Department of Biostatistics Seminar/Workshop Series

An empirical Bayes approach to estimate associations between air pollution and cause-specific cardiorespiratory emergency department visits

Jenna R. Krall, PhD, Assistant Professor Department of Global Health and Community Health, College of Health and Human Services, George Mason University

Multicity epidemiologic studies of air pollution and health are critical for understanding city-to-city heterogeneity in health effects and for estimating health effects to inform regulation. Bayesian hierarchical models are commonly used to estimate health effects associated with short-term air pollution exposure in multicity studies; however, these methods generally require data from a large number of cities. We developed an empirical Bayes framework for estimating associations between pollution and health in small multicity studies. Our approach facilitates comparisons across cities by incorporating additional information, such as estimated health effects for several related health outcomes. In five US cities, we estimated associations between 12 air pollutants and cause-specific emergency department (ED) visits for cardiorespiratory diseases. Additionally, we extended our approach to a multipollutant framework by estimating associations between air pollution mixtures and cause-specific ED visits. Our empirical Bayes approach synthesizes evidence across exposure-outcome combinations for multiple pollutants and cause-specific ED visits in multicity studies.