A discussion of competing events in epidemiologic studies of air pollution

Abstract

BACKGROUND AND AIM: Adverse health impacts of air pollution are well-documented in epidemiologic studies. However, one inferential challenge in estimating the effect of air pollution on chronic conditions arises from competing events. In observational studies, other health outcomes may preclude or “compete” with the outcome of interest. Although a variety of approaches and frameworks to consider competing events have been described in epidemiologic literature, they are underutilized in studies of air pollution. This research aims to present case studies demonstrating competing events and provides solutions to address inferential questions. METHODS: Three approaches were demonstrated to account for competing events, each with unique assumptions and target estimands. First, a controlled direct effect, not mediated by the competing event, was estimated with a cause-specific hazard ratio. Next, the total effect, including pathways through the competing event was estimated with a subdistribution hazard ratio. Finally, inverse probability weights were applied to correct for time-varying informed censoring due to a competing event and estimate a weighted cause-specific hazard ratio. Various sensitivity analyses will also be discussed. RESULTS:Findings from this study underscore the limited consideration of competing events in epidemiologic studies of air pollution and highlight three approaches to account for competing events. CONCLUSIONS:Consideration of competing events will allow for more robust inferences from epidemiologic studies. We provide recommendations for future investigators considering competing events in epidemiologic studies of air pollution and chronic conditions. KEYWORDS: Air pollution, epidemiology, causal inference