|
Abstract:
|
Short-term exposure to air pollution has been associated with combined cardiorespiratory diseases, however determining the specific diagnoses, for example asthma, associated with pollution can help inform public health recommendations. Identifying associations between air pollution and cause-specific morbidity in time series studies can be challenging because small daily counts for specific diagnoses lead to low statistical power. We developed a Bayesian hierarchical modeling framework for conducting multicity studies of air pollution and cause-specific morbidity. Across 5 US cities, we first estimated city- and cause-specific associations between air pollution and emergency department (ED) visits with time series regression models. Next, we applied Bayesian hierarchical models that borrow information across diagnoses to estimate both multicity and city-specific associations between air pollution and cause-specific ED visits. This Bayesian modeling approach yields estimated associations that are attenuated relative to standard approaches, leading to more conservative effect estimates that better reflect the information available for each city and diagnosis.
|
