US heat deaths tripled among the least-educated adults, and barely moved at the top

Most reporting on temperature-related death asks how many. A study published in March in Preventive Medicine asks a harder question — who — and finds that the answer has been getting more lopsided for over a decade.

Researchers used United States national mortality data from 2010 to 2023 to identify all heat- and cold-related deaths among adults aged 25 and over, then sorted them by educational attainment: high school or less, at least some college, or more than college [s1]. Population denominators came from the American Community Survey, and rates were expressed as age-adjusted mortality per 100,000 adults [s1].

The gradient

Among adults with a high school education or less, the age-adjusted heat mortality rate rose from 0.5 per 100,000 (95% CI 0.5–0.6) to 1.8 per 100,000 (95% CI 1.7–1.9) across the study period [s1].

Among the most-educated group, the same rate moved from 0.1 per 100,000 (95% CI 0.1–0.2) to 0.2 per 100,000 (95% CI 0.2–0.3) [s1].

Both groups saw heat deaths increase. Only one saw them increase by more than a full death per 100,000 adults. The least-educated group both started with the highest rates and increased fastest throughout the study period [s1].

Cold went the other way — for one group

The cold figures are the more revealing half of the paper, because they move in opposite directions.

For adults with a high school education or less, cold-related mortality rose from 1.0 per 100,000 (95% CI 1.0–1.1) to 1.8 per 100,000 (95% CI 1.7–1.9) [s1]. For the most-educated group, it was 0.4 per 100,000 (95% CI 0.3–0.5) in 2010 and 0.3 per 100,000 (95% CI 0.3–0.4) in 2023 [s1].

Warming winters are frequently offered as a partial offset to warming summers — fewer cold deaths against more heat deaths. In this dataset that offset materialised for the most-educated Americans and did not materialise for the least-educated, whose cold deaths nearly doubled over the same 14 years [s1].

The authors report that findings were consistent across sub-groups of sex, region, and race/ethnicity [s1]. That consistency matters: it argues against the gradient being an artefact of educational attainment tracking something regional, such as the southern states where heat is most severe.

What education is standing in for

Educational attainment is not a cause of death from heat. It is a marker, and the study is explicit that it is being used as a means of evaluating socioeconomic disparities [s1].

What it plausibly marks is the set of things that determine whether a hot day is dangerous: whether a job is outdoors, whether a home has working air conditioning and whether its electricity stays on, whether chronic illness is managed, whether a person lives alone. The study does not measure any of those, and does not claim to. Anyone reading the education gradient as an explanation rather than a signal is reading past the paper.

The same caution applies to the death certificates underneath it. These are deaths identified as heat- or cold-related in national mortality records — the case-ascertainment approach, not a statistical excess-mortality estimate. That approach is known to capture only part of the burden, because heat most often kills through cardiovascular and respiratory failure that gets recorded as cardiovascular and respiratory failure. The absolute rates here are therefore floors, not totals.

Why the trend is the finding

The absolute numbers are small — under two deaths per 100,000 adults in the worst-affected group [s1]. Temperature is not a leading cause of death in the United States by any accounting.

The finding worth carrying is the direction of travel. Over 14 years in which national attention to extreme heat rose sharply, in which heat warning systems were built out and cooling assistance programmes expanded, the disparity in temperature-related mortality by educational attainment widened rather than narrowed [s1]. Whatever adaptation has occurred has not reached the people dying most.

What to watch

The study ends in 2023, which excludes the record heat seasons that followed. Whether the gap continued to widen through 2024 and 2025 is the obvious next question, and national mortality data will answer it on the usual two-to-three-year lag.

The second thing to watch is whether the cold divergence holds. If cold deaths keep falling for higher-education groups while rising for lower-education groups, the common framing of milder winters as a partial health benefit of warming will need to be stated with a distributional caveat attached — because in this dataset, that benefit went to one group only.