East Coast Ballroom
Optimal matching approaches in health policy evaluations under rolling enrollment (307809)
*Samuel D. Pimentel, University of California, BerkeleyLauren Vollmer Forrow, Mathematica Policy Research, Inc.
Jonathan Gellar, Mathematica Policy Research, Inc.
Jiaqi Li, Booz Allen Hamilton, Inc.
Keywords: Matching, propensity score, rolling enrollment, network optimization, causal inference, health policy evaluation
Comparison group selection is paramount for health policy evaluations, where randomization is seldom practicable. Rolling enrollment is common in these evaluations, introducing challenges for comparison group selection and inference. We propose a novel framework, GroupMatch, for comparison group selection under rolling enrollment, founded on the notion of time-agnosticism: two subjects with similar outcome trajectories but different enrollment periods may be more prognostically similar, and produce better inference if matched, than two subjects with the same enrollment period but different pre-enrollment trajectories. We articulate the conceptual advantages of this framework and demonstrate its efficacy in a simulation study and in an application to a study of the impact of falls in Medicare Advantage patients.