Abstract:
|
Propensity score (PS) and disease risk score (DRS) are popular covariates balancing techniques for observational studies. However, few has been reported on the performance of prognostic propensity score (DRS-PS) method which combines the two techniques together in covariate adjustment setting. We conducted simulations to evaluate the performance of three DRS-PS models, a probability-based PS models, and an inverse probability treatment weighting using PS model (IPTW). We observed, in general, IPTW method out-performed other methods models, particularly in the scenarios with rare outcome, lower exposure rates and high relative risk (RR). Among three DRS-PS models, the full-cohort DRS-PS method usually performed better than the other two methods. It sometimes even out performed PS and IPTW. DRS-PS performance generally improved when outcome and exposure rates increased, and RR decreased. IPTW and the full-cohort DRS-PS method usually had better mean squared errors than PS and other two DRS-PS methods. Users need to be aware of outcome and exposure rate before choosing an appropriate method to estimate RR.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.