Heat waves have killed millions. Here’s how scientists tally lives lost

Extreme heat is killing hundreds of thousands of people every year — and the annual toll is growing.

An estimated 210,000 more people died of heat-related causes each year from 2012 to 2021 than from 1990 to 1999¹, on average (see ‘Heat’s toll’) . Some 75,000 deaths per hot season have been linked to heatwaves in Asia from 1990 to 2019², and nearly 14,000 people are estimated to have died in June this year in Europe during a record-breaking heatwave³. This week, extreme heat has once again blistered parts of the continent and East Asia.

So how do scientists know how many people are killed by heat? Creating an accurate tally is not straightforward. Heat increases the risk of heart attack and exacerbates many serious health conditions, but death certificates rarely cite heat as the cause of death.

To avoid relying on death certificates for information, researchers have developed methods to isolate heat-related deaths from total deaths. The methods can’t give a precise death count, but researchers say that estimates of a heatwave’s death toll are necessary to develop measures to prevent future deaths. “Nobody needs to die in a heatwave,” says Kristie Ebi, a climate epidemiologist at the University of Washington in Seattle. “If you don’t know how many people died in a heatwave, you’re not going to take appropriate preventive action.”

How are heat deaths tallied?

Most public-health agencies estimate heat-related mortality using the ‘excess deaths’ method, which compares the total number of deaths during a heatwave with the figure for a baseline period without extreme heat. Subtracting the baseline death count from the heatwave death count yields the number of people assumed to have died from heat.

This method is straightforward and fast, says Antonio Gasparrini, a biostatistician and epidemiologist at the London School of Hygiene and Tropical Medicine, but it can’t distinguish between excess deaths caused by the heatwave and those caused by other factors. The method also can’t be used until after a heatwave has passed, limiting its effectiveness in forecasting the death toll for an upcoming heatwave.

A more sophisticated approach involves a tool called an exposure-response function. Such functions describe the correlation between temperature and the number of people who die at that temperature. Researchers use these functions in models that estimate the number of deaths for a particular rise in temperature. A model might predict, for example, that a 5 °C increase in temperature could result in a 3% increase in a city’s weekly mortality rate.

Unlike the excess-deaths method, such models can be applied in real time, such as during a heatwave, or to future scenarios. But accurate predictions of deaths require accurate temperature forecasts. And non-heat-related changes in mortality can decrease the accuracy of a model’s predictions.

Do the results of the two methods agree?

Last week, the UK Health Security Agency (HSA) announced that an estimated 2,877 people had died of heat-related causes in England during heatwaves in May and June. The agency used the excess-deaths method, comparing the average number of deaths on each day of a heatwave with the average number on days that did not have extreme heat.

Gasparrini and his colleagues also analysed the May and June heatwaves but used the exposure-response method instead of the excess-deaths method. Even so, their estimate of a total of 2,736 deaths in England and Wales closely matched the HSA’s estimate. They also found that almost half of the deaths can be attributed to human-induced climate change. The alignment between the methods inspires confidence in the exposure-response approach, says Gasparrini.