A Bayesian hierarchical model of environmental impact on human mortality and its spatial variation in the United States 2000-2005
Many environmental factors influence human mortality simultaneously. However, assessing their cumulative effects remains a challenging task. In this study we used the Environmental Quality Index (EQI), developed by the U.S. EPA, as a measure of overall environmental exposure for 2000-2005. We investigated the relative contributions of five different environmental domains (air, water, land, sociodemographic and built environment) on human mortality and their spatial variations across the U.S. The domain-specific effects were simultaneously modeled on county-level using a Bayesian hierarchical structure. The 2000-2005 mortality data for all U.S. counties (n = 3141) were obtained from the National Center for Health Statistics. We additionally stratified by Rural-Urban Continuum Codes (RUCC), Koppen climate region, states and their combinations, to account for spatial variations in the effect. Separate models were built for all-cause (non-accidental) and cause-specific death rates (neoplasms, circulatory diseases and respirational diseases). Models were adjusted for county-level characteristics including median age, percent white and population density obtained from the U.S. census.
The estimated effects of environmental quality on mortality varied by spatial location and by domain. We found that stratification by the combination of RUCC and climate regions to be most informative. Among the five domains, air had the strongest effects on all-cause death rate, where 1 standard deviation increase in air-EQI, a worsening of air quality, was associated with a 0.32±0.07 increase in deaths per 1000 people on average. Spatially, the highest adverse effects of worse air (0.84±0.22 deaths/1000) and water (0.23±0.07 deaths/1000) environmental quality were found in the northwestern U.S. The adverse effects of worse land and built environmental quality were found in most parts of the U.S. except in the western mountainous area. Finally, adverse effects of poor sociodemographic quality were found mostly in urbanized areas, e.g., in these areas higher poverty, lower education, and larger non-English speaking population were associated with higher mortality rate. Similar results were observed in the cause-specific models. This study assessed the cumulative effect of overall environmental quality on human mortality rates across the U.S. The relative impact of the effects and their spatial variation can provide further reference for environmental management of cumulative exposures. This abstract does not represent EPA policy.