Heat-related mortality and adaptation to heat in the United States

Abstract

Background: In a changing climate, increasing temperatures are anticipated to have profound health impacts. These impacts could be mitigated if individuals and communities adapt to changing exposures; however, little is known about the extent to which the population may be adapting.

Objective: We investigated the hypothesis that if adaptation is occurring, then heat-related mortality would be decreasing over time.

Methods: We used a national database of daily weather, air pollution, and age-stratified mortality rates for 105 U.S. cities (covering 106 million people) during the summers of 1987-2005. Time-varying coefficient regression models and Bayesian hierarchical models were used to estimate city-specific, regional, and national temporal trends in heat-related mortality and to identify factors that might explain variation across cities.

Results: On average across cities, the number of deaths (per 1,000 deaths) attributable to each 10°F increase in same-day temperature decreased from 51 [95% posterior interval (PI): 42, 61] in 1987 to 19 (95% PI: 12, 27) in 2005. This decline was largest among those ≥ 75 years of age, in northern regions, and in cities with cooler climates. Although central air conditioning (AC) prevalence has increased, we did not find statistically significant evidence of larger temporal declines among cities with larger increases in AC prevalence.

Conclusions: The population has become more resilient to heat over time. Yet even with this increased resilience, substantial risks of heat-related mortality remain. Based on 2005 estimates, an increase in average temperatures by 5°F (central climate projection) would lead to an additional 1,907 deaths per summer across all cities.

Figure 1

Estimates of the change, from 1987 to 2005, in the excess number of deaths (per 1,000 deaths) attributable to each 10°F increase in the same day’s summer temperature. Intervals correspond to 95% confidence intervals (city-specific estimates) or PIs (regional and national estimates).

Figure 2

Temporal trends, from 1987 to 2005, in the excess number of deaths (per 1,000 deaths) attributable to each 10°F increase in the same day’s summer temperature, nationally in the United States, on average across age groups (A), and stratified by age group (B), cause of death (C), and geographical region (D). Abbreviations: IM, Industrial Midwest; NE, Northeast; NW, Northwest; UM, Upper Midwest; SC, Southern California; SE, Southeast; SW, Southwest. Asterisks in the key denote statistically significant trends.

Figure 3

Effect modification by local climate. City-specific (posterior mean) estimates of the change, from 1987 to 2005, in heat-related mortality risk (A) and baseline (1987) heat-related mortality risk (B), plotted against the average temperature over the study period. (C) Temporal trends, from 1987 to 2005, in heat-related mortality risk for cities at the 25th percentile (52.0°F) and 75th percentile (64.5°F) of average temperature. Heat-related mortality risk is defined as the excess number of deaths (per 1,000 deaths) attributable to each 10°F increase in the same day’s summer temperature. The solid lines show the estimated associations, and the shaded bands denote 95% PIs.

Figure 4

Effect modification by central AC prevalence. City-specific (posterior mean) estimates of the change, from 1987 to 2005, in the excess number of deaths (per 1,000 deaths) attributable to each 10°F increase in the same day’s summer temperature, plotted against the change in central AC prevalence over that period. The solid line shows the estimated associations, and the shaded band indicates 95% PIs.