Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Jan 8.
Published in final edited form as: Addiction. 2025 Oct 6;121(2):275–285. doi: 10.1111/add.70191

Heat Exposure and Drug Overdose Mortality in the United States

Julia M Dennett 1,4, Daniel Carrión 2, David A Fiellin 3, Gregg S Gonsalves 1
PMCID: PMC12776456  NIHMSID: NIHMS2127334  PMID: 41048198

Abstract

Background and aims:

Understanding the relationship between heat and drug overdose mortality is critical given the ongoing overdose crisis and rising global temperatures driven by climate change. Evidence from local jurisdictions indicates that exposure to heat is associated with increased drug overdose fatalities, but the widespread impact of heat on national drug overdose deaths is undetermined. We aimed to determine the effect of heat exposure on drug overdose mortality in the United States.

Design:

An observational study using monthly, county-level data from 1999 to 2020 and a fixed effect model estimated with linear regression. The model accounts for time-invariant county-month factors and state-level time-specific shocks.

Setting:

All counties in the continental U.S. during the months of June to September.

Participants:

Decedents in the National Center for Health Statistics restricted Vital Statistics mortality data with a cause of death of a drug overdose, aggregated to county-level mortality rates.

Intervention:

The monthly average maximum heat index (measured in degrees Celsius) collected from the Centers for Disease Control and Prevention National Environmental Public Health Tracking Network

Measurements:

All drug overdose mortality rate per 100,000 population.

Findings:

The county-level, monthly analytic sample represented 3,108 counties in the continental U.S. between 1999 and 2020 from June to September. A one-degree Celsius increase in the heat index was associated with an increase in all drug overdose mortality by 0.0098 deaths per 100,000 population (95% CI 0.0057–0.014, p<0.001), with significant effects for deaths related to opioids (0.0060 deaths, 95% CI 0.0026–0.0095, p<0.001), cocaine (0.0028 deaths, 95% CI 0.0012–0.0045, p<0.001), and psychostimulants (0.0028 deaths, 95% CI 0.0015–0.0041, p<0.001). In further analyses, larger impacts were observed after 2013, in counties with greater levels of social vulnerability, and in suburban and urban counties. Estimates suggest that approximately 150 excess drug overdose deaths occurred per year during the hottest periods due to heat exposure.

Conclusions:

Exposure to heat was associated with increases in drug overdose deaths in the United States in recent decades. These findings stress the need for interventions to prevent drug overdoses during periods of heat and highlight the global consequences of climate change for drug overdose mortality.

Keywords: Heat, drug overdose, mortality, climate change, opioids, cocaine, psychostimulants

INTRODUCTION

Rising temperatures due to climate change represent a substantial threat to global public health (1). Identifying the relationship between heat and specific health outcomes will be critical to shield people who may be at-risk of harmful effects. Exposure to extreme heat is linked to numerous adverse health events, including increased all-cause mortality, suicide rates, cardiovascular and respiratory diseases, and violent crime (2–7).

Heat exposure may also be particularly harmful to people who use drugs. Drug use can directly cause the body to overheat (8, 9) and can reduce a person’s ability to sense and respond to overheating (9, 10). Moreover, heat may exacerbate the risks of drug use; for example, respiratory depression due to opioid use (11) may interfere with respiratory compensation needed for cooling and both heat and stimulant use are associated with cardiovascular issues (5, 12). Polysubstance use, especially the co-use of opioids and stimulants, may be especially hazardous. Stimulant use is associated with increased body core temperature, and their combination in people who use opioids may increase risk. In addition, people who use drugs may experience additional risk factors that increase vulnerability to heat, including low income (e.g., less access to air conditioning), homelessness (e.g., greater outdoor exposure), residing in a city (e.g., the urban heat island effect, where cities experience elevated heat from less green space and increased heat absorption from impervious surfaces), and the use of psychotropic medications for comorbid conditions, which may also affect thermoregulation (13, 14).

Understanding the effect of heat on drug overdoses is additionally important given the ongoing overdose crisis, which has exacted a staggering toll in North America. Overdose deaths have more than doubled since 2015 (15), although recent surveillance shows a modest decrease in annual overdose mortality. Over 105,000 Americans died from a drug overdose in 2023 (16). Most deaths are the result of highly potent synthetic opioids such as fentanyl, although stimulants and polysubstance use are important contributors (15, 16).

Observational research has identified an association between heat and fatal drug overdoses using data from local jurisdictions. Two studies in New York City and one in Quebec, Canada using data from before the mid-2010s documented a relationship between high temperatures and fatal cocaine overdoses, but not deaths from opioids or other drugs (17–19). A study that includes more recent data from British Columbia, Canada linked heat to deaths related to cocaine alone or with polysubstance use (20). Beyond mortality, additional observational studies find that heat is associated with increased morbidities related to drug use, including emergency department visits, hospital visits, and acute poisonings (21–23). Further research shows that heat increased deaths due to unintentional injuries in the United States, a broad outcome that includes drug overdoses as well as traffic mortality, poisonings, and accidents in the workplace, as part of an investigation into climate change adaptation (24). Despite these compelling findings, however, evidence of the effects of heat on overdose deaths at a national level are unavailable.

In this study, we investigated the impacts of heat exposure on all drug overdose mortality using monthly (June to September), county-level mortality data for the continental U.S. spanning 1999 to 2020. This analysis is the first to address this research question by applying fixed effect methods that address both location-specific and period-specific factors on a national, county-level panel dataset.

METHODS

Data sources

To represent heat exposure, we used the monthly average maximum heat index (measured in degrees Celsius) (25). The heat index incorporates both temperature and relative humidity. A high relative humidity reduces the rate at which sweat evaporates, which makes it more difficult for the body to stay cool (26) and makes the ambient temperature feel hotter than the dry bulb temperature would indicate. As a result, the heat index better captures the health risks of heat exposure than temperature alone. Beyond the heat index, we additionally used alternative definitions of heat exposure from several data sources in robustness tests, including temperature, precipitation, and pollution variables.

Our primary outcome was the all drug overdose mortality rate per 100,000 population. County-level death counts were calculated from the National Center for Health Statistics (NCHS) restricted Vital Statistics multiple cause of death data (27), which includes information on the month of death but not the date. We identified drug overdoses using ICD 10 codes X40-X44 (unintentional), X60-X64 (suicide), X85 (homicide), or Y10-Y14 (undetermined) as the underlying cause of death (28). Counts were converted to mortality rates using population data (29).

We also examined multiple secondary outcomes. First, we analyzed overdose mortality rates by drug type, which we defined as a drug overdose with a multiple cause of death code of T40.0-T40.4 or T40.6 (any opioid), T40.1 (heroin), T40.2 (natural and semisynthetic opioids), T40.4 (synthetic opioids), T40.5 (cocaine), and T43.6 ( “psychostimulants with abuse potential” other than cocaine, e.g., methamphetamines; we refer to this category throughout the study as “psychostimulants”) (28). We also examined overdose mortality by single- vs. polysubstance use for combinations of any opioid, cocaine, and psychostimulants (e.g., any opioid without cocaine or psychostimulants; any opioid and cocaine; etc.), as well as by demographic characteristics. In addition, we analyzed other potentially heat-sensitive mortality outcomes to place our results for drug overdose mortality into a wider context. Finally, we used cancer mortality rates in placebo tests; these endpoints were selected because they were unlikely to exhibit temporal variation and should not be sensitive, or be comparatively less sensitive, to heat exposure.

To investigate the heterogeneous effects of heat exposure in our sample, we used data on several county features, including median household income in 2000, social vulnerability (Social Vulnerability Index, SVI) in 2000 (defined as how susceptible a population’s health may be to external shocks like natural disasters or pandemics), and urbanicity (urban, suburban, and rural) in 2006. We selected data from years earlier in our study period to minimize the influence of the opioid crisis on these values (e.g., counties more adversely impacted by overdose deaths may have lower income later in the study period), though we also used classifications defined later in the sample (2020, 2020, and 2013, respectively) as a robustness check.

See Section 1 of the Methods Appendix for additional details on all data sources and variable construction.

Analytic sample

Our final analytic sample was a monthly, county-level panel data set spanning 1999 to 2020 for the continental U.S. We focused on June to September, which are the hottest months of the year in this region (Figure S1). Consistent with other heat exposure studies, we excluded counties in Alaska and Hawaii (30–32), which may have uniquely different climates and differing periods of elevated heat in comparison to the contiguous U.S. We also took several steps to account for historical county changes across data sets and maintain consistent linkages over time (specific details on the construction of the analytic sample are available in Section 2 of the Methods Appendix).

Statistical analysis

We employed a fixed effects model to analyze the effect of heat exposure on drug overdose mortality, which we estimated using linear regression. Our model includes county-by-month fixed effects to account for time-invariant county-month factors (e.g., geography and fixed regional adaptations to heat). As a result, our analysis identified effects using fluctuations in the heat index within a county-month across years (e.g., New Haven County, CT in July for 1999, 2000, etc.). We also included state-by-year fixed effects to account for state-level time-specific shocks (e.g., implementation of state drug policies and secular trends in temperature and drug use). Standard errors were clustered by county-month and regressions were weighted using the county population. See Section 3 of the Methods Appendix for additional information on the statistical analysis and specifications.

Our estimate represents the effect of a one-degree Celsius increase in the heat index on all drug overdose deaths under the identifying assumption that, conditional on fixed county-month characteristics and year-specific factors that impact all counties in a state, we are identifying effects using random variation in the heat index. Other studies analyzing the impacts of heat exposure on health have used this or similar approaches (3, 4).

In further analyses, we investigated heterogeneous effects of heat exposure by time period (before versus after the 2013 diffusion of fentanyl and increase in polysubstance use) (15) and by geography (east versus west of the Mississippi River, which provides a historical division in heroin and fentanyl availability) (33). We additionally examined heterogeneity by three county characteristics that may interact with heat to influence drug overdose mortality. First, we evaluated median household income to investigate if populations living in wealthier areas were more insulated from heat than populations living in less wealthy areas (e.g., due to improved access to resources like air conditioning and green spaces). We classified counties into having above vs. below the median value of median household income to allow for a broad comparison across all income levels. Second, we looked at social vulnerability to determine if populations living in areas with greater susceptibility to external shocks were more exposed to heat compared to populations living in areas with less susceptibility. Similar to median household income, we classified counties into having above vs. below the median value of social vulnerability to broadly compare high vs. low levels of social vulnerability. Finally, we analyzed urbanicity to assess if suburban and urban counties have magnified heat exposure compared to rural areas due to the urban heat island effect (13) or if they have distinctive drug supplies (34) that may interact differentially with heat. See Section 1 of the Methods Appendix for additional details on the construction of these variables. We stratified our fixed effects model by these features and used interaction models to determine if the difference across counties were statistically significant.

We conducted multiple analyses to examine the validity of our findings. These include using alternative definitions of heat exposure; modifying our regression specification by using different time fixed effects, not weighting the regression by population, and applying higher levels of standard error clustering; incorporating lagged overdose rates as controls; excluding the year 2020 (and thus the COVID-19 pandemic) from the analytic sample; dropping individual Census Divisions; testing for a nonlinear relationship between heat and overdose mortality; including the lagged heat index as an exposure variable; and lastly conducting placebo tests using cancer mortality as an outcome. See Section 4 of the Methods Appendix for detailed information on our robustness and placebo tests.

Finally, to better understand the magnitude of our results, we estimated the excess number of drug overdose deaths due to heat exposure during our sample period using our main estimate, deviations in heat exposure relative to earlier years, and county populations. A study examining the impacts of heat exposure on all-cause mortality conducted similar calculations (3). See Section 5 of the Methods Appendix for detailed information on our estimation.

This study was determined to be not human subjects research by the Yale University Institutional Review Board. The analysis plan was not pre-registered, and our analyses are therefore exploratory.

RESULTS

Sample characteristics

The analytic sample includes 273,488 observations (representing 3,108 counties) and spans 1999 to 2020 from June to September for each year. During this study period, the mean monthly average maximum heat index value was 30.3 degrees Celsius (standard deviation is 5.1 degrees) and the median monthly drug overdose mortality was 0.9 deaths per 100,000 (IQR is 1.2 deaths per 100,000) (Table 1; summary statistics for all variables presented in Tables S1a and S1b). Over half of the counties included in the sample were located east of the Mississippi River and had a mean median household income of approximately $44,000 and a median SVI ranking of 0.55 in 2000. Roughly 65 percent of the counties were classified as rural in 2006, while two percent were classified as urban. We additionally calculated summary statistics for counties with heat index values above versus below the median value to document nationwide differences by level of heat exposure. Counties with a heat index above the median value were slightly more likely to be located west of the Mississippi River, have lower mean median household income, have greater social vulnerability, were slightly more likely to be classified as suburban and less likely to be classified as urban, have a younger population, and have a greater share of non-Hispanic Black and Hispanic people and a smaller share of non-Hispanic White people.

Table 1: Summary statistics of the analytic sample and county-level characteristics.

For the analytic sample summary statistics, rows labeled with (Median, IQR) represent the median and interquartile range; otherwise summary statistics represent means and standard deviations. Means and standard deviations (for variables with normal distributions) and medians and interquartile ranges (IQR, for variables with other distributions) were calculated using county-level, monthly data (3,108 counties representing June to September spanning 1999 to 2020) and were weighted using population. For county characteristics, rows labeled with (No., %) are reported as the number of counties and corresponding percent of total counties with that feature; rows labeled with (Median, IQR) represent the median and interquartile range; otherwise summary statistics represent means and standard deviations. These means and standard deviations (for variables with normal distributions) and medians and interquartile ranges (IQR, for variables with other distributions) were calculated using county-level data and were weighted using the mean annual county population. County-level data for demographic features (sex, age group, and race/Hispanic origin) were generated as the share of the total annual population during the study period with that characteristic. The heat index is the average maximum heat index in degrees Celsius. In the analytic sample (in the county-level data), the median heat index value is approximately 30.63 degrees (30.67 degrees). Mississippi River is abbreviated as “MS River”. Social Vulnerability Index is abbreviated as “SVI”.

Full Sample Below or equal to median heat index Above median heat index
(N=273,488) (N=136,748) (N=136,740)
(3,108 counties) (1,554 counties) (1,554 counties)
Mean SD Mean SD Mean SD
(No.) (%) (No.) (%) (No.) (%)
(Median) (IQR) (Median) (IQR) (Median) (IQR)
Analytic sample
 Avg. max. heat index (degrees C) 30.3 5.1 26.4 2.8 34.9 2.7
 All drug overdoses (deaths per 100,000; Median, IQR) 0.9 1.2 0.8 1.1 0.9 1.3
County characteristics
 East of MS River (No., %) 1,606 51.7 859 55.3 747 48.1
 Median household income, 2000 ($) 44,332 11,068 47,224 11,591 40,363 8,896
 SVI, 2000 (Total percentile ranking; Median, IQR) 0.55 0.47 0.46 0.46 0.67 0.38
 Rural, 2006 (No., %) 2,023 65.1 1,013 65.2 1,010 65.0
 Suburban, 2006 (No., %) 1,022 32.9 505 32.5 517 33.3
 Urban, 2006 (No., %) 63 2.0 36 2.3 27 1.7
 Male (%) 49.2 1.2 49.1 1.1 49.2 1.4
 Female (%) 50.8 1.2 50.9 1.1 50.8 1.4
 Aged 0 to 29 (%) 40.6 4.4 39.9 4.0 41.5 4.9
 Aged 30 to 59 (%) 40.5 2.6 41.0 2.4 39.8 2.6
 Aged 60 and older (%) 19.0 4.3 19.2 3.6 18.7 5.1
 Black (non-Hispanic) (%; Median, IQR) 9.0 15.5 6.9 10.8 12.4 17.2
 Hispanic (any race) (%; Median, IQR) 9.2 19.3 8.5 16.1 10.5 23.9
 White (non-Hispanic) (%; Median, IQR) 68.2 35.3 73.9 36.7 60.4 31.2
Open in a new tab

The average maximum heat index has increased over time in our analytic sample (Figure S2). Geographically, the counties with the greatest heat index were further south, whereas the counties with the greatest variation in the heat index across years (measured using standard deviation) were in the Midwest, centered around Missouri, Iowa, Illinois, and surrounding states (Figure S3).

Drug overdose mortality outcomes

We find that heat exposure was associated with an increase in the rate of all drug overdose deaths. An increase of one-degree Celsius corresponded to an increase in all drug overdose deaths within a county-month by 0.0098 per 100,000 population (95% CI 0.0057–0.014, p<0.001) (Figure 1). When examining drug overdose deaths related to select drug types, we find that a one-degree Celsius increase in heat exposure was associated with increases in deaths from any opioid by 0.0060 deaths (95% CI 0.0026–0.0095, p<0.001), cocaine by 0.0028 deaths (95% CI 0.0012–0.0045, p<0.001), and psychostimulants by 0.0028 deaths (95% CI 0.0015–0.0041, p<0.001) per 100,000. When focusing on opioid subtypes, we find that the estimates for heroin and synthetic opioids were positive and statistically significant, while the estimates for natural and semisynthetic opioids were also positive, but not significant.

Figure 1: Effect of heat exposure on drug overdose mortality.

Figure 1:

Open in a new tab

This figure plots estimates with 95% confidence intervals. Heat exposure is the average maximum heat index (in degrees Celsius). N=273,488 (county-year-months).

Moreover, we find statistically significant, positive effects when a death is related to any single selected drug type alone (any opioid, cocaine, or psychostimulants), as well as when deaths are related to any opioid in combination with cocaine or any opioid in combination with psychostimulants. Deaths related to both cocaine and psychostimulants and deaths related to all three drug categories were not statistically significant, though there were comparatively few deaths from these combinations in our sample (both had a mean of 0.01 deaths per 100,000 in the analytic sample compared to a mean of 0.13 deaths for opioids and cocaine, and 0.07 deaths for opioids and psychostimulants, respectively; Table S1a).

As an additional step, we calculated the relative effect associated with these estimates using the analytic sample mean for each drug type category to determine which drug types were most affected by heat exposure (Table S2). We find that an increase of one-degree Celsius was associated with an increase in mortality from any opioid by 0.9 percent (with synthetic opioids increasing by 1.7 percent), cocaine by 1.4 percent, and psychostimulants by 2.0 percent. For single and polysubstance use, there were also large impacts for psychostimulants only, cocaine only, and opioids and psychostimulants (2.7 percent, 1.9 percent, and 1.4 percent, respectively). These findings suggest that stimulants like cocaine or methamphetamines may be comparatively more sensitive to heat than opioids, excluding synthetic opioids like fentanyl.

By demographic characteristics

When examining the effects of heat exposure on the drug overdose mortality of specific demographic groups, we find large, statistically significant effects for men, people aged 30 to 59, and the non-Hispanic White population (Figure S4). This result is consistent with the demographic groups that have been frequently associated with fatal overdoses during the opioid epidemic (35). We also find statistically significant effects for women and people aged 0 to 29, though the magnitudes of the estimates are smaller compared to the estimate for men and people aged 30 to 59, respectively. Heat exposure had a positive, but not statistically significant, effect for people aged 60 and older, as well as non-Hispanic Black and Hispanic populations.

Heat-related mortality outcomes

To better understand our results, we compared the impacts of heat exposure on drug overdose deaths to the impacts of heat exposure on other causes of death that may be related to heat, which we estimated using our framework (Figure 2). Consistent with the literature, we find statistically significant, positive effects for all-cause mortality, all cardiovascular causes, and suicides, though we do not see effects for chronic lower respiratory causes or assault fatalities (2–7). We also find statistically significant positive effects for natural heat-related causes, which includes hyperthermia, heatstroke, and heat exhaustion, but not for dehydration. Beyond the comparison to our drug overdose mortality results, these cause-specific findings should be separately considered in the context of their literatures, as our estimates may differ in terms of study setting, methods, or data.

Figure 2: Effect of heat exposure on other potential heat-sensitive causes of mortality and cancer deaths.

Figure 2:

Open in a new tab

This figure plots estimates with 95% confidence intervals. Heat exposure is the average maximum heat index (in degrees Celsius). N=273,488 (county-year-months).

Heterogeneous effects

Next, when considering heterogeneous effects, we show that heat exposure was associated with increases in all drug overdose mortality both before and during the period in which fentanyl began to dominate the opioid supply beginning in 2013, as well as both east and west of the Mississippi River (Figure 3). However, we also find that while the period 2013 to 2020 was statistically significantly larger than the period pre-2013 (p=0.032), geographic differences were not (p=0.62) (Table S3).

Figure 3: Heterogeneous effects of heat exposure on all drug overdose mortality.

Figure 3:

Open in a new tab

This figure plots estimates with 95% confidence intervals. Regressions were stratified by listed time periods, geographies, and county characteristic categories. Heat exposure is the average maximum heat index (in degrees Celsius). In county-year-months, the number of observations for each stratified regression are as follows: Main results, N=273,488; Years pre-2013, N=174,032; Years 2013+, N=99,456; West of the MS River, N=132,160; East of MS River, N=141,328; Below median 2000 income, N=136,752; Above median 2000 income, N=136,736; Least vulnerable 2000 SVI, N=136,736; Most vulnerable 2000 SVI, N=136,752; Rural in 2006, N=178,024; Suburban in 2006, N=89,920; and Urban in 2006, N=5,544. Income represents median household income. Mississippi River is abbreviated as “MS River”.

Moreover, we find that statistically significant effects of heat exposure are concentrated in counties with higher levels of median household income in 2000, both the least and the most socially vulnerable counties in 2000, and suburban and urban counties in 2006 (Figure 3; we find similar results using county characteristic definitions from later in our sample, Figure S5). According to our interaction models, however, there were not any significant differences by median household income (p=0.47), while the most vulnerable SVI counties had significantly larger effects than the least vulnerable (p=0.0046) and both suburban and urban counties had significantly larger effects than rural counties (p<0.001 and p<0.001, respectively) (Table S3).

Validity of results

Our estimates were robust to using alternative definitions of heat exposure (Figure S6) and modifying our specification (Figure S7). Furthermore, our results were not affected by including previous overdose death rates as controls (Figure S7), focusing on the period before the COVID-19 pandemic (Figure S7), or dropping individual Census Divisions from our analytic sample (Figure S8). This suggests that neither spurious correlations between heat and drug overdose deaths over time nor single regions of the nation, respectively, are driving the results. Next, when modeling the heat exposure using dummies for ranges of the heat index, we observed a linear relationship between heat and drug overdose deaths, especially above 19 degrees Celsius (Figure S9). Similarly, our plots of binned regression residuals showed a linear association between the heat index and overdose deaths fatalities (Figure S10).

Additional tests indicated that the effect of the heat exposure from the previous month on current drug overdose mortality was not statistically significant and had a limited impact on the current month’s heat exposure (Figure S11). Moreover, our placebo tests show that heat exposure did not have a statistically significant effect on our unrelated cancer outcomes (Figure 2), as expected.

Excess mortality

Finally, we approximated the excess number of drug overdose deaths related to heat exposure during our sample period to clarify the magnitude of our estimates. We find that heat exposure was associated with a total of 3,310.5 excess drug overdose deaths (lower estimate: 1,940.6, upper estimate: 4,680.3) during our sample period. This corresponds to an average of 150.5 excess deaths during the hottest months of each year (lower estimate: 88.2, upper estimate: 212.7). Estimated excess deaths were generally increasing over time (Figure 4), corresponding to increases in the heat index.

Figure 4: Estimated number of excess drug overdose deaths due to heat exposure by year (June to September).

Figure 4:

Open in a new tab

Shaded bands represent lower and upper estimates based on 95% confidence intervals of our main estimate.

DISCUSSION

Using national data from the U.S. spanning 1999 to 2020 for June to September and a fixed effects strategy, we find that heat exposure was associated with an increase in all drug overdose mortality. This includes overdoses related to opioids, cocaine, and psychostimulants (including methamphetamines). Our results align with documented effects of heat on all-cause mortality, all cardiovascular causes, and suicides (2–6), as well as natural heat-related causes. We also find that effect sizes were larger during the period of rising prevalence of fentanyl in the drug supply and in counties with greater levels of social vulnerability and in suburban and urban counties. Our results were robust to multiple modeling choices and consistent with the findings of placebo tests. We estimated that heat exposure corresponded to approximately 150 excess drug overdose deaths during the hottest months of each year, with a greater number of deaths occurring in more recent years.

Policy implications

Our findings have important policy implications. A greater understanding of the risks of drug use in heat can provide policymakers with information around which to plan and implement additional strategies to prevent drug overdoses. Some potential interventions may reduce heat exposure for all populations, including cooling stations, access to water, and improvements to the built environment like expanding available vegetation and parks. Others may be more specific, such as targeted outreach to people who use drugs and their health care providers to increase awareness of heat risks. Furthermore, enhanced public health surveillance related to heat and drug use may help local stakeholders quantify and prevent overdose deaths. For example, the Department of Public Health in Maricopa County, AZ issued a report on heat deaths that found that more than half of heat-related fatalities in the county were associated with drug use (36).

Policymakers should also be aware that ongoing global warming will have consequences for substance use and the opioid epidemic. Our estimates suggest that the effect of heat exposure on drug overdose deaths is similar to its effect on other causes of mortality linked to climate change, such as suicides (4). People who use drugs should be included when considering populations at risk of adverse health consequences from heat and when designing interventions to address the harms of climate change.

Finally, while a large, quasi-experimental literature has analyzed factors that may have contributed to or mitigated the opioid crisis in the U.S., including pharmaceutical industry marketing, economic conditions, and drug policy changes (37), environmental exposures are less studied. Our results provide evidence that, beyond supply and demand factors, environmental determinants can also shape the overdose crisis.

Limitations and future research

Mechanisms

This study has several important limitations with implications for future research. First, we were unable to determine the specific mechanisms linking heat and drug overdose deaths. There are multiple possible channels, including biological pathways, behavioral changes, and correlations with other heat-sensitive medical conditions. For example, our results are consistent with interactions between heat exposure and physiological responses to drug use (8, 9), the negative impacts of heat on mental health (38), and increased drug use during elevated periods of heat within a given county and month (though a general lack of seasonality in drug overdose deaths suggests the latter is less likely to be driving our results – see Figure S1). These relationships are further complicated by evidence that cold temperatures are also associated with increased fatal opioid overdoses (39). Future research should examine these potentially complex pathways. Given increasing rates of polysubstance use (15), understanding the potential impacts of different types of drugs both individually and in combination is particularly important.

Heat exposure granularity

Second, while an important strength of our study is the use of national data and fixed effect methods, this also limited our analysis to the use of monthly, county-level data. As a result, we were unable to conduct a more detailed investigation into the timing, magnitude, and locality of the heat exposure. For example, our analysis cannot disentangle the timing of a heat exposure relative to a death within a given month (though our result showing a limited impact of lagged heat may suggest a narrow window – see Figure S11). Similarly, this study cannot separate the effects of a single, particularly hot day versus an extended, extreme heat wave; effects by neighborhoods or areas with specific characteristics like green space; or effect sizes earlier versus later in the summer for a specific area. Future studies should consider how to take advantage of daily environmental exposure data available at finer geographic levels.

Similarly, we were restricted to using a common heat exposure for the full population, although there is likely to be substantial differences in the heat experienced across demographic and socioeconomic groups. For example, there are important inequities in heat vulnerability by residential location and race/ethnicity (40). Future research can further explore these disparities to best identify and protect the areas and people at the highest risk of drug overdoses due to heat exposure.

Alternative methods

Analyzing the relationship between heat and overdose deaths using additional empirical strategies may also help refine or clarify our results. One improvement could include accounting for time-varying, county-level characteristics; for example, we were unable to explicitly model changing access to naloxone or the uneven geographic diffusion of fentanyl during our study period due to data limitations. While our inclusion of state-by-year fixed effects may help address these issues (e.g., by adjusting for the implementation of state-level naloxone access laws), future research could provide better insight into their influence.

Future research could also consider alternative empirical approaches, including correlated time series methods and generalized difference-in-differences designs, which may provide further novel evidence into this research question.

Additional research could explore the economic implications of heat-related overdose deaths and their contribution to the overall costs of the drug crisis, as well as conduct cost-effectiveness analyses to investigate potential interventions. For example, a wide range of prevention strategies may help mitigate heat exposures in general (41), but how to best tailor these strategies to prevent drug overdoses is less clear.

Generalizability

This study focuses on the period spanning 1999—2020. Since 2020, however, important events may have shaped the course of the opioid crisis and thus the relationship between heat and drug overdose fatalities. These events include the full trajectory of the COVID-19 pandemic and an increase and subsequent, modest decline in overdose mortality (15). As a result, future research could help clarify the ongoing relationship between heat and drug overdose deaths over time.

Our analysis additionally focuses on the effect of heat exposure during the hottest months of the year, corresponding to other studies of heat exposure (30, 42). However, it is possible that heat, or comparatively extreme hot or cold temperatures more generally, may affect drug overdose mortality at different times of the year. Future studies could explore these relationships.

Finally, given the variety of climates and patterns and types of drug use, our overall results may also not be generalizable to every area of the U.S. or other nations. As a result, our findings should be interpreted in combination with previous studies using data from local areas (17–20) and future work should consider the potential for varying impacts across the globe.

CONCLUSION

In this study, we find that heat exposure was associated with increases in drug overdose fatalities, including deaths related to opioids, cocaine, and psychostimulants, in the continental United States. Our study stresses the need for interventions to mitigate the risks of drug overdoses during heat and suggests that fatal drug overdoses may increase globally due to ongoing climate change.

Supplementary Material

Appendix

Acknowledgements

The authors thank Robert Dubrow, Laura Forastiere, and participants of the Health Policy and Management Research in Progress and Gonsalves Lab Research in Progress seminars at the Yale School of Public Health for helpful suggestions on preliminary results. The authors also thank Thomas Thornhill for assistance with data acquisition and several anonymous reviewers for their helpful comments and suggestions.

Declaration of Competing Interest

Dr. Fiellin’s wife is Founder of Playbl, Inc., which makes serious videogames for substance use prevention.

Primary Funding

Julia M. Dennett and Gregg S. Gonsalves reported receiving grant funding from the National Institute on Drug Abuse (Grant No. DP2 DA049282). Gonsalves reported receiving grant funding from the National Institute on Drug Abuse (Grant No. R37DA15612). The funding sources played no role in this study.

Data Availability Statement

Restricted Vital Statistics multiple cause of death data that support the findings of this study are available from the National Center for Health Statistics. Restrictions apply to the availability of these data. Data are available from the National Center for Health Statistics (see https://www.cdc.gov/nchs/nvss/nvss-restricted-data.htm). All other data are publicly available. The analysis code that supports the findings of this study are openly available in GitHub at https://github.com/gregggonsalves/heat-overdose.

REFERENCES

  • 1.IPCC. Summary for Policymakers. In: Climate Change 2023: Synthesis Report. In: Core Writing Team HLaJRe, editor. Contribution of Working Groups I, II and III to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change Geneva, Switzerland: IPCC; 2023. p. 1–34. [Google Scholar]
  • 2.Dell M, Jones BF, Olken BA. What Do We Learn from the Weather? The New Climate-Economy Literature. J Econ Lit 2014;52(3):740–98. [Google Scholar]
  • 3.Khatana SAM, Werner RM, Groeneveld PW. Association of Extreme Heat With All-Cause Mortality in the Contiguous US, 2008–2017. JAMA Network Open. 2022;5(5):e2212957–e. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Burke M, González F, Baylis P, Heft-Neal S, Baysan C, Basu S, et al. Higher temperatures increase suicide rates in the United States and Mexico. Nature Climate Change. 2018;8(8):723–9. [Google Scholar]
  • 5.Liu J, Varghese BM, Hansen A, Zhang Y, Driscoll T, Morgan G, et al. Heat exposure and cardiovascular health outcomes: a systematic review and meta-analysis. The Lancet Planetary Health. 2022;6(6):e484–e95. [DOI] [PubMed] [Google Scholar]
  • 6.Lin S, Luo M, Walker RJ, Liu X, Hwang S-A, Chinery R. Extreme High Temperatures and Hospital Admissions for Respiratory and Cardiovascular Diseases. Epidemiology. 2009;20(5):738–46. [DOI] [PubMed] [Google Scholar]
  • 7.Mahendran R, Xu R, Li S, Guo Y. Interpersonal violence associated with hot weather. The Lancet Planetary Health. 2021;5(9):e571–e2. [DOI] [PubMed] [Google Scholar]
  • 8.Matsumoto RR, Seminerio MJ, Turner RC, Robson MJ, Nguyen L, Miller DB, et al. Methamphetamine-induced toxicity: An updated review on issues related to hyperthermia. Pharmacol Ther 2014;144(1):28–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Crandall CG, Vongpatanasin W, Victor RG. Mechanism of Cocaine-Induced Hyperthermia in Humans. Ann Intern Med 2002;136(11):785–91. [DOI] [PubMed] [Google Scholar]
  • 10.Ryus C, Bernstein SL. A New Syndemic: Complications of Opioid Use Disorder During a Heat Wave. J Health Care Poor Underserved. 2022;33(3):1671–7. [DOI] [PubMed] [Google Scholar]
  • 11.White JM, Irvine RJ. Mechanisms of fatal opioid overdose. Addiction. 1999;94(7):961–72. [PubMed] [Google Scholar]
  • 12.Schwartz BG, Rezkalla S, Kloner RA. Cardiovascular Effects of Cocaine. Circulation. 2010;122(24):2558–69. [DOI] [PubMed] [Google Scholar]
  • 13.United States Environmental Protection Agency. Climate Change and Extreme Heat: What You Can Do to Prepare. EPA, 2016. October 2016. Report No. [Google Scholar]
  • 14.National Integrated Heat Health Information System. Who Is Most At Risk To Extreme Heat? Heat.gov [November 14, 2023]. Available from: https://www.heat.gov/pages/who-is-at-risk-to-extreme-heat. [Google Scholar]
  • 15.National Institute on Drug Abuse. Drug Overdose Death Rates Gaithersburg, MD: National Institute on Drug Abuse; [updated May 14, 2024July 11, 2024]. Available from: https://nida.nih.gov/research-topics/trends-statistics/overdose-death-rates. [Google Scholar]
  • 16.Garnett MF, Miniño AM. Drug Overdose Deaths in the United States, 2003–2023. NCHS Data Brief. 2024;No. 522. [Google Scholar]
  • 17.Auger N, Bilodeau-Bertrand M, Labesse ME, Kosatsky T. Association of elevated ambient temperature with death from cocaine overdose. Drug Alcohol Depend 2017;178:101–5. [DOI] [PubMed] [Google Scholar]
  • 18.Bohnert ASB, Prescott MR, Vlahov D, Tardiff KJ, Galea S. Ambient temperature and risk of death from accidental drug overdose in New York City, 1990–2006. Addiction (Abingdon, England). 2010;105(6):1049–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Marzuk PM, Tardiff K, Leon AC, Hirsch CS, Portera L, Iqbal MI, et al. Ambient Temperature and Mortality From Unintentional Cocaine Overdose. JAMA 1998;279(22):1795–800. [DOI] [PubMed] [Google Scholar]
  • 20.Henderson SB, McLean KE, Ding Y, Yao J, Turna NS, McVea D, et al. Hot weather and death related to acute cocaine, opioid and amphetamine toxicity in British Columbia, Canada: a time-stratified case-crossover study. CMAJ Open. 2023;11(3):E569–E78. [Google Scholar]
  • 21.Chang HH, Zhang H, Latimore AD, Murray BP, D’Souza RR, Scovronick N, et al. Associations between Short-term Ambient Temperature Exposure and Emergency Department Visits for Amphetamine, Cocaine, and Opioid Use in California from 2005 to 2019. Environ Int 2023:108233. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Hensel M, Stuhr M, Geppert D, Kersten JF, Lorenz J, Kerner T. Relationship between ambient temperature and severe acute poisoning by alcohol and drugs. Alcohol. 2021;94:57–63. [DOI] [PubMed] [Google Scholar]
  • 23.Parks RM, Rowland ST, Do V, Boehme AK, Dominici F, Hart CL, et al. The association between temperature and alcohol- and substance-related disorder hospital visits in New York State. Communications Medicine. 2023;3(1):118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Burke M, Zahid M, Martins MC, Callahan C, Lee R, Avirmed T, et al. Are We Adapting to Climate Change? National Bureau of Economic Research Working Paper Series. 2024;No. 32985. [Google Scholar]
  • 25.Centers for Disease Control and Prevention. National Environmental Public Health Tracking Network 2024 [updated May 31, 2023January 2, 2024]. Available from: https://ephtracking.cdc.gov/indicatorPages?selectedContentAreaAbbreviation=35&selectedIndicatorId=173. [Google Scholar]
  • 26.National Weather Service, National Oceanic and Atmospheric Administration. What is the heat index? Amarillo, TX: National Weather Service,, National Oceanic and Atmospheric Administration; 2024. [January 24, 2024]. Available from: https://www.weather.gov/ama/heatindex. [Google Scholar]
  • 27.National Center for Health Statistics. Restricted-Use Vital Statistics Data - multiple cause of death 1999–2000, as compiled from data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative Program.: Centers for Disease Control and Prevention. [Google Scholar]
  • 28.Ahmad F, Cisewski J, Rossen L, Sutton P. Provisional drug overdose death counts: National Center for Health Statistics; 2023. [updated January 17, 2024January 25, 2024]. Available from: https://www.cdc.gov/nchs/nvss/vsrr/drug-overdose-data.htm. [Google Scholar]
  • 29.Surveillance Epidemiology and End Results Program. Download U.S. County Population Data - 1969–2020: National Cancer Institute; 2023. [March 20, 2023]. Available from: https://seer.cancer.gov/popdata/download.html. [Google Scholar]
  • 30.Klompmaker JO, Laden F, Dominici F, James P, Rimm E, Roscoe C, et al. Seasonal Average Temperature Differences and CVD Incidence: Results from the US-Based Nurses’ Health Study, Nurses’ Health Study II, and the Health Professional Follow-Up Study. Environ Health Perspect 2025;133(5):057027. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Cleland SE, Steinhardt W, Neas LM, Jason West J, Rappold AG. Urban heat island impacts on heat-related cardiovascular morbidity: A time series analysis of older adults in US metropolitan areas. Environ Int 2023;178:108005. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Yoo E-H, Sergi C, Senders A, Kim H. Regional Variability of Extreme Heat and Cold Risk Among Dual-Eligible Individuals. JAMA Health Forum 2025;6(1):e245134–e. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Centers for Disease Control and Prevention. Rising Numbers of Deaths Involving Fentanyl and Fentanyl Analogs, Including Carfentanil, and Increased Usage and Mixing with Non-opioids 2018. [updated July 11, 2018May 27, 2024]. CDCHAN-00413:[Available from: https://emergency.cdc.gov/han/han00413.asp. [Google Scholar]
  • 34.Spencer MR, Garnett MF, Miniño AM. Urban–Rural Differences in Drug Overdose Death Rates, 2020 Hyattsville, MD: National Center for Health Statistics; 2022. [updated July 21, 2022May 27, 2024]. Available from: 10.15620/cdc:118601. [DOI] [Google Scholar]
  • 35.Altekruse SF, Cosgrove CM, Altekruse WC, Jenkins RA, Blanco C. Socioeconomic risk factors for fatal opioid overdoses in the United States: Findings from the Mortality Disparities in American Communities Study (MDAC). PLoS One. 2020;15(1):e0227966. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Maricopa County Department of Public Health. 2022 Heat Deaths Report. In: Informatics DoEa, editor. 2023. [Google Scholar]
  • 37.Maclean JC, Mallatt J, Ruhm CJ, Simon K. Economic Studies on the Opioid Crisis: A Review. National Bureau of Economic Research Working Paper Series. 2020;No. 28067. [Google Scholar]
  • 38.Nori-Sarma A, Sun S, Sun Y, Spangler KR, Oblath R, Galea S, et al. Association Between Ambient Heat and Risk of Emergency Department Visits for Mental Health Among US Adults, 2010 to 2019. JAMA Psychiatry. 2022;79(4):341–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Goedel WC, Marshall BD, Spangler KR, Alexander-Scott N, Green TC, Wellenius GA, et al. Increased Risk of Opioid Overdose Death Following Cold Weather: A Case–Crossover Study. Epidemiology. 2019;30(5):637–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Manware M, Dubrow R, Carrión D, Ma Y, Chen K. Residential and Race/Ethnicity Disparities in Heat Vulnerability in the United States. Geohealth 2022;6(12):e2022GH000695. [Google Scholar]
  • 41.Jay O, Capon A, Berry P, Broderick C, de Dear R, Havenith G, et al. Reducing the health effects of hot weather and heat extremes: from personal cooling strategies to green cities. The Lancet. 2021;398(10301):709–24. [Google Scholar]
  • 42.Vaidyanathan A Heat-related emergency department visits—United States, may–september 2023. MMWR Morbidity and Mortality Weekly Report. 2024;73. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Appendix
NIHMS2127334-supplement-Appendix.docx (5.9MB, docx)

Data Availability Statement

Restricted Vital Statistics multiple cause of death data that support the findings of this study are available from the National Center for Health Statistics. Restrictions apply to the availability of these data. Data are available from the National Center for Health Statistics (see https://www.cdc.gov/nchs/nvss/nvss-restricted-data.htm). All other data are publicly available. The analysis code that supports the findings of this study are openly available in GitHub at https://github.com/gregggonsalves/heat-overdose.

RESOURCES