A Heat Death Has to Be Named Twice
A new review in Forensic Science, Medicine and Pathology argues that heat mortality is measured by two systems — individual medicolegal investigation and population-level statistical attribution — whose numbers overlap but are not interchangeable. British Columbia's 2021 heat dome shows what that means in practice.
In the last week of June 2021, coroners in British Columbia investigated more than 800 deaths in seven days. Six hundred and nineteen of them were later identified as heat-related, following a heat dome that pushed temperatures above 40°C across much of the province between 25 June and 1 July. When the provincial death review panel reported the following June, one finding reframed how the event is usually told: 98% of those deaths happened indoors. Most of the people who died were older adults with several chronic conditions, and 56% lived alone — in homes, the panel noted, without adequate cooling.
The number is well known in Canada. What is less often discussed is that a death like those is one of the hardest things a mortality system can be asked to record, and that the two systems which attempt it are not measuring the same object. A review published on 28 October in Forensic Science, Medicine and Pathology makes that case in unusually plain terms. It is a structured narrative review — literature to 17 July 2026, reported against the SANRA checklist, no new data — and its author, Ignacio Coll-Orduña, frames the problem as a collision between “two incompletely connected systems”: the medicolegal investigation of an individual body, and the statistical attribution of deaths across a population.
An exclusionary diagnosis
The review’s clinical core is that fatal environmental hyperthermia is normally diagnosed by exclusion. Gross and microscopic findings are non-specific, and they can be altered by cooling, by how long the person survived, by the interval since death and by decomposition. No biochemical, immunohistochemical, proteomic or lipidomic marker independently confirms it. What is left is scene reconstruction, the ruling out of competing causes, and a judgement about whether heat was sufficient to kill someone whose heart or lungs were already failing.
Out of that uncertainty the review draws a three-way distinction that is easy to collapse and should not be. There are direct heat deaths, where heat is the principal cause. There are heat-contributory deaths, where heat acts through a natural disease already present. And there are population-attributable deaths — people whose deaths would not have occurred in that week had the temperature been ordinary, but whose individual certificate records something else entirely. The categories overlap; the review insists they are not interchangeable.
That distinction has consequences for a form. Whether a certifier writes heat into the principal causal sequence or as a contributory condition, and whether the local coding rules then send the death to an external-cause code, determines whether the case enters any heat statistic at all.
The rule that decides who is counted
British Columbia shows the mechanism. Under ICD-10, deaths from environmental heat sit among the external causes and are therefore treated as accidental, which in that province makes them reportable to the coroners service. Recognising that clinicians might not make the connection during an unprecedented event, the service distributed a notice on 2 July 2021 reminding them of the reporting requirement — and reports of deaths from exposure to excessive natural heat went on arriving for days afterwards, well after the temperature had fallen. Case ascertainment lagged the hazard.
The indoor setting compounds it. A death in a stuffy apartment has no scene in the ordinary sense: no collapse in a field, no exertion, no obvious exposure. It has an absence — of air conditioning, of a fan, of someone to notice. The panel found that more than 60% of the people who died had seen a medical professional within the month before their death, which is to say the health system had recently had them in view and still could not have predicted which of its patients would be killed by the weather.
The other ledger
The population-level count works differently, and its numbers look nothing like 619. In a study published in Nature Medicine in 2023, Ballester and colleagues analysed the Eurostat mortality database — 45,184,044 death counts from 823 regions in 35 European countries, a population of more than 543 million — and estimated 61,672 heat-related deaths (95% CI 37,643–86,807) in Europe between 30 May and 4 September 2022. That figure is a modelled excess, derived from how far weekly deaths rose above what the same weeks would have produced at ordinary temperatures. No certificate anywhere corresponds to it.
The review is careful about what follows. Matched comparisons of the same events, it notes, show that modelled excess mortality can exceed the number of deaths individually attributed to heat — without the two measures becoming interchangeable. One counts bodies a pathologist could defend in an inquest; the other counts the deaths a season produced. Neither is a better version of the other, and reporting one as though it were the other is where the trouble starts.
Where there is no coroner
For most of the world’s deaths, neither ledger is available. Where deaths happen at home with no physician in attendance, there is no scene for a forensic pathologist to reconstruct and no certificate to code, and the only route from a household’s bereavement to a cause in the national table is an interview conducted months later. That is the problem the WHO verbal autopsy instrument was built for, and the reason verbal autopsy has become the default source of cause-of-death data in settings where registration is thin.
Verbal autopsy inherits exactly the difficulty the review describes, at a different resolution. A probabilistic model such as InterVA, developed at Umeå University, assigns causes from symptom histories by symptom–cause probabilities and returns population-level fractions, not verdicts about individuals — a discipline the field had to learn. Our earlier report on the uMkhanyakude surveillance site in KwaZulu-Natal described what happens when the two ledgers meet: a cohort with a well-run verbal autopsy programme and a clean excess-mortality estimate for the COVID years, and still only a partial answer to what people actually died of during the peaks.
The review’s prescription is procedural rather than technological. Standardised scene data. Toxicology appropriate to the case. Explicit certification of contributory heat exposure rather than silence. Linkage between medicolegal and vital-statistics records. And joint interpretation of forensic and epidemiological counts, published as complementary measures instead of competing ones.
Read as advice to a mortality system, that is not so different from what the verbal autopsy literature has been saying for two decades: record the circumstances, not only the symptoms, and be honest about which question the number answers. The 619 is not a failure. Nor is the 61,672. What fails is the presentation of either as the count.
Sources
- Coll-Orduña I. Heat-related deaths: forensic diagnosis, death certification and reconciliation with excess-mortality estimates. Forensic Science, Medicine and Pathology, published 28 October 2026. doi.org
- BC Coroners Service Death Review Panel. Extreme Heat and Human Mortality: A Review of Heat-Related Deaths in B.C. in Summer 2021. Report to the Chief Coroner, released 7 June 2022. www2.gov.bc.ca
- Ballester J, Quijal-Zamorano M, Martínez-Turrubiates RF, et al. Heat-related mortality in Europe during the summer of 2022. Nature Medicine 2023;29:1857–1866. pmc.ncbi.nlm.nih.gov
- World Health Organization. 2022 WHO verbal autopsy instrument. who.int