The Effects of Heat Exposure on Human Mortality Throughout the United States

Abstract

Exposure to high ambient temperatures is an important cause of avoidable, premature death that may become more prevalent under climate change. Though extensive epidemiological data are available in the United States, they are largely limited to select large cities, and hence, most projections estimate the potential impact of future warming on a subset of the U.S. population. Here we utilize evaluations of the relative risk of premature death associated with temperature in 10 U.S. cities spanning a wide range of climate conditions to develop a generalized risk function. We first evaluate the performance of this generalized function, which introduces substantial biases at the individual city level but performs well at the large scale. We then apply this function to estimate the impacts of projected climate change on heat-related nationwide U.S. deaths under a range of scenarios. During the current decade, there are 12,000 (95% confidence interval 7,400-16,500) premature deaths annually in the contiguous United States, much larger than most estimates based on totals for select individual cities. These values increase by 97,000 (60,000-134,000) under the high-warming Representative Concentration Pathway (RCP) 8.5 scenario and by 36,000 (22,000-50,000) under the moderate RCP4.5 scenario by 2100, whereas they remain statistically unchanged under the aggressive mitigation scenario RCP2.6. These results include estimates of adaptation that reduce impacts by ~40-45% as well as population increases that roughly offset adaptation. The results suggest that the degree of climate change mitigation will have important health impacts on Americans.

Keywords: climate change; heat exposure; heat‐related mortality.

Figure 1

Relative risk functions for all‐cause all‐age mortality as a function of daily temperature for 10 U.S. cities as documented in Weinberger et al. (2017). Summer mean temperatures for these cities are Atlanta 25.7°C, Boston 21.2°C, Chicago 22.5°C, Dallas 28.9°C, Houston 28.4°C, Los Angeles 21.8°C, Miami 28.4°C, New York City 23.1°C, Philadelphia 23.9°C, and Washington DC 24.4°C.

Figure 2

Coefficient values for the exposure‐response function binomial fits for each of the 10 cities and the regression across those points. R values for the regression are (a) 0.51 and (b) 0.62.

Figure 3

Heat‐related deaths for the indicated cities near the end of the century under Representative Concentration Pathway (RCP) 8.5 without population changes or adaptation. Values from Weinberger et al. (2017; filled bars) are annual averages for 2085–2095, whereas values from this study (open bars) are for 2090–2099, normalized to match county‐level population used in the earlier work. Uncertainties for Weinberger et al. (2017) include exposure‐response function (ERF) and temperature projections (and extend off the chart to −1,500 for Miami), whereas for this evaluation of the generalized equation we show ranges across the five models used for temperature projections.

Figure 4

State‐level heat‐related mortality. Values are shown in deaths/year per million persons by state for the indicated years and for the indicated future scenarios including (b–d) no‐adaptation, (e–g) adaptation to summer mean temperatures, or (h–j) adaptation to the change in temperature distribution (“lagged adaptation to OT”). Note the change in scales between near‐present (a) and future (b–j).

Figure 5

(top) Total and (bottom) per capita increases in nationwide U.S. heat‐related deaths over the 21st century under the three scenarios. Values include projected increases in population and account for potential adaptation.