CMStatistics 2020: Start Registration
View Submission - CMStatistics
Title: Robust causal inference for point exposures with missing confounders Authors:  Alexander Levis - Harvard T.H. Chan School of Public Health (United States) [presenting]
Sebastien Haneuse - Harvard TH Chan School of Public Health (United States)
Abstract: The gold standard inferential target in comparative effectiveness research is a causal treatment effect. When clinical trials are infeasible, observational cohorts can be used to estimate these effects, but statistical methods are needed to control for confounding factors. Such cohorts, especially when extracted from large observational databases, are often subject to large amounts of missing data, and methods must simultaneously handle confounding and missingness. We propose robust, semiparametric efficient estimators of average treatment effects from cohort studies when confounders are missing at random. The approach is based on a novel factorization of the likelihood that, unlike alternative methods, facilitates flexible modelling of nuisance functions while still maintaining consistency at nominal rates of convergence. Simulations, derived from an EHR-based study of the long-term effects of bariatric surgery on weight outcomes, verify the robustness properties of the proposed estimators in finite samples. Extensions of the methods to the matched cohort study design are discussed.