Estimating Direct Effects on a Dichotomous Outcome Using Logistic Structural Direct Effect Models

• Activity Number: 451
• Type: Topic Contributed
• Date/Time: Wednesday, August 4, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Stijn Vansteelandt*+
• Companies: Ghent University
• Keywords: causal inference ; mediation ; structural nested model ; logistic regression ; intermediate confounding ; time-varying confounding

Abstract

We consider the problem of assessing whether an exposure affects a dichotomous outcome other than by modifying a given mediator. The standard approach is logistic regression, adjusting for both exposure and mediator, but is known to be biased in the presence of confounders for the mediator-outcome relationship. Because additional regression adjustment for these confounders is only justified when they are not affected by the exposure, marginal structural modeling has been advocated as a preferred approach, but is not ideally tailored to the analysis of mediators that are continuous or have strong measured predictors. In this presentation, we overcome this limitation by developing inference for a novel class of causal models, which are closely related to Robins' logistic structural direct effect models, but do not inherit their difficulties of estimation.