Online Program

Methods for Comparative Effectiveness Analyses in a High-Dimensional Covariate Space with Few Events
View Presentation View Presentation

*Jessica Myers Franklin, Brigham and Women's Hospital and Harvard Medical School 
Wesley Eddings, Brigham and Women's Hospital and Harvard Medical School 
Peter Austin, Institute for Clinical Evaluative Sciences 
Elizabeth Stuart, Johns Hopkins Bloomberg School of Public Health 
Robert Glynn, Brigham and Women's Hospital and Harvard Medical School 
Sebastian Schneeweiss, Brigham and Women's Hospital and Harvard Medical School 

Keywords: propensity score, causal inference, simulation, boosting, trimming, epidemiology

Nonrandomized studies of comparative effectiveness and safety evaluate treatments as used in routine care by diverse patient populations and are therefore critical for producing the information necessary to making patient-centered treatment decisions. Most nonrandomized studies based on administrative health care data use a propensity score (PS) to control hundreds of measured covariates and to estimate the causal effect of treatment. Even in large studies, high-dimensional confounder control can lead to problems in causal inference due to unstable estimation of the PS model or inappropriate use of observations with extreme PS values. However, in studies with few outcome events, each observed event is highly influential, and potential problems are exacerbated. We used a novel “plasmode” simulation framework to evaluate and improve methods for estimation of high-dimensional PS models and estimation of treatment effects controlling for the PS when there are few observed outcome events. Plasmode creates realistic simulated data sets with hundreds or thousands of potential covariates based on an observed cohort study. Methods were also evaluated in three comparative effectiveness studies.