Learning to interpret data on ‘heat-related deaths’ to counter alarmism

Half a century of cognitive science has catalogued, with almost notarial precision, the ways in which our minds deceive themselves. That catalogue was originally intended as a warning. The Italian press seems to have treated it as a manual. Journalistic coverage of heatwaves is one of the most instructive examples.

It is always the use of words that causes problems. When newspapers talk about "heat", they use at least four different concepts as synonyms. In physics, heat is a transfer of energy and temperature is a state variable. In physiology, what matters is what is known as thermal strain – that is, the body’s inability to dissipate metabolic heat, which also depends on humidity, radiation, wind, physical activity and clothing. In psychophysics, on the other hand, what matters is the thermal sensation – that is, the perceived discomfort. Finally, epidemiology does not measure any of these phenomena: it uses the temperature recorded by the weather station, a proxy for actual exposure.

This is where the catchphrase "45 degrees felt" comes from. The figure is derived from heat indices, designed to approximate what we individually "feel". It is always difficult to identify an objective parameter for subjective perceptions. It is misleading to think that an indicator incorporating the subjective dimension, however flawed, is nonetheless the best description of reality available to us. The first major international comparison of heat stress metrics and mortality showed that the best measure varies from country to country, and that the intuitively plausible idea that temperature and humidity together always predict mortality better than air temperature alone has little basis in fact. This means that the ‘perceived’ temperature adds very little. A 2010 study, which examined over a hundred American cities, had already suggested that temperature alone was a better predictor of mortality than temperature and humidity combined.

Perception is not the same as exposure. It is a defence mechanism. Feeling hot prompts one to drink, seek shade, open a window or slow down physical activity. Thermal discomfort triggers behavioural thermoregulation. Heat-related deaths occur mainly among people in whom this system has deteriorated. Ageing impairs both the perception of heat and physiological and behavioural responses. In 2018, a study was conducted on older adults exercising in a progressively heated environment: whilst body temperature rose significantly, the perceived discomfort increased far less than expected. If one does not perceive the strain, one does not alter one’s behaviour to protect oneself. This is yet another example demonstrating the usefulness of states of discomfort.

Here, then, is the first paradox of communication. Newspapers scream ‘45 degrees felt’ to convey the risk. But those who die most frequently belong to the section of the population in which thermal perception is most impaired. The index describes the reader’s discomfort, not that of the victim. It is not perception that is associated with mortality. It is, often, its absence.

Next come the statistical misunderstandings. The first is the neglect of the denominator. If the number of people aged over 80 increases, the absolute number of vulnerable people automatically increases as well. Italy has one of the oldest populations in the world. Epidemiological studies standardise for age. Press releases almost never do. No less problematic is the selection of the dependent variable. If only heatwaves are studied, only the right-hand side of the curve linking temperature and mortality is observed – a U-shaped relationship, with a minimum at an intermediate temperature and increasing towards both colder and warmer temperatures. This means that both the left-hand side of the curve, relating to cold weather, and the reference point against which excess mortality is measured are lost from view.

Our argument is not a ‘denialist’ one. It is simply in line with the findings of one of the most extensive epidemiological analyses. Antonio Gasparrini and colleagues, writing in "The Lancet" in 2015, analysed over 74 million deaths across 13 countries and estimated that 7.71 per cent of mortality was attributable to sub-optimal temperatures. However, of this figure, around 7.3 per cent was associated with cold weather and only 0.4 per cent with hot weather. Furthermore, the majority of deaths were not caused by extreme weather events, but by moderately sub-optimal temperatures. Heatwaves make the news because they have a name and a date, a ‘before’ and an ‘after’. Winter does not: its mortality is spread over months, without a single headline-grabbing event. It is always the availability bias: what is vivid and localised carries more weight, in our perception, than what is widespread and prolonged.

There are many issues where comparisons are essential. And others where comparisons are not informative. As in this case. Comparing 2025 with 1960 without taking into account population ageing, air-conditioning in homes, improvements in healthcare and early-warning systems means attributing to the climate effects that also depend on societal change. Similarly, in Gasparrini’s study, the proportion of mortality attributable to temperature varies enormously from one country to another, a clear sign that temperature alone explains only part of the phenomenon. This is demonstrated by a 2016 study by Alan Barreca and colleagues. Over the course of the 20th century, the impact of the hottest days on mortality in the United States fell by around 75 per cent, almost entirely after 1960. The authors attribute much of this result to the widespread use of air conditioning. The relationship between temperature and mortality is not a natural constant: it depends on the degree of technological and social adaptation.

At the same temperature, with fewer elderly people, greater poverty, no weather warnings, no air conditioning, no A&E, no August holidays and no medication, more people were dying. The dose-response function is not a natural constant. It depends on the stock of capital – that is, on the "things" and technologies at our disposal. This renders the projections for 2050 inconsistent, as they factor in climate change and an ageing population but freeze the dose-response curve at the levels seen in the 2000s.

Millions of people are exposed to the same temperature without any ill effects. Mortality depends on the interaction between age, cardiovascular frailty, medication, social isolation, housing quality and energy poverty. To say that a person has died ‘from the heat’ does not describe an observable fact: it describes a counterfactual estimate, that is, what would probably not have happened in the absence of that exposure.

This does not mean that the heat is irrelevant. In public health, the relevant question is not what the ‘real’ cause of a death is. It is which cause we can influence. There is little we can do about someone being eighty-five years old. But when it comes to being eighty-five years old in an uninsulated loft, without care and without air conditioning, there is a great deal we can do. This should be at the heart of the public debate, far more so than summers portrayed as if the thermometer, on its own, were a death sentence.