Publication Highlights

This study leverages Georgia fetal records, which included those with gestational length less than 20 weeks. We found exposure to traffic-related air pollution during early pregnancy was associated with increased risks of miscarriages.

This study utilizes ICD10 "Y" code that indicates location of injury for heat-related emergency department visits. We examined how increasing temperature exposure results in visits more or less likely to occur at 11 locations (e.g., street, recreational areas and public) compared to home residence. 

The case-crossover design is popular for estimating short-term health effects of environmental exposures. Here we extend the Bayesian additive regression tree methodology to help identify sub-populatons with different risks defined by multiple potential effect modifiers. 

Pregnant people represent a vulnerable population for severe outcomes of respiratory infections. Population-based studies on exposure to influenza infection and pregnancy outcomes are challenging due to exposure assessment and low incidence rate issues. Here we leverage the time-series design and report short-term associations between community-level influenza infection and risks of preterm.  

In a time-series analysis of count data and environmental exposures, the dispersion parameter is universally assumed to be constant over the study period. Here we use a double GLM framework and show that accounting for time-varying and covariate-depedent dispersion can result smaller standard error for the exposure of interest. 

Emergency department visits during influenza seasons represent a critical yet less examined indicator of the acute burden of influenza. Here we estimate ED visits attributable to influenza using hospital surveillance and hospital discharge data. We found that ED visit rate for respiratory diseases was almost 6 times larger compared to the subset of ED visits that resulted in hospitalization. This difference was particularly large for the 0-4 years age group. 

Estimating mortality due to respiratory infection is challenging because not all cases are recorded. We describe Bayesian spatial–temporal model that uses measures of infection activity to estimate excess deaths. The model allows for time-varying coefficients to better characterize associations between infection activity and mortality counts time series.