Cold and heat both raised out-of-hospital cardiac arrest risk in Sweden

Days at either end of the local temperature range carried a higher risk of out-of-hospital cardiac arrest than mild days across Sweden between 2010 and 2019 [s1]. In a national case-crossover analysis of 29,604 arrests, the risk was lowest around the 86th percentile of each area's temperature and rose at both extremes — by about 20% on the coldest days and about 16% on the hottest, relative to that low point [s1].

The finding sits inside a larger, counterintuitive fact about temperature and health: across the world, cold days kill more people than hot ones, even as heat draws most of the attention [s2]. This study adds a specific acute event — sudden cardiac arrest outside hospital — to that picture, and does it with a design that controls tightly for who the patients are.

What was analysed

The authors used records from the Swedish Registry for Cardiopulmonary Resuscitation for 2010 to 2019 [s1]. The analysis included 29,604 out-of-hospital cardiac arrests; 32% of the patients were women and the median age was 74 [s1]. Out-of-hospital cardiac arrest is common and usually fatal — the paper notes roughly 300,000 cases a year in Europe with overall survival of about 11% [s1].

Daily mean temperature was estimated on a 1×1 km grid using a machine-learning model, then converted into location-specific percentiles for each of Sweden's 290 municipalities [s1]. That last step matters. Rather than asking whether a fixed absolute temperature is dangerous, the study asks whether a day that is extreme for that place is dangerous — which is how heat and cold health effects actually behave.

The design was a time-stratified case-crossover, analysed with conditional logistic regression [s1]. A case-crossover compares each arrest against control days for the same person in the same month, so anything that does not change within a person over a few weeks — age, sex, chronic illness, where they live — is controlled by construction. It is a strong design for short-term environmental triggers and a weak one for anything that acts slowly.

What was found

The reference point, where cardiac-arrest risk was lowest, fell at the 86th percentile of local temperature [s1]. Compared with that point, the odds ratio at the 1st percentile — the coldest days — was 1.197 (95% CI 1.066 to 1.344) [s1]. At the 99th percentile — the hottest days — the odds ratio was 1.156 (95% CI 1.058 to 1.263) [s1].

Both intervals sit clearly above 1, so neither association is a chance finding at conventional thresholds, but neither is large. The cold estimate is modestly higher than the heat estimate, and their confidence intervals overlap, so the study supports "both extremes raise risk, cold a little more" rather than any sharp ranking between them.

The absolute temperatures behind those percentiles varied widely by location, which is the point of using percentiles at all. Across municipalities the 1st percentile ranged from −32.7°C to −5.98°C, and the 99th percentile ranged from 16.6°C to 22.8°C [s1]. A day of 20°C is unremarkable in much of the world; in the coolest parts of northern Sweden it lands near the top of the local distribution, and the analysis treats it accordingly.

Why "local extreme" is the useful frame

The study's own conclusion is that risk tracked locally relative temperature extremes, "even when temperatures do not represent extreme temperature events by absolute metrics" [s1]. This is the recurring lesson of temperature-health research: populations adapt — physiologically, behaviourally, and through housing and heating — to their usual climate, so departure from the local normal predicts harm better than any universal threshold. A heat-health or cold-health warning system calibrated to national absolute cut-offs will misfire in both directions.

How it fits the wider evidence

The most comprehensive accounting of temperature-related death, a modelling study across 750 locations in 43 countries, estimated 5,083,173 deaths a year associated with non-optimal temperatures between 2000 and 2019 — 9.43% of all deaths [s2]. Of that, 8.52% of deaths were cold-related and 0.91% heat-related [s2]. Cold's share of the temperature burden is roughly nine times heat's in that global tally, a proportion that public attention does not reflect.

The Swedish arrest data do not reproduce a nine-to-one ratio — the two odds ratios here are close [s1] — and they should not be read as measuring the same thing. One counts total attributable mortality across all causes and both temperature tails [s2]; the other measures the per-day trigger risk for a single acute event in one high-latitude country [s1]. What they share is direction: cold is not the lesser hazard, and a warming climate does not simply subtract cold deaths as it adds heat deaths.

Limits

The authors flag the main one: ambient temperature is not the same as what any individual actually experienced [s1]. Someone who collapses indoors, in a heated or air-conditioned space, was not exposed to the grid-cell outdoor temperature assigned to them, and that mismatch biases estimates toward the null. The study is also confined to Sweden, a wealthy country with widespread heating and a cold-skewed climate; the balance between the cold and heat tails would likely look different in a hot country with limited air conditioning. And a case-crossover measures short-term triggering, not the slow physiological toll of living through repeated extremes.

What to watch

Whether comparable registry analyses in hotter countries, or in places with weaker cold-weather infrastructure, shift the balance between the two tails — and whether heat-health warning systems, which are proliferating faster than cold-weather ones, are being calibrated to local distributions rather than absolute numbers. On this evidence, the temperature that raises cardiac-arrest risk is the one that is unusual for where you live, not the one that reads as extreme on a national map.

This article is informational and does not constitute medical advice.

Sources

• [s1] Yamron E, Dahlquist M, Jonsson M, et al., "Ambient air temperature and out-of-hospital cardiac arrest: A national case-crossover study," Journal of Internal Medicine, vol. 300, no. 4, published online 3 August 2026. https://doi.org/10.1111/joim.70144
• [s2] Zhao Q, Guo Y, Ye T, et al., "Global, regional, and national burden of mortality associated with non-optimal ambient temperatures from 2000 to 2019: a three-stage modelling study," The Lancet Planetary Health, vol. 5, no. 7, July 2021. https://doi.org/10.1016/S2542-5196(21)00081-4