The study, led by environmental epidemiologist Amruta Nori-Sarma of Harvard T.H. Chan School of Public Health in collaboration with colleagues at Truveta, Harvard, and Boston University, exploited a technological capability that earlier generations of heat researchers lacked: continuously refreshed electronic health record data. Rather than waiting months for claims data to be processed and aggregated, the team drew on state-day counts of emergency department visits from 19 eastern states, allowing them to quantify the health burden of the summer 2025 heatwaves while the season was still underway. In total, the analysis encompassed 8,169,637 emergency department visits recorded during the warm seasons of 2023 through 2025, providing a robust historical baseline against which the 2025 episodes could be compared.

The two heatwaves under scrutiny were stark in their timing and severity. The first gripped the region from 22 to 25 June 2025, and the second returned from 10 to 13 August 2025, engulfing the eastern United States in conditions that triggered National Weather Service heat alerts across much of the study area. To estimate how many visits these episodes caused, the researchers compared observed visit counts during the heatwave windows with counterfactual predictions of what visits would have been under typical temperature conditions, using temperature-health associations estimated from the prior three warm seasons.

Technically, the analysis rested on a conditional Poisson model, a statistical framework well suited to time-series data on disease counts. The researchers stratified time by year, calendar month, and day of week, with stratum-specific fixed effects that absorb routine patterns in healthcare utilization, such as weekday and seasonal variation. This design effectively compares hot days with cooler days occurring under otherwise similar conditions, a strategy closely related to the time-stratified case-crossover approach widely used in environmental epidemiology. Temperature exposure was modeled using distributed lag non-linear models, or DLNMs, which capture two crucial features of heat’s health effects simultaneously: the non-linear dose-response relationship between temperature and risk, and the fact that effects unfold over multiple days, with heat stress on one day influencing emergency visits over a window of zero to four days afterward.

The headline numbers were striking. During the June 2025 heatwave, the researchers estimated that 9.2 percent of all-cause emergency department visits were attributable to heat over the zero-to-four-day lag window, with a 95 percent confidence interval of 6.2 to 11.6 percent. During the August heatwave, the attributable fraction was 8.3 percent, with a confidence interval of 6.0 to 10.1 percent. In plain terms, roughly one in every eleven or twelve people who walked into an emergency department in these nineteen states during those weeks was there, at least in part, because of extreme heat. Similar patterns emerged when the team examined cumulative lags and cause-specific outcomes, indicating that the association was not an artifact of a single modeling choice.

Because the underlying data included diagnostic information, the researchers could decompose the burden by cause. Beyond direct heat-related illness diagnoses, they tracked emergency visits for kidney disease and for mental health conditions, both of which have well-documented physiological links to heat exposure. Kidney disease is acutely sensitive to dehydration and thermal strain, while a growing body of literature connects high ambient temperatures to psychiatric crises, including exacerbations of mood disorders and substance-related emergencies. The lag-specific analyses showed that the association between the 99th percentile of daily maximum temperature, about 36.9 degrees Celsius, and the sample median of 27 degrees Celsius was strongest on the same day for direct heat illness, while kidney and mental health outcomes followed somewhat different lag structures, consistent with the diverse physiological pathways through which heat injures the human body.

Those pathways are worth spelling out. Extreme heat forces the cardiovascular system to work harder, shunting blood toward the skin to dissipate warmth and increasing cardiac output, changes that laboratory studies involving more than 400 controlled heat exposure experiments have documented in detail. In people with preexisting heart disease, this added strain can precipitate decompensation. Dehydration concentrates the blood and promotes the formation of kidney stones and acute kidney injury, particularly in older adults and in people taking common medications that impair thermoregulation or fluid balance. Meanwhile, heat disrupts sleep, alters neurotransmitter function, and can increase impulsivity and aggression, mechanisms that plausibly underlie the observed rise in mental health emergency visits during hot weather.

The heterogeneity analyses added an important dimension to the findings. By fitting subgroup models for six age groups, three US census regions, and both sexes, the researchers probed whether the temperature-emergency visit association varied across populations. The results, visualized in forest plots comparing rate ratios at lag zero, showed differences by age, region, and sex, echoing decades of evidence that the very young, older adults, and residents of areas unaccustomed to extreme heat bear disproportionate risk. The state-specific exposure-response functions also revealed geographic variation: states with historically milder summers tended to show steeper increases in emergency visits per degree of warming, a pattern consistent with the idea that populations adapt, physiologically and infrastructurally, to the temperatures they usually experience, and that anomalous heat is most damaging where it is least familiar.

Methodological sensitivity checks strengthened the study’s credibility. The team replicated the primary analysis using daily average temperature instead of daily maximum temperature, and separately extended the lag window from four to seven days; both variations reproduced the central findings. Exposure data were drawn from the Open-Meteo weather API and National Weather Service records, and heat alert dates by state were compiled from National Weather Service advisories, allowing the investigators to anchor their heatwave definitions in the same official warnings that reached the public during those weeks. All analytic code, including the data extraction scripts, DLNM fitting routines, meta-analysis, and attributable burden estimation, has been made publicly available on GitHub, an unusual and welcome degree of transparency for a rapidly produced surveillance analysis.

The broader significance of the study lies less in any single percentage than in what it demonstrates about the future of climate-health surveillance. Traditional epidemiology delivers its verdicts long after the disaster has passed, when the findings can inform the next heat season but not the current one. By contrast, a near-real-time pipeline built on continuously refreshed electronic health records can tell public health officials, while a heatwave is still unfolding, how many excess patients are arriving, which conditions are driving the surge, and where resources are most strained. The authors argue that this capability should become a standing component of heat preparedness, complementing heat action plans and warning systems whose effectiveness, demonstrated in cities from Philadelphia to Montreal, depends on knowing when and where the danger peaks. As climate change renders summers like 2025 more frequent and more intense, the study suggests that the emergency department itself, properly instrumented, can serve as an early-warning sensor for the health impacts of a warming world, and that the true toll of extreme heat is far larger than the narrow category of heatstroke diagnoses would suggest. Nearly a tenth of all emergency care during those scorching weeks was heat’s shadow, and counting it in real time may be the first step toward preventing it.

Subject of Research: Attribution of summer 2025 US heatwave emergency department visits using electronic health records

Article Title: Impact of the summer 2025 heatwaves on emergency department visits in the USA

Article References: Nori-Sarma, A., Cartwright, B. M., Zanobetti, A., Pastwa, A., Stucky, N., & Willis, M. D. (2026). Impact of the summer 2025 heatwaves on emergency department visits in the USA. Nature Health. https://doi.org/10.1038/s44360-026-00194-y

Image Credits: AI Generated

DOI: 10.1038/s44360-026-00194-y

Keywords: heatwaves, emergency department visits, electronic health records, extreme heat, public health surveillance, climate change, epidemiology, distributed lag non-linear models, heat-related illness, kidney disease, mental health, United States