. The findings are even stronger when we restrict the sample to episodes involving individuals less than 25 years of age. These results provide convincing evidence that there are real acute health consequences associated with marijuana use.

Permalink

Fatal and Nonfatal Motorcycle Injuries: Can Alcohol Policies Influence Rider Safety and Reduce Accidents?

Presenter:

Michael French

Authors:

Michael T. French

Chair: Jody L. Sindelar; Discussant: Willard Manning Mon June 5, 2006 15:30-17:00 Room 332

Numerous studies have established the effectiveness of alcohol policies (e.g., blood alcohol content for driving under the influence convictions, minimum legal drinking age, beverage taxes) in reducing rates for automobile accidents, injuries, and fatalities. Only a few studies have analyzed the impact of alcohol policies on motorcycle accidents and none within the past 5 years. Given the growing popularity of motorcycle riding among all age groups and the inherent safety risks, a new and comprehensive investigation is warranted. The present study uses state-level longitudinal data from 1994 to 2003 to determine whether a variety of different alcohol and motorcycle safety policies have a significant impact on both fatal and nonfatal motorcycle injuries. Besides state-level policies, the analysis controls for numerous demographic and environmental factors in each state through the estimation of state fixed-effects models. The results lead to direct policy recommendations for how states can reduce motorcycle injuries and enhance overall motorcycle safety.

Permalink

The Economics of Racial Health Disparities

Presenter:

Amitabh Chandra

Authors:

Amitabh Chandra, Douglas Staiger

Chair: Jonathan Skinner; Discussant: Mon June 5, 2006 15:30-17:00 Room 335

Permalink

Identifying Inefficiency in the Diffusion of Medical Care Practices

Presenter:

Wesley Yin

Authors:

Wesley Yin

Chair: Jonathan Skinner; Discussant: Mon June 5, 2006 15:30-17:00 Room 335

Permalink

Diffusion and the Geography of Health Care Productivity

Presenter:

Jonathan Skinner

Authors:

Jonathan Skinner, Douglas Staiger

Chair: Jonathan Skinner; Discussant: Mon June 5, 2006 15:30-17:00 Room 335

Permalink

Cohort Turnover and Productivity: The July Phenomenon in Teaching Hospitals

Presenter:

Robert Huckman

Authors:

Robert Huckman, Jason Barro

Chair: Joel Hay; Discussant: TBA Mon June 5, 2006 13:45-15:15 Room 121

We consider the impact of cohort turnover-the simultaneous exit of a large number of experienced employees and a similarly sized entry of new workers-on productivity in the context of teaching hospitals. In particular, we examine the impact of the annual July turnover of house staff (i.e., residents and fellows) in American teaching hospitals on levels of resource utilization (measured by risk-adjusted length of hospital stay) and quality (measured by risk-adjusted mortality rates). Using patient-level data from roughly 700 hospitals per year over the period from 1993 to 2001, we compare monthly trends in length of stay and mortality for teaching hospitals to those for non-teaching hospitals, which, by definition, do not experience systematic turnover in July. We find that the annual house-staff turnover results in increased resource utilization (i.e., higher risk-adjusted length of hospital stay) for both minor and major teaching hospitals and decreased quality (i.e., higher risk-adjusted mortality rates) for major teaching hospitals. Further, these effects with respect to mortality are not monotonically increasing in a hospital’s reliance on residents for the provision of care. In fact, the most-intensive teaching hospitals manage to avoid significant effects on mortality following this turnover. We provide a preliminary examination of the roles of supervision and worker ability in explaining why the most-intensive teaching hospitals appear able to reduce turnover’s negative effect on performance.

Permalink

The Strange Case of the Negative Resident Wage

Presenter:

Jerry Cromwell

Authors:

Jerry Cromwell

Chair: Joel Hay; Discussant: TBA Mon June 5, 2006 13:45-15:15 Room 121

Authors: Jerry Cromwell (jcromwell@rti.org) and Ed Drozd (edrozd@rti.org), Research Triangle Institute (RTI)

Title: The Strange Case of the Negative Resident Wage

Rationale: Economic theory rarely encounters an input with a negative wage, or supply price. Theory also consistently rejects the notion of a “Giffin” input exhibiting an income effect wherein an employer seeks to buy more of the input when its price rises. Medical residents hired by non-profit teaching hospitals may exhibit both of these strange features.

Objectives: The paper derives estimates of the marginal subsidized wage, or shadow price, of residents in teaching hospitals at differing levels of Medicare financial dependence. The goals are (a) to quantify the degree to which residents are a “line-of-business” for hospitals, (b) to explore the role of Medicare payment factors in generating negative resident wage rates, and (c) to explain the continued growth in residents in the face of Congressional constraints on Medicare Direct and Indirect Medical Education (DME & IME = total GME) payments.

Methodology: Teaching hospital utility functions are maximized with respect to community services and size of teaching programs subject to revenue, cost, and patient care production functions. First-order conditions produce resident demand curves dependent on stipends that are offset by marginal Medicare DME and IME subsidies as well as preferences for higher patient volumes. Adjusted resident shadow prices are simulated by varying key Medicare payment and hospital demand parameters. Next, subsidies are calculated for roughly 1,000 teaching hospitals for 2001 and financial “bite” estimates quantified as a result of the 1997 Balanced Budget Act (BBA). Descriptive statistics and OLS regressions show trends in resident growth post-BBA, financial impacts of GME rollbacks, size distributions of subsidies, and the ultimate effective marginal resident wages by hospital characteristic.

Results: Post-BBA, 1996-2001, FTE resident counts in teaching hospitals increased roughly 7,000, or 9%, despite a 17-20% reduction in expected resident subsidies. Teaching hospitals easily absorbed such reductions that were only 0.5% of total revenues. Simulations indicated negative marginal effective wages for over 90% of teaching hospitals in 2001, encouraging further resident hires. With Medicare subsidies alone, resident shadow wages averaged -$57,000, implying that Medicare payments, at the margin, more than doubled the entire resident stipend (about $40,000). Econometric results showed a direct relationship between Medicare bite and resident growth, with hospitals facing negative marginal wage rates continuing to increase demand holding bite constant.

Conclusions: While the federal government has enacted wage subsidies for selected adults in the past, none compare with the enormous subsidies afforded residents in teaching hospitals. Congressional attempts to rein in resident growth through “soft” payment caps proved futile; persistent negative resident shadow prices remain for many hospitals. Extremely low relative wages, coupled with high resident patient care productivity, guarantees continued training slots for all U.S. medical graduates, large influxes of foreign graduates, and teaching hospital overpayments.

Permalink

Which is a Better Investment for Health Workforce Growth: Expanded Educational Space or Financial Support to Students?

Presenter:

Joanne Spetz

Authors:

Joanne Spetz, Susan Chapman, Jean Seago

Chair: Joel Hay; Discussant: TBA Mon June 5, 2006 13:45-15:15 Room 121

Rationale: The past eight years have been marked by a severe, acute shortage of registered nurses in the United States. In response to this shortage, federal and state governments have provided funding to nursing education programs to expand the supply of RNs. Most of these programs either provide financial support to currently-enrolled students, with the goal of improving the productivity of education programs, or expand the number of education slots available for students. There has been no research of the relative effectiveness of these strategies.

Objectives: This study compares the costs and results of programs to provide financial support to nursing students and expand nursing education slots, using new data from California.

Methodology: Twenty-two regional groups were awarded a total of $28 million to expand RN supply, with projects ending in 2005. Some of these groups used the funds to provide financial support to students, some expanded education slots, and some did both. In addition, some programs received private funds from hospitals and other local agencies to support their regional efforts. We are using data provided by the California Board of Registered Nursing, California Community Colleges Chancellors Office, the funded groups, and other state government agencies. To estimate the effectiveness of the financial support programs, we estimate a multivariate equation in which the graduation rate is the dependent variable, and the key explanatory variable is the percent of students receiving financial support. The coefficients from this equation are then used to estimate the net increase in graduating RNs resulting from the financial support programs. To estimate the effect of the slot-expansion programs, we multiply the number of new slots by the share of students expected to graduate. These two numbers of “net new nurses” are compared with the costs of each strategy.

Results: Over 1500 RN students received financial support to improve program productivity, and over 1800 new education spaces were created. Preliminary data suggest that the productivity-improvement programs had a substantial effect on student attrition. However, the programs that used funds to expand slots have allowed for the education of many additional nurses. It appears that the slot expansion programs provide more nurses per dollar. Final data are now being received by the research team and a final report will be written by March 31, 2006.

Permalink

Patients' Willingness to Accept Risk-Benefit Tradeoffs in Treating Multiple Sclerosis

Presenter:

George van Houtven

Authors:

George van Houtven

Chair: Peter Zweifel; Discussant: Frank Sloan Mon June 5, 2006 13:45-15:15 Room 213

Permalink

How Does Price Matter in Healthcare Stated-Choice Surveys?

Presenter:

F. Reed Johnson

Authors:

F. Reed Johnson

Chair: Peter Zweifel; Discussant: Mark Pauly Mon June 5, 2006 13:45-15:15 Room 213

A literature review of published stated choice (SC) studies related to health care indicates that all published studies estimate willingness to pay (WTP) by dividing the estimated utility difference between an intervention or outcome and a reference condition by the marginal utility of income (MUY). However, estimates of the marginal utility of income are only valid if subjects accept price levels as income-constrained, out- of-pocket expenses of obtaining the indicated outcome and if MUY is approximately constant over the relevant range of prices. There currently is little empirical evidence on how SC subjects evaluate price information when health care is partly or fully insured. SC subjects may ignore prices altogether in evaluating tradeoffs, discount prices because they are accustomed to paying only a fraction of actual prices, or recode price levels to categories such as low, medium, and high that ignore actual differences in price levels. Recoding may be a symptom of simplifying a less important attrib?????here patients currently pay for certain healthcare-related services out of their own pocket, and “cheap talk” strategies that alert subjects to common response errors in SC surveys.

Permalink

Age and Choice in Health Insurance: Evidence from Switzerland

Presenter:

Peter Zweifel

Authors:

Karolin Becker, Peter Zweifel

Chair: Peter Zweifel; Discussant: TBA Mon June 5, 2006 13:45-15:15 Room 213

TBA

Permalink

What matters: Reality or Perception? The impact of peer binging on college student drinking behaviors

Presenter:

Amy Wolaver

Authors:

Amy Wolaver, Lisa Powell, Frank Chaloupka, Christina Ciecierski, Henry Wechsler

Chair: Catherine McLaughlin; Discussant: Jeff DeSimone Mon June 5, 2006 13:45-15:15 Room 225

The objective of this paper is to determine the relative importance of actual and perceived rates of heavy drinking on individual behavior. We will also test whether the accuracy of students’ perceptions of heavy drinking matter by examining the difference in the student perceptions from the actual rates of heavy drinking on campus has an effect on the individual’s propensity to consume.

Methodology: Because peer effects potentially suffer from multiple sources of endogeneity problems, we use instrumental variable techniques and the 1997, 1999 and 2001 College Alcohol Studies to examine how the peer rate measures affect individual behavior. We assume that no contextual effects exist, which allows us to use average parental characteristics of peers to identify the effect of the level of heavy drinking on the probability that an individual engages in heavy episodic drinking. The CAS surveys ask students the question, “Based on what you have heard or experienced, approximately what proportion of [all students/ your friends] do you think are heavy or problem drinkers at this school?” We will examine whether there are differences in the peer effects based on the actual and perceived measures (based on the above self-reports) of peer problem drinking. These measures include: percent of all students estimated who are heavy or problem drinkers, percent of friends that are heavy or problem drinkers, the difference between their estimate of problem drinkers and the actual rate of binge drinking. We will also test for differences based on whether individual’s over- or under-estimate the peer rate of drinking.

Preliminary Results: Heavy episodic drinkers estimate slightly higher average rates of heavy/problem drinking at schools and much higher rates of heavy drinking among their friends than non-heavy episodic drinkers. Student perceived rates of heavy drinking are positively correlated with the actual rates of drinking at their schools. The perceived rate of heavy drinkers has a larger marginal effect on the probability of binge drinking by the individual than the actual rate at the school; but the difference between the actual and perceived rates of heavy drinking reduces the probability of binge drinking.

Permalink

Identifying Peer Effects in Substance Abuse Treatment

Presenter:

Tracy Falba

Authors:

Tracy Falba, Patrick Bayer, Jody Sindelar

Chair: Catherine McLaughlin; Discussant: Phil DeCicca Mon June 5, 2006 13:45-15:15 Room 225

Background: A wide variety of studies have examined the impact of peers on outcomes including education, crime, health, and labor market activity. Peer effects in drug and alcohol use (particularly among teens) have received considerable academic attention because peers are often thought to be important both for usage and abstinence. Research to date has not adequately isolated causal effects due to the fact that observed peers are typically not randomly assigned but chosen by the individuals studied.

Aims of Study: This study examines the impact of peers in drug and alcohol treatment programs in Connecticut on a variety of outcomes including successful completion of the program, drug usage during and after the program, relapse into drug or alcohol treatment (e.g., future admission for detoxification), crime, and labor market activity. We study the impact of a large set of peer characteristics related to demographic characteristics, past drug usage (by intensity and drug category), criminal activity, and labor market activity on these outcomes.

Methods: Using data from all drug treatment facilities in Connecticut, we use the methodology developed in Bayer et. al. (2004) to isolate causal effects of peers on outcomes. This methodology uses only the within-program variation in peer characteristics and therefore eliminates any variation associated with the selection or assignment into particular facilities. Because individuals spend reasonably short amounts of time in treatment and are admitted and released from programs on a regular basis, considerable variation exists in the characteristics of one’s peers in terms of past drug use, criminal activity, labor market activity, and age depending on the exact period of time that individuals receive treatment.

Results: To date, we have explored the features of the institutional setting in Connecticut. Our primary sample consists of all admissions for drug treatment (over 150,000) in Connecticut from 1995-2003. Of these about 60 percent are outpatient, 15 percent residential-short term, 15 percent residential long-term, and 10 percent methadone maintenance. Individuals were treated in a total of approximately 200 programs over this period. The data reveals a great deal of variation both across and within facilities for the wide set of peer characteristics and outcomes mentioned above, making the institutional setting appropriate for our methodology.

Discussion: This analysis provides a wide set of results that characterize a number of causal channels through which peers affect outcomes. Consequently this study has a number of important implications for how individuals should best be assigned to programs and groups within programs while receiving drug treatment. Additionally, this study informs our understanding of the nature of peer effects in drug use and treatment. The size of these effects have far-reaching policy implications ranging from the optimal timing of drug interventions, to the optimal range of treatment options, to the identification of individuals particularly at risk of relapse.

Permalink

The Determinants of Gambling Behavior

Presenter:

John Nyman

Authors:

John Nyman, John Welte, Bryan Dowd

Chair: Catherine McLaughlin; Discussant: Robert Kaestner Mon June 5, 2006 13:45-15:15 Room 225

Gambling has become an important public health issue. Gambling has been associated with suicides, marital violence, substance abuse, increased morbidity, and violent crime. Although gambling is primarily an economic activity, no single theory of the demand for gambles has gained wide-spread acceptance among economists. This paper presents a new model of the demand for gambling that is based on the standard economic assumptions that resources are scarce and consumer’s utility increases with income at a decreasing rate. It suggests that those who gamble have a labor market perspective, and as such, they regard the utility gain from gambling not only as related to the income gained but also to the work avoided.

This model is tested using the 2000 Survey of Gambling in the U.S., a nationally representative survey of 2,631 respondents. The existence of relationships that were suggested by the theory were tested for, using a system of 3 regression equations. It was found that whether a person gambled or not depended on whether they had a labor market orientation, as determined by their labor market experience. It was also found that the amount of gambling was determined by education and factors that were associated with labor market imperfections-such as, the respondent’s race, his health, whether he was obese, and the unemployment level in the respondent’s local market. Both sets of results are generally consistent with the theory.

This study has implications for public policy. It suggests that the findings from other studies that pathological gambling is more likely among certain racial groups may instead reflect the fact that certain races are more likely to engage in non-pathological gambling because of labor market imperfections, and among those who engage non-pathological gambling, these racial groups are no more likely than members of other races to become pathological gamblers. This has implications for programs that intervene to reduce pathological gambling.

By showing the connection between the labor market and gambling, this study has also demonstrated that states that use lotteries to obtain revenues are methodically and disproportionately drawing on the resources of those who are least advantaged in society. This has implications for horizontal equity and tax justice.

Permalink

Alcohol Consumption and Unemployment

Presenter:

Gulcin Gumus

Authors:

Gulcin Gumus

Chair: Edward Norton; Discussant: Euna Han Mon June 5, 2006 13:45-15:15 Room 226

This paper analyzes how alcohol consumption affects the probability of being unemployed using data from the 2001-2002 National Epidemiologic Survey on Alcohol and Related Conditions (NESARC). In order to address the simultaneity between labor market performance and alcohol use, I estimate a maximum likelihood instrumental variables (IV) probit model. The age of drinking onset is used as an IV which proves to be a strong and valid identifying variable for alcohol use in the context of employment outcomes. Results indicate that there are substantial differences between standard probit estimates and the IV estimates. Controlling for all other factors, standard probit results suggest that men who drink more are more likely to be unemployed. However, this effect becomes statistically insignificant in the MLE results once I control for endogeneity. Finally, I also show that some IVs commonly used in the past studies fail to pass validity tests indicating that they might not be ideal candidates for an IV method.

Permalink

Employment and Income Effects Related to Drinking with Controls for Drug Abuse and Smoking: Analysis Using the NESARC Database

Presenter:

Allen Goodman

Authors:

Allen Goodman, Janet Hankin

Chair: Edward Norton; Discussant: Gary Zarkin Mon June 5, 2006 13:45-15:15 Room 226

Authors. Allen C. Goodman (allen.goodman@wayne.edu) and Janet R. Hankin (janet.hankin@wayne.edu), Wayne State University

Title: Employment and Income Effects Related to Drinking with Controls for Drug Abuse and Smoking: Analysis Using the NESARC Database

Rationale: A considerable literature has examined impacts of drinking on employment and income, but existing analyses have generally used databases that do not reflect the general population. In addition, existing analyses reflect data collected no later than the early to mid 1990s. This study seeks to update analyses with a new database that addresses both issues, with methods that reflect selection effects.

Objectives: The objective is to examine both direct and indirect impacts of alcohol consumption on labor force participation and income, with controls for drug abuse and smoking.

Methods: Multiple regression and maximum likelihood methods are used to estimate impacts of alcohol consumption on labor force participation and income, with controls for drug abuse and smoking. The study uses the recently collected National Epidemiologic Survey on Alcohol and Related Conditions (NESARC), designed to be the primary source for information and data on the U.S. population for alcohol and drug use. The NESARC provides multi-dimensional information on alcoholism, drug abuse, and cigarette smoking, as well as excellent individual and labor market information.

Results: Separate analyses by gender show that alcohol use significantly impacts the choice between part-time and full-time employment. Heavy drinkers are more likely to be unemployed or part-time workers compared to lighter drinkers or abstainers. In contrast, alcohol use has an insignificant impact on income conditional on employment. Because better health, education, and part-time v. full-time status are related to higher annual income, increased drinking indirectly lowers annual income by influencing these factors, but it has little direct impact on income outside of these indirect effects.

Conclusions: The labor force impacts of alcohol and substance abuse are complex. The alcohol impacts are most pronounced with respect to labor force participation, and full- v. part-time employment. Drug use has its major impact on men with respect to labor force participation (less likely to be full-time) and health (worse health). Women who use drugs are also less likely to work full-time and will have worse health. Current smoking has negligible effects on labor force status, but it is inversely related to health and to education, and thus impacts income indirectly in a negative manner. It has modest direct impacts on part-time worker incomes, but significantly negative direct impacts on income for both men and women.

Permalink

The Impact of Parental Drinking on Children's Utilization of Health Care Services

Presenter:

Ana Balsa

Authors:

Ana Balsa

Chair: Edward Norton; Discussant: Hua Wang Mon June 5, 2006 13:45-15:15 Room 226

Approximately one in four children in the United States are exposed to alcohol abuse or dependence in the family by the age of 18. While prior research has underscored the detrimental effects that parental drinking inflicts upon children, there are no large scale studies analyzing the magnitude of these effects relative to the costs directly incurred by those affected by problem drinking. Using the National Health Interview Survey, a nationally representative data set, this paper analyzes the economic consequences of parental risky drinking on children’s utilization of healthcare. We hypothesize that children with parents who misuse alcohol are more likely to suffer from certain health conditions (behavioral health problems, injuries, stress related conditions and physical abuse) that lead them to use more acute healthcare. On the other hand, negligence by a problem drinking parent can result in a lower demand for services that are not immediately needed by the child, such as dental or preventive care. The analysis is addressed in two ways. First, instrumental variables are used to analyze the impact of parental yearly consumption of alcohol on children’s use of healthcare services. Results indicate that higher amounts of parental drinking raise children’s utilization of mental healthcare, specialty care, and hospital services. Parental drinking also adds to the probability that a child goes without needed care and decreases the likelihood that a child has regular check-ups and dental visits. Children in non-intact or lower income families, girls and Hispanics are more likely to suffer the effects of parental risky drinking. A second analysis quantifies the effects of a change in a policy variable, beer tax, on children’s healthcare costs, controlling for various other state-related characteristics. Children’s use of acute healthcare services decrease substantially as the beer tax increases, and use of preventive services increase. Because female drinking is more elastic to beer tax than male drinking, it is possible that these differences are being led by variation in maternal alcohol consumption. Overall, the analysis suggests that the costs of parental alcoholism on children’s health and healthcare are not negligible and should be taken into consideration when designing policies and treatments, and when evaluating programs and outcomes.

Permalink

Do HMOs Provide Better Quality Primary Care to Medicare Beneficiaries?

Presenter:

Jayasree Basu

Authors:

Jayasree Basu, Lee Mobley

Chair: Albert Okunade; Discussant: TBA Mon June 5, 2006 13:45-15:15 Room 235

Managed care plans can directly reduce preventable hospitalizations by increasing primary and preventive services available to constituents. Accordingly, a higher rate of preventable hospitalization has been proposed as an indicator of poorer health plan performance. The literature suggests that there is greater use of preventive care in HMOs. To date, no study has evaluated whether Medicare HMO enrollees experience differential preventable hospitalization rates than those in traditional Medicare (FFS), or whether there would be beneficial spillovers from managed care penetration in the overall market that would reduce preventable hospitalizations for the Medicare population. The purpose of this study is to assess both direct and indirect effects of managed care on preventable hospitalizations among the elderly in four states.

This research uses the complete 2001 hospital discharge abstracts for elderly Medicare enrollees (age 65 and above) who reside in four states (NY, PA, FL, CA), from the Healthcare Cost and Utilization Project (HCUP-SID) database of the Agency for Healthcare Research and Quality. These states exhibit high but varying degrees of managed care experience and penetration. We use a multivariate cross sectional design with patient level data, modeling each state’s cohort separately due to heterogeneity across states. The logistic model controls for socioeconomic and demographic variables, patient severity of illness, and county resources. We use admissions for persons with “marker” conditions as a comparison group for the observed admissions for persons with preventable conditions, following methods used in the literature.

We find no evidence of beneficial spillovers from HMO penetration onto area Medicare populations. However, holding other factors such as demographics and illness severity constant, we find that in 3 out of 4 states, Medicare HMO patients had lower odds of a preventable admission versus marker admission than Medicare FFS patients. The odds in CA, NY, and FL were respectively 17%, 12%, and 9% lower for HMO as compared to FFS patients. There was no significant difference for HMO versus FFS patients in PA, the state with the tardiest managed care presence and lowest penetration by Medicare HMOs. Moreover, in the two states with longest tenure and greatest Medicare HMO penetration, CA and FL, the reduction in preventable admissions among Medicare HMO patients was mainly concentrated among more ill patients. That is, the predicted marginal probability of preventable vs marker admission was higher in FFS plans than Medicare HMO plans, and the gap widens across the patient severity index.

These findings suggest that Medicare HMO enrollees may have experienced better quality primary care than their FFS counterparts in these three states. The largest difference in quality between HMO and FFS patients was found in CA, the state with the earliest managed care penetration and largest Medicare HMO presence. Because we control well for patient illness severity, these findings add to the evidence that managed care outperforms traditional care among the elderly, rather than simply skimming off the healthiest populations. Although the study uses one outcome indicator for the quality of primary care, it is quite important for total social cost of care and a signal of effective management of chronic illness.

Permalink

HMO Selection Incentives and Underprovision of Mental Health Care

Presenter:

Zhun Cao

Authors:

Zhun Cao

Chair: Albert Okunade; Discussant: TBA Mon June 5, 2006 13:45-15:15 Room 235

Authors: Zhun Cao (zcao@charesearch.org), Thomas G. McGuire (mcguire@hcp.med.harvard.edu)

Title: HMO Selection Incentives and Underprovision of Mental Health Care

Rationale: Adverse selection, a well-known problem in insurance markets, interferes with the provision of fair and efficient coverage for mental health (MH) services. Managed care plans paid by capitation have an incentive to attract “good” risks (e.g. people more likely to use preventive care) and avoid “bad” risks (e.g. people with MH or other chronic conditions). The problem for MH care arises because people likely to use MH coverage are generally considered “bad risks” with higher overall health care costs than other potential enrollees. Competing managed care plans may therefore structure their product in a way to discourage enrollment by persons with mental illness. This type of selection could take the form of a limited MH provider network or especially strict requirements for referral to MH specialty care.

Objectives: The purpose of this paper is to test for the presence of service-level selection against MH care in the pattern of choice of Medicare managed care plans. We improve on existing tests of service-level selection in choice models by explicit control for health status variables, by use of HMO expenditures directly, and by following beneficiaries over time.

Methodology: Using an individual choice model, we assume that an individual’s utility is affected by the benefit expected from the plan once she/he is sick, personal preferences and needs, and cost of joining a plan. Empirically, we employ a random effect logit model, where the dependent variable is a dummy indicating if the individual is enrolled in a managed care plan, versus a traditional fee-for-service plan. The independent variables include service-level medical spending, demographic and socio-economic variables, mental and physical health status, county characteristics, etc. We also assume a random disturbance that is invariant for the individual across time. Negative signs of the coefficients for previous MH spending would imply MH services are possibly underprovided.

Data: Data from Medicare Current Beneficiary Survey (MCBS, 1996-2001) are used. The MCBS provides extensive information on Medicare HMO enrollment, personal characteristics, health status, etc. The longitudinal feature of the data allows us to follow an individual’s choice over years. We also merge MCBS with Area Resource File, Medicare Market Penetration File and Census Data to obtain information on market characteristics.

Results: We find that people with higher previous expenditure in MH services are less likely to enroll in a Medicare HMO plan, even after controlling for health status. The results imply that MH services may be underprovided by Medicare HMOs, compared to physical health care services.

Conclusions: The findings in this paper provide evidence that Medicare HMOs ration care at the service level, and that MH services are underprovided by HMO. In order to ensure sufficient MH care to the elderly people, service-level risk adjustment should be adopted in the Medicare payment policy to offset the HMOs’ incentive to select against people with higher MH risk.

Acknowledgement: Research support from NIMH grant R03 MH071602-01A1 is gratefully acknowledged.

Permalink

Propensity Score and Instrumental Analysis of Mortality Differences Between Medicare HMO Enrollees and FFS Beneficiaries

Presenter:

Matthew Maciejewski

Authors:

Matthew Maciejewski, Song Wang, Andrew Zhou

Chair: Albert Okunade; Discussant: TBA Mon June 5, 2006 13:45-15:15 Room 235

Rationale: Identification of the causal effect of an intervention is challenging in observational data. Propensity score and instrumental variable analysis has been used in health services applications to reduce confounding, but it is unknown whether these two methods are likely to generate similar results.

Objectives: To adjust for selection bias in estimating the effect of Medicare HMO enrollment on mortality for Medicare beneficiaries with diabetes comparing instrumental variables (IV) and propensity score matching (PS) approaches.

Methodology: Medicare claims data were obtained on a 2% sample (N=59,089) of Medicare beneficiaries with diagnosed diabetes in 1995. Managed care benefits and out-of-pocket premiums were also obtained from Medicare to generate county-level measures of plan generosity. We performed cross-sectional analysis of 59,089 patients. IV and PS using radius matching are used to generate the average effect of HMO enrollment on mortality across the entire sample and average treatment effects within quintiles. A non-parametric bootstrap is performed to calculate confidence intervals for the treatment effects by quintile.

Results: The average treatment effect from a PS matching model and a simple probit model suggested that HMO enrollment is associated with a 4% reduction in mortality for Medicare beneficiaries with diabetes in 1995. The mortality reduction from the IV model was much smaller (0.4%). The quintile-specific treatment effects were also very different been PS matching and IV, and also varied across quintiles. Improving balanced in observed covariates via PS were much less effective in reducing bias in the effect of HMO enrollment on mortality than was statistical correction for unobserved confounding via IV. When such a correction is applied, the mortality of HMO enrollees and FFS beneficiaries are almost equivalent.

Conclusions: Propensity score and instrumental variables analysis provided different treatment effect estimates in this study. Propensity score adjustment for imbalance in observed covariates was not able to address bias due to unobserved covariates in instrumental variables, which generated a null effect of HMO enrollment on mortality. Researchers interested in applying methods for identifying causal effects of treatment must clearly understand the nature of the bias that needs to be addressed.

Permalink

Determinants of catastrophic health expenditures in Brazil: the impact of the public system and private insurance

Presenter:

Antonio Bos

Authors:

Antonio Bos, Hugh Waters

Chair: Jon Christianson; Discussant: Jean Abraham Mon June 5, 2006 13:45-15:15 Room 309

Despite the fact that Brazil has a health system with universal access and care free of charge at point of delivery, several previous studies have shown that Brazil has a high number of households with large health expenditures. The primary objective of this research is to assess the effectiveness of the public health system and private health insurance in providing financial protection in health care to the Brazilian population.

This research utilizes data from a large national survey of households. The determinants of catastrophic health expenditures (CHE) were estimated by a probit regression with a Heckman selection adjustment, controlling for the need for health care.

According to our findings, the Brazilian health system provides a significant reduction (41%) in the probability that a household has CHE. On the other hand, private health insurance did not reduce out of pocket expenditures and, given that the premiums are quite expensive, made CHE more likely - by 37%.

Recommendations include further improvements in the quantity, accessibility, quality and reliability of public health providers and more extensive and appropriate provision of medicines by the public system. A tighter regulation of private health plans is also necessary, both on the high cost of the insurance premiums and on the restricted nature of the benefits.

Permalink

Ranking the health system efficiency among Canadian provinces and American states

Presenter:

Chunping Liu

Authors:

Chunping Liu, Brian Ferguson, Audrey Laporte

Chair: Jon Christianson; Discussant: Jennifer S. Schultz Mon June 5, 2006 13:45-15:15 Room 309

Rationale: The World Health Report 2000 ranks the performance of the health systems on 191 countries. In many countries such as Canada, health systems are actually managed at sub-national levels, provinces or states. In addition, most efficiency studies restrict themselves to a single output measure. We consider the multiple outputs case and apply the methodology, where data permits, at sub-national levels.

Objectives: The objective of this paper is to compare the efficiency of health systems amongst Canadian provinces and American states using non-parametric Data Envelopment Analysis (DEA) approach and a multiple inputs and outputs system.

Methodology: We employ DEA to analyze the technical efficiency of health systems, using both input-oriented and output-oriented approaches to fitting the efficient frontier and comparing the relative efficiency of units within frontier with their best practices peers on the frontier. As inputs we use information on physicians, nurses, hospitals and pharmaceuticals. We fit two models, one measuring inputs in quantity terms and the other measuring them in real expenditure terms, and investigate the implications of these two approaches for measures of relative efficiency. Of particular interest is what measures of expenditure efficiency relative to quantity efficiency can tell us about the degree to which different countries health expenditures go towards increasing factor income rather than increasing output. Another contribution of this paper is to include output variables other than the most commonly used Life Expectancy at Birth variable. Low birth weight infant mortality and life quality variables are very important output variables as the outcomes of the health system performance. We also consider measures of self-reported health status and Potential Year of Life Lost (PYLL) for stroke and cancer. Sensitivity analysis is performed using different sets of output variables.

Results: Half the Canadian provinces and eight American states are technically efficient in both the quantity and monetary models. The analysis suggests that Canadian provinces distributed resources relatively more efficiently overall in comparison with American states. It was determined that American states pay their inputs at levels above those in Canadian provinces’.

Conclusions: The analysis in this paper suggests that Canadian provinces distributed the resources relatively efficiently in comparison with American states. It is more reasonable to compare the health system efficiency at sub-national level. Since the efficiency scores for Canadian provinces are close to one, it is hard to squeeze more output for those inefficient Canadian provinces and we will have to invest more on resources in order to produce more outputs.

Permalink

Federal Health Care Funding in Canada: Is It Allocated According to Need?

Presenter:

Yukiko Asada

Authors:

Yukiko Asada, George Kephart

Chair: Jon Christianson; Discussant: Jon Christianson Mon June 5, 2006 13:45-15:15 Room 309

Background: Equity is one of the defining values of the Canadian health care system. While equity in the access to or utilization of health care has often been the focus of passionate public and policy debate, equity in federal health care funding has been largely neglected. The Canadian government currently distributes targeted health funding to provinces largely on an equal per capita basis. Thus, differences in average per capita need resulting from differences in health status or demographics are not taken into account. The objectives of this study were to assess potential magnitude of unequal need for health services between provinces and to compare different indicators and approaches to measuring need for health services.

Methods: We estimated relative per capita need for general practitioner, specialist, and hospital services by province using the following three approaches: (1) estimating relative per capita need for health services based on demographics (age and sex), (2) estimating relative per capita need for health services based on demographics, socioeconomic status, and health status, and (3) estimating relative per capita need for health services based on demographics and premature mortality. For all three approaches, we first fitted regression models to estimate standardized utilization of each of three types of health services by indicators of need. The standard was assumed to be average levels of utilization by needs indicators in the national sample. Two-part regression models were employed: logistic regression for use and nonuse, and negative binomial regression for frequency of use. Subsequently, the standard was used to estimate “expected” per capita utilization of each type of health services in each province. In addition, for (3) we combined expected per capita utilization of health services based on (1) and the standardized mortality ratio for people younger than 75 years of age. Data on health care utilization, demographics, socioeconomic status, and health status came from the 2000/2001 Canadian Community Health Survey. Data necessary to calculate the standardized premature mortality ratio for each province in 2001 came from E-STAT and CANSIM. The sample size for all analyses in this study was 111,249.

Results: All three approaches suggested that expected relative per capita utilization for general practitioner, specialist, and hospital services was not uniform across provinces, ranging from -1.07% to 24.31% of the Canadian average utilization. Higher need provinces indicated by all three approaches were Nova Scotia, New Brunswick, and Saskatchewan for general practitioner services, Nova Scotia and New Brunswick for specialist services, and Prince Edward Island, Nova Scotia, New Brunswick, Manitoba, and Saskatchewan for hospital services. Different approaches, however, yielded inconsistent results for Newfoundland and Quebec. For all three types of services, these two provinces had higher than average need based on (3) and lower than average need based on (2).

Conclusions: This study suggests that there may be substantial differences in need for health services across provinces. Different approaches provided different results, and the assessment of needs for health services requires further methodological refinement.

Permalink

Affordability of health insurance: Do assets and net wealth explain the demand for health insurance better than income?

Presenter:

Didem Bernard

Authors:

Didem Bernard, Jessica Banthin, William Encinosa

Chair: TBA; Discussant: TBA Mon June 5, 2006 13:45-15:15 Room 313

Understanding the affordability of coverage is important for evaluating the role of policy in reducing the number of uninsured workers. We study worker health insurance take-up and coverage decisions using data from the Medical Expenditure Panel Survey (MEPS) from 1997 to 2003. Unlike previous studies which control only for current income, we include information on the presence and value of family-level assets such as home ownership, vehicles, savings accounts, stocks, bonds, retirement accounts, as well as liabilities, to estimate the effect of net wealth on insurance purchase decisions. Unlike most studies which focus on “worker” take-up, we also take into account the availability of insurance offers through the spouse’s employer in estimating enrollment decisions of workers and their families.

We estimate worker demand for employer sponsored health insurance as a function of the premium as well as worker, family, employer and plan characteristics. We use two approaches to deal with the lack of premium data for workers who decline coverage. The first approach uses a sub sample in MEPS from 1997-1999 with linked data from the Household Component (HC) and the Insurance Component (IC) which is a survey of employers. Although the HC-IC link sample is not nationally representative, it contains data on the premiums for takers and decliners as well as the availability of choice of health plans, and types of plans offered. The second approach uses simulated premiums from the Insurance Component List Sample. Using the nationally representative sample of employers in the MEPS-IC, we estimate average plan premiums as a function of predictor variables available on both the employer and household surveys, including location (state, MSA), firm size, industry, and plan types offered. We then use this model to predict premiums for workers in the household survey.

Among adults living in families with health insurance offers in 2001 and 2002, 7.6 percent did not take up private insurance. As expected, probability of take up declined with income: 8.6 percent of adults with middle income, 19.9 percent of adults with low income, and 32.4 percent of poor and near poor adults did not take up private insurance. (Bernard and Selden, 2005) Preliminary work based on this sample, suggests that assets and net wealth play a significant role in insurance coverage decisions. Controlling for income, adults who did not take up health insurance were significantly less likely to have assets. For example, among poor and low income adults, the decliners were less likely to own homes (46% vs. 54%), less likely to have cars (78% vs. 86%), less likely to have checking accounts (37% vs. 51%), less likely to have stocks (1% vs. 5%), and less likely to have individual retirement accounts (13% vs. 22%). Research using affordability thresholds based on income has shown that health insurance was affordable to between 25% to 75% of the uninsured in 2000. (Bundorf and Pauly, 2002) Our preliminary results suggest that in explaining health insurance purchase decisions, affordability thresholds based solely on income may be inadequate.

Permalink

Insurance Turnover as an Impediment to Improving Health Care Quality

Presenter:

Mark Votruba

Authors:

Mark Votruba, Randall Cebul, James Rebitzer, Raymond Herschman

Chair: TBA; Discussant: TBA Mon June 5, 2006 13:45-15:15 Room 313

Patients with chronic illnesses (e.g. diabetes, congestive heart failure, coronary heart disease) account for about 75% of medical care spending in the United States. A growing body of evidence suggests that these costs could be reduced significantly if chronic diseases were managed better. Insurers could therefore potentially profit by financing formal disease management programs for policyholders with chronic diseases. However, the financial incentive to do so is undermined by high insurance turnover rates when the returns from such investments accrue in the future or over many years. Based on membership records from a large regional insurer, we find that turnover rates are quite high. Almost 50% of members had discontinued coverage with our insurer after two years following their enrollment date, and fewer than 15% remained after eight years. The high rate of insurance turnover is primarily driven by turnover in employment. We show that turnover rates of this magnitude negate the financial incentive insurers have to finance disease management programs shown (in previous studies) to be effective at reducing costs. While the existence of longer-term insurance contracts could potentially mitigate this problem, a simple model of insurance competition in the employer group market suggests that adverse selection problems hinder the adoption of such contracts. Another interesting finding of this model is that increased insurance competition, while driving down employer group premiums, hinders adoption of cost-reducing disease management programs by increasing turnover. Finally, we show why state mandates for insurance coverage of disease management programs are likely to prove disappointing, as the underlying incentives of insurers are left unchanged.

Permalink

The Effects of Partial-Year Insurance Coverage on the Healthcare Utilization of Low-Income Children

Presenter:

Lindsey Leininger

Authors:

Lindsey Leininger, Willard Manning

Chair: TBA; Discussant: TBA Mon June 5, 2006 13:45-15:15 Room 313

Title: The Effects of Partial-Year Insurance Coverage on the Healthcare Utilization of Low-Income Children

Rationale: Studies on the dynamics of the insurance coverage of children demonstrate that treating insurance as a static measure provides an incomplete representation of the coverage of low-income children. Insurance coverage for this group is especially dynamic throughout the course of a year, with more children spending part of the year uninsured than spending the full year uninsured. The vast majority of research on the effects of insurance coverage on children’s access to care estimates the effects of point-in-time insurance coverage on access and utilization outcomes. With few exceptions, existing research does not examine the relationship between the amount of the year spent with insurance coverage and children’s healthcare utilization.

Objective: This work examines the effects of partial-year insurance coverage on the healthcare utilization of low-income children, accounting for the endogeneity of insurance status.

Data and Methods: The relationship seen in observational studies between health insurance and healthcare utilization is biased by unobservable characteristics that affect both insurance status and healthcare usage. Existing work on the effects of partial-year coverage on utilization measures is observational and does not account for unobserved heterogeneity among insured and uninsured children. This study uses instrumental variable techniques appropriate for limited dependent variables to provide evidence regarding the causal impacts of partial-year insurance coverage on the healthcare utilization of children.

We employ various instruments seen in the existing literature as well as additional instruments suggested by theory to estimate the insurance-utilization relationship for children. Several previous studies have utilized the variation in eligibility across states and years as an instrument with which to identify the relationship of interest. In addition to using this measure, we also use family-level and state-level instruments to estimate the effects of partial-year coverage on utilization measures.

The dependent variables of interest are the number of doctor visits received in the past year, whether a child has a usual source of care, whether a child was admitted into the hospital in the past year, and whether the child visited the emergency room in the past year. A sample of approximately 35,000 low-income children drawn from the 1997, 1999 and 2002 waves of the National Survey of America’s Families (NSAF) is used for the analysis.

Preliminary Results: Children with partial-year coverage have lower rates of healthcare utilization than those with full-year coverage but use more care than children who spend the entire year uninsured. The effect of the number of months of coverage on utilization exhibits a dose-response relationship, with each additional month of coverage lowering the probability of using care. Estimates from instrumental variables specifications are larger than those of ordinary least squares specifications.

Conclusions: This paper provides new evidence on the relationship between partial-year insurance status and the healthcare utilization of children. Its hope is to contribute to the currently small literature on this topic through employing a quasi-experimental design that accounts for non-random selection into insurance status.

Permalink

Generic Scrip Share and the Price of Brand-name Drugs: The Role of Consumer Choice

Presenter:

John Rizzo

Authors:

John A. Rizzo, Richard Zeckhauser

Chair: Avi Dor ; Discussant: Avi Dor Mon June 5, 2006 13:45-15:15 Room 325

Pharmaceutical expenditures have grown rapidly in recent decades, and now total nearly 10% of health care costs. Generic drug utilization has risen substantially alongside, from 19% of scrips in 1984 to 47% in 2001, thus tempering expenditure growth through significant direct dollar savings. However, generic drugs may lead to indirect savings as well if their use reduces the average price of those brand-name drugs that are still purchased. Prior work indicates that brand-name producers do not lower their prices in the face of generic competition, and our study confirms that finding. However, prior work is silent on how the mix of consumer choices between generic and brand-name drugs might affect the average price of those brand-name drugs that are purchased. We use a nationally representative panel of data on drug utilization and costs for the years 1996-2001 to examine how the share of an individual’s prescriptions filled by generics (generic scrip share) affects his average out-of-pocket cost for brand-name drugs, and the net cost paid by the insurer. Our principal finding is that a higher generic scrip share lowers average brand-name prices to consumers, presumably because consumers are more likely to substitute generics when brand-name drugs would cost them more. This effect is substantial: a 10% increase in the consumer’s generic scrip share is associated with a 15.6% decline in the average price paid for brand-name drugs by consumers. This implies that the potential cost savings to consumers from generic substitution are far greater than prior work suggests. In contrast, the percentage reduction in average brand costs to health plans is far smaller, and statistically insignificant.

Permalink

How Does Direct-To-Consumer Advertising Influence the Demand for Mental Health Treatment?

Presenter:

Samuel Zuvekas

Authors:

Chad D. Meyerhoefer, Samuel H. Zuvekas

Chair: Avi Dor ; Discussant: David Bradford Mon June 5, 2006 13:45-15:15 Room 325

Pharmaceutical manufacturers continue to increase their spending on direct-to-consumer (DTC) advertising, with DTC expenditures jumping from $0.6 billion in 1996 to almost $4 billion in 2004 (TNS Media Intelligence). Rosenthal et al (2003) suggest that such advertising is an important driver of rising prescription drug expenditures. These expenditure increases have been especially rapid among several classes of drugs used to treat mental health problems. For example, Zuvekas (2005) found that spending on mental health related prescription drugs increased 20 percent a year in real terms between 1996 and 2001, more than doubling from $6 billion to almost $15 billion. In addition, one-third of the Top 25 drugs ranked by sales are either antidepressants or antipsychotics (NIHCM, 2002). It is unclear how much of the increase in the demand for mental health treatment is due to increased advertising by the pharmaceutical industry. DTC advertising may cause both a shift from traditional forms of treatment to pharmacotherapy, as well as an increase in the aggregate demand for mental health drugs. Indeed, research by Iizuka and Jin (2003) and Iizuka (2004) suggests that increased DTC expenditures have a market-expanding effect. We seek to better understand and quantify the extent to which DTC advertising explains increased demand for mental health treatment and changes in how it is delivered. We derive a model of the joint demand for mental health pharmacotherapy and behavioral therapy, which incorporates the relevant costs influencing consumption decisions, including outof-pocket payments (cost-sharing) for ambulatory services, out-of-pocket prescription drug costs, insurance premiums, and the out-of-pocket costs for non-mental health ambulatory treatment and drugs. We then merge quarterly data on direct-to-consumer advertising for the 100 largest markets obtained from TNS Media Intelligence to data from the 1996-2003 Medical Expenditure Panel Survey (MEPS), a nationally representative survey of the U.S. civilian, noninstitutionalized population. These 100 largest markets account for approximately 86 percent of the television-owning population (80-82 percent of the total population). By exploiting regional market-level variation in DTC advertising for local (spot) television, local (spot) radio, and newspaper media, as well as variation over time in national and local media spending, we are able to identify the effect of advertising levels on the demand for mental health treatment. This is done using two different measures of consumer demand: 1) The number of ambulatory mental health visits and prescription fills; and 2) The level of consumer expenditures on ambulatory treatment and prescription pharmaceuticals. The former is estimated via a Zero-Inflated Ordered Probit specification, and the latter using a two-part model of medical expenditures. Both models deviate from the conventional demand literature by making use of a correlated random effects specification (Chamberlain, 1982) to control for individual-level heterogeneity in preferences and the endogeneity of model regressors. Estimates from these empirical models are used to forecast future advertising-attributable demand in light of advertising trends in the pharmaceutical industry for both public and private payers.

Permalink

Do Prescription Drugs Reduce Hospitalizations and Costs?

Presenter:

Bill Encinosa

Authors:

Bill Encinosa, Didem Bernard, Avi Dor

Chair: Avi Dor ; Discussant: Jian Li Mon June 5, 2006 13:45-15:15 Room 325

As health care costs continue to soar, Medicare, employers, and insurers are now designing pharmaceutical benefit plans to require greater consumer cost sharing. In fact, between 2000 and 2004, copayments for preferred drugs increased 62%, and copayments for non-preferred drugs increased 94%. In most circumstances, economists would conclude that increased patient cost sharing reduces moral hazard and excessive medical consumption, thereby improving social welfare. However, the case of prescription drugs is more complex. Often drugs are associated with preventive efforts to reduce further illness and complications, but the patient might not internalize these long run benefits of drugs. As a result, increased cost-sharing may lead to underuse of drugs. In fact, Dor and Encinosa (NBER, 2004) recently showed that copayment increases reduced patient compliance with anti-diabetic medications. In this paper, we examine the effect of drug noncompliance on patient outcomes and costs. We focus on diabetes since it is one of the most common chronic conditions, with 18.2 million affected in 2004 (6.3% of US population). It is also the leading cause of adult blindness, kidney failure, lower extremity amputations, and a leading cause of heart disease. It is the 6th leading cause of death. The prevalence of diabetes increased by more than 30% over the past 10 years. We examine the impact of anti-diabetic drug purchase compliance on hospitalizations and costs among persons with type 2 diabetes. We use Medstat’s Marketscan claims database for 45 large employers for 2001 and 2002. Since unobservable patient severity may determine both hospitalization and drug compliance, we control for unobservable severity by using GMM and AGLS Probit instrumental variable techniques. Our results show that compliance with drug treatment regimes lead to a lower probability of hospitalizations and to lower health care expenditures. Using the AHRQ HCUP Prevention Quality Indicators, we find that the probability of preventable diabetic hospitalizations under perfect drug compliance is 0.5%, compared to 8.8% under 50% compliance. More generally, under perfect compliance the probability of any hospitalization is 2.8% with an average length of stay of 3.6 days, compared to 38.6% under 50% compliance, with an average length of stay of 6.2 days. Under perfect compliance the probability of having any ER visits is 6.4%, compared to 24.5% under 50% compliance. In all IV models, patients with more severe unobservable severity were more apt to comply with their medications. As expected, the annual prescription drug costs are higher for patients who perfectly comply with drug treatment. The annual drug costs under perfect compliance are $3,683, and $1,727 under 50% compliance. It is not clear a priori whether there would be outpatient cost offsets associated with drug compliance, since getting prescriptions requires physician visits. We find that compliance is associated with more office visits, but fewer ER visits. The expected outpatient expenditures under perfect compliance are $3,116, compared to $592 for 50% compliance. However, the reduction in hospitalizations associated with drug compliance lead to large cost offsets…

Permalink

Cost-effective treatment for adolescents with opioid dependence

Presenter:

Daniel Polsky

Authors:

Daniel Polsky

Chair: William S. Cartwright; Discussant: Kathryn McCollister Mon June 5, 2006 13:45-15:15 Room 326

Optimal treatment strategies for drug dependence is rarely based on analysis of economic outcomes, yet limited financial resources for treatment frequently inhibit greater access to treatment services. A clinical trial in the National Drug Abuse Treatment Clinical Trials Network (CTN) is testing the effectiveness of Buprenorphine/Naloxone-facilitated rehabilitation (Bup/Nal) in opioid dependent adolescents/young adults. Bup/Nal in combination with psychosocial treatment has already been shown to improve treatment outcomes in adults. This new combined therapy offers hope for opiate dependent young people who have been growing in number and have very few prospects for treatment or recovery. Yet the tradeoff between short-term treatment and maintenance is acutely felt by young adults who may obtain better outcomes from maintenance therapy, but face the high costs of a lifetime of maintenance. If economic analyses with relevant and useful outcome measures could be performed on these treatment alternatives, a preferred treatment strategy for these patients could be ascertained. We are conducting a cost-effectiveness analysis to compare the incremental economic outcomes between short term Buprenorphine/Naloxone treatment and maintenance for treating opioid dependence in adolescents and young adults. The EuroQol EQ5D will be used to assess the patient’s preference for their health state. Urine analysis will determine whether subject is drug free at 3, 6, 9, and 12 months. The incremental medical costs (including both substance abuse treatment costs, the costs of other medical therapies, and costs of medical care born by patients) associated with the BUP/NAL intervention will be estimated. The analysis will estimate cost-effectiveness ratios (CER) which will represent the incremental medical costs of BUP/NAL per drugfree addict and per QALY; and estimate net social benefits of BUP/NAL by estimating the value to society of a drug-free addict in terms of reduced morbidity, crime, and improved productivity net of medical costs. Data collection within the clinical trial will involve the integration of instruments to collect counts of medical service use both within and outside the study protocol. A modified client-DATCAP will be used to determine price weights as proxies for social costs to value these medical services. The clinical trial is ongoing but a preliminary cost effectiveness analysis at three months postrandomization will be completed on the first 140 subjects by June 2006.

Permalink

Patient and Societal Utilities For A Spectrum Of Opiate, Cocaine, And Marijuana Problems

Presenter:

Kevin Kraemer

Authors:

Kevin L. Kraemer, Mark S. Roberts, Ihsan Salloum

Chair: William S. Cartwright; Discussant: Kathryn McCollister Mon June 5, 2006 13:45-15:15 Room 326

The lack of empirical data on societal and patient preference-based utility weights for drug-related health states and consequences limits the application of costutility analysis (CUA) to these common problems. To address this issue, we developed and pilot-tested a spectrum of opiate, cocaine, and marijuana related health state descriptions. Using a fractional factorial design, we then presented the health state descriptions to study participants recruited from the general population and from the drug treatment population for completion of Visual Analogue Scale, Time Trade-Off, and Standard Gamble exercises. We will present the results of these assessments and compare the advantages and disadvantages of these direct methods to indirect methods (e.g. EQ-5D, SF-6D) that rely on multi-attribute utility theory. In addition, we will present data on the test-retest reliability, consistency, and discrimination of the direct utility measurements. The results of this project will be useful for quality-adjusted life-years calculation in future CUA of drug abuse treatment programs and for burden of disease estimates in populations.

Permalink

Sensitivity of Preference-Weighted Health-Related Quality of Life Measures and Substance Use Severity

Presenter:

Jeffrey Pyne

Authors:

Jeffrey M. Pyne, Brenda Booth, Shanti Tripathi, Richard Rapp, Michael French, Kathryn McCollister

Chair: William S. Cartwright; Discussant: Kathryn McCollister Mon June 5, 2006 13:45-15:15 Room 326

Cost-effectiveness analyses using generic preference-weighted health-related quality of life (PWHRQL) measures to calculate cost per quality-adjusted life year (QALY) ratios are useful for making decisions about how to most efficiently allocate limited healthcare resources. To date, cost per QALY analyses have not been widely reported for substance abuse interventions, however. One of the first steps towards supporting the validity of cost per QALY analyses for substance use disorder interventions is examining the relationship between substance use severity and PWHRQL. We will present data on this relationship for subjects referred to substance abuse treatment programs in Ohio. The generic PWHRQL measures used in this study include the self-administered Quality of Well-Being (QWB-SA) scale (originally designed for use in cost per QALY analyses) and the standard gamble-weighted Medical Outcomes Study SF-12 (SF-12 SG). The substance use severity measure was the Addictions Severity Index (ASI). Three groups of patients were identified: substance abuse, drug dependence only, and alcohol and drug dependence. Our current sample includes 238 subjects and we expect to add approximately one hundred additional subjects by the time of presentation. Using the N=238 sample, we found significant QWB-SA, SF-12 SG, and ASI subscale differences between the substance dependent and substance abuse groups (p<0.01). The QWB-SA and SF-12 SG scores for the substance dependent groups were similar to those found among chronic physical health disorders. In multivariate analyses predicting the PWHRQL measures and controlling for sociodemographic variables; the ASI medical, psychiatric, and drug use subscales were significant predictors of the QWB-SA (R2=32%) and these predictors plus the ASI family/social subscale were significant predictors of the SF-12 SG (R2=48%). We will also present results of the effect of substance dependence related to specific substances in the presentation. Based on our current findings, it appears that both the QWB-SA and SF-12 SG are sensitive to cross-sectional substance use severity in a treatment-referred sample. Ultimately, cost per QALY analyses can provide an additional metric for communicating the value of substance abuse interventions and will allow for the direct comparison with existing cost per QALY ratios for other physical and mental health interventions.

Permalink

Is the Impact of Managed Care on Hospital Prices Decreasing?

Presenter:

William White

Authors:

William D. White, David Dranove, Richard Lindrooth, Jack Zwanziger

Chair: Frank Sloan; Discussant: Kathleen Carey Mon June 5, 2006 13:45-15:15 Room 332

Rationale: The introduction of managed care was followed by a sharp slow down in the growth of healthcare spending. However, recently expenditures have surged, raising the question of whether the impact of managed care has weakened. Previous research finds a strong inverse relationship between the level of concentration and prices in hospital markets under managed care. Two widely discussed hypotheses about why the effects of managed care may be decreasing are: 1) On the demand side of the market, relaxation of constraints on provider choice by Managed Care Organizations (MCOs) is reducing ability to steer patients and engage in price sensitive shopping to win discounts, weakening the relationship between concentration and price; and 2) on the supply side, downsizing and provider consolidation have led to increased market concentration, enabling providers to exercise growing countervailing market power. Objective: To examine how interactions between demand and supply factors have affected prices for hospital inpatient care for privately insured patients. Methods and Data: This research uses multiple regression analysis and pooled cross section time series data for Florida and California to test hypotheses regarding the effects of changes in concentration and the price sensitivity of shopping on hospital prices. Florida and California are selected for study because of long histories with managed care and data availability. Data are drawn from state and AHA hospital data files, the ARF, and unique data collected by Co-PI Jack Zwanziger on MCO network size. Pricing patterns are compared in five time periods: i) 1983 before the growth of managed care; ii) 1990, after the initial growth of managed care, but early in the process of hospital consolidation; iii) 1995 as MCO began to reduce restrictions on consumer choice of providers; iv) at the end of the 1990s (1999); and v) using the most recent data available (2001). To disentangle the effects of supply and demand side changes, two questions are considered. First, what would have happened to prices in more recent periods if measures of hospital market concentration and/or factors associated with the price sensitivity of shopping had remained fixed at previous levels? Second, have the magnitudes of coefficients associated with measures of concentration and/or the price sensitivity changed? Results: Our initial results indicate that that there was a growing inverse relationship between concentration and price in hospital markets until 1999, but between 1999 and 2001 this relationship plateaued. At the same time, concentration is increasing in many markets, consistent with a reduced ability to win price concessions. We are continuing to explore the possible effects of changes in constraints on consumer choice and the price sensitivity of shopping. Disclosure Information: This research is supported by a grant from the Robert Wood Johnson Foundation.

Permalink

Do the Financial Incentives Linked to Ownership of "Limited-Service" Hospitals Affect Physicians Practice Patterns?

Presenter:

Jean Mitchell

Authors:

Jean M. Mitchell

Chair: Frank Sloan; Discussant: Kevin Volpp Mon June 5, 2006 13:45-15:15 Room 332

Under existing federal law it is illegal for physicians to refer Medicare and Medicaid patients to a health care facility in which the physician has an ownership interest. Some states have enacted similar physician self-referral prohibitions that apply to privately insured patients. The federal law and most state self-referral prohibitions do not apply to “whole” hospitals and ambulatory surgical centers (GAO, 2003). Because hospitals are typically multi-product firms that offer a wide array of services, it was rationalized that any referral made by an individual physician investor of a “whole” hospital would in theory result in only small financial gains for the individual physician owner. Yet, under existing federal law it is illegal for a physician to refer patients to a hospital department in which he/she has an ownership interest (GAO, 2003). The “whole” hospital exception is one of the major factors that spurred the recent emergence and growth of physician-owned “limited-service” or “specialty” hospitals. Since the late 1990s, these facilities have become commonplace in states where certificate-of-need laws do not exist. Such facilities typically specialize in the provision of profitable services, most notably orthopedic, spine, cardiac or surgical procedures (GAO, 2003; Iglehart, 2005). Proponents contend that physician-owned limited-service hospitals are “focused factories”, that is, specialization results in economies of scale and lower production costs. In other words, high volume specialization enhances productivity. Advocates also maintain that because physician owners have direct control over management decisions, quality of care is better and patient satisfaction is higher (MedPAC, 2005). Opponents argue that physician ownership creates an inherent conflict of interest for referring physicians. Physicians receive a professional fee for performing surgery. However, physician owners also have a share of any profits generated from facility fees paid to the hospital. Thus, physician ownership has the potential to influence physicians’ referral behavior and practice patterns. Specifically, critics argue that the financial incentives linked to ownership result in the steering of patients to the physician-owned facility, increased utilization of more profitable services and selection of low-acuity and well-insured patients ( Iglehart, 2005; MedPAC, 2005). They contend that such practices by referring physician investors will undermine the ability of competing community hospitals to provide money-losing services such as emergency medical care. Despite concerns regarding the increasing number of physician-owned limited-service hospitals, nearly all empirical evidence to date employs a case study approach that focuses on the effects of limited-service hospitals on competing community hospitals…

Permalink

Safety Net Activities and Hospital Profitability During the 1990s

Presenter:

Jack Zwanziger

Authors:

Jack Zwanziger, Anil Bamezai

Chair: Frank Sloan; Discussant: Richard C. Lindrooth Mon June 5, 2006 13:45-15:15 Room 332

Background: Safety net (SN) hospitals in the US, also known as “hospitals of the last resort” provide health care services to the uninsured, low income, underinsured, to Medicaid beneficiaries, to patients who are “undesirable”. They have survived through a variety of direct subsidies from public funds and cross-subsidies from private payers. Both were imperiled in the 1990s with budget cuts and price competition. This study examined the financial performance of these hospitals over the 1990-2000 period. Data and Methods: We combined data (revenue, expenses, uncompensated beds size and teaching) from the Medicare Cost Report data sets with a range of hospital characteristics from the American Hospital Association Annual Survey of Hospitals for all urban general acute care hospitals in the US. We created a set of safety net measures including a factor characterizing the socio-economic status of the population living in the hospital’s service area (using Medicare patient origin data and the 1990 and 2000 census), the proportion of Medicaid patients and the uncompensated care burden (the proportion of a hospital’s expenses accounted for by uncompensated care). We then modeled revenue, expenses and profit margin as a function of time varying hospital and market characteristics (outputs, the competitiveness of the hospital market, HMO penetration) year dummies and interactions (with the year dummies) using hospital fixed effects specifications. We also tested for the potential endogeneity of the percent Medicaid and the uncompensated care burden using a 2SLS specification Finally, we estimated a logistic regression for the probability a hospital had a negative profit margin in a given year, as a marker of financial distress. Results: Profit margins were consistently lower at safety net hospitals but the gap between these profit margins and those of non-safety net hospitals did not increase during the 1990s. These differences are primarily due to lower revenue although expenses are somewhat higher as well. The models show essentially the same pattern. Hospitals with higher levels of safety net measures tend to have somewhat lower profits but the gap did not change significantly over the decade. Implications: Hospitals with higher safety net measures have consistently tended to have lower profit margins and as profit margins have slid may be the ones in the greatest distress. There is a need to focus subsidies properly to ensure the survival of the safety net in an environment where budget constraints and price competition are likely to persist and even to increase.

Permalink

Cigarette Advertisements and Smoking Cessation

Presenter:

Don Kenkel

Authors:

Donald Kenkel, Rosemary Avery, Dean Lillard, Adan Mathios

Chair: Sara Markowitz; Discussant: Hope Corman Mon June 5, 2006 13:45-15:15 Room 335

The academic and policy debate on cigarette advertising has mainly focused on whether cigarette advertising encourages children and adolescents to start smoking. Our econometric approach uses microdata to examine a related but neglected question: Does cigarette advertising maintain the size of the cigarette market by discouraging current smokers from quitting? We use variation in exposure to magazine advertisements as a natural or quasi-experiment to identify the influence of cigarette advertising on smokers’ attempts to quit smoking. Our preliminary results indicate that smokers whose reading habits expose them to more cigarette advertisements are less likely to attempt to quit and to successfully quit. The profits from discouraging cessation appear large enough to explain some, but not all, of the industry’s expenditures on advertising.

Permalink

Smoke and/or Drink? The Effect of Alcohol Control on Smoking by Teenagers

Presenter:

Philip Cook

Authors:

Philip Cook, Christopher Carpenter

Chair: Sara Markowitz; Discussant: John Tauras Mon June 5, 2006 13:45-15:15 Room 335

This is an analysis of whether different types of addictive goods are substitutes or complements in consumption. The behaviors considered are cigarette smoking and alcohol consumption by teenagers. Previous research has shown that zero tolerance laws, which set very low legal blood alcohol limits for individuals under the age of twenty-one, have reduced underage drinking. The current paper focuses on the effects of these laws on cigarette smoking by the affected group.

Permalink

An Exploration of the Relationship Between Risky Sexual Behavior and Substance Use by Teenagers and Young Adults

Presenter:

Michael Grossman

Authors:

Michael Grossman, Freddy Siahaan

Chair: Sara Markowitz; Discussant: Sara Markowitz Mon June 5, 2006 13:45-15:15 Room 335

We evaluate the extent to which the relationship between the use of such substances as marijuana and alcohol and various aspects of sexual behavior is causal. That is, does the use of marijuana and alcohol cause young people to initiate sexual intercourse at an earlier age, to be more likely to engage in sexual intercourse in the past month or past year, to be less likely to use condoms or other methods of birth control, and to have had more sexual partners? Establishing a causal effect of substance use on sexual behavior is essential to the design of effective public policies targeted at improving public health by affecting sexual behavior. A simple statistical association between substance use and sexual behavior is not sufficient evidence to determine causality because the observed relationship between substance use and sexual behavior may reflect causality in both directions (i.e., structural endogeneity) and may also reflect the influence of an omitted “third variable” (i.e., statistical endogeneity)—for example, “a thrill-seeking personality.” Thus, to obtain policy relevant causal estimates of the effect of substance use on sexual behavior, it is necessary to use statistical procedures that address these two sources of endogeneity. In previous research economists have used substance use control policies as instruments for use, but these attempts suffer from biases due to weak instruments, especially since the control policies do not vary much within areas over time.
Using panel data from National Longitudinal Survey of Youth 1997 with four observations on each person in the period from 1997 through 2000, we take two new approaches to establish causality. The first is a Granger causality model. The idea here is to see whether past substance use influences current sexual behavior, with past sexual behavior held constant. Conversely, we also regresses current substance use on lagged substance use and lagged sexual behavior. We take first differences to eliminate unobserved individual-specific and time-invariant fixed effects. Since first differences of lagged substance use and lagged risky sex are correlated with the disturbance term, consistent estimates of the first-differenced equation cannot be obtained by ordinary least squares. Therefore, we employ an instrumental variables estimation methodology that has been developed for panel data. In this methodology, second lags and longer lags of levels of substance use and sexual behavior serve as instruments for the lagged first differences. Unlike substance use control policies, these lags have significant explanatory power. Our second approach allows current substance use to have an impact on current sexual behavior. The equations specified in this model omit the lagged dependent variable as a regressor. Once again, first differences are obtained to eliminate fixed effects, and second and longer lags of levels of substance use serve as instruments for the lagged first difference.

Permalink

Effect of Drug Manufacturers' Rebates on Ohio Health Plans Pharmaceutical Expenditures

Presenter:

Enrique Seoane Vazquez

Authors:

Enrique Seoane Vazquez, Rosa Rodriguez-Monguio, Jay Visaria

Chair: Marisa Domino; Discussant: TBA Mon June 5, 2006 10:45-12:15 Room 121

Background: Medicaid federal drug rebates (i.e. retroactive discounts) were implemented in 1991, and Ohio Health Plans (OHP) -the program that manages Ohio Medicaid- supplemental rebates were implemented in April 2003. Manufacturers that sign a federal rebate agreement are eligible for federal Medicaid reimbursement for their products. Manufacturers of originator drugs that negotiate supplemental rebates with OHP are included in Ohio’s Medicaid prefer drug list (i.e. these drugs are covered without prior authorization). Objectives: The objective of the study is to evaluate the effect of manufacturers’ drug rebates on OHP drug expenditures. Material and Methods: The sources of data used in the study included OHP drug claim database, CMS, and other public sources. Drug expenditures trends for 1982-2003 were calculated and projected for 2004-2005. Drug expenditures for 2003 were calculated from the OHP claim database, and estimated rebates were subtracted from the fee for service (FFS) drug reimbursement in order to estimate the net cost of the pharmacy program.

Results: OHP’s FFS drug expenditures reached $1.7 billion in 2003. Originator drugs represented 85.7% and generic drugs 14.3% of the drug expenditures. The average expenditure per claim was $96.74 for originator drugs and $16.13 for generic drugs.

Manufacturers’ rebates provided an estimated reduction in drug expenditures of $407 million, representing 23.9% of FFS drug expenditures. OHP’s rebates included federal rebates (providing a reduction of 20.6% in drug expenditures) and state supplemental rebates (providing a reduction of 3.3%). Medicaid rebates favorable compare with rebates for the largest PBM in the U.S. that represented 8.4% of drug expenditures in 2003.

The average rebate for originator drugs was estimated at 27.33% of expenditures in those drugs, in contrast with the 5.5% estimated rebate for generic drugs. The implementation of the federal rebate generated an 8.8% reduction in the cost per Rx in 1991. Nevertheless, the cost per Rx had increased 41.2% during the three-year period (1988-1990) prior to the rebate, especially in the 1989, when prices went up by 44.6%. The implementation of the state rebate generated an increase of 0.1% in the cost per Rx in 2003. The cost per Rx had increased 34.1% during the three-year period (2000-2003) prior to this rebate, with an increase of 13.4% in 2002.

Conclusions: Ohio Medicaid pharmaceutical expenditures have doubled every five years during the period 1990-2003. Manufacturers’ rebates have not curved the upward trend in OHP drug expenditures. Medicaid federal and state rebates reduce the prices paid by the Medicaid program in relation to the prices listed by pharmaceutical companies. Nevertheless, the rebate, which is based on the AMP, creates an incentive for manufacturers to raise their prices before the rebate implementation.

In spite of the large rebates from utilization of originator drugs, substantial savings would derive from increasing the market share of generic drugs.

Permalink

Is drug coverage a free lunch? Cross-price elasticities and the design of prescription drug benefits

Presenter:

Jian Li

Authors:

Martin Gaynor, Jian Li, William Vogt

Chair: Marisa Domino Mon June 5, 2006 10:45-12:15 Room 121

Recently, many US employers have adopted more stringent prescription drug benefits. In addition, the U.S. will offer prescription drug insurance to approximately 42 million Medicare beneficiaries in 2006. These changes in drug prices can be expected to change not only the consumption of drugs but also the consumption of other medical goods, if those goods are complements or substitutes for drugs.

We use data from Medstat on individual claims and benefit design from 1997-2003 to examine the dynamic structure of demand for health care, focusing specifically on whether prescription drugs are (dynamic) complements or substitutes for inpatient and outpatient care. The Medstat data contain inpatient, outpatient, and pharmaceutical claims for a population of employees of a number of large, US employers. Our analytic sample contains more than 520,000 individuals each followed for at least three years and at most seven.

We study the effect of changing consumers’ copayments for prescription drugs on the quantity demanded of and expenditure on prescription drugs, inpatient care and outpatient care, allowing for effects both in the year of the copayment change and in the year following the change and controlling for changes in inpatient and outpatient deductibles, copayments, etc. Since many of our dependent variables have nonstandard distributions in our main analyses we employ fixed-effect count data models for our quantity measures: outpatient visits, inpatient stays, number of days supply of prescription drugs. We employ fixed-effect Tobit models for our various spending measures. In addition to fixed effects, we use block-bootstrap techniques to correct our standard errors for any serially correlated individual random effects.

Our results show that a one dollar increase in the consumer copayment for drugs has the following effects. In the first year after the change, drug spending falls by $20.6, but the outpatient spending increases by $10.9. In the second year after the change, drug spending continues to fall by $12.8, while outpatient spending rises by $15.3. There are no significance changes in inpatient spending in both years. In sum the total spending decreases by $9.5 in the first year after the price changes and increase by $1.5 in the second.

As is evident, there are both dynamic and instantaneous effects of changes in consumer drug copayments. Moreover, there is clear substitution effect between the consumption of prescription drugs and other medical services. The cost-savings on prescription drugs are largely offset by the increases in outpatient spending. These dynamic and substitution effects should be taken into account when estimating the expected cost of programs like the Medicare prescription drug benefit. The findings from this study also provide useful insights into the optimal design of insurance benefits.

Permalink

The Effect of Insurance on the Demand for Prescription Drugs by the Elderly and Near-Elderly

Presenter:

Merrile Sing

Authors:

Merrile Sing, Edward Miller, Jessica Banthin

Chair: Marisa Domino Mon June 5, 2006 10:45-12:15 Room 121

Rationale: Many people under age 65 have drug coverage through private group health insurance plans or Medicaid. Medicare beneficiaries are less likely to have drug coverage, and some obtain coverage through less generous, individually-purchased private plans. Medicare begins covering outpatient prescription drugs in 2006, when Medicare Part D becomes effective. The institution of Part D will increase access to drugs for some beneficiaries, but some beneficiaries may lose more generous drug coverage as some employer-sponsored retiree health plans discontinue coverage.

Objectives: To provide insight into the possible effects of Medicare Part D, this paper examines how prescription drug expenditures for the elderly and near elderly (those aged 55 to 64) are affected by different types of prescription drug coverage. We include the near-elderly because they will soon become Medicare beneficiaries, and policymakers are also interested in their demand for drugs.

Methods: We examine the effects of drug coverage on drug expenditures for elderly and near-elderly with econometric models estimated with data from the Medical Expenditure Panel Survey (MEPS), a nationally-representative database of the civilian non-institutionalized population in the U.S. Our models control for a wide variety of observed characteristics that differ across insurance groups, such as income, age, gender, race/ethnicity, education and health status. We distinguish between those with relatively generous drug coverage from Medicaid, employer-sponsored plans and Medicare HMOs, those with less generous coverage from individually-purchased plans, and those who have no coverage because they are uninsured or are covered only by Medicare (before 2006). We use two approaches to correct for the endogeneity of insurance status. First, we control for an important source of omitted variables bias by including variables that measure individuals’ attitudes toward risk and health care. Second, we develop several instruments for insurance coverage, such as Medicare HMO penetration rates, various measures of assets, and Food Stamp eligibility. We also use a therapeutic classification scheme to examine the demand for a few large classes of drugs of policy interest, such as statins. Although other studies have examined these issues, none have included the near-elderly in their analysis, used MEPS data, or included variables that measure individuals’ attitudes toward risk.

Preliminary Findings: At a point in time, approximately 70 percent of the near-elderly have prescription drug coverage through a private group health insurance plan or Medicaid. In contrast, approximately one-third of Medicare enrollees have prescription drug coverage through these relatively generous sources. From 1996 to 2002, the average prescription drug expenditure for the near-elderly increased from approximately $590 to $1,215 in constant 2002 dollars. During this same period, prescription drug expenditures increased from approximately $795 to $1,507 for Medicare enrollees.

Permalink

Clinical Trial Participation and Prescription Drug Use

Presenter:

Meredith Kilgore

Authors:

Meredith Kilgore, Dana Goldman

Chair: John Rizzo; Discussant: David Bradford Mon June 5, 2006 10:45-12:15 Room 213

Title: Clinical Trial Participation and Prescription Drug Use

Objective: To estimate the effect of cancer clinical trial participation on prescription drug utilization, drug costs, and out-of-pocket expenditures.

Methods: As part of the Cost of Cancer Treatment Study (CCTS), a national probability sample was drawn of patients participating in cancer clinical trials (n = 781) and a matched cohort of patients not enrolled in trials (n = 595) receiving treatment for the same cancers from the same providers as the trial participants. All CCTS subjects were interviewed about their prescription drug utilization and out-of-pocket drug expenditures. Treatment costs were estimated based on a large pharmacy transactions database. Multivariate regression was used to estimate the effects of trial participation on drug costs and out-of-pocket expenditures. Standard errors were corrected to account for the clustering of observations within trials and institutions. Weights were constructed using both propensity scores and sampling probabilities to correct for selection bias and to reflect the sampling design.

Results: Trial participation was associated with a $131 increase in prescription drug costs over a six-month period prior to the time of the interviews (p < 0.05), but no significant difference was found in out-of-pocket expenditures. These results were robust to a variety of model specifications.

Conclusions: Trial participation is associated with an increase in prescription drug utilization and costs, but these costs do not necessarily impose an economic burden on cancer trial participants.

Permalink

Cost-Effectiveness of Early-Stage Lung Cancer Adjuvant Treatments in Practice Using Instrumental Variable Estimates

Presenter:

John Brooks

Authors:

John Brooks, Elizabeth Chrischilles, Eun Cho, Shari Chen-Hardee, Shane Scott

Chair: John Rizzo; Discussant: James Henderson Mon June 5, 2006 10:45-12:15 Room 213

Rationale: Treatment “effectiveness” is generally defined as the effect of a treatment used in practice. Yet the usual approach in cost-effectiveness analysis is to use treatment efficacy and cost estimates from controlled trials to estimate what should be called “cost-efficacy” ratios. In an earlier paper (Journal of Econometrics,vol (77), 1997, pp 39-64) McClellan and Newhouse proposed using instrumental variable (IV) methods to estimate treatment effectiveness and treatment cost in practice to estimate true cost-effectiveness ratios. We apply their approach here. Our cost-effectiveness ratio estimates are from Medicare’s perspective and can help policy-makers judge whether treatments are over or underused in practice.

Objectives: To estimate cost-effectiveness ratios for adjuvant treatments (first course chemotherapy and radiation therapy after surgery) in practice for patients with non-small cell early-stage lung cancer (ESLC).

Methodology: We used patients with ESLC that had surgery for tumor removal in the SEER-Medicare linked database, 1992-1999 and assessed whether patients had adjuvant chemotherapy and radiation after surgery. Outcomes measured for each patient were three-year survival and total three-year Medicare reimbursements post surgery. Separate but identically specified instrumental variable models were estimated for each outcome variable. Both treatment variables (chemotherapy and radiation) were specified separately in each model. Instruments included local area (50-mile radius around patient residence) treatment rates over the period 1992-1999 and local patient concentration indices among essential providers (medical oncologists, radiation treatment centers). All models controlled for tumor stage, grade and site, patient age, gender, and race, and the socioeconomic characteristics of the patient’s zipcode. Models were re-estimated with different instrument specifications to assess the robustness of the estimates. Cost-effectiveness ratios were estimated as the ratio of the treatment-specific IV three-year cost estimate to the treatment-specific IV three-year survival estimate.

Results: Instruments had statistically significant effects on the choice of each treatment. Using IV analysis, adjuvant chemotherapy had positive and statistically significant effects on three-year survival and Medicare reimbursements. The statistical significance of these estimates varied somewhat with the instrument specification but the effects remained positive. No statistically significant relationships were found between adjuvant radiation and three-year survival or three-year Medicare reimbursements. Using different instrument specifications we found three-year adjuvant chemotherapy cost-effectiveness ratios ranging from $37,854 to $113,829.

Conclusions: Using the rationale within McClellan and Newhouse, our estimates suggest that increasing the rate of adjuvant chemotherapy for ESLC patients would cost Medicare between $37,854 to $113,829 for each additional three years of life saved. If a year of human life is valued around $75,000, these results suggeste that adjuvant chemotherapy was underused for ESLC patients. As radiation therapy did not increase three year survival this suggests that radiation treatment was potentially overused.

Disclosure Information: This research was supported by a grant from the National Cancer Institute.

Permalink

The Decision to Conduct a Head-to-Head Comparative Trial: A Game-Theoretic Analysis

Presenter:

Edward Mansley

Authors:

Edward Mansley, Elamin Elbasha, Steven Teutsch, Marc Berger

Chair: John Rizzo; Discussant: John Rizzo Mon June 5, 2006 10:45-12:15 Room 213

Recent Medicare legislation calls on the Agency for Healthcare Research and Quality to conduct research related to the comparative effectiveness of health care items and services, including prescription drugs. This reinforces earlier calls by government officials for “practical clinical trials” involving clinically relevant treatment alternatives. Using a game theoretic model, we explore the decision of pharmaceutical companies to conduct such “head-to-head” comparative trials. The model suggests that an important factor affecting this decision is the potential loss in market share and profits following a result of inferiority or comparability. This “hidden cost” is higher for the Market Leader than the Market Follower, making it less likely that the Leader will choose to conduct a trial. The model also suggests that in a full-information environment it will never be the case that both firms choose to conduct such a trial. Furthermore, if market shares and the probability of proving superiority are similar for both firms, it is quite possible that neither firm will choose to conduct a trial. Finally, our results indicate that incentives that offset the direct cost of a trial can prevent a “no-trial equilibrium”, even when both firms face the possibility of an inferior outcome.

Permalink

Effect of antidepressants on employment of women with HIV in the HAART era: 1995-2004

Presenter:

Omar Galarraga

Authors:

Omar Galarraga

Chair: Randall Ellis; Discussant: Cynthia Perry Mon June 5, 2006 10:45-12:15 Room 225

Background: Since the mid-1990s highly active antiretroviral treatment (HAART) has extended the lives, and the ability to work for individuals living with HIV in the United States. During the same period, antidepressants have become some of the most widely prescribed drugs. Employment of persons living with HIV depends on several factors including physical and mental health status, as well as local area labor market characteristics.

Data: This paper investigates the effect of antidepressants on the chances of being employed, using data from the Women’s Interagency HIV Study (WIHS). Started in 1995, the WIHS is a prospective cohort of 3,768 participants to study comprehensively the effects of HIV on women. About 80 percent of the participants are from racial minority groups in Bronx and Brooklyn (New York), Chicago (Illinois), Los Angeles and San Francisco (California), and Washington, D.C. Participant follow-up visits are scheduled every six months.

Rationale: There may be two channels for depression treatment to affect the labor market outcomes of persons living with HIV. First, a direct channel by which antidepressants may reduce depression symptoms and thus may improve the likelihood of being employed. Second, an indirect channel by which antidepressants may improve adherence to HAART regime, improving physical health, and then lead to better employment outcomes.

Methods: The methodological challenge is that more depressed persons are more likely to receive treatment, but they are also more likely to be unemployed. This paper uses the econometric method of instrumental variables to identify the treatment effect of antidepressants on employment. The main instrument is a measure of the generosity of the Medicaid program at the State level in terms of coverage of a particular class of antidepressants: selective serotonin reuptake inhibitors (SSRIs). The instrument helps to predict treatment choice independently of individual confounders, and thus reduce the bias in estimation.

Results: The empirical results from linear, non-linear, and auto-regressive models (under fixed and random effects assumptions) for panel data suggest that antidepressants may have a positive effect on the employment probability for the women’s sample. Conditional on receiving HAART, and controlling for individual and local area labor market characteristics, women who use antidepressants may have at least a seven-percentage-point higher probability of being employed than women who do not use antidepressants.

Policy Implications: The findings suggest that increasing efforts to improve screening, diagnosis, and treatment of depression in specific high-risk groups may be warranted not only for the physical and mental health benefits, but also as an avenue to increase employment. Furthermore, since the majority of the HIV care is financed by public funds, expanded access to depression treatment through the Medicaid and Ryan White CARE Act formularies may potentially improve labor market outcomes of persons living with HIV.

Permalink

The Influence of Mental Health Status on Employment for HIV-Positive Individuals

Presenter:

Stephanie Bernell

Authors:

Stephanie Bernell

Chair: Randall Ellis; Discussant: Didem Bernard Mon June 5, 2006 10:45-12:15 Room 225

Objective: In this study, we evaluate whether persons dually diagnosed as HIV-positive and having mental health problems have differential labor market outcomes than those without mental health problems. The underlying assumption of this study is that mental health can influence employment directly or indirectly through a decreased likelihood of using recommended treatment regimens that could stave off the onset of AIDS.

Empirical Model: In this paper, we specify a bivariate probit model of (1) the decision to work, which we represent as y1, and (2) the decision to use HAART, which we represent as y2, allowing for correlation (ρ) between the two decisions. The underlying assumption of this model is that two related decisions are considered by the same individual. Normally, for a two equation probit model it is assumed that the errors in the two equations (ε1 and ε2) are distributed normally N(0,1) and that the two errors are independent of one another (Cov(ε1,ε2) = 0). When there is reason to believe that the error terms are possibly related, then the bivariate model is appropriate. For example, there may be an underlying factor such as risk preferences that may affect HAART use and employment.

The dependent variables in the model are the following: (1) y1, which equals 1 if employed and 0 otherwise, and (2) y2, which equals 1 if the individual uses HAART and 0 otherwise. The independent variables in the model include demographic measures, exposure route, health measures, mental health status, and usual source of care.

Data: Data from the HIV Cost and Services Utilization Study (HCSUS) were used in this study. HCSUS is a survey of adults with known HIV infection who made at least one visit to a non-military, non-prison medical provider other than an emergency department in the United States during first two months of 1996. HCSUS collected data on 2864 individuals in 180 clinics, hospitals and private practices in 28 urban areas and 24 clusters of rural counties. For this analysis, we used data from the first follow-up interview to categorize mental health and data from the second follow-up interview to obtain data on HAART use, employment status and other demographic information. The appropriate sampling weights were used in all study analyses.

Results: Preliminary evidence demonstrates an interrelationship between HAART use, employment and mental health. More than half (57.7%) of those unemployed at the time of the baseline survey and about one-third (33%) of those employed at the time of the baseline survey had mental health composite score less than the median. However, almost 50% of those using HAART at the time of the baseline survey and 50% of those not using HAART at the time of the baseline survey had a mental health score below the median. The bivariate probit model results indicate that HIV-positive individuals with poor mental health have a reduced probability of working, but do not have generally have a reduced probability of using HAART. Individual with general anxiety disorder do, however, have an increased probability of using HAART.

Permalink

Labor Market Outcomes of Persons with Mental Disorders

Presenter:

Marjorie Baldwin

Authors:

Marjorie Baldwin, Steven Marcus

Chair: Randall Ellis; Discussant: Richard Scheffler Mon June 5, 2006 10:45-12:15 Room 225

Background: Mental disorders are common and associated with high costs and substantial levels of work disability. The poor labor market outcomes of persons with mental disorders are surely attributed, in part, to the association between mental disorders and deficits in higher-order social and cognitive skills that are important determinants of workplace productivity, yet pervasive stigma against mental illness may also play a role. This is the first study to apply the analytical tools economists have used to study other disadvantaged groups in the labor market specifically to persons with mental disorders, a group subject to some of the most intense stigma and poorest outcomes. Objectives: We use data from the 1999 Medical Expenditure Panel Survey to analyze employment outcomes among persons with mental disorders. The objectives are to compare the mean wages and employment rates of persons with and without mental disorders; and estimate the extent to which the poor outcomes for persons with mental disorders may be attributed to stigma.

Methods: We decompose wage and employment differentials between workers with and without mental disorders into an explained component, attributed to differences in the average productivity of the two groups, and an unexplained component, some part of which may be attributed to the effects of stigma against mental illness. The decompositions are based on wage and employment functions estimated separately for persons with and without mental disorders. Key explanatory variables include workers’ demographic and human capital characteristics, non-wage incomes, job characteristics, and functional limitations (physical/cognitive/social). We provide estimates of productivity-adjusted employment and wage differentials for persons with mental disorders overall, and for subgroups of persons with mood, anxiety, adjustment, or psychotic disorders.

Results: Employment rates are 85% for persons with no disorder vs. 70% for persons with mental disorders. The differential is explained, in part, by differences in health status, non-wage incomes, and gender, but approximately 20% remains unexplained. Among the employed, mean hourly wages are $14.85 for persons with no disorder vs. $14.09 for persons with mental disorders. About 70% of the wage differential is explained by differences in functional limitations, gender, part-time employment, and health insurance coverage. The remaining 30% is unexplained. The results reveal distinctly different patterns of labor market outcomes across the subgroups, consistent with a severity gradient of mental disorders such that persons with adjustment disorders experience the most favorable outcomes, while persons with psychotic disorders experience the least favorable. Implications: The project is timely because of the employment mandates of the ADA, and because recent advances in medications for serious mental disorders enable many persons with these illnesses to function at a level where they are able to work. The results identify productivity-related characteristics that contribute to the low wages and employment rates of persons with mental disorders, but also suggest that stigma may play an important role. By helping to identify barriers that impede success in the labor market, the results have important implications for mental health services directed toward improving employment outcomes among persons with mental disorders.

Permalink

Time Series Issues in the Estimation of the Rational Addiction using Micro Data

Presenter:

Brian Ferguson

Authors:

Brian Ferguson, Audrey Laporte, Frank Windmeijer

Chair: Susan Ettner; Discussant: Bill Crown Mon June 5, 2006 10:45-12:15 Room 226

Increased availability of micro panel data sets make time series econometric issues relevant in the estimation of RA models. In particular, the theoretical RA model possesses a saddlepoint equilibrium, meaning that the time series properties of the data are represented by a second order difference equation with one stable and one unstable root. The unstable root is expected to introduce nonstationarity problems akin to those found in the unit root macroeconometrics literature, and is also expected to dominate, and bias, the estimation of the coefficients of the RA equation as the time dimension of the data grows. In this paper we use Monte Carlo simulation methodology to illustrate the effects of the presence of an unstable root on the estimation of the coefficients of an RA model, and in particular on the estimated coefficients on lead and lag consumption. We show that as the time dimension of the data increases, the unstable root comes to dominate the estimation of RA-type models, to the point where the RA form can be decisively rejected even when it is the true data generating process. We also show that the larger the unstable root, the smaller the length of the time series required for the unstable root to come to dominate the estimation. Conclusions: The earlier applied RA literature paid little attention to time series issues, probably because the lack of individual level panel data sets made those issues moot. As, increasingly, micro panel data becomes available, researchers will need to take account of the implications of the dynamic structure of the RA model for its estimation, and in particular for the likelihood of significant bias to the estimated coefficients.

Permalink

Health Expenditure Estimation and Functional Form: Applications of Generalized Gamma and Extended General Linear Models

Presenter:

Edward Miller

Authors:

Edward Miller, Steven Hill

Chair: Susan Ettner; Discussant: Joel Hay Mon June 5, 2006 10:45-12:15 Room 226

Rationale: Health care expenditure regressions are used in a wide variety of economic analyses including risk adjustment and program and treatment evaluations. Two recent articles have demonstrated that generalized gamma models with heteroskedasticity (GGM-het) and extended general linear models (EGLM) provide flexible approaches to deal with a variety of data problems commonly encountered in expenditure estimation. To date, however, there have been few empirical applications of these models to expenditures.

Objective: We use nationally representative data from the first six panels of the U.S. Medical Expenditure Panel Survey (MEPS) to compare the bias and predictive accuracy of GGM-het and EGLM models with other regression models in a cross-validation study design.

Methodology: We estimate models of prescription drug, ambulatory and total health care expenditures conditional on having any expenditure. Models are estimated separately for the elderly and other privately insured adults. Since expenditure distributions vary by type of service and population, the appropriate functional form is also likely to vary. In estimating expenditures, we focus on two recently developed modeling approaches that flexibly accommodate skewness, kurtosis, heteroskedasticity and other data problems. The GGM-het model, proposed by Manning, Basu, and Mullahy (2005), uses a log-link like many standard GLM models. However, the GGM-het model is more flexible than standard models because the generalized gamma distribution has a scale parameter and two shape parameters and variance is explicitly modeled as a function of explanatory variables. In the EGLM model, proposed by Basu and Rathouz (2005), the link function is not specified prior to estimation. Instead, both the link and variance functions are simultaneously estimated along with the coefficients.

Our models use socioeconomic characteristics and condition information from the first year of each MEPS panel to predict expenditures in the second year. We use a split-sample cross validation design to compare results from GGM-het, EGLM, log OLS with heteroskedasticity (log-het), linear OLS, Poisson and Gamma models. We use the validation sample to test for over-fitting and to examine predictive ratios and mean prediction errors in the entire sample, in the tails of the distribution and for persons with chronic conditions.

Results: In our preliminary analysis we focused on total expenditures and estimated all types of models except EGLM. We found that the expenditure distribution for the elderly was more kurtotic than the distribution for other adults and the distributions varied in the extent of heteroskedasticity beyond simple functions of the mean. Overall, the GGM-het and log-het models fit the data for privately insured adults very well. However, none of our models was clearly superior for the elderly.

Conclusions: Our preliminary analysis confirms that GGM-het models are robust to a wide variety of common data problems. For some distributions, however, an even more flexible estimator, such as the EGLM model, may be required.

Permalink

Properties of treatment effect estimates in switching regime models: impacts of sample size and strength and dirtiness of instruments.

Presenter:

Henry (Joe) Henk

Authors:

Henry (Joe) Henk, William Crown, David Vanness

Chair: Susan Ettner; Discussant: Michael French Mon June 5, 2006 10:45-12:15 Room 226

Instrumental variables methods are widely used to adjust for unobserved variables that are correlated with both treatment selection and patient outcomes. However, there is a growing literature showing that such models often perform more poorly than ordinary least squares regression. The bias and precision of instrumental variables estimators have been shown to be sensitive to comparatively weak correlations of the instrumental variable with the residuals of the outcome equation. The bias and precision of instrumental variables estimators have also been found to be sensitive to the strength of the instruments. Larger sample sizes cannot overcome problems of bias but are helpful in improving precision.

To date, the examination of the properties of IV estimates has been confined to the single equation case. Single equation estimates of treatment effects do not account for the interactions of treatment cohorts with observed covariates. Nor do they account for differences in the distributions of patient characteristics between treatment cohorts. Switching regime models offer the potential to control for both of these factors-potentially enabling a more precise estimate of treatment effects. However, we know little about the sensitivity of treatment effects estimates obtained from switching regime models to issues such as the “dirtiness” of the IV, strength of the IV, or sample size. In this paper, we conduct simulation analysis to examine these properties for IV estimation of switching regime models when these factors are varied.

We present the results of a simulation study to demonstrate estimation error caused from correlation between the residual and the instrumental variable when IV estimation of switching regime models are employed to estimate of regression parameters in presence of an endogenous regressor.

Permalink

Do Residential Segregation and Economic Inequality Explain Race Disparities in Health Services Use?

Presenter:

Darrell Gaskin

Authors:

Darrell Gaskin

Chair: James Marton; Discussant: Patricia Ryan Mon June 5, 2006 10:45-12:15 Room 235

Authors: Darrell J. Gaskin (dgaskin@jhsph.edu), Adrian Price (adrian.price@ccaix.jsums.edu), Dwayne T. Brandon (dbrandon@jhsph.edu) Thomas A. LaVeist (tlaveist@jshph.edu)

Title: Does Residential Segregation and Economic Inequality Explain Race Disparities in Health Services Use?

Rationale: Nationally, disparities in health care utilization between African Americans and whites are well documented. Even after controlling for health status and health insurance coverage, African Americans still have lower rates of use of medical services. However, national analyses are unable to adequately control for the impact of racial segregation on geographic access and often compare African American samples that are disproportionately low income to White samples that have higher percentages of moderate/middle and high income respondents.

Objective: To determine if racial disparities in health care use can be attributed to residential segregation and economic inequality, we compared disparities in health care use in a national sample of adults to a sample of adults from a low income racially integrated community.

Methods: We estimated models of health care use using data from Medical Expenditure Panel Survey (MEPS), a national sample of adults conducted by the Agency for Health Research and Quality, and data from the Exploring Health Disparities in Integrated Communities Project (EHDIC). The EHDIC data is a 2003 survey of residents from a low income urban community in a northeastern state. This community has equal numbers of white non-Hispanic and African American residents. Census data shows that racial groups have similar median income and educational attainment. We conducted analyses of the full MEPS sample and a matched subsample, created by matching MEPS respondents to EHDIC respondents by race, gender, income and educational attainment. The MEPS and the EHDIC databases contain information on health services utilization: number of health care visits, emergency room use, usual source of care and whether the respondent has a regular doctor. For each dataset, we estimated the effects of race on each utilization measure and controlling for general health status, presence of chronic conditions, age, gender, marital status, insurance status, employment status, income and education.

Finding: We found differences in the race disparities in health care use across the datasets but some similarities too. In the MEPS data, African Americans were 15% less likely to have a health care visit compared to whites. However, in the EHDIC data, African Americans were 40% more likely to have a health care visits than whites. However, in the MEPS and EHDIC, African Americans were about 10% less likely to have a regular doctor compared to whites. In MEPS data, African Americans were 25% less likely to multiple medical visits compared to whites. In the EHDIC data, African Americans were 20% less likely to have multiple medical visits.

Conclusion: Segregation and income inequality may explain substantial proportion of the observed race disparities in the initiation of health care use. However, differences in the amount of services use and whether individual have a regular doctor are probably due to factors related to the physician-patient interactions and individuals’ experiences in the health care system.

Permalink

Disparities in Late-Stage Cancer Diagnoses in 1990 and 2000: A Comparative Study of Three American Cities

Presenter:

Janis Barry

Authors:

Janis Barry, Nancy Breen

Chair: James Marton; Discussant: Sylvia Brand Mon June 5, 2006 10:45-12:15 Room 235

Rationale: Previously, we reported that women living in poor, medically underserved, inner-city areas were at significantly increased risk of late-stage cancer diagnosis in 1990. Breast and cervical screening are widely used in the US to detect early stage cancer. We found that in neighborhoods where the odds of late-stage diagnoses are high, women face access barriers to screening.

Objectives: We use data on inner-city markets and residents to model economic changes over a 10 year period. Other researchers found independent effects of neighborhood unemployment, racial composition, environmental threats, and low socioeconomic status on a range of health outcomes. Analysis of the supply of health services, including the location of hospitals, clinics and drugstores, has been largely absent from the neighborhoods and health literature. We will examine these variables in our follow-up analysis.

Methodology: For both the initial 1989-1990 study and the 1999-2000 follow-up, we used variables from Surveillance, Epidemiology, and End Results (SEER) cancer registry data (race, age, marital status and city of residence) and linked them with 1990 and 2000 Census data at the tract level. We linked a measure of medical underservice (MUA) developed by the Health Resources and Services Administration (HRSA), and a measure of extreme poverty, defined as a Census tract in which more than 40% of the population lives in poverty. We analyze Atlanta, Detroit, and San Francisco, three major metropolitan areas with both high and low-income tracts and racially and ethnically diverse populations.

Results: In 1989-1990 we found that residence in an extremely poor or medically underserved area increased the likelihood of a late-stage cancer diagnosis. City of residence, and race/ethnicity also were major determining factors. Detroit represented the clearest case of market failure in the delivery of cancer screening services. Our preliminary studies for the 1999-2000 period show the same set of factors is associated with late-stage diagnosis. Despite a significant decline in late-stage diagnoses between 1990 and 2000; and a dramatic (24%) decline in the nationwide number of residents in extremely poor neighborhoods, our findings confirm that health care markets are underserving economically disadvantaged locations. Further, many poor neighborhoods without the federal MUA designation in 1990 still had not received it in 2000.

Conclusion: Our results identify areas in which women are not receiving preventive cancer services. In these areas, the combination of low income and social disadvantage act to reduce the health care choices available to women. We evaluate whether the same neighborhoods we examined in 1990 have improved their ability to provide residents access to services in 2000. For 2000, we analyze the location of health care delivery factors in the three inner cities, especially how spatial and racial divides foster late-stage diagnoses in the clearest case of market failure, Detroit.

Permalink

Location decisions of mammography facilities: racial/ethnic disparities across zip-codes

Presenter:

Jonathan Sunshine

Authors:

Mythreyi Bhargavan, Jonathan Sunshine

Chair: James Marton; Discussant: TBA Mon June 5, 2006 10:45-12:15 Room 235

Rationale: Among women age 40 and older, minority women have been found to have mammograms at a lower rate than white women. One potential cause may be that minority women live in neighborhoods with no mammography facilities in proximity to them, or because facilities in their neighborhoods are inadequately committed to patient service.

Study Objectives: To measure the effect of characteristics of a zip-code (and the surrounding area) on two sets of outcomes: (a) whether mammography facilities locate in the zip-code, and (b) commitment to patient service of mammography facilities in it. In particular, to measure if high percentage of minority population influences facility location decisions, above and beyond socio-economic factors.

Data: Data on demographic and socio-economic characteristics of zip-codes are obtained from the US Census. The list of mammography facilities and their zip-codes is obtained from the American College of Radiology (ACR) and Food and Drug Administration (FDA) databases of accredited mammography facilities. Patient-service commitment was measured using a pilot phone survey of 105 mammography facilities.

Methods:
The unit of analysis is the zip-code, referred to here as the target zip-code. Target zip-codes are categorized as high-white, high-black, high-Hispanic, and high-Asian based on the race and ethnicity of the largest sub-population in the zip-code. Quantitative outcomes of interest for the target zip-code are (i) presence of a mammography facility, (ii) the number of facilities per 100,000 women age 40 and older, and (iii) distance to nearest facility if the target zip-code has none. The explanatory variables are socio-economic characteristics of the target and surrounding zip-codes. Outcome (i) will be estimated using simple logistic regression and (ii) and (iii) with linear and log-linear regression.

In addition, we will use logistic regression to compare mammography facilities in high-minority and high-white zip-codes in terms qualitative factors, such as whether the facility accepts Medicaid and charity patients, language skills and minority composition of front office and technologist staff, types of outreach activities, etc.

Preliminary Results: We find no large differences in the minority composition of the population (31% vs. 30% minority respectively) between zip codes with and without mammography facilities. However, controlling for zip-code socio-economics and population density, high black and Hispanic zip-codes are much less likely to have a mammography facility in them than predominantly white zip-codes (for example, 10% increase in the black percentage of the population decreases the odds of a facility in the zip-code by 14%). There is no significant effect of percent of minority populations on the number of facilities per 100,000 women age 40 and older.

Facilities in high minority zip-codes seem to be more committed to patient service than those in high-white areas, for example, one-half of the facilities surveyed in high-Hispanic and three-fifths of those in high-black zip-codes use posters and flyers for outreach into the community to educate women about mammograms, whereas less than one-third of the facilities in the high-white zip-codes do.

Conclusion:
High-minority zip-codes are less likely to have mammography facilities, but facilities in high-minority zip-codes may be more committed to patient care than those in high-white zip-codes.

Permalink

The Effect of Health Insurance Characteristics on Outpatient Mental Health Care and Substance Abuse Treatment Utilization among Privately-Insured Employees and their Dependents

Presenter:

Laura Dunlap

Authors:

Laura Dunlap, Edward Norton, Gary Zarkin

Chair: Michael Grossman; Discussant: Adetokunbo Oluwole Mon June 5, 2006 10:45-12:15 Room 309

Rationale: Although many studies exist that have examined the effect of private health insurance on demand for medical care, few studies have examined this relationship for mental health care or substance abuse treatment demand. Those studies that have examined MH/SA treatment demand have often focused on health insurance per se rather than examining specific effects of MH/SA plan coverage. Furthermore, many of these studies use data limited to service-using individuals and are unable to examine the effect of health insurance on any use of MH/SA care as well as level of use among service users.

Objective: The objective of this study is to estimate the effect of specific MH/SA health insurance characteristics on outpatient utilization of mental health and substance use treatment services for privately-insured employees and their dependents.

Methodology: We use a two-part model to estimate the effect of MH/SA health insurance characteristics on the likelihood of any use of outpatient MH or SA services and, conditional on use, the number of outpatient days. Our models include 4 variables representing MH/SA health insurance coverage, including three different measures of coinsurance rates and whether the health plan required precertification by the company’s employee assistance program (EAP) prior to MH/SA service use. Any use is modeled using a random-effects logit model and days of use is modeled using a random-effects negative binomial model. Both models are estimated using an instrumental variable approach to deal with potential endogeneity of the health insurance characteristics. Data used are private insurance enrollment and claims data for 1997-1998 from MEDSTAT’s Marketscan® database.

Results: We find that the outpatient coinsurance rate for in-network MH/SA is positively and significantly associated with any outpatient MH use, although the estimated marginal effect is quite small. Contrary to our expectations, employees’ number of outpatient MH care days received decreases as the in-network coinsurance rate increases. But again, this effect is extremely small. Among health insurance characteristics, the strongest predictor of MH use is the EAP precertification requirement which is negatively associated with MH use. Our analysis of substance use treatment finds little significant association between MH/SA health insurance variables and outpatient SA treatment utilization.

Conclusions: Our findings suggest that employees and their dependents do respond to expected out-of-pocket expenses for outpatient MH care, but this response is very small. Furthermore, MH/SA health characteristics appear to have little or no effect on SA treatment utilization. Finally, our results indicate that the role of an EAP is not straightforward. Rather than facilitating treatment access, EAP precertification may create an obstacle to treatment and discourage utilization. However, it is also possible that EAP precertification may decrease utilization through the formal health care system by providing some MH/SA services. Individuals with milder conditions may receive an adequate dose of services through the EAP and, therefore, not need additional services.

Permalink

Health Care Costs Associated with Autism among Privately Insured Children

Presenter:

Douglas Leslie

Authors:

Douglas Leslie, Andres Martin

Chair: Michael Grossman; Discussant: Laura Dunlap Mon June 5, 2006 10:45-12:15 Room 309

Rationale: Costs associated with the treatment of autism and other autism-spectrum disorders (ASDs) are largely unknown, but are likely considerable. If public and private health insurance programs are not designed so that coverage for individuals with ASDs is appropriate, access to care for these patients may be compromised.

Objectives: The objective of this paper is to compute health care cost estimates for children with autism, and to compare the costs associated with autism to those of other mental health disorders among children.

Methodology: Data for the study come from the MarketScan database, which includes private health insurance claims data for over 2.5 million covered lives across the U.S. Data for calendar years 2000 through 2003 were included in the analyses. Individuals aged 17 and under with a diagnosis of autism (ICD-9 code 299.0x) or any ASD (ICD-9 codes 299.0x, 299.1x or 299.8x) were identified, and costs of all health care services were calculated. In addition to total health care costs, costs associated with inpatient, outpatient and prescription drug care were also computed. Costs were also computed for children with other mental health disorders and compared with those for autism. Costs were adjusted for inflation using the medical care component of the consumer price index.

Results: The average total annual cost per autism patient ranged from $2,468 in 2000 to $3,562 in 2002, and among patients with any ASD ranged from $2,211 in 2000 to $3,133 in 2003. On average, approximately 3.5% of patients with autism or any ASD had an inpatient stay during the year. Most patients received outpatient or pharmacy care (96% and 83% on average, respectively). Inpatient costs among hospitalized autism patients averaged $9,927 per treated patient, but varied considerably over the period. Average outpatient and pharmacy costs for autism patients were more stable over time, and averaged $1,533 and $1,414 per treated patient over the period, respectively. In general, costs associated with autism or ASDs increased over the period. Costs associated with autism or ASDs were generally higher than those for other mental health disorders, although outpatient costs per treated patient were generally lower. The lower outpatient costs may be due to the fact that many patients with ASDs receive services through the school system, whereas patients with other mental health disorders may not.

Conclusions: These data suggest that ASDs are among the more costly mental health disorders in privately insured children, and are increasing over time. Given that the prevalence of ASDs is also increasing over time, it is clear that autism is an expensive disorder. It is important to note that the data presented in this study reflect only paid health care claims. To the extent that health insurance benefits for patients with autism are more limited than they are for other mental health conditions, as some studies suggest, these results likely underestimate the costs associated with ASDs relative to other disorders.

Permalink

Did Medicaid/SCHIP Crowd-Out Private Insurance Coverage among U.S. Low-Income Children?: A Multilevel Approach

Presenter:

Adetokunbo Oluwole

Authors:

Adetokunbo Oluwole, Dennis Shea

Chair: Michael Grossman; Discussant: TBA Mon June 5, 2006 10:45-12:15 Room 309

Authors: Adetokunbo Oluwole (axo129@psu.edu) and Dennis Shea (dgs4@psu.edu)

Rationale: The implementation of the Medicaid/SCHIP, which began shortly after the major welfare reform of 1996, represented a more recent effort to expand health insurance among lower income populations in the U.S. Previous studies have shown that earlier Medicaid expansions significantly increased public program participation and have reduced the number of uninsured individuals. However, empirical evidence indicates that the reduction in the number of uninsured has been offset by declines in private insurance coverage. Because the Medicaid/SCHIP expansions targeted a different population than earlier expansions and took a different form, a detailed study of these more recent efforts is needed.

Objectives: The objective of this paper is to measure the extent of crowd-out of private insurance coverage, among American low-income children, that is attributable to Medicaid/SCHIP expansion.

Methodology: We combine event history analysis and multilevel modeling techniques to analyze data from the 1996 panel of the Survey of Income and Program Participation (SIPP), which covers the period 1996 to 2000. The combined approach will estimate discrete-time logistic regression models. Events to be modeled are the following transitions in insurance coverage: (1) private coverage to Medicaid; (2) private to uninsured; (3) uninsured to private coverage; (4) uninsured to Medicaid; (5) Medicaid to private; and (6) Medicaid to uninsured. In addition to addressing the question of whether or not children have particular types of health insurance coverage, event history analysis also incorporates the timing of such coverage and transitions. Multilevel modeling accounts for cross-state variations in coverage transition rates as well as corrects for within-group correlation and non-constant error terms. State-level data from other sources such as the Area Resource Files, Regional Economic Information Services, and the Centers for Medicare and Medicaid Services will be appended to the SIPP dataset. The sample for the study consists of unmarried children nineteen or younger, and living in families with incomes at or below 300 percent of the federal poverty level (FPL). The analytic sample is expected to comprise about 8,000 low-income children.

Expected Results and Relevance: This study will provide some insight into how health insurance coverage policies might be directed to have maximum positive effect on the targeted (uninsured) population. This is especially important if there is some evidence of crowd-out of private insurance. Preliminary analyses show that the conditional probability of transition from private insurance to Medicaid/SCHIP started to increase shortly after states began to implement their Medicaid/SCHIP expansion programs.

Conclusions: Given the controversial estimates of crowd-out in previous studies, this study’s use of combined techniques of event history and multilevel modeling will provide more accurate and updated estimates using the 1996 panel of the SIPP dataset, which no other study has analyzed. In addition to personal and family characteristics, the models estimated in this study also control for some state- and program-specific factors, which many previous studies failed to account for.

Permalink

Involuntary Psychiatric Commitment and the Incidence of Suicide: An Economic Analysis

Presenter:

Martin Zelder

Authors:

Martin Zelder

Chair: David Cutler; Discussant: Alison Cuellar Mon June 5, 2006 10:45-12:15 Room 313

State laws permitting involuntary psychiatric commitment of those who constitute a risk to themselves, such as suicide attempters, are presumably intended to decrease suicide rates. Whether such laws have this intended effect has, however, not been previously investigated. Hence, this paper examines the relationship of involuntary commitment laws to state-level suicide rates drawn from the CDC’s Injury Mortality Reports data for 1999-2002. The empirical analysis is grounded in a discrete-choice expected-utility-maximization model in which an individual chooses whether or not to attempt suicide, given that the ‘success’ of an attempt is uncertain and that unsuccessful attempts are associated with involuntary psychiatric commitment. Within this model, the comparative-static impact of an increase in the term of involuntary commitment comprises two opposing effects. One effect is a standard deterrent effect: a longer commitment term reduces utility for those who attempt and fail, thus discouraging attempts and reducing the probability of a ‘successful’ suicide. The other, opposing effect entails escalation in the intensity of attempt made: to reduce the chance of ‘failure’ (remaining alive), an individual faced with a longer commitment term will increase the lethality of his attempt, thereby increasing the probability of ‘success’. Consequently, the impact of an increase in the commitment term is ambiguous—the suicide rate will fall or rise, respectively, depending on whether the deterrent effect or the escalation effect dominates. To test whether longer commitment terms discourage or encourage suicides, I have compiled data on state commitment laws, focusing in particular on the statutory maximum commitment term allowed (for an initial term of commitment). This data is then merged with state-level gender- and age-specific suicide rates obtained from the CDC’s WISQARS database. Other covariates, in addition to the commitment term measure and standard demographics, include indices of binge drinking, mental distress, disability, religiosity, and gun control. Potential endogeneity of commitment term laws is partially addressed by including the state’s lagged suicide rate (for 1960) as a regressor. Preliminary analysis of the 2002 data for a variety of age-gender cells finds dominant deterrent effects for women aged 65-69 and men aged 55-64, but a dominant escalation effect for men aged 70-79. This sort of heterogeneity is predicted by the model, in that the deterrent effect is predicted to be relatively small for the oldest men (who have the highest reported probabilities of success). This analysis will be expanded for the entire period, 1999-2002, for which WISQARS permits age-adjusted suicide rates to be generated. Also, data on commitment rates will be collected in order to reflect the true expected penalty posed by commitment. Finally, the issue of legal endogeneity will be explored via instrumental variables and longitudinal analysis extending well beyond the period of the WISQARS data.

Permalink

Antidepressants and Youth: Healing or Harmful?

Presenter:

Sara Markowitz

Authors:

Sara Markowitz, Alison Evans Cuellar

Chair: David Cutler; Discussant: Judy Shinogle Mon June 5, 2006 10:45-12:15 Room 313

Since 1987 some important changes have taken place in the development of antidepressants, including the introduction of several selective serotonin reuptake inhibitors and other new medications that reduce the likelihood of overdose and offer fewer negative side effects. There has been a tremendous growth in the use of the new antidepressants, and while these new drugs are safe and effective for treating adults with depression, there is growing concern about the risks of prescribing antidepressants to children. Recent reports link the use of antidepressant drugs to an increased risk of suicidal behaviors in teenagers. In the fall of 2004, the U.S. Food and Drug Administration made a controversial decision directing manufacturers of antidepressant drugs to include warning labels about the increased risk of suicidal thinking and behavior. This paper seeks to examine the relationship between antidepressants and suicide among youth. They main hypothesis to be tested is whether variations in the number of prescriptions written for antidepressants can explain differences in completed and attempted suicide rates at the Metropolitan Statistical Area (MSA) level. Information on prescriptions will be obtained from the National Ambulatory Medical Care Survey. This individual level data will be aggregated to the MSA level in order to make inferences about the trends in prescription practices. Linear regression techniques will be used to correlate the prescription data with attempted and completed suicide rates among youth under the age of 19. A host of state and area level factors will be included in order to account for potentially confounding effects of other variables. Results from this paper will contribute to the academic debate on the harmfulness of prescribing antidepressants to youth.

Permalink

Suicide and the Evolution of the Market for Anti-Depressants: A Test of Rational Choice Theory

Presenter:

Dean Lillard

Authors:

Dean R. Lillard

Chair: David Cutler; Discussant: Pinka Chatterji Mon June 5, 2006 10:45-12:15 Room 313

In early work, Hammermesh and Soss (1974) use the rational choice framework of economics to ask whether patterns in suicides can be explained by models of utility maximization. Despite interesting findings, the empirical literature on the economics of suicides has received little attention. Recently, however, Marcotte (2003) extended the role suicide plays by examining not successful suicides but both attempted and successful suicides. He first observes an empirical regularity that income of suicide attempters is higher than people who consider but do not actually attempt to commit suicide. He then incorporates this finding into a utility maximization model by positing state dependent utility functions. In this paper, I take a different approach to investigate whether suicide attempts and successes change in response to the cost of mental states of being. In particular, I take advantage of the fact that, since 1958, new pharmacological breakthroughs have led to the development of new drugs that treat depression. Three waves of drug development have occurred over this time period, each successive development yielding more effective treatments with fewer side effects. As shown by Berndt and coauthors (Berndt, Busch, and Frank 2001, Berndt, Frank, and McGuire 1997, Frank, Busch, and Berndt 1998) the market for these drugs grew phenomenally over the 1990s. In this empirical analysis I investigate how the probability of suicide attempts and completed suicides vary with the availability and cost of anti-depressant drugs. I use data from the National Comorbidity Survey (used also by Marcotte), the National Health Interview Surveys, and data on aggregate rates of suicides by state. To these individual and aggregate data I merge data on the number and type of anti-depressant drugs on the market in each year, the average list price of those drugs, and the flow of advertising for those drugs in 27 nationally distributed consumer magazines. Data on availability and list prices are taken from the Red Book from 1959-2004. Data on advertising of these products come from a print magazine advertisement archive collected and compiled at Cornell University. I then investigate whether the probability of attempted or successful suicide varies with these factors. I also investigate whether the same patterns are present in aggregate (state level) suicide rate data. The estimation strategy is to treat the arrival of the anti-depressant drugs as an event that treats individuals born in different years exogenously. I empirically compare decisions to attempt/complete suicide across age, race, income, and demographic characteristics.

Permalink

Use of instrument variables in the presence of heterogeneity and self-selection: An application in breast cancer patients

Presenter:

Anirban Basu

Authors:

Anirban Basu, James J. Heckman, Salvador Navarro-Lozano

Chair: Bruce Stuart; Discussant: Partha Deb Mon June 5, 2006 10:45-12:15 Room 325

Background: Instrumental variables (IV) are widely used in health economics literature to adjust for hidden selection biases in observational studies and estimate causal treatments effects. However, less attention is paid to the proper use of instrumental variables if treatment effects are heterogeneous across subjects and more importantly, if individuals select treatments based on expected idiosyncratic gains or losses from treatments. In the context of this interaction of heterogeneity and self-selection, we highlight the role conventional instrumental variable analysis and alternative approaches using instrumental variables in estimating causal treatment effects on 5-year costs in breast cancer patients. Given the similarities in 5-year mortality rates between breast conserving surgery with radiation therapy (BCSRT) and mastectomy (MST), estimation of treatment effects on costs is important in evaluating the cost-effectiveness of alternative treatments. Data: We use the Center for Medicare and Medicaid Services national claims database of a 5% random sample of all Medicare beneficiaries. The data were collected as part of the Outcomes and Preferences in Older Women Nationwide Survey (OPTIONS) project, and was used by other researchers to study costs of breast cancer patients under alternative treatments. Outcome: Average Treatment Effect (ATE) on 5-year direct costs. Treatments: BCSRT versus MST Alternative Methods: Traditional covariate adjustments, traditional IV analysis, control function approach using ordinary least-squares regressions. Results: We find that covariate adjustments estimates a treatment effect of $12,829 (se=1,639). Using two instruments, the treatment effect is estimated to be $32,136 ($15,005). Using one instrument at a time and also the propensity score as an instrument give estimates of treatment effects ranging from $8,421 ($8,828) to $44, 921 ($21,790), where some of these estimates are significantly different from each other. Employing a newly proposed test, we find evidence of self-selection based on idiosyncratic differences in outcomes between treatments. This implies the treatment effects estimated by traditional IV analysis may not be consistent estimates of ATE. We therefore estimate the marginal treatment effects (MTEs) using the control function approach with the polynomials of propensity scores. However, in this application, we find that it is not possible to estimate the MTE over its full support; hence ATE cannot be estimated with this data. Any attempt to estimate ATE with these data would require extrapolation of the control function beyond the range of observed support for the propensity scores. Consequently, depending on the function form of the control function we get estimates of ATE ranging from -$46,698 ($69,571) to $88,727 ($34,435). Conclusions: Estimation of causal ATE with instrumental variables faces several limitations. Traditional IV analysis do not provide consistent estimates of ATE if heterogeneity in treatment effects exist and patients self-select into treatment based on idiosyncratic effects. Marginal treatment effects can be estimated and used to overcome this limitation of traditional IV analysis. However, one may still fail to estimate an ATE, if the data do not provide full support for the propensity scores. Consequently, development and application of instrument-free semi-parametric methods to estimate causal treatment effects can be extremely valuable.

Permalink

More Ado About Two: Endogenous Switching, Sample Selection, Endogenous Treatment Effects and the Modified Two-Part Model

Presenter:

Joseph Terza

Authors:

Joseph V. Terza

Chair: Bruce Stuart; Discussant: Willard Manning Mon June 5, 2006 10:45-12:15 Room 325

Applications and theoretical assessments of the two-part model (2PM) of Duan et al. (1983) abound in the health economics, econometrics and biostatistics literatures. Mullahy (1998) shows that the conventional 2PM specification and estimator will likely be biased when used as a basis for policy analysis via estimated marginal effects or elasticities derived from the conditional mean function. At issue here is the fact that the conventional 2PM is estimated under log transformation of the outcome variable, and that so-called retransformation to express the model in untransformed conditional mean form, often requires that additional structure be imposed on the model. Mullahy (1998) shows that a modified 2PM (M2PM) which obviates retransformation is, therefore, more appropriate for policy analytic purposes. In the present paper, we derive a version of the M2PM that accommodates endogenous switching or one of its two more common incarnations – sample selection or endogenous treatment effects. There are numerous contexts in which such a model is applicable. For example, Welch et al. (1984) and Parente and Evans (1998) apply standard methods for dealing with endogenous switching to the conventional 2PM in the context of estimating the impact of health insurance on utilization and expenditure. As a by-product of the present research, we show that these rote applications of extant methods to the standard 2PM formulation will be biased. We review the 2PM and M2PM approaches and introduce the new estimator in the context of the estimation of average treatment effects. Details of the model and estimator pertaining the sample selection and full endogenous switching applications are also given. The performance of the estimator is then evaluated in the sample selection context using simulated data. We assess the accuracy and efficiency of the estimator for policy analysis relative to more conventional approaches (Heckman, 1976; Terza, 1998) which are biased in the M2PM context. The method is applied to data from the 2002 National Survey on Drug Use and Health in an analysis of the effect of drug abuse on earnings.

Permalink

Health Care Use and Endogenous Insurance Decisions of the Elderly: An Econometric Analysis of HRS Panel Data Using Latent Factor Models

Presenter:

Pravin Trivedi

Authors:

Fei Liu, Pravin K. Trivedi

Chair: Bruce Stuart; Discussant: William Greene Mon June 5, 2006 10:45-12:15 Room 325

With the growth of the elderly population in the United States, a major concern is the coverage of their health insurance plans. The elderly may acquire private insurance to supplement their Medicare benefits because the program does not cover commonly used services such as outpatient prescription drugs and nursing home service. These observations motivate this paper,which fills the gaps in research on the private insurance selection of elderly and the associated impacts on health care utilizations. A series of questions arise when we research this area: What are the causes and consequences of enrolling in supplementary insurance plans? Does unobserved heterogeneity, such as personal risk attitudes and life style, play a role in insurance decision and associated health use? Supplementary plans include Medigap plans and other supplemental policies, which can cover specific benefits such as long-term care. It is not hard to find that insurance enrollment decision is dependent with health care use. On one hand, these two variables are influenced by observable information, such as health status, income etc. On the other hand, latent factors, or unobservable heterogeneity, such as attitudes toward risks and life style, may also have a substantial impact on both enrollment decision and utilization. Thus the econometric modeling of the effect of insurance on health care utilization faces three challenges. First, to control for the endogeneity, we use maximum simulated likelihood estimation with latent factors. Second, the dependence between two decisions causes a simultaneity problem. The simultaneity is corrected by jointly modeling insurance enrollment decision and health care utilization. Third, this paper analyzes a five-wave panel data from Health and Retirement Study (HRS). The analysis of longitudinal data presents a challenge both to methodology and empirical application. This paper applies the joint maximum simulated likelihood estimation approach within a panel framework. We use the original HRS, which started in 1992 and surveyed subsequently every two years. However, Wave 1 (1992) questions ask about health care utilizations in the last 12 months, while questions in subsequent waves ask about the period since the previous interview or the last 2 years. In order to obtain consistent measurement of medical care utilization, we use data from wave 2 to 6 and restrict the attention to individuals who have Medicare. The final data consists of 4,541 individuals with 11,775 observations. This is an unbalanced panel because some individuals, who appear in the initial waves but not in all subsequent waves, do so due to death. Although follow-up response rates in HRS are high, attrition due to death for the elderly sample is unavoidable. And new people join HRS survey later mainly because of family reconstruction.

Permalink

Panel

Presenter:

Carol Medlin

Authors:

Chair: TBA Mon June 5, 2006 10:45-12:15 Room 326

Permalink

Panel

Presenter:

Ethan Yeh

Authors:

Chair: TBA Mon June 5, 2006 10:45-12:15 Room 326

Permalink

Panel

Presenter:

Julian Jamison

Authors:

Chair: TBA Mon June 5, 2006 10:45-12:15 Room 326

Permalink

Economic projections of heath care spending and the demand for physicians

Presenter:

Mark Henk

Authors:

Mark Henk

Chair: Richard Cooper Mon June 5, 2006 10:45-12:15 Room 332

Permalink

Expenditures necessary to increase the production of physicians within the US

Presenter:

Gail Wilensky

Authors:

Gail Wilensky

Chair: Richard Cooper Mon June 5, 2006 10:45-12:15 Room 332

Permalink

Impact of physician shortages on the delivery of services

Presenter:

Richard Cooper

Authors:

Richard Cooper

Chair: Richard Cooper Mon June 5, 2006 10:45-12:15 Room 332

Permalink

Economic impact of medical migration from developing to developed countries

Presenter:

Marko Vujicic

Authors:

Marko Vujicic

Chair: Richard Cooper Mon June 5, 2006 10:45-12:15 Room 332

Permalink

The Fragile Foundations of Regional Scientific Advantage: The Impact of US Stem Cell Policy on the Geography of Scientific Discovery

Presenter:

Scott Stern

Authors:

Jeffrey L. Furman, Fiona Murray, Scott Stern

Chair: Ernst R. Berndt Mon June 5, 2006 10:45-12:15 Room 335

This paper evaluates the impact of restrictions on embryonic stem cell use in US Federally funded scientific research on the geography of scientific discovery. In order for localized knowledge spillovers to be translated into sustained scientific leadership, researchers in close proximity to an original discovery must be able to exploit that discovery more rapidly and more intensively than distant researchers. Local researchers must be able to take scientific advantage of a discovery more quickly than their more distant competitors are able to catch up. This paper exploits an exogenous shock to the process of step-by-step scientific discovery to assess the sensitivity of regional scientific agglomeration to a temporary shift in the knowledge production process. Specifically, this paper examines the impact of the Bush Administration’s policy of limiting the scope of Federally funded human embryonic stem cell research to experiments using a already existing stem cell lines. Over the past several years, research into the biological foundations of stem cells has been described by biologists as one of the most promising areas of scientific progress, with rapid advances using both embryonic and non-embryonic stem cells, human as well as non-human sources. Moreover, at least in the initial period after the key discoveries of the 1990s, stem cell research has tended to be geographically localized, with a small number of locations and institutions accounting for a very large fraction of the overall discoveries. In August, 2001, the Bush Administration enacted a policy that placed a subtle but substantive restriction on the freedom of Federally funded researchers; it limited experiments using human embryonic stem cells and conducted with Federal funds to only a small number of possible stem cell lines that had been developed prior to the date of the policy change. While researchers were free to use these specific stem cell lines or to seek alternative funding sources, the policy seems to have placed significant restrictions on academic researchers dependent on Federal funding, and that adapting to the policy required a period of adjustment and exploration. Observers have suggested that this unexpected delay in the activity of those at the scientific frontier provided an opportunity for less well-positioned researchers to catch up and for equally well-positioned researchers to forge ahead…

Permalink

Does Financing Have a Real Effect on Biotech Drug Development?

Presenter:

Sean Nicholson

Authors:

Patricia Danzon, Andrew Metrick, Sean Nicholson

Chair: Ernst R. Berndt Mon June 5, 2006 10:45-12:15 Room 335

In the 1990s the U.S. government instituted three policies that simplified the approval process and/or shortened the review time for drugs targeting life-threatening illnesses and have unmet medical needs, thereby increasing the expected profit of developing specific types of drugs. This paper examines whether the accelerated review (1991), fast track (1997), and priority review policies (1997) increased innovation by biotech and pharmaceutical firms in the targeted disease areas, as measured by pre-clinical research activity, human/clinical trials, and approved compounds. We also examine whether companies responded to the policies differently based on the availability of financing and their prior therapeutic focus. We use the PharmaProjects data base of drugs in development between 1989 and 2004, and apply a propensity score method to predict the types of compounds that would have received accelerated review, fast track, or priority review status prior to the implementation of the three policies. We predict the propensity to receive a special status based on attributes of the compound and the number of therapies on the market that are approved to treat a particular condition.

Permalink

Biotechnology Drugs, Traditional Pharmaceuticals and International Price Controls

Presenter:

John Calfee

Authors:

John E. Calfee, Elizabeth DuPre, Mario Villarreal

Chair: Ernst R. Berndt Mon June 5, 2006 10:45-12:15 Room 335

In recent years the U.S. share of worldwide pharmaceutical expenditures has increased from about one-third to one-half, partly because of the imposition of price controls in all advanced economies other than the U.S. This has raised questions about the rewards to new drug development, the bulk of which is performed by profit-seeking private firms. At the same time, the advent of biotechnology has generated new drugs that in comparison to traditional pharmaceuticals are more tightly targeted at biological mechanisms, less susceptible to competition, and possibly less susceptible to price controls. Using a data based that includes prices and volumes of individual drugs (both biotech and traditional) across six industrialized nations, this paper explores the impact of biologically targeted drugs on foreign price controls and foreign adoption of new drugs. We also suggest implications for pharmaceutical reimbursement and costs in the U.S.

Permalink

Exploring Behavioral/Economic Interventions to Improve Health Outcomes among Adolescents in Poor Countries

Presenter:

Cynthia Perry

Authors:

Cynthia Perry

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Rationale: While the medical literature has previously shown an association between maternal depression and poor pediatric asthma outcomes in children, it has not yet addressed the causal question of whether treating maternal depression leads to an improvement in child health management in the case of a chronic disease such as asthma. There are in fact many reasons (e.g., poor social networks) why we might find maternal depression in the same households with poor child health management, and most of these would not be expected to change with the treatment of depression.

Objectives: The objective of this paper is to compare asthmatic children whose mothers are offered treatment for depression with asthmatic children whose mothers are not offered treatment for depression to determine whether there is evidence of improvements in child health management following the offer of treatment.

Methodology: This project uses Florida Medicaid Claims data. A concern with these data is that mothers are not randomized into a depression diagnosis. In particular, child health utilization can positively affect the probability that maternal depression is identified by a physician. Because this could bias the results of a direct pre-post analysis, I use an instrumental variables technique that uses differential propensities to diagnose depression among primary care physicians to predict whether a child’s mother is diagnosed.

Results: OLS results suggest no improvement in child health management with the offer of depression treatment. But these results likely confound women who are diagnosed for depression precisely because their children have high health care utilization with women who are diagnosed for reasons unrelated to their child’s health. The high health care utilization (that is unrelated to maternal depression) of the former group can swamp the possible improvements of the latter group. Using the instrumental variables strategy to predict which mothers are likely to be diagnosed with depression, I find a reduction in ER visits for asthma care in the post-diagnosis period. I also estimate an overall reduction in total asthma spending in the post-diagnosis period.

Conclusion: These results suggest that in the case of mothers of asthmatic children, improving the ability of primary care physicians to detect depression could yield a cost offsets from better asthma management. Further research is necessary to determine whether similar benefits are present in the case of other chronic childhood diseases.

Permalink

Diffusion of prescription drugs for Alzheimer's disease (AD) among older adults

Presenter:

Chandrakala Ganesh

Authors:

Chandrakala Ganesh, Dennis Shea

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Rationale: AD is projected to affect 14 million Americans by 2050. Technological advances have led to the approval of five drugs to treat AD. While the demand for these drugs has been rising rapidly since their introduction there is little nationally representative research available on individuals who use them. Findings from this study would help in integrating aspects of access, quality and costs of health care services such as prescription drugs for elderly with Alzheimer’s disease.

Objectives: This paper has three objectives: (1) to examine the trends in diffusion of AD drugs among community residing Medicare beneficiaries from 1993 to 2002, (2) to determine the racial, ethnic and socioeconomic disparities in diffusion of AD drugs, (3) to identify determinants of current and projected future costs of AD drugs.

Methodology: The paper uses data on community dwelling beneficiaries from the Medicare Current Beneficiary Survey, years 1993 to 2002, and data on health care systems from the 2003 Area Resource File. Analysis procedures would be conducted on three samples: beneficiaries with only a claims diagnosis of AD, beneficiaries with a claims diagnosis or a survey report of AD, and beneficiaries with a claims diagnosis of AD, survey report of AD, or other dementia diagnosis. This categorization would help in determining how the results would change based on varying sample definitions. Analyses procedures will include conducting basic descriptive analysis on the numbers and types of drugs that are used by elderly with AD. Hazard models will be used to estimate the trends and disparities in diffusion of AD drugs. Finally, costs would be estimated by a generalized linear model using gamma regression functions.

Results: Preliminary studies of the 2001 MCBS data show that 3.5 percent of the community-dwelling Medicare beneficiaries suffered from AD. In 2001, 31 percent of Medicare beneficiaries used AD drugs, compared to 1.4 percent in year 1999. In general, beneficiaries who were white, lived in urban areas, had higher education, with drug coverage had greater utilization rates of AD drugs than those who were black, in rural areas, with lower education, without drug coverage. In 2001, the mean annual total drug and out-of-pocket expenditures of beneficiaries using AD drugs were $2317 and $850 respectively, versus $1749 and $663 for beneficiaries not using AD drugs.

Conclusion: The introduction of several new drugs for the treatment of AD makes it an ideal candidate for the study of diffusion. Preliminary studies suggest that new drug technology for AD has a significant impact on health care utilization and health care costs

Permalink

Impact of supplemental insurance on access, expenditures and health: Evidence from Europe & the U.S.

Presenter:

Jennifer Schultz

Authors:

Jennifer Schultz, Bedassa Tadesse

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Context: Health insurance reform has been attempted many times in the U.S. and it will continue to be a heavily debated issue. Thus, it will be important to learn from the experiences of countries with social or tax-financed health insurance systems experimenting with different payment arrangements to help inform policy decisions in the U.S. on the appropriate mix of public and private health services and health insurance. To date, there have been a limited number of studies analyzing the impact of supplemental health insurance and out of pocket expenditures on health care utilization and health status in countries outside of the U.S.

Objectives: To study the purchase of supplementary insurance and evaluate health care expenditures, health care utilization, ex post static moral hazard and health status of older Europeans from 10 countries with social or tax-financed health insurance and varying private payment arrangements. To compare results to the experiences of Medicare beneficiaries in the U.S.

Methodology: We investigate our study objectives using data from the 2003 Survey of Health, Aging, and Retirement in Europe, which is a representative sample of the non-institutionalized population aged 50 and older in each country (Austria, Denmark, France, Germany, Greece, Netherlands, Italy, Spain, Sweden, & Switzerland). We model the probability of purchasing supplemental health insurance as a function of socioeconomic and health variables. We also model the number of outpatient visits (over one year) and total out-of-pocket health expenditures as a function of supplemental health insurance, characteristics of health care systems in each country, and individual characteristics. To address potential adverse selection (endogeneity of insurance) we rely on extensive health-related control variables (e.g. chronic conditions, diseases, smoking status, ADLs, mental health, parental health, self-reported health status) and multiple regression strategies (instrumental variables) to reduce the bias of unobserved heterogeneity. The existence of moral hazard is investigated by analyzing the relationship between the use of physician services and the price of care, conditional on health insurance characteristics, health status and other control variables.

Results: In all of these countries, individuals face some out of pocket payments, even though most are living in countries with public insurance. The two exceptions are Germany and the Netherlands, where most are privately insured (based on income thresholds). Only Switzerland has a private, mandatory insurance system financed through premiums. Out of pocket payments range from 11% (Germany) to 42% (Italy) of health expenditures, the majority going toward outpatient care and medications. The percent of survey respondents facing positive out of pocket expenditures ranges from 36.8% (France) to 88.7% (Sweden). 9.1% (Germany) to 86% (France) of the population purchases supplementary insurance. Deductibles, coinsurance, and copayments vary across countries. In most countries, the proportion of individuals with positive out of pocket expenditures is not significantly different between those who have supplementary insurance and those who do not. Early results from the multivariate models suggest that supplemental insurance has a significant effect on the utilization of physician services. Results from other models are pending.

Permalink

The effect of health care expenditures on sickness absence

Presenter:

David Granlund

Authors:

David Granlund

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Rationale: Previously, the effects of individuals’ health status on sickness absence have been estimated. From a public policy perspective it is also helpful to know how a policy variable like public expenditures on health care can influence the absence rate. The paper also relates to the still open question of whether increased expenditure on health care on an aggregated level actually does improve the health statues of the population.

Objective: The objective of this paper is to estimate the impact of public health care expenditures on absence from work due to sickness or disability.

Methodology: Using panel data on the Swedish municipalities during the period of 1993-2003, the effect of public health care expenditures on the sickness absence in the municipalities is estimated. Demographic characteristics and educational levels are included as proxies for health. The model also incorporates proxies for working conditions and economic incentives to remain at work. Finally, municipality and year specific fixed effects are included to take account of time-invariant heterogeneity between the municipalities and changes over time common to all municipalities in Sweden. An instrumental variable method is used to avoid the potential problem of endogeneity. The main sources of data are Statistics Sweden, The Swedish Social Insurance Agency and The Swedish Association of Local Authorities and Regions.

Results: Preliminary results indicate that the public expenditures on health care have a significant negative effect on the absence days. According to the estimate, an increase of the operating cost on health care with 1000 SEK (~$125 USD) per inhabitant reduces the absence with approximately 0.6 days per insured. Since there are roughly two inhabitants per insured, this indicates that the cost of avoiding one day of absence by increasing operating cost on health care accounts to approximately 3 300 SEK (~$413 USD). This can be compared with the GNP per working day in Sweden, which is roughly 3 000 SEK (~$375 USD).

Conclusions: The results from this paper indicate that public expenditures on health care have an effect on the number of absence days. These results give support for a positive causality between health care expenditures and the health of the population.

Permalink

The Effect of Managed Care on the Diffusion of Psychotropic Medications

Presenter:

Marisa Domino

Authors:

Marisa Domino

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Background: The last decade has witnessed unprecedented growth in new technologies in mental health treatment. The availability of new pharmacological agents has had an enormous impact on the treatment of mental health disorders, enabling recipients of these medications to experience relief of many symptoms and improve their levels of functioning and quality of life. The diffusion of these new behavioral health technologies, or rate at which these products have spread through the market, has been very uneven. Some psychotropic medications, such as Selective Serotonin Reuptake Inhibitors (SSRIs), now considered first line treatments for depression, have diffused very quickly, while others, such as venlafaxine, also a treatment for depression with equal efficacy have not had much success developing a substantial market share. In order to advance models of best practice, it is important to understand the factors that underlie these varying rates of diffusion. Differences in adoptions and diffusion rates of psychotropic medications across insurance settings, geographic regions, or subpopulations defined by age, gender, or racial or ethic groups has important implications for the quality of care received by persons with mental illnesses. As the tools of managed care increasingly penetrate behavioral health insurance, the diffusion of new psychotropic products into existing treatment options is likely to be strongly affected. The direction of this effect is ambiguous due to the complexity and differing incentives inherent in these tools.

Objectives: The purpose of this paper is to examine whether managed care penetration is associated with different rates of psychotropic drug diffusion in fee-for-service Medicaid programs.

Methodology: Data on all psychotropic medications reimbursed through fee-for-service Medicaid programs in almost all states (n=49) in the US were obtained from 1991-2003. Prescriptions were converted to daily dose units in order to pool information from medication classes. Two of the major classes of psychotropic medications, antidepressants and antipsychotics were examined separately. State-level fixed effect regressions were run on the dependent variable of daily dose units, with state-level Medicaid managed care penetration as the explanatory variable of interest. Diffusion curves were also modeled using a classic logit framework as well as with hyperbolic secant function.

Results: Results indicate a negative association between the managed care penetration level in a state and the number of daily dose units of psychotropic medication filled.

Conclusion: Although managed care emphasizes outpatient treatment over more costly inpatient care, the tools used by many managed care programs, including prior authorization and drug formularies may delay the receipt of treatments and these effects may be spilling over to non-managed settings. Further research to examine this effect is warranted.

Permalink

The Empirical Relationship between Community Social Capital and the Demand for Cigarettes

Presenter:

Timothy Brown

Authors:

Timothy Brown

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Authors: Timothy T. Brown (tbpetris@berkeley.edu), Richard M. Scheffler (rscheff@berkeley.edu), Sukyong Seo (sue_seo@berkeley.edu), and Mary Reed (maryreed@berkeley.edu). Nicholas C. Petris Center on Health Care Markets and Consumer Welfare, School of Public Health, University of California at Berkeley.

Title: The Empirical Relationship between Community Social Capital and the Demand for Cigarettes.

Rationale. Community social capital has been theorized to positively affect individual health behaviors through its positive effects on information flows and societal health norms. Smoking behavior is potentially one of these health behaviors.

Objectives: The objective of this paper is to estimate the association of community social capital with the demand for cigarettes. This association should be strongest in areas where there is a larger concentration of groups with strong non-smoking norms, such as religious groups.

Methodology: We estimate a two-part demand function focusing on the association of community-level social capital (CSC) and cigarette smoking. We use a new validated measure of community-level social capital, the Petris Social Capital Index, which is based on an actual ecological measure of CSC rather than on an aggregation of individual survey responses. We measure CSC at the level of the Metropolitan Statistical Area (MSA) in order to capture the effects of CSC in both the areas of an individual’s residence and employment. We link our measure of CSC to individual-level data on 38,807 adults in 48 MSAs with a population of 1 million or more for the years 1998-2000. We also split our measure of CSC into overall CSC and the percentage of CSC attributable to religious groups. We control for sex, age, race/ethnicity, marital status, education, cigarette prices, potential smuggling, income, nonsmoking regulations, and MSA-level fixed effects. Participation in smoking is estimated using a linear probability model and the number of cigarettes smoked is estimated using a negative binomial model.

Results: We find that the percentage of CSC attributable to religious groups is inversely related to the number of cigarettes demanded by smokers, but is not related to smoking participation.

Conclusions: Community social capital is a potentially powerful inhibiter of the smoking behavior of current smokers.

Disclosure information: This research was supported by the Centers for Disease Control and Prevention, Atlanta, GA through Cooperative Agreement Number U48/cCU 909706. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention. The authors are employees or students of the University of California at Berkeley and have no conflict of interest with regard to this project.

Permalink

The Impact of Health on Academic Performance: New Evidence Using Genetic Markers

Presenter:

Steven Lehrer

Authors:

Steven Lehrer, Weili Ding, J. Niels Rosenquist

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

The causal effect of health on education outcomes remains an open question since numerous unobserved factors may explain the correlation between health and education. Using a very rich data set that tracks 900 students through high school and an instrumental variables strategy we estimate the causal impact of obesity, ADHD and depression on adolescent academic performance by exploiting variation in genetic markers. Our data contains direct measures of four genes, whose strong associations with the health outcomes considered in our study are well established in the scientific literature. We also account for endogenous health behaviors such as regular smoking that may influence the health outcomes observed. We present evidence that these genetic markers and their interactions are valid instruments with good statistical properties, revealing a new dynamism from health to education. We find substantial heterogeneity in the effects of health on academic performance. Poor physical and mental health status have a strongly perverse effect on the academic performance of female students which, while of great policy importance, requires additional investigations.

Permalink

(In)Consistency of Preference Ratings: Results Using Three Methods of Preference Elicitation

Presenter:

Kathleen Gillespie

Authors:

Kathleen Gillespie, Elena Andresen, Sarah Boslaugh, Angela Recktenwald

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Background. The direct elicitation of preferences for different health states is one method suggested for measuring quality of life. Direct elicitation methods include the standard gamble (SG), time trade-off (TTO), and visual analog scale (VAS). Though theoretically appealing, in practice these methods produce some troubling results. First, the distribution of elicited preferences can be both left and right censored, with many persons rating health scenarios with a zero or one value. Second, health states that should have a logical ordering may receive reverse rankings. Finally, little is known about how the direct elicitation methods correlate with quality of life ratings derived from scales such as the Quality of Well-Being (QWB) scale.

Setting. Eighty-eight persons rated health states using the SG, TTO, and VAS methods and completed the Quality of Well-Being (QWB) scale. Thirty-four of the persons were randomly recruited; the other 54 were persons with disabilities. All persons rated 9 health states (the subject’s own health, single-eye blindness, double-eye blindness, and 6 scenarios that included various disabilities) using the 3 methods. Each of the 6 disability scenarios had two variants that differed only in health status; respondents were randomly assigned one of the two versions. Preferences were elicited using computerized software.

Methods. The mean preferences from the SG, TTO, and VAS methods were compared and the distributions were graphed. Multinomial logistic regression to predict zero, intermediate, and one responses used patient demographics and preference elicitation method as explanatory variables. The number of preference reversals within and across methods was counted. Correlation with preference imputation was examined by comparing the SG, TTO, and VAS utility values for own health with the QWB score.

Results. The SG and TTO methods gave higher mean preferences than the VAS method. However, the higher means were due primarily to a large number of ones. There were also many zeros, so that the distributions were both left and right censored. The VAS method yielded smoother single-peaked distributions of preferences across all health states. Inconsistent preferences occurred relatively infrequently. Blindness health states were rated inconsistently by 9% of respondents using the SG, 3% using the TTO, and 11% using the VAS. Across disabilityscenarios, persons with disabilities gave preference ratings that were 0.02 (SG) to 0.09 (VAS) higher than randomly selected participants, although among 36 options of health status and rating technique, there were several comparisons that were reversed. Randomly selected subjects were more consistent in giving the higher health state scenario a higher rating. None of the three methods were significantly correlated with the QWB scores for own health, the VAS method had the highest correlation (r=.179; p=.108).
Discussion. The distributions of preferences suggest that use of the SG and TTO methods is problematic. The large numbers of zeros and ones could be valid responses, or could indicate difficulty understanding the methods or protest valuations. The number of preference reversals was a small minority of all possible comparisons, but is troubling nonetheless. None of the three methods was correlated with QWB scores when rating own health.

Permalink

Bayesian Analysis of Health Care Provider Choice in Rural India

Presenter:

Bijan Borah

Authors:

Bijan Borah

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

This paper uses a Bayesian mixed multinomial logit (MMNL) framework to analyze the determinants of health care provider choice in rural India. Classical framework often becomes infeasible when random coefficients in MMNL model are allowed to be correlated and/or when distributions other than normal are assigned to the random coefficients. Allowing correlation between the random coefficients increases the number of parameters to be estimated, which often becomes difficult using classical methods. Normal distribution for the random coefficients is appropriate when there is some theoretical basis for the coefficient to be any real number. However, on many occasions, a random coefficient is expected to be either positive or negative. For example, price coefficient in the utility specification is expected to be always negative as people dislike higher price. In such cases, distributions with positive support (e.g., log-normal distributions) are more appropriate. However, when the distributions are assumed to be non-normal, the resulting likelihood surface often becomes very irregular, leading to failure in finding the maximum. Bayesian framework provides a natural way to deal with these complexities. In this paper, we assume both normal and log-normal distributions for the price coefficients. We use a Gibbs sampler with a layer of Metropolis-Hastings algorithm embedded within it to make random draws from the resulting joint posterior distribution. Our findings show that models with normal coefficients fit the data better. Price and distance are found to be important determinants of health care provider choice in rural India. Moreover, when individual-specific characteristics are interacted with provider attributes, model fit improves, indicating that these characteristics also influence health care provider choice decisions.

Permalink

Controlling for Quality in Hospital Inefficiency Estimation

Presenter:

Ryan Mutter

Authors:

Ryan Mutter, Michael Rosko

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Rationale: Recent increases in the costs of healthcare in the U.S. apparently are not accompanied by commensurate increases in quality, leading many to speculate that inefficiencies exist. In addition, the growing recognition of the important role that resource allocation decisions play in quality improvement has augmented interest in healthcare inefficiency measurement. Conceptually, healthcare quality improvement can reduce resource use by eliminating medical errors and unnecessary procedures. Alternatively, quality improvement can result in greater resource use because it requires more or better services. While stochastic frontier analysis (SFA) has become a popular method to estimate the inefficiency of hospitals, and it may be a powerful analytic tool, existing applications of SFA may not adequately control for the quality in analyses of hospital inefficiency. Without proper control, inefficiencies may be over- or under-estimated. This study assesses the role of quality in hospital inefficiency estimation.

Methodology: We identify existing strategies to control for hospital quality that have been employed in SFA estimation of hospital inefficiency. Employing SFA methods, we compare hospital inefficiency estimates with no hospital quality controls to the estimates with the following controls for hospital quality: 1) structural measures only (e.g., hospital teaching status), 2) structural and aggregated outcome measures (e.g., risk-adjusted, all-causes mortality), and 3) structural measures and multiple outcome measures designed to capture the multi-dimensional nature of hospital quality (e.g., risk-adjusted mortality rates for a variety of conditions and procedures and patient safety event rates).

Our sample is comprised of over 1,200 U.S., urban, general hospitals for the period 1999-2002. The quality variables are obtained by applying the Inpatient Quality Indicator (IQI) and Patient Safety Indicator (PSI) modules of the Agency for Healthcare Research and Quality (AHRQ) Quality Indicator (QI) software to 20 Healthcare Cost and Utilization Project (HCUP) State Inpatient Databases (SID). Other control variables in our hospital cost function include the input, output, price, and product descriptor variables commonly used in hospital applications of SFA from data files compiled by the American Hospital Association (AHA), the Center for Medicare & Medicaid Services (CMS), and Solucient.

Results: Preliminary analysis with 2001 data suggests that it is important to employ controls for quality in hospital applications of SFA. Although the average estimated inefficiency measure across hospitals is unaffected by the inclusion of quality controls, a number of the hospital quality variables entered the model significantly, and the rankings of hospitals by inefficiency scores was impacted by the choice of quality control strategy. Moreover, an inaccurate assessment of the impact of quality on cost can result from the use of a single outcome measure only.

Conclusions: We provide a rationale and empirical support for the inclusion of controls for quality in hospital applications of SFA. Researchers able to include variables that control for the multi-dimensional nature of hospital quality should do so. Researchers faced with data constraints that limit the quality variables they can include should be aware of the limitations of their analysis when interpreting and applying their results.

Permalink

Effects of Population Movement on the Risk of Chagas Disease in the California Blood Supply: A Preliminary Model

Presenter:

Leslie Wilson

Authors:

Leslie Wilson, Janine Ramsey

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Rationale: Chagas disease is a parasitic disease found predominantly in Latin America, including Mexico. Mexico routinely screens for Chagas disease in their blood supply, however there is currently no screening of the United States blood supply for Chagas disease. Although there are only a few documented cases of chagas attributed to blood transfusions, symptoms of this disease often don’t manifest until 10-30 years after contracting the disease. In addition with the increase in migration from Mexico to parts of the United States, the risk may be increasing.

Objectives: The objectives of this paper is to present a preliminary model of chagas disease transmission risk in the California and Mexican blood supply based on both the migration within Mexico from endemic to non-endemic areas in the migration from Mexico to California.

Methodology: We developed a model of migration and transmission risk for Chagas disease in Mexico and California. We used existing prevalence of chagas in the migrating population and probability estimates of blood donation, likelihood of having Chagas disease, infectivity depending on blood fractionation type, and other factors, to estimate the rate of seropositivity and risk of contracting Chagas from donated blood. We conduct a sensitivity analysis for the uncertain parameters.

Results: The base case California analysis we demonstrated a seropositivity rate of 0.015, and an immigrant donation rate of 0.01232. We estimated that the current verbal screen would eliminate only 0.000542 donations and that each unit given would be fractionated to go to 2.5 recipients. Sensitivity analysis showed that if the immigrant population continues to increase and the current efforts continue to increase blood donation from 1 to 10% in Hispanic communities, then the risk of contracting Chagas disease from blood transfusion in California increases from about 112 to 913 individuals. If the seropositivity rate of Mexican Immigrants is varied from the low rate of 0.000222 to the highest rate of 0.023, then the number of Californians contracting Chagas goes from about 1 person to 172 persons. When the infectivity is varied from a low of 0.01 to a high of 0.34, then the annual number contracting Chagas disease in California goes from only 11 to 384.

Conclusion: We suggest that more data collection is needed so that a more refined model can be made to estimate risk of Chagas disease to Californians receiving blood. However, from the current estimates, we feel that the risk of contracting Chagas disease from the blood supply is high enough to warrant testing, at least in the high risk areas with high population movement from endemic areas.

Permalink

Food for Thought: Food Stamps, Food Insecurity and Disability

Presenter:

Brian Armour

Authors:

Brian Armour, M. Melinda Pitts

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Food insecurity, defined as having insufficient funds to enable families to purchase enough food to maintain an active and healthy lifestyle, is particularly problematic in impoverished households. Disabled impoverished persons, in particular those who live alone, may be more susceptible to food insecurity as functional limitations may impede their ability to purchase and prepare food. The association between food insecurity and disability among the impoverished was assessed using data from the Panel Study of Income Dynamics (PSID) for 2003. The PSID is a longitudinal study of a representative sample of U.S. individuals and the family units in which they reside. The study collected economic and health information on 7822 families in 2003. Household food insecurity was calculated from the USDA 18 core item food security module. The household head or spouse was defined as disabled if they reported having difficulty with one or more activities of daily living or instrumental activities of daily living. Food stamp participation information was used to identify 728 families receiving food stamps in 2003. Results from our analysis suggested that among households receiving food stamps in 2003, 41.3% of disabled households were food insecure relative to only 26.3% of non-disabled households. Results from our multivariate analysis revealed that households whose head and or spouse was disabled were at increased risk of being food secure in comparison to households whose head and/or spouse had no disability. Food insecurity may be reduced by the Food Stamp Program; however, the program is significantly less able to minimize the probability of food insecurity in disabled households. These results suggest that the food stamp program is in need of reform in order to be better able to address the needs of the disabled.

Permalink

Replace Pharmaceutical Patents Now!

Presenter:

James Henderson

Authors:

Earl Grinols, James Henderson

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Pharmaceutical patent protection places Americans on the horns of an untenable dilemma. On one hand, monopoly protection limits the treatment options for individuals without prescription drug coverage and causes American consumers - many of whom are sick and not working - to pay too much for their medications. On the other hand, providing insurance for prescription drugs brings into play the monopolist’s response which is to raise the drug’s price in proportion to the inverse of the plan co-pay. A 20% co-pay, for example, creates an overwhelming incentive that leads the supplier to raise price to 5 times the initial level and reap the profit windfall. The firm adjusts its price-output pair from the (5p 0, q0) point only if profits are thereby raised further.

In this paper we discuss a socially preferable alternative to the patent system as a mechanism protecting intellectual property rights. Three important issues must be addressed. 1) How can we deal effectively with the above-mentioned monopoly pricing response to a prescription drug co-pay? 2) What is the best way to reward innovators to insure that the benefits of innovation are spread quickly and widely (at marginal cost prices)? 3) How do we go about calculating the reward for innovation?

The only way out of the unacceptable quandary is to sever the link between monopoly power created by the patent system and drug distribution. In fact, patent protection for new inventions is an anachronistic holdover compromise solution from an earlier era. Modern theory and tools imply an entirely different approach. This paper reviews historical policies toward research and development in light of the intervention principles of public finance. It concludes that an intertemporal bounty satisfying certain conditions is more efficient than current law and also dominates other structures including patent buyouts, except for a full-economy Ramsey pricing solution. Since Ramsey pricing is unlikely ever to be seen in practice, the investigation of efficient replacements for pharmaceutical patenting is timely and significant. The investigation identifies a mechanism for marginal cost drug pricing consistent with inducement to innovation as strong as provided by patents, consistent with increased social efficiency and practical implementation. The latter is argued from positive consideration of the fact that a similar system is successfully in place and has been used for many years in another sector of high innovation, the music industry.

The present research raises the possibility of change in the pharmaceutical drug industry as part of the larger movement in the United States to find a more desirable social ethic that can drive and sustain our health care system.

Permalink

Smoking and Ratings of Perceived Health Status

Presenter:

Kirsten Yaffe

Authors:

Kirsten Yaffe

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Introduction: The perceived health status of smokers has been shown to be lower than that of nonsmokers; however, causality is not clear. Smoking could lower someone’s self-rated health status physically, due to the adverse effects of smoking; psychologically, due to feelings of guilt or of weaker self control; or due to other factors. An external factor that could increase the propensity to smoke may also cause other health-related issues, lowering perceived health status. The biochemical effect of nicotine on the nicotinic receptors in the brain suggests that smoking can also be a form of self-medication. The causative agent, which evokes the need to self-medicate, could also lead to a lower health status.

Purpose: The purpose of this study is to examine the extent to which reported quality of life differs between smokers and nonsmokers when other physical and mental health status factors are controlled for.

Methods: Data from the 2002 Medical Expenditure Panel Survey (MEPS) was used in this analysis. The dataset represented a population of 97,221,295 non-institutionalized adults in the US, 22% of which were current smokers.

To address confounding of explanatory variables by smoking, logistic regressions were run using smoking status as the dependent variable. Explanatory variables included demographics, smoking-unrelated diseases, smoking-related diagnoses, and mental/emotional characteristics. Variables that were significantly associated with smoking and smoking status were used as the explanatory variables in a linear regression with perceived health status as the dependent variable.

Results: Even after adjusting for the other significant measures of emotional and physical health, smoking remained a significant factor affecting perceived health status, although the influence of smoking on health status decreased by more than half. In the univariate model, the coefficient on smoking was 0.40. When other explanatory variables were included in the model, the coefficient on smoking status was 0.18.

The only smoking-related diagnosis significantly associated with smoking in the logistic regressions was emphysema. In the linear regression, the three factors with the largest effects on health status were perceived mental health status (0.50), diabetes (0.46), and emphysema (0.46). Mental health status also has a much larger t-value than any others do in the model; it is likely that the health and mental health status measures are correlated strongly enough that a multivariate model should be used instead with mental health status included as a dependent variable.

Discussion: A limitation of this model is the accuracy of the outcome variable, which is measured on a five-point scale, from 1 (“excellent”) to 5 (“poor”). Although this scale is treated as an interval scale, it is unlikely that ratios between adjacent values are the same throughout the scale. (The difference between “excellent” and “very good” is likely not as large as between “fair” and “poor.”) It is therefore harder to determine the exact effects on health status of a change in the explanatory variables (i.e., the coefficients are harder to interpret); however, inferences can be drawn about the relative effects of each factor in comparison with others.

Permalink

Smoking cessation and health

Presenter:

Yu-whuei Hu

Authors:

Yu-whuei Hu, Dong Hwa

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Rationale: Medical and public health studies have found that smoking may increase the likelihood of diseases occurring and causing premature death. Many developed countries, including Taiwan, have implemented a variety of policies designed to raise the costs of smoking in order to reduce cigarette consumption. Most studies have provided evidence that smoking may harm one’s health. However, not enough studies have shown the benefits to health from a cessation of smoking.

Objectives: In this paper, we empirically study the health effect of smoking cessation on the elderly male (aged sixty and over) in Taiwan.

Methodology: We derive data from the “Survey of Health and Living Status of the Middle Aged and Elderly in Taiwan: A Longitudinal Study Since 1989” which provides information regarding the subjects’ smoking history before 1989 and smoking behavior from 1989 to 1999. We define the health effect by the following two measures. One is survival years, given survived at age 60 and the other is self-assessed health status, which is categorized into five degrees: excellent, very good, good, fair and poor. We adopt Cox proportional hazard model to estimate the effect of early smoking cessation on the hazard rate of one’s life and ordered probit model to analyze the health effect of the length of sustaining smoking cessation.

Results: The empirical results show that those who quit smoking before the age of 54 had a lower hazard ratio and a higher survival probability. In addition, the results show that, among those who quit smoking before 1989, the self-assessed health status was better the longer the smoking had ceased. Besides, the results also show that the younger a person is when he becomes addicted, the more likely he is to die at each age.

Conclusions: Our empirical analysis implies that early smoking cessation is positively associated with longer life and better health status.

Permalink

Tiered Co-payments and Cost-Related Medication Underuse?

Presenter:

Todd Wagner

Authors:

Todd Wagner, Michele Heisler, John Piette

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Introduction Many insurance plans use tiered co-payments, in which patients pay less for generic medications than for branded prescriptions, to encourage the use of generic medications. However, co-payments may force some patients to forgo prescription drugs due to costs. Research indicates that approximately 20% of adults with a chronic illness will underuse prescription drugs due to cost pressures. With cost-sharing on the rise, many decision makers are concerned because underuse of medication has been linked to worse health outcomes and to greater rates of hospitalization. In this study, we analyze how tiered co-payments are associated with medication underuse.

Methods We fielded a national survey of chronically ill adults over age 50 (n=4085). Respondents reported whether they had been given a prescription for each of 16 chronic health problems. On a condition-by-condition basis, participants were asked: “In the past 12 months, have you ever taken less of this medication than prescribed by your doctor because of the cost?” If they said yes, they were asked whether they (1) did not fill a prescription at all, (2) took fewer pills or a smaller dose, (3) put off or postponed getting a prescription filled, (4) used herbal medicines or vitamins when they felt sick rather than take their prescription medication, and (5) took the medication less frequently than recommended to ‘stretch out’ the time before getting a refill. For each of these five questions, responses were collecting using a 5-point ordinal scale: never, less than every two or three months, about every two or three months, at least once a month, at least once a week. We used linear probability models to estimate the probability of underuse as a function of co-payments among those with tiered co-payments (n=2771). We created dummy variables for co-payments: <$5, >$5 and <=$15, >$15 and <=$50, or >$50, do not know, or are not sure. In the multivariate analyses, we controlled for age, gender, income, and chronic conditions, and we used sampling weights and corrected the standard errors for clustering.

Results Cost-related underuse was more consistently associated with higher co-payments for generic medications than for branded medications, suggesting that people were less-price sensitive to their branded medications. The association between co-payments and cost-related underuse was strongest for not filling a prescription and postponing the filling of a prescription; the association was considerably weaker for stretching out the dose, taking fewer doses or substituting herbal medications.

Conclusions A substantial minority responds to co-payments by underusing their medication, and the effect is greatest at the cash register where people either do not fill the prescription or postpone filling the prescription. Although underuse is associated with co-payments, the effects are not the same for generic and brand name medication. People were more likely to forego medication when prescribed a generic than a brand name medication. Given that most patients do not talk to their clinician about medication underuse, clinicians need to explain the use of generic medications to their patients so that patients do not always equate value with price.

Permalink

Demand for Private Health Insurance in Germany

Presenter:

Majed Elhewaihi

Authors:

Majed Elhewaihi

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Undoubtedly, demand for private health insurance in Germany is growing, alone in 2004 around 400.000 Germans have chosen one of the private coverage schemes. Between 2002 and 2003, about 300.000 of statutory insured persons have taken out outpatient supplementary private insurance, while in the same period, around 242.000 have taken out optional hospital benefits. We argue that perceived quality of private and public health care, insurance premiums and income are among the determinants of this growing demand. This study is the first in Germany that moves from theoretical to empirical framework. It uses economic modelling to analyze the effects of these determinants on the probability of purchasing private health insurance. Economic modelling states, demand for private health insurance is demand for high health care quality. The model was developed taking into consideration the unobserved endogeneity and heterogeneity. Results support the hypothesis that the demand for private health insurance is driven by the quality gap between both insurance sectors.

Permalink

Mixed Ownership Markets: Theory and Evidence from Hospitals in the US and China

Presenter:

Karen Eggleston

Authors:

Karen Eggleston

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

A mixture of government and private ownership, including both not-for-profit and for-profit firms, is an abiding feature of healthcare delivery in many established market economies and is emerging in many developing and transitional economies. Yet virtually no existing theories model endogenous co-existence of government, private nonprofit and for-profit ownership in a market, or focus on the normative question of whether the co-existence of all three forms is socially desirable. This paper aims to fill this gap through a parsimonious theoretical “merger” of the noncontractible quality theory of ownership (Hart, Shleifer and Vishny 1997; Glaeser and Shleifer 2001) and the theory of soft budget constraints as a dynamic commitment problem (Kornai, Maskin and Roland 2003). I argue that endogenous soft budget constraints are a more compelling explanation of the soft incentives of government managers than that presented by Hart, Shleifer and Vishny (1997), that government managers are more replaceable than private managers. In fact, it seems to be quite the opposite: government employment is often more stable than that in the private sector, frequently entailing a lower wage in exchange for more job security. Consistent with empirical evidence from US hospitals (e.g., Duggan 2000; Baicker and Staiger 2005), I allow for both a predatory principal (the ratchet effect) and soft principal (the soft budget constraint) in the same context, operating on different margins.

The model helps to explain why most industries are dominated by private for-profit firms; why some sectors such as healthcare tend to be dominated by private not-for-profit and government providers; and what factors drive the changing patterns of ownership mix across industries over time. In the model, private entrepreneurs are most responsive to market conditions; government provision assures supply (access) but is differentially prone to soft budget constraints and slower innovation. Mixed ownership lessens the social cost of government provision because competitive discipline and private R&D leads to contractibility - with a lag - for many aspects of innovation. I present evidence consistent with the model from two contrasting contexts: a meta-analysis of hospital ownership and performance in the US; and recent data on government and private hospitals in China.

Permalink

Exploring Deductibles: Do they Matter in the Era of Managed Care?

Presenter:

Maia Platt

Authors:

Maia Platt

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Rationale: Copayments / coinsurance and medical costs have been researched extensively, but the similar effects of deductibles have not yet been comprehensively established.

Objectives: To establish and quantify the effect of deductibles on the health care costs and utilization of privately insured non-elderly population in the USA.

Methodology: Medical claims, enrollment records and health benefits design from 40 large health care plans in 1998, obtained through MEDSTAT Group Inc. provided financial and demographic information on more than one million enrollees. Translog model with interaction terms and polynomials was used to obtain derivatives at various values of deductibles. Marginal effects are modeled as projected changes in average enrollee’s annual costs or utilization, per $1 change in deductible, in 1998 dollars.

Results: Contrary to the initial expectations, the relationship between deductibles and medical costs was found to be predominantly positive. It is our hypothesis that a mix of plans with both positive and negative marginal effects is due to self selection: people with low medical costs tend to move away from plans with high deductibles, resulting in higher mean costs for the remaining enrollees. The range of derivatives variation is established as (-$5) to (+$13) for plans with existing deductibles, and is further expanded up to $30 for plans with no existing deductibles: “sticker shock” effect. Changes in total costs are primarily driven by the change in the outpatient utilization and costs: both sign and value. Inpatient care demand curve was found to be perfectly inelastic, indicating that deductibles affect discretionary care only. The effect of deductibles does not seem to be mitigated by isolated application of managed care tools (such as precertification requirements). However, a systematic concept of managed care applied by PPO plans seems to be capable of reducing their impact on medical costs under certain conditions.

Conclusions: Even in the presence of managed care, deductibles continue to be a barrier, restricting access to medical care. However, higher deductibles in most cases fail to restrict spending. The predominantly positive relationship between deductibles and medical costs indicates, that if the deductible is lowered, there is a higher probability that insurer’s average costs per enrollee will go down, due to the inflow of people with low expected costs (i.e. healthy population). However, this probability varies substantially across the geographic states, most likely due to the states’ different regulations of the health insurance and health care industries, as well as different wages and / or health care prices across the states. Thus, opportunities exist for improving access to health care by lowering deductibles without imposing heavy financial burdens. Government involvement is strongly encouraged.

Permalink

The Effects of Competition and Multi-Hospital Systems on Hospital Prices

Presenter:

Emmett Keeler

Authors:

Glenn Melnick, Emmett Keeler

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Hospital prices in the US are rising again following years of little growth in which managed care plans had been able to leverage competition among hospitals to negotiate lower prices. Two explanations for the apparent shift in bargaining power from managed care plans to providers are that managed care may be weakened from consumers demanding broader networks, or that providers may be strengthened by the formation of larger scale multi-hospital systems.

Our study investigates how competition and the formation of hospital systems has affected the balance of power in hospital-health plan bargaining as reflected by hospital pricing behavior over time. We analyze trends in adjusted per diem hospital prices in California during a period of significant price growth (1995 through 2003) to assess whether the price constraining effects of hospital competition have weakened or if the ability of hospitals that are part of systems to charge higher prices explains the increase in hospital prices in that period. We find that competition in local markets remains an important factor in controlling hospital prices. However, the formation of hospital systems, both nonprofit and for-profit, has allowed member hospitals to charge higher prices. System effects have gotten larger over time, and apply even in competitive markets. This shift could be a permanent source of hospital price inflation, both for hospitals that are part of systems as well as for non system hospitals that compete in the same local markets.

Permalink

Applying Lean Toyota Production System Methods to Improving Health Care Systems

Presenter:

Sheri Eisert

Authors:

Sheri Eisert

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Rationale: While health care has seen strides in technology and pharmaceutical development, there has been little change in the processes of health care provision in the last 30 years. There are significant redundancies in the processes of care, numerous patient hand-offs between health care providers and a “silo mentality’ in care provision. Although many health care processes resemble “assembly-line like processes” health care systems have not widely utilized the knowledge of other industries in improving efficiencies, quality, customer satisfaction and workforce development.

Objectives: To identify the usefulness in applying Lean Toyota Production System principles and techniques to improving efficiency, effectiveness and quality in health care systems.

Methodology: Through case study analysis, “real” health care system processes are mapped to identify current and operationally agreed upon future state process activities, distinguishing between value added and non-value added activities. The implemented future state process and related improvements (efficiency, effectiveness and quality) are evaluated, including follow-up for improvement sustainability. The primary lean principles and tools that are used to create the current and future state include value stream mapping, standard work and elimination of the eight wastes (underutilized people, waiting, inventory, transport, defects, motion, overproduction, excess processing). The approach to improvement implementation will be the use of Kaizen (continuous improvement) through Rapid Improvement Event weeks (five day work week cycles) and selected operational teams that are involved in the particular process (including upstream and downstream).

Results: The use of Lean Toyota Production System principles and techniques are readily adapted to health care systems. The degree to which health care systems, patients and providers could benefit through becoming “lean” are significant. Through the elimination of waste, improvements in quality of care, patient safety and satisfaction (both patient and workforce) will follow. The primary challenge in implementing Lean Toyota Production System principles in health care is convincing individual health care providers that their participation in eliminating waste and standardizing processes are of benefit to their individual work processes and outcomes.

Conclusion: With ever increasing health care costs and continued concerns for patient safety, quality of care, patient satisfaction and workforce shortages, health care systems are of need of major transformation. The application of Lean Toyota Production Systems is one approach to viewing health care systems through “new eyes”. Health care systems have much to learn from other industries surviving and thriving in competitive markets.

Permalink

New cost sensitivity index based on cognitive cost cues for an application of the Lens model on physicians' choices

Presenter:

Christine Huttin

Authors:

Christine Huttin

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Rational: This paper presents a generalization of the construction of a cost sensitivity index that analyzes the influence of implicit financial and economic information on physicians. The use of an adaptation of the Lens model on physicians’ choices allows describing how such type of cost information could interact with clinical information. It may be controversial to integrate implicit financial or economic information and cognitive knowledge in current measurement of price and cost in health care. However, the application of the Lens model on physicians’ choices from a European pilot study presented at the Brunswik society tended to show that it may capture significant behavioral changes (for up to 1/3 of the sampled physicians in some of the surveyed countries). The use of such information can also increase the explanatory power of current demand models for health care services and current measured effects of out of pocket payment and prices.

Objective: This paper will present the construction of this new type of cost index, based on cost cognitive cues and will discuss how to combine this type of measure with current information on price and cost. Traditional measures used to assess responses of agents to monetary value of goods and services are prices and costs. These prices can be changed in order to take into account psychological effects of a measure on a consumer, but they will still be measured by traditional objective price and cost information collected through official agencies or company surveys. In health care, net price to consumer reflects very poorly the information used by either the patient or his physician.

Methodology: The type of information used in this index is described in verbal terms, can use monetary values, but not always. Measures are based on thresholds and intervals that represent decisions’ shifts, not represented by traditional distance measurement. New weighting systems within cost cues and between modules of cost cues are also used, based on market survey research techniques such as conjoint analysis (or other statistical procedures that are used to test judgment analysis designs)

Results: The current development of the index is based on results from the pilot study from the allergy study on a European health care system and a disease economic model tested on NAMCS US data. The methodology integrates stated and revealed physicians’ preference with the new type of cost information. Validation stages are currently in process in the US with new sets of data and some results on sources of biases will be presented (for instance, the existence of systematic bias between internet and mail surveys).

Conclusions: This paper aims to propose an integration of implicit cost information in current measurement systems of price and cost of health care, when current accounting systems are insufficient to forecast health care expenditures. Such conditions for instance refer to economic change or periods of transitions that totally modify the relative cost structure and economics of both providers and patients.

Permalink

The Flip-Side of Social Capital: The Distinctive Influences of Trust and Mistrust on Self-Reported Health in Rural China

Presenter:

Hongmei Wang

Authors:

Hongmei Wang, Mark Schlesinger, Hong Wang, William Hsiao

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Rationale: Health researchers have observed stagnant and sometimes declining health outcomes in rural China despite the fact that absolute income is increasing constantly. The elimination of inequality in health is seen as an urgent public health policy issue, but the determinants of these disparities are poorly understood once economic deprivation has been reduced. Few studies have focused on social capital and health in China. In the case of China, a relevant question to ask is whether social capital, measured as interpersonal trust, has changed due to economic reforms and whether the erosion of social capital is related to the worsening of the health status of rural residents.

Objectives: In the setting of rural China: (1) Is social capital measured as trust related to individual’s health? If yes, how does social capital impact on health? (2) How does social capital compare to other socio-economic factors, such as income and education, as predictors of physical and psychological well-being? (3) How does social capital differentially affect an individual’s health status in developing countries compared to advanced economies?

Data and Methods: This paper employed data from the household survey of the Rural Mutual Healthcare Project, jointly conducted by Harvard China Health Policy Center and China Health Economics Institute. The survey was conducted in two provinces in China in year 2002 using stratified clustered sample. The final sample for this paper consisted of 9608 participants. The outcome variables included a 5-point scale self-reported general health and a 3-point scale self-reported mental health measure. Social capital was measured by 10 questions measuring both trust and mistrust levels. A trust index and a mistrust index were constructed and both measures were also aggregated at village level for analysis. This paper adopted multivariate ordered logistic regression analyses, adjusting for survey design effects, for main analyses. A rich set of control variables including demographic, health risk measures, social economics status variables, and sanitary measures were controlled in the models.

Conclusions: (1) Social capital measured as interpersonal trust is significantly associated with an individual’s health in rural China. Social capital appears to impact on both mental health and physical health, but to different degrees. Both individual-level social capital and village-level social capital have a significant impact on health. Individual trust and mistrust have opposite impact on a person’s health as expected. However, results from stratified analyses suggest that mistrust operates on health separately from trust. (2) The overall magnitude of effects of social capital on health is substantial, comparable to social economic status factors such as education and health insurance. (3) Compared to that of advanced economies, the magnitude of influences of social capital on health appears to be bigger in rural China.

Disclosure information: The authors are either faculty members or student at Yale University or Harvard University and have no conflicted interest with regard to this project.

Part of the abstract has been presented in ASSA 2005 annual meeting.

Permalink

Institutional Aspects of Medical Technology Assessment

Presenter:

Robert Kemp

Authors:

Robert Kemp

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Economic evaluation is being prescribed for improving allocation of health care resources, for marketing technologies, and as an academic endeavor unto itself. The institutional arrangements for the delivery of health care technologies are often assumed to be exogenous to the technological assessment. However, for MTA to be a valuable tool, the institutional aspects of the assessment have to be made explicit. The first aspect to be considered is the nature of the patient-practitioner transaction. The decision on the use of a technology may be bound by institutional rules, a specific practice environment, or be relatively open to clinical imperative. The decision may have a directly measurable pecuniary impact, generate a charge, or may be made in a framework of a priori allocation of health care resources. Agency issues need to be addressed. Secondly, the decision framework may be “ex-post” or “ex-ante”; ‘ex-ante’-undertaken prior to the implementation to estimate net outcome in relation to costs or “ex-post’ subsequent to knowing a program’s net outcome effects. The ex-ante analysis may be based principally on an experimental design where as the ‘ex-post’ may be based principally on retrospective data. Ex-post data may be based on multiple experiments introducing the ambiguity of common comparator trials. Ex-ante evaluations are required for technologies that will be difficult or impossible to control after they are diffused. Third, the values used in economic evaluation may be determined by the institutional rules used to allocate resources in an area, and may be determined by a precedent of using cost-effectiveness in such allocations, leading to the impossibility of finding a set of preferred programs. Non-transparency or ambiguity about these institutional aspects will limit the usefulness of the subsequent evaluation. Non-transparency or institutional ambiguity allows for the publication of essentially useless studies and to the ability of vested interests to use economic evaluation for non-substantive purposes, such the marketing of a technology. Critical appraisals from treatments for dementia, chemotherapy, and diabetes are used to illustrate the importance of the specification of these institutional aspects.

Permalink

Obesity and Nutritional Knowledge

Presenter:

Irina Grafova

Authors:

Irina Grafova

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Rationale: Currently, obesity presents one of the biggest public health problems. From a public policy perceptive, it is important to know whether propagation of nutritional knowledge could be used as an effective public policy tool in reducing the prevalence of obesity.

Objective: The objective of this paper is to examine the relationship between nutritional knowledge and obesity.

Methodology: This study uses the 1999 wave of the Panel Study of Income Dynamics (PSID). PSID is a longitudinal study of a representative sample of U.S. individuals and the family units in which they reside. The 1999 wave of PSID used in this study contains questions on height, weight, nutritional knowledge and attitude toward importance of balanced diet of the respondents. The econometric technique used is ordered probit with known cut points, i.e. interval regression model.

Results: The main result is that there are two distinct patterns in the relationship between nutritional knowledge obesity. First, for individuals attaching greater importance to a balanced diet, greater nutritional knowledge corresponds to a lower probability of being obese. This pattern is consistent with the notion of preventive nutritional knowledge: persons who are aware of the connection between poor nutrition and certain health conditions have higher expected costs associated with obesity and are more likely to follow a balanced diet and avoid obesity. Second, for individuals who attach low importance to balanced diet, greater nutritional knowledge corresponds to greater probability of being obese. This pattern is consistent with the notion of reactive nutritional knowledge: obese persons, especially those with obesity-associated health conditions are more likely to be aware of health problems caused by dietary imbalances and obesity and would be more likely to possess greater nutritional knowledge due to his/her condition. Empirical analysis reveals that consistent with the pattern of preventive nutritional knowledge the negative effect of nutritional knowledge on the probability of being obese or overweight for women without obesity-related conditions is double that for women with such conditions. Likewise, consistent with the pattern of reactive nutritional knowledge the positive relationship between nutritional knowledge and the probability of being obese without obesity-related conditions is half that for men with such conditions.

Conclusion: The analysis shows that there are two distinctive patterns in the relationship between nutritional knowledge obesity. Nutritional knowledge could be preventive to obesity, in which case greater nutritional knowledge leads to lower probability of being obese. Nutritional knowledge could also be reactive to obesity, in which case obesity leads greater nutritional knowledge. Ignoring the reactive effect of nutritional knowledge, while estimating the size of its preventive effect on obesity, may lead to biased and insignificant estimates.

Permalink

Impact of public health program and maternal education on immunisation behaviour of children in rural Bangladesh

Presenter:

M. Zia Sadique

Authors:

M. Zia Sadique, Mohammad Asadullah

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Rationale: Vaccine preventable diseases account for considerable proportion of child deaths and morbidity in many developing countries. Understandably so, the expanded program of immunization (EPI), features one of the corner stones of public health program of most developing courtiers. These programs have brought considerable direct and indirect benefits to health and nutritional status of children. Despite these programmatic interventions, many developing countries have yet to achieve the target of Universal Childhood Immunisation of 80 percent, and in many situation immunisation uptake remains low posing the threat of disease outbreak very likely. It is extremely important for policy makers in developing countries to evaluate the contribution of public health programs which will help developing programs that are more effective in achieving their targets.

Objective:

The objective of the paper is to evaluate the contribution of a public health program on the demand for childhood vaccination by decomposing the direct and indirect channels through which public health program influences household demand for preventive health care.

Methodology: The paper aims to estimate household demand for child immunization in a rural area of Bangladesh where a maternal health care program known as Maternal and Child Health (MCH) Program is in operation. We exploit the existence of an exogenously assigned maternal health care program in the study area where households in the treatment area received important health facilities. Households in the control area were exposed to pubic health facilities only.

We focus on three immunisation/vaccination outcomes for the most recently born child for a sample of 1033 children: (i) no immunisation, (ii) partial immunisation and (iii) full immunisation. The empirical analysis employs Ordered Probit regressions where the dependent variable is defined jointly for all possible outcomes mentioned above.

While the multivariate analysis accounts for a host of demand side determinants of immunisation outcomes in the treatment and control areas, special attention is given to mother’s education and risk attitude, and their interaction with the program effect on immunisation behaviour.

Results:

The MCH program has significantly enhanced immunisation status of children - both directly and indirectly. Mother’s education has shown important influence over child health care choices related to immunisation. This positive effect is purely driven by the knowledge and awareness associated with maternal education. The results lead to confer that the MCH program acts as a substitute for maternal education as it has found to compensate low endowment of mothers education. Prenatal care visit has also significantly enhanced household’s demand for childhood immunization, where prenatal care proxies for both risk awareness of pregnant mothers and knowledge about health care choices. Public health program has reinforced the household’s risk awareness towards child health choices. These findings remain robust to control for household income leading to the conclusion that influence of maternal education and pre-natal health care choice are not proxying for an income effect.

Conclusion: Public health program has important spillover effects on child health and welfare which should be included in their evaluation.

Permalink

Continuous Enrolment Requirement: Do we really need it?

Presenter:

Onur Baser

Authors:

Onur Baser

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Rationale: Continuous Enrolment is a “must requirement” for the analyses based on claims data since failure to control for different variable length creates a bias in our estimators. This requirement, however, may decrease the sample sizes substantially. Relaxing of continous enrolment requirement and the appropriate adjustment to the models are necessary.

Objective: We proposed a two stage estimation method which can be used to estimate our outcomes consistently without imposing continous enrolment requirement.

Method: We provide systematic treatment of the correction for possible selection bias of any claims data where the selection rule is described by a censored regression model. We first use the duration of time a patient is tracked for selection. Second using Tobit residuals in the structural equation, we showed that we removed possible selection bias due to continuos enrolment requirement. We used every patient in our first stage estimation, continously enrolled and not continously enrolled and used this infromation to estimate our second stage model over continously enrolled patients. We also derive a simple test to determine possible selection bias due to continous enrolment requierment.

Results: We proved that the resulting estimators are consistent and asymptotically normal. Simulation studies are comfirmed our results.

Conclusion: Continous enrolment requirement can be very restrictive in certian cases, and failure to adjust for information from droped out patients may lead to bias. Adjustment is required to control for possible selection bias.

Permalink

One way to improve two part modeling of longitudinal cost data - using latent class clustering of cost casemix groups

Presenter:

Jeonghoon Ahn

Authors:

Jeonghoon Ahn

Mon June 5, 2006 9:30-10:45 Room Alumni Lounge

Rationale: Though two part model has prevailed for the last two decades as the methodology to analyzing cost data in health economics, the advantages of using longitudinal (panel) data received little attention.

Objectives: The objective of this paper is to highlight the advantage of using a clustering technique in longitudinal two part model.

Methods: A longitudinal medical cost data of 3,260 patients for two years was first clustered by latent class cluster analysis by cost characteristics - categorical indicator variables with 5 categories of the each year’s cost, dichotomous indicator variables with 0 for zero cost and one for positive cost for each year. Latent class cluster analysis identifies cost casemix groups using the correlation structure of these cost indicator variables and these indicator variables are mutually independent in each identified casemix group. To determine the optimal number of casemix groups, Sample Size Adjusted Bayesian Information Criteria and Consistent Akaike Information Criteria was compared. For each group of identified cost casemix group, panel data two part model (random effect and fixed effect) for cost estimation was applied and then compared with the pooled data result, which only models heterogeneity in the intercept.

Results: The two results with casemix control and without casemix control were significantly different. This difference is mainly resulted from the significant differences in characteristics among the casemix groups. The higher level of heterogeneity control in latent class clustering can significantly improve two part model in longitudinal analysis.

Conclusions: When two part model is applied in longitudinal data, there is an advantage of heterogeneity control whether assuming individual heterogeneity intercept as random effect or fixed effect. This advantage can be further extended by using a clustering technique such as latent class clustering specifically on longitudinal cost variable.

Permalink