Continuous Enrolment Requirement: Do we really need it?
- Presenter:
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.
