Estimation of excess mortality due to long-term exposure to PM_2.5 in continental United States using a high-spatiotemporal resolution model

Highlights

• Reducing PM2.5 to levels already seen in the cleanest part of each county would lower US deaths by 112,000 per year.

• CCapturing fine scale (1 km) PM2.5 levels and ZIP code baseline mortality rates results in larger estimated risks than coarser scale approaches.

• Almost all of the benefits occur in counties already meeting EPA’s standards, indicating tightened standards are needed to achieve these results.

• Fine scale mapping of attributable risk can identify hot spots for monitoring and intervention within urban areas.

Abstract

Background

Exposure to fine particulate matter (<2.5 mm in aerodynamic diameter, PM_2.5) pollution, even at low concentrations is associated with increased mortality. Estimates of the magnitude of the effect of particulate air pollution on mortality are generally done on a coarse spatial scale, such as 0.5°, and may fail to capture small spatial differences in exposure and baseline rates, which can bias results and impede the ability to consider environmental justice. We estimated the burden of mortality attributable to long-term exposure to ambient PM_2.5 among adults in the Continental United States on a 1 km scale, in order to provide information for decision makers setting health priorities.

Methods

We conducted a health impact assessment for 2015 using a model predicting U.S. PM_2.5 concentrations at a spatial resolution of 1 km cells. We applied a concentration-response curve from a recently published meta-analysis of long-term PM_2.5 mortality association which incorporates new findings at high and low PM_2.5 concentrations. We computed the change in deaths in each grid cell, based on its grid cell population, Zip code level baseline mortality rates, and exposure under two scenarios; a decrease of PM_2.5 exposure levels by 40% and a decrease of PM_2.5 exposure levels to the county minimum PM_2.5 concentrations.

Results

We estimated the deaths would fall by 104,786 (95% CI 57,016–135,726) and 112,040 (95% CI 63,261-159,116) attributable to 40% reduction and reduction to the county minimum PM_2.5 concentrations_, respectively. The greatest mortality impact due to 40% reduction in PM_2.5 was observed in California with; 11,621 (95% CI; 7156-15,989) and Texas with; 9616 (95% CI; 5798–13,352) excess deaths attributable to annual mean PM_2.5 concentrations of 9.54 and 9.12 μg m⁻³, respectively. Within city analyses showed substantial heterogeneity in risk. The estimated Attributable fraction (AF %) in locations with high PM_2.5 levels was 8.6% (95% CI 5.4–11.7) compared to the overall AF% of 4.9% (95% CI; 2.9–6.8). In comparison, results using county average PM_2.5 were smaller than the estimates from the 1 km PM_2.5 datasets. Similarly, estimates using county-level mortality rates were smaller than estimates based on Zip-code level mortality rates.

Conclusions

Our study provides evidence of major health benefits expected from reducing PM_2.5 exposure, even in regions with relatively low PM_2.5 concentrations. Spatial characteristics of exposure and baseline mortality (e.g., accuracy, scales, and variations) in disease burden studies can significantly impact the results.

Introduction

The negative health effects of ambient fine particulate matter air pollution (<2.5 μm in aerodynamic diameter, PM_2.5) are well established. Recent evidence show that even at very low concentrations, PM_2.5 is associated with increased mortality(Di et al., 2017a, Di et al., 2017b; Wei et al., 2020).

In the United States (US), air pollutant concentrations have been generally decreasing, but with substantial differences in reductions across metropolitan areas. On Dec. 14, 2012 the U.S. Environmental Protection Agency (EPA) strengthened the legally enforceable National Ambient Air Quality Standards (NAAQS) for fine particle pollution by revising the primary annual PM_2.5 standard to 12 μg per cubic meter (μg/m³), less stringent than the 10 μg/m³ guideline recommended by the World Health Organization (WHO). In a 2010 report, the EPA estimated that 62 U.S. counties did not meet the revised PM_2.5 standard, out of 3141 counties/equivalents (Schmidt et al., 2010). Hence most of the risk of particle-associated mortality occurs in counties meeting current standards.

Many health impact assessments have primarily used exposure estimates at a resolutions of 50 or 12 km, and population and estimates of baseline rates at national or county levels. Since areas with higher population density tend to have higher exposure, and may have different baseline mortality and disease rates, this can bias the results of the health impact assessments, and impede identification of vulnerable communities (Liu et al., 2019).

Accurate exposure assessment of PM_2.5 is essential of to explore its adverse health effect on the population. A number of models have been developed to estimate PM_2.5 exposure, including satellite-based aerosol optical depth (AOD) models, land-use regression or chemical transport model simulation (Just et al., 2015; Marais et al., 2019; Tsai et al., 2015). We recently published a hybrid model which incorporates multiple chemical transport models, land use, and AOD with ensemble averaging of machine learning, which predicts daily PM_2.5 on a 1 km grid across Continental United States (Di et al., 2019).The high spatial resolution of this model also enables urban health impact assessments to account for within-city variability in PM_2.5 concentrations.

Similarly, we have addressed the error in how many people are exposed by using the gridded population of the globe (SEDAC) to produce population estimates for the same grid cells (1 km) as the exposure model, and have refined the baseline rate estimates by downscaling county level mortality rates with Zip code level mortality rates from Medicare.

The aim of this study is to determine the burden of chronic exposures to PM_2.5 on all-cause mortality across Continental United States by taking advantage of recent advancements in high resolution model of PM_2.5 concentrations and fine scale of mortality statistics, to explore the patterns within major cities, and assess the difference from what would have been obtained using more traditional spatial scales.