Thousands of Americans die every year from a chronically underreported cause: heat.
As summer heat intensifies because of human-driven climate change, the United States' official tally of heat-related deaths does not come close to capturing the full human toll. The data feeding into nationwide death counts is collected inconsistently and rarely thoroughly.
The result is a vast undercount of the number of people who die because of heat. And experts say failing to recognize and address this may cost more people their lives.
So NPR set out to create a more comprehensive national count.
We partnered with Boston University to find the total number of deaths influenced by heat every year in the United States. The investigation took two years and involved building a statistical model that assessed nearly every county in the country. The full analysis will soon be published in a scientific journal.
An accurate assessment of the true scope — as well as understanding who is affected, where and why — is critical for developing lifesaving solutions, says Quinn Adams, a research scientist at Boston University's Center for Climate and Health and the lead partner in the joint NPR-BU investigation.
Beyond the statistical assessment, NPR set out to find people whose deaths slipped through the cracks — those who might not have died if heat hadn't pushed them beyond their breaking point — to understand exactly how heat could be missed as a factor. We will be bringing you their stories in the coming days.
But first, here are three key takeaways from our research — and why it matters.
How big is the undercount?
The scale of the undercount is significant.
The primary keeper of cause-of-death records in the U.S. is the Centers for Disease Control and Prevention (CDC). It keeps an annual tally of heat-related deaths in the U.S., which is compiled from state reports based on whether heat is noted on individual death records.
From 2004 to 2018, the national number of heat-related deaths compiled by the CDC was low: on average, around 700 a year. The number has crept up in recent years, averaging around 1,700 but sometimes topping 2,000, as heat risks grow and counting strategies evolve.
But even that number is an underestimate, according to NPR and BU's analysis. On average between 2018 and 2025, we estimate that some 9,000 deaths each year in the country are impacted by heat — a factor of five higher than the official average over that same period.
Some years, heat-related deaths can top 10,000.
The true figure could be even higher, says Adams, because of some limitations in the data and the conservative nature of our analysis.
The dangers from heat are growing every year because of human-caused climate change. The heat wave season has extended by a month and a half since the 1960s. In 50 of the country's biggest cities, the number of heat waves has doubled since the 1980s. Heat domes that settle over the country regularly expose tens of millions of Americans — sometimes well over 100 million at a time — to dangerous heat.
"The frequency and the duration of these heat events has been expanding very rapidly," says Adams. "That's what we can expect moving forward into the next few decades: having more heat waves, and having them last longer, and having them be more intense such that people don't get the escape from the heat that they need to physiologically reset."
The reasons for the undercount are complex. But a crucial issue, says doctor and researcher Sameed Khatana of the University of Pennsylvania's Perelman School of Medicine, is that heat's role in a death is often masked by other more obvious causes of death, like a heart attack.
"Heat is what people often call the silent killer, compared to things like hurricanes and typhoons," says Khatana.
Heat, he says, can often be the trigger that pushes someone's health condition — like cardiovascular disease — from manageable to fatal. "The heart starts beating faster and harder," he explains, to shunt blood to the skin to cool the body. And eventually, if someone's heart is already weak, it may not be able "to keep up with the demand that was required of it," he says. That can lead to something like a heart attack.
But often, Khatana says, because heat leaves no obvious mark behind for doctors or coroners to examine, the cause of death gets marked down only as a heart attack. Heat isn't noted on the death certificate, and therefore it is not included in the tallies reported to the municipality, the state and eventually the federal recordkeepers.
Such omissions are understandable, Khatana says. But repeated thousands of times, in all parts of the country, they create a blind spot that obscures the scale of heat's danger.
National risks — even in unexpected places
The NPR and BU investigation found that heat is killing people in nearly every corner of the country.
Our analysis finds that Florida and Texas have the most heat-related deaths overall. That is in part due to their large populations. But those states remain among the most impacted even when adjusted for their size.
But the losses are not limited to states historically considered hot.
Our model finds that more than 100 people die from heat-related causes in Michigan every year, 10 times the official reported numbers. In Massachusetts, official records say only about 5 people die from heat each year, but our estimate suggests that number could be more than 60.
And in Florida, the difference is even more dramatic: We estimate more than 1,900 deaths each year, about 40 times as many as reported.
Heat can be particularly dangerous for people who are not already used to it. And in many parts of the country that were not historically hot, infrastructure to keep cool is less common. Nearly all homes in Alabama have some form of air conditioning, for example. But in Washington state, only 53% of homes do. In Maine, 70% do. Yet the risks of heat exist, and are growing, in those states as well as those where it is a more entrenched problem.
Counts are getting more accurate — but not everywhere
The official national counts from the CDC have trended upward in recent years, rising from an average of roughly 700 between 2004 and 2018 to about 1,700 between 2018 and 2025.
That increase is due in large part to two things: Heat is simply getting worse, according to the most recent National Climate Assessment, and also there is better counting of heat-related deaths in some parts of the country.
Arizona's Maricopa County — home to Phoenix — has developed sophisticated protocols for sussing out the subtle influences of heat on people's deaths. Maricopa County trains "death investigators" to look for signs of heat stress, such as asking about the state of the deceased's air conditioning system, their ability to pay electric bills to run ACs, their use of medications that increase heat sensitivity, and more.
But Maricopa County is something of an outlier. In contrast, the counties home to Los Angeles and Houston report very few heat-related deaths. Yet our model finds hundreds of people likely die in each of those counties because of heat each year.
Los Angeles recognizes the problem, says Nichole Quick, the chief science officer at the Los Angeles County Department of Public Health. The county's official numbers "are vast undercounts," Quick says. "The true impact of this is higher than I think a lot of the numbers show."
But solving the counting problem isn't simple. It took Maricopa County years to fine-tune its strategies for more accurately assessing heat's impact on deaths. And many doctors who certify deaths in emergency rooms and hospitals, or for patients who die at home, don't have protocols for confidently identifying heat impacts, says Khatana.
Until such issues are addressed, says Adams, the undercount will remain, and heat's true toll will remain obscured — making it harder for policymakers, planners and families to understand the urgency to protect themselves.
"Most people do not see themselves as vulnerable to heat. They're like, 'Oh, that's someone else. It's not me,'" says Adams. But the data is clear: Heat can take anyone, anywhere in the country. "But then, you know, push comes to shove and it actually is you."
NPR will be continuing this coverage in the coming days, exploring why deaths slip through the cracks and how lives can be saved.
Credits
- This reporting was supported by a Nova Institute for Health Media Fellowship. It was also supported by the Dennis A. Hunt Fund for Health Journalism.
- Nick McMillan contributed to this report.
Data sources and methodology
Mortality data
We analyzed weekly all-cause mortality from 2018 through 2025 for 530 U.S. counties with populations of at least 50,000. These counties represented approximately 73% of the U.S. population in 2024. We excluded data from 2020 because the COVID-19 pandemic substantially disrupted typical patterns of mortality and health care. For comparison, we also queried the Centers for Disease Control and Prevention's WONDER database to compile death data for which "heat exposure" was listed as a primary or contributing cause on the official death record.
Temperature data
We obtained temperature estimates from ERA5-Land, a global weather dataset produced by the European Centre for Medium-Range Weather Forecasts that combines observations with a weather model to estimate conditions across areas without monitoring stations. For each county and week, we calculated a population-weighted average of daily maximum temperature, giving greater weight to temperatures in more heavily populated areas. Because the same temperature may have different health effects in different parts of the U.S., we additionally ranked temperatures relative to each county's usual climate, allowing us to consider local differences in acclimatization.
Heat-related mortality estimates
We used space-time stratified case-crossover methods to estimate how unusually hot weeks affected deaths in each county during the same week and the following week. We allowed the effect of temperature to be nonlinear, meaning that mortality could rise more sharply at very high temperatures. We then combined county-level estimates within each state to identify the temperature associated with the lowest mortality, commonly referred to as the minimum mortality temperature (MMT). State-level MMT estimates were used because sparse weekly mortality counts in some counties made county-specific estimation of the MMT unreliable. This approach improved statistical reliability while still allowing regional differences in climate and heat adaptation to be reflected. County-level attributable mortality estimates were extrapolated to the full population of the contiguous United States by applying the estimated attributable mortality rate from the included counties to the total contiguous United States population. We defined heat-attributable deaths as deaths associated with temperatures above the state-specific minimum mortality temperature. We conducted all analyses using the open-source cityClimateHealth R package.
Copyright 2026 NPR
