B0773
Title: Mediation analysis with unmeasured treatment-induced confounding
Authors: Fan Xia - University of California San Francisco (United States) [presenting]
Abstract: In causal mediation analysis, covariates affected by the treatment or exposure can be a source of confounding between the mediator and the outcome. Like any type of confounders, the treatment-induced confounders can be mismeasured or unmeasured, which could lead to invalid causal inference. Treatment-induced confounders, even when correctly measured, are especially challenging because it mediates part of the exposure effect while confounds the exposure effect through the mediator. As a result, the identification and estimation of natural direct and indirect effects of the exposure to treatment-induced confounders deviate from those of the well-studied average treatment effect. We use variables associated with the treatment-induced confounders as their proxies to account for confounding for the identification of the natural direct and indirect effects. We develop semiparametric theory for estimation and propose estimators that are robust to different types of model misspecifications. We use simulation studies to evaluate the performance of the proposed method.
