Trends in Rates of Heat-Related Deaths Across Population Groups in the United States, 2000-2023

Abstract

Heat-related mortality is a growing public health concern as global temperatures continue to rise, yet little is known about how trends differ across various population groups in the United States. Using data from the Centers for Disease Control and Prevention Wide-ranging Online Data for Epidemiologic Research from 2000 to 2023 and joinpoint regression models, we examined heat-related mortality trends across major demographic population groups. Our analysis revealed increasing mortality rates across all groups, with steep rises since 2018. American Indian/Alaska Native populations had the most pronounced increase: the average annual percentage change (AAPC) was 8.7% from 2011 to 2023, accelerating to 27.8% during 2018-2023. Heat-related mortality rates per 100 000 population increased during 2019-2023 among populations that were Hispanic (AAPC = 28.7%) and non-Hispanic Black (AAPC = 28.6%), and the AAPCs were significantly higher than among non-Hispanic White people (AAPC = 5.8% overall and 23.9% during 2019-2023). Asian American/Pacific Islander people had the least pronounced overall increases in heat-related mortality rates but had significant increases recently (AAPC = 25.2% during 2020-2023). These findings suggest the importance of considering differential patterns in heat-related mortality across population groups.

Heat-related mortality is a growing public health concern as global temperatures continue to rise.^1-3 While overall trends in heat-related deaths have been examined, ⁴ few studies have analyzed how these trends differ across racial and ethnic groups in the United States. ⁵ Understanding such differences and potential variations is crucial for developing targeted interventions and public health policies.^5,6

Some racial and ethnic minority groups may be disproportionately affected by extreme heat events because of multiple intersecting factors, including residential segregation and occupational exposure.^5-10 Additionally, differences in the prevalence of heat-sensitive conditions (eg, cardiovascular disease, respiratory disease),^5,9,11 as well as demographic factors (eg, age distribution, pregnancy rates),^3,12 may amplify vulnerability to heat-related illness. Limited access to health care, ⁹ reduced health literacy on symptoms of heat-related illness, and barriers to using community cooling resources can compound these risks.^7,10 These geographic distributions of environmental conditions and occupational factors can lead to varying levels of heat exposure across populations, which may contribute to different patterns of chronic health conditions and ultimately to variations in heat-related mortality.^5,9 For example, Hispanic workers are more likely than workers in other racial and ethnic groups to be employed in industries with high heat exposure, such as agriculture or construction, which could increase their vulnerability to heat-related health issues.^5,13,14

While international studies have examined variations in heat-related mortality,^15,16 analysis of trends across racial and ethnic groups in the United States is limited, particularly in recent years, which have been among the warmest years on record.^1,2 Such an analysis is essential to understand how various social and environmental factors manifest in health outcomes and for developing evidence-based public health strategies in response to rising temperatures in the United States.

The present study examined trends in heat-related mortality rates across racial and ethnic groups in the United States from 2000 to 2023. We aimed to identify variations in heat-related vulnerability and provide timely evidence to inform an effective public health response to rising temperatures.

Methods

We analyzed data on heat-related mortality from 2000 to 2023 using the Centers for Disease Control and Prevention’s Wide-ranging Online Data for Epidemiologic Research (CDC WONDER) database.^17,18 CDC WONDER includes deaths from all 50 US states and the District of Columbia, including those occurring on Indigenous reservations and military bases, but it excludes deaths in US territories. ¹⁸ Deaths were identified by using the International Classification of Diseases, 10th Revision codes for heat-related causes (X30, T67, and P81.0) ¹⁹ listed as the underlying or contributing cause of death from the multiple cause-of-death files.^3,4 We calculated age-adjusted mortality rates per 100 000 population by using direct standardization to the year 2000 US standard population for groups as categorized in the database: non-Hispanic White, non-Hispanic Black, Hispanic, non-Hispanic Asian American/Pacific Islander, and non-Hispanic American Indian/Alaska Native. ¹⁷ We consider these demographic categories as reported on death certificates to examine patterns in heat-related mortality across population groups. ²⁰ This approach allows for the identification of geographic variations in heat-related mortality, acknowledging that reported demographic characteristics often correlate with regional differences in environmental exposures and socioeconomic factors that may influence health outcomes. ²⁰

To assess trends in heat-related mortality, we conducted joinpoint regression analysis using the Joinpoint Regression Program version 5.2.0 (National Cancer Institute). This method identifies points where significant changes occur in temporal trends and calculates the average annual percentage change (AAPC) for each identified segment. The AAPC represents the weighted average of the individual annual percentage changes during a fixed interval, providing a summary measure of the trend during the study period. ²¹ We conducted all analyses in July and September 2024. We considered P < .05 to be significant. Following CDC WONDER criteria, ¹⁸ death rates calculated with ≤20 deaths are noted as unreliable estimates, and data were suppressed for any group with <10 deaths in a given period. The University of Florida Institutional Review Board determined this study to be exempt from review as nonhuman subjects research because we used publicly available deidentified data.

Results

From 2000 to 2023, heat-related mortality rates increased significantly among populations with complete data coverage. Heat-related mortality rates among non-Hispanic White people increased, with an AAPC of 5.8% (95% CI, 3.5%-7.9%; P < .001) from 2000 to 2023, with age-adjusted mortality rates rising from 0.14 (95% CI, 0.12-0.15) to 0.55 (95% CI, 0.52-0.58) per 100 000 population (Table). Similarly, heat-related mortality rates among Hispanic populations increased significantly (AAPC = 5.4%; 95% CI, 2.9%-8.1%; P = .002), with age-adjusted mortality rates increasing from 0.18 (95% CI, 0.12-0.25) in 2000 to 0.76 (95% CI, 0.68-0.83) in 2023. Rates among non-Hispanic Black populations increased overall during the study period from 0.50 (95% CI, 0.41-0.59) to 0.94 (95% CI, 0.85-1.03) but not significantly (P = .09). Among groups with limited historical data, American Indian/Alaska Native populations (data available from 2011) had the most pronounced increase, with age-adjusted mortality rates increasing from 1.07 (95% CI, 0.69-1.6) in 2011 to 2.51 (95% CI, 1.91-3.24) in 2023 (AAPC = 8.7%; 95% CI, 3.2%-14.8%; P < .001). The overall trend in heat-related mortality rates during the period in which data were available for Asian American/Pacific Islander populations (data from 2017) was not significant (P = .24).

Table.

Heat-related deaths, age-adjusted mortality rates per 100 000 population, and average annual percentage change, by population groups, United States, 2000-2023 ^a

	Hispanic		Non-Hispanic American Indian/Alaska Native		Non-Hispanic Asian American/Pacific Islander		Non-Hispanic Black		Non-Hispanic White
Year	No. of deaths	Age-adjusted mortality rate (95% CI)	No. of deaths	Age-adjusted mortality rate (95% CI)	No. of deaths	Age-adjusted mortality rate (95% CI)	No. of deaths	Age-adjusted mortality rate (95% CI)	No. of deaths	Age-adjusted mortality rate (95% CI)
2000	44	0.18 (0.12-0.25)	<10	Unreliable	<10	Unreliable	132	0.50 (0.41-0.59)	296	0.14 (0.12-0.15)
2001	42	0.13 (0.08-0.19)	<10	Unreliable	<10	Unreliable	109	0.42 (0.34-0.50)	373	0.16 (0.14-0.17)
2002	68	0.24 (0.18-0.32)	<10	Unreliable	<10	Unreliable	152	0.56 (0.47-0.65)	390	0.17 (0.16-0.19)
2003	55	0.18 (0.13-0.25)	11	Unreliable	<10	Unreliable	79	0.26 (0.20-0.33)	257	0.11 (0.09-0.12)
2004	53	0.15 (0.10-0.21)	<10	Unreliable	<10	Unreliable	41	0.13 (0.09-0.18)	199	0.08 (0.07-0.09)
2005	94	0.30 (0.24-0.38)	10	Unreliable	14	Unreliable	152	0.51 (0.43-0.60)	420	0.19 (0.17-0.21)
2006	119	0.40 (0.32-0.48)	16	Unreliable	<10	Unreliable	224	0.72 (0.62-0.82)	628	0.27 (0.25-0.29)
2007	57	0.18 (0.13-0.24)	<10	Unreliable	10	Unreliable	132	0.43 (0.35-0.50)	336	0.15 (0.14-0.17)
2008	55	0.13 (0.10-0.18)	12	Unreliable	<10	Unreliable	112	0.35 (0.28-0.41)	311	0.14 (0.12-0.15)
2009	86	0.24 (0.18-0.30)	13	Unreliable	11	Unreliable	95	0.30 (0.24-0.37)	334	0.14 (0.12-0.16)
2010	74	0.21 (0.16-0.27)	12	Unreliable	11	Unreliable	196	0.58 (0.50-0.67)	500	0.23 (0.21-0.25)
2011	108	0.27 (0.22-0.33)	26	1.07 (0.69-1.6)	10	Unreliable	223	0.64 (0.55-0.73)	637	0.26 (0.24-0.29)
2012	101	0.26 (0.20-0.31)	17	Unreliable	12	Unreliable	155	0.42 (0.35-0.49)	544	0.21 (0.19-0.23)
2013	91	0.20 (0.16-0.25)	19	Unreliable	10	Unreliable	107	0.27 (0.22-0.32)	419	0.17 (0.16-0.19)
2014	63	0.14 (0.10-0.18)	18	Unreliable	<10	Unreliable	61	0.15 (0.11-0.19)	257	0.08 (0.07-0.10)
2015	68	0.15 (0.12-0.20)	15	Unreliable	14	Unreliable	99	0.26 (0.21-0.32)	370	0.14 (0.13-0.16)
2016	136	0.30 (0.25-0.36)	26	1.00 (0.65-1.48)	15	Unreliable	121	0.31 (0.25-0.36)	504	0.20 (0.18-0.22)
2017	133	0.29 (0.23-0.34)	22	0.86 (0.52-1.32)	36	0.17 (0.12-0.24)	111	0.27 (0.22-0.33)	549	0.23 (0.21-0.25)
2018	125	0.24 (0.19-0.28)	25	0.91 (0.58-1.37)	21	0.10 (0.06-0.16)	157	0.39 (0.33-0.45)	673	0.28 (0.26-0.31)
2019	150	0.28 (0.23-0.33)	21	0.73 (0.44-1.12)	24	0.12 (0.07-0.19)	154	0.34 (0.29-0.40)	540	0.22 (0.20-0.24)
2020	208	0.38 (0.32-0.43)	35	1.22 (0.84-1.71)	21	0.09 (0.05-0.14)	165	0.38 (0.32-0.44)	700	0.28 (0.26-0.31)
2021	196	0.36 (0.31-0.41)	62	2.50 (1.9-3.22)	29	0.13 (0.09-0.20)	236	0.56 (0.49-0.63)	1034	0.41 (0.38-0.43)
2022	273	0.48 (0.42-0.54)	58	2.43 (1.83-3.15)	39	0.19 (0.13-0.27)	248	0.57 (0.50-0.65)	1056	0.39 (0.37-0.42)
2023	411	0.76 (0.68-0.83)	62	2.51 (1.91-3.24)	42	0.17 (0.12-0.24)	407	0.94 (0.85-1.03)	1408	0.55 (0.52-0.58)
AAPC (95% CI) [P value] b	5.4 (2.9 to 8.1) [.002]		8.7 (3.2 to 14.8) [<.001]		4.4 (−3.8 to 15.2) [.24]		2.2 (−0.5 to 4.7) [.09]		5.8 (3.5 to 7.9) [<.001]

Abbreviation: AAPC, average annual percentage change.

^aData source: Centers for Disease Control and Prevention Wide-ranging Online Data for Epidemiologic Research.^17,18 Data are suppressed for any group with <10 deaths in a given period. Death rates are flagged as “unreliable” when calculated with a numerator of ≤20 deaths.

^bUsing joinpoint regression, with P < .05 considered significant.

The joinpoint analysis identified significant changes in trends across groups, particularly in recent years (2018-2023; Figure). A notable joinpoint occurred in 2018 among American Indian/Alaska Native people, marking the beginning of a steep increase in heat-related death rates (AAPC = 27.8%; 95% CI, 17.0%-40.9%; P < .001). The pattern of changes among Hispanic and non-Hispanic Black people was similar in 2019, with both groups having sharp increases in heat-related mortality rates (AAPC = 28.7%; 95% CI, 11.0%-67.1%; P < .001; and 28.6%; 95% CI, 6.9%-40.9%; P = .002, respectively). These recent increases were significantly higher than increases among non-Hispanic White people (AAPC = 23.9%; 95% CI, 8.5%-59.8%; P < .001 for trend, P = .04 for comparisons), which showed a joinpoint in 2019-2023. A significant trend change occurred among Asian American/Pacific Islander people in 2020 (AAPC = 25.2%; 95% CI, 2.0%-75.1%; P = .03).

Figure.

Age-adjusted mortality rates per 100 000 population for heat-related deaths, by population groups, United States, 2000-2023. Data source: Centers for Disease Control and Prevention Wide-ranging Online Data for Epidemiologic Research.^17,18

Discussion

Our findings reveal significant increases in heat-related mortality rates across all population groups in the United States, ⁴ with notable variations in absolute rates and temporal trends. While our analysis was limited in explaining observed differences, these trends align with studies documenting rising heat-related mortality as temperatures reach record highs.^1,2 As climate change continues to drive increasing temperatures, differences in environmental exposure based on geographic location and socioeconomic factors appear to influence heat vulnerability patterns across population groups.^5,9 Our results must be interpreted through the lens of race and ethnicity as social constructs, as our findings may reflect the complex interplay of social, economic, and environmental factors that vary geographically across the United States.^5,9

The consistently higher age-adjusted mortality rates among American Indian/Alaska Native populations (2-3 times higher than other groups) and the recent sharp increases across multiple groups warrant particular attention. While this broad category encompasses populations with distinct geographic distributions and climate-related risks, the elevated rates likely reflect multiple overlapping vulnerabilities: geographic concentration in regions experiencing severe heat events (particularly for American Indian populations in the Southwest), historical patterns of land dispossession leading to residence in areas with limited infrastructure, and reduced access to cooling resources and health care services.^3,5-8,13 The higher baseline rates observed among non-Hispanic Black people as compared with other racial and ethnic groups at the start of our study period are consistent with previous research suggesting associations with residential patterns and environmental exposures, although our study cannot determine causal factors.^6,8 Similarly, the marked recent increases among Hispanic populations may reflect occupational exposures, particularly in states with large agricultural and construction sectors where Hispanic workers are disproportionately represented in outdoor labor.^5,13,14

This study had several limitations. First, our analysis was constrained by differential data availability across racial and ethnic groups in CDC WONDER, with shorter time series for populations that were American Indian/Alaska Native (from 2011) and Asian American/Pacific Islander (from 2017) as compared with the full 2000-2023 period available for non-Hispanic White and non-Hispanic Black populations; moreover, given the data suppression policy, ¹⁸ trend analysis of these groups was limited. While our use of age-adjusted rates helps standardize age distributions across populations, the underlying US Census–based population estimates used as denominators may be affected by undercounting and changes in racial and ethnic identification between censuses. ¹⁸ Second, the accuracy of race and ethnicity categorization on death certificates may have affected our observed disparities, ²² with misclassification particularly affecting American Indian/Alaska Native and Hispanic populations, likely leading to an underestimation of heat-related mortality rates for these groups. Third, heat-related deaths may be underreported, particularly when heat is a contributing cause rather than the primary cause of death, potentially underestimating the true burden of heat-related mortality across all groups. Last, the ecologic nature of our study limits causal inferences about individual risk factors, because heat-related vulnerabilities and exposures may vary considerably within each racial and ethnic group. Because our analysis used aggregate population-level data, we could not examine individual factors or make detailed geographic (eg, state) or seasonal (eg, summer season) comparisons, particularly for smaller racial and ethnic populations where data were insufficient for analysis. ¹⁸

Despite these limitations, our study provides important insights into trends in disparities in heat-related mortality at a national level. Future research using more granular data sources could examine how these disparities intersect with regional and socioeconomic factors, particularly focusing on understanding the environmental and occupational drivers of observed differences in heat vulnerability and the recent accelerations in mortality rates. This approach would provide a more nuanced understanding of how climate, geography, and environmental conditions intersect to inform comprehensive public health strategies addressing climate-related health challenges.

Our findings suggest substantial and increasing differences in heat-related mortality patterns across population groups in the United States. Our results should be interpreted within the context of socioeconomic, environmental, and health care access factors that vary regionally and may contribute to observed differences between groups. Addressing these outcome differences requires a comprehensive understanding of how climate, geography, and regional population characteristics intersect. Future studies and interventions should focus on geographically specific analyses and develop strategies that enhance community resilience to and public health preparedness for extreme heat events.