Quantifying the Number of Heat-Related Deaths in the United States: An Estimate for 297 Populous Counties

Exposure to high ambient temperature (i.e., “heat”) is a widely recognized threat to public health. Using data on cause of death abstracted from death records, the Centers for Disease Control and Prevention estimates that an average of 658 deaths are directly attributable to heat each year in the United States. This estimate is frequently cited in the academic literature and in the popular media. However, given the absence of standardized criteria for classifying a death as “heat-related”, estimates derived from death records likely miss many cases where heat contributed to death from another cause (e.g., cardiovascular disease). Thus, estimates derived from death records likely underestimate the burden of heat-related mortality. In contrast, regression-based approaches allow for a more comprehensive assessment of the impact of heat on population health by estimating the number of deaths for which heat may have been a contributing factor. However, previous regression-based estimates have been geographically limited to locations in close proximity to weather stations.

Here, we used a regression-based approach in combination with temperature estimates from a high-resolution, gridded meteorological dataset (PRISM) to estimate the number of deaths attributable to heat in 297 counties representing 62% of the US population, 1997-2006. Specifically, we used distributed lag non-linear models with a quasi-Poisson distribution to estimate the 21-day cumulative association between daily mean temperature and all-cause mortality (excluding external causes) in each county, controlling for seasonal and long-term time trends, day of week, and federal holidays. We then used these county-specific associations to calculate the number of deaths attributable to heat in each county.

Between 1997 and 2006 a total of 12,992,584 deaths occurred in the study counties. Across all 297 counties, we estimate that an average of 5313 deaths per year (95% empirical confidence interval: 4502, 5972) are attributable to heat (defined as all temperatures greater than the location-specific minimum mortality temperature). Extreme heat (defined as temperatures greater than the 97.5^th percentile of the local daily mean temperature distribution) contributed to an average of 2218 deaths per year (95% empirical confidence interval: 1985, 2379), while moderate heat (defined as temperatures between the location-specific minimum mortality temperature and the 97.5^th percentile of the local daily mean temperature distribution) contributed to an average of 3094 deaths per year (95% empirical confidence interval: 2420, 3689). These results suggest that the number of heat-related deaths occurring in the United States is substantially larger than is often reported, and highlight the importance of moderately hot temperatures in driving the burden of heat-related mortality.