Causal Inference in Pharmacoepidemiology

Abstract

The causal inference framework can be employed to quantify causal effects in both randomized and non-randomized settings. Often in pharmacoepidemiology research, study designs lack randomized interventions that allow for causal inference. Yet, there are important and meaningful causal questions to address for non-randomized interventions. In this chapter, we provide an overview of the causal inference paradigm, review current methodology, and discuss applications of these concepts to strengthen and improve pharmacoepidemiologic studies. Specifically, we introduce marginal structural models fit using inverse probability weights and discuss additional advanced topics, including survival analyses, time-varying exposure, and instrumental variables. We conclude the chapter with references to relevant software packages.