CMStatistics 2022: Start Registration
View Submission - CMStatistics
Title: Methods to improve the efficiency of RCTs using observational studies Authors:  Amir Asiaee - Vanderbilt University Medical Center (United States) [presenting]
Abstract: The aim is to develop a framework for precisely estimating the Conditional Average Treatment Effect (CATE), which characterizes treatment effect heterogeneity. Due to cost constraints, RCTs are often small in size and scope and are often significantly underpowered to detect treatment effect heterogeneity; on the other hand, the estimated effect from one population may not be transportable to another. The transportability of RCT findings has been a subject of methodological and practical interest, and much progress has been made. In contrast, the integration of observational studies (OS) in the analysis of RCT for variance reduction is under-explored. The target population of heterogeneous data integration is the RCT, and the goal is to utilize large observational studies, usually from a non-comparable population, to compensate small sample size of RCTs for CATE estimation. We address a few critical theoretical challenges presented in real-world scenarios. First, we study the settings where the set of measured covariates in the RCT and OS are not entirely identical. Further, almost all of the current work in transportability assume that the CATE is invariant across populations. The CATE transportability assumption or its more stringent forms (trial participation ignorability) are cornerstones of current theoretical results. Still, they are rarely satisfied, e.g., where there are systematic, unrecorded differences in those who participate in RCTs versus the OS.