Propensity Score Approaches to Estimating Treatment Effect on Censored Cost

• Activity Number: 493
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Health Policy Statistics Section
• Author(s): Jiaqi Li* and Nandita Mitra and Elizabeth Handorf and Justin Bekelman
• Companies: University of Pennsylvania and University of Pennsylvania and Fox Chase Cancer Center and University of Pennsylvania
• Keywords: cost estimation ; observational studies ; propensity score ; treatment effect ; doudbly robust ; super learner

Abstract

Medical cost estimation is vital to health economics evaluation and decision-making. Common challenges include non-independent censoring and right skewness of cost data. Lin (2000) and Bang and Tsiatis (2000) have developed linear regression and weighting techniques to model medical cost from trial data. However, medical costs are often collected from observational claims data which are subject to confounding. We propose common propensity score (PS) methods for cost estimation including covariate adjustment, stratification, inverse probability weighting and doubly robust weighting (DR). We also use Super-Learner to 1) choose among regression models (linear, lognormal, glm with gamma variance) to estimate cost in DR and 2) choose among various model specifications for PS. Simulation studies show that when PS model is correctly specified, weighting and DR perform well. When the PS model is misspecified, DR with SL can still provide unbiased estimates. We apply these approaches to a cost analysis of two bladder cancer treatments, cystectomy versus bladder preservation therapy, using SEER-Medicare data.