Semiparametric causal mediation analysis with unmeasured mediator-outcome confounding

TL;DR

This paper introduces a semiparametric method for causal mediation analysis that accounts for unmeasured confounding between mediator and outcome, using heteroskedasticity restrictions to identify effects.

Contribution

It develops novel semiparametric estimators for natural direct and indirect effects under unmeasured confounding, with robustness to model misspecification.

Findings

Estimators perform well in simulations.

Application reveals effects of self-efficacy on fatigue.

Method effectively handles unmeasured confounding.

Abstract

Although the exposure can be randomly assigned in studies of mediation effects, any form of direct intervention on the mediator is often infeasible. As a result, unmeasured mediator-outcome confounding can seldom be ruled out. We propose semiparametric identification of natural direct and indirect effects in the presence of unmeasured mediator-outcome confounding by leveraging heteroskedasticity restrictions on the observed data law. For inference, we develop semiparametric estimators that remain consistent under partial misspecifications of the observed data model. We illustrate the proposed estimators through both simulations and an application to evaluate the effect of self-efficacy on fatigue among health care workers during the COVID-19 outbreak.