B1005
Title: Characterizing heterogeneity of causal effects in air pollution epidemiology via Bayesian causal inference
Authors: Falco Joannes Bargagli Stoffi - Harvard University (United States) [presenting]
Abstract: Several epidemiological studies have provided evidence that long-term exposure to fine particulate matter (PM2.5) increases mortality risk. Furthermore, some population characteristics (e.g., age, race, and socioeconomic status) might play a crucial role in understanding vulnerability to air pollution. To inform policy, it is necessary to identify mutually exclusive groups of the population that are vulnerable to air pollution. In the causal inference literature, the conditional average treatment effect (CATE) is a commonly used metric designed to characterize the heterogeneity of a treatment effect based on some population characteristics. A novel confounder-dependent Bayesian mixture model (CDBMM) is introduced to characterize causal effect heterogeneity. More specifically, the method leverages the flexibility of the dependent Dirichlet process to model the distribution of the potential outcomes conditionally to the covariates, thus enabling to: (i) estimate individual treatment effects, (ii) identify heterogeneous and mutually exclusive population groups defined by similar CATEs, and (iii) estimate causal effects within each of the identified groups. Through simulations, the effectiveness of the method is demonstrated in uncovering key insights about treatment effects heterogeneity. The method is applied to claims data from Medicare enrollees in Texas. Seven mutually exclusive groups are found where the causal effects of PM2.5 on mortality are heterogeneous.
