Causal Inference in the Estimation of the Effects of Psychoactive Medications on Delirium

It is a standard practice that causality should only be drawn from a randomized control trial, and not from an observational study. However, in observational studies, the intention is often to interpret associational effects as causal effects. In pharmacoepidemiology, one is often interested in the causal effects of the drugs on a particular health outcome. In addition, in many of these studies, exposure to drugs is time-varying, and the health outcome of interest is measured longitudinally. In such a setting, the health outcome in one time window affects the prevalence and dose level of the drugs given in the next time window, which in turn, affect the health outcome in subsequent time window. This results in a causal pathway which has the intermediate health outcome acting as both a confounder and a mediator. Traditional modeling methods cannot be used to deal properly with this situation. In this talk, I will present a class of models for causal inference in observational studies, and I will demonstrate the application of these models in the estimation of the effects of five groups of psychoactive medications on delirium persistence in a cohort of elderly patients.

Lunch will be provided.
Hosted by: Dr. Nananda Col