Figure 1a shows that heat-related mortality (red lines, with red shading indicating its 95% confidence interval) in England and Wales primarily occurred between July and September during the study period of 30 January 2020 to 31 December 2022. A total of 8,481 excess deaths (95% confidence interval: 6,387–10,493) were attributable to high temperatures, with daily heat-related mortality peaking at 580 deaths (95% confidence interval: 484–670) on 19 July 2022, when England recorded 40.3°C unprecedented extreme heat. Figure 1b zooms in on the time evolution of mortality during this heatwave for greater legibility. Although the UK Met Office and UK Health Security Agency (UKHSA) had issued Level 2 (yellow) and Level 3 (amber) Heat Health Alerts for all regions except North East England since 11 July 2021 to warn the public about this heatwave, which they subsequently raised to the highest Level 4 (red) alert for all regions on 18 July 2021 to trigger a national emergency response 26, peak heat-related mortality during this heatwave exceeded the maximum daily level from the previous ten years (i.e., 2010-2019; red error bar in Figure 1a), which was 376 deaths (95% confidence interval: 304–445).
In months other than July, August and September, cold-related mortality (blue lines, with blue shading indicating its 95% confidence interval) dominated over heat-related mortality. Over the study period, a total of 128,533 excess deaths (95% confidence interval: 107,430–153,642) were attributable to low temperatures, indicating a fifteen-fold larger cold than heat mortality burden. These results are consistent with the literature, which found that most days of the year are considered moderately cold in England and Wales in distributed lag non-linear models (DLNMs) for health modelling, resulting in a large number of cold-related deaths 27 . Daily cold-related mortality peaked at 531 excess deaths (95% confidence interval: 493–574) on 15 December 2022, but this falls within the range from the previous ten years (maximum: 691 deaths, 95% confidence interval: 643–743).
These temperature-related deaths are theoretically independent of COVID-19 deaths because they are calculated from DLNMs that describe the relationships between daily mean temperature and daily all-cause mortality, after COVID-19 mortality has been removed (see Methods). For comparison, Figure 1 also shows the time evolution of daily COVID-19 mortality according to death certificates (purple lines). Distinct surges in COVID-19 deaths were seen soon after the first emergence of COVID-19 in early 2020 and the domination of the Alpha variant in December 2020 (dotted vertical line in Figure 1a) 28, with the highest daily mortality level being 1,382 deaths on 19 Janurary 2021. The emergence of the Delta and Omicron variants were not followed by as large a surge in deaths, likely because COVID-19 vaccination had become more common by then (see Figure S1), with about half of the population above the age of 12 having had two doses of COVID-19 vaccines by 1 July 2021 (solid vertical line in Figure 1a), in all regions in England and Wales except London (see Table S1). Over the whole study period, 194,480 COVID-19 deaths were reported on death certificates in England and Wales.
Figure 1a shows that from June to October 2020, March to August 2021, and from September 2021 to the end of 2022, temperature-related deaths (the sum of heat- and cold-related deaths) exceeded COVID-19 deaths. These exceedances were driven by heat-related mortality spikes when COVID-19 mortality was relatively low, e.g., during the July 2022 heatwave (Figure 1b), as well as cold-related mortality dominating in the colder months after COVID-19 vaccination was introduced. To further examine the respective mortality impacts of non-optimal temperatures and COVID-19, Figure 2 shows the ratios of cumulative deaths from these two causes for each region in England and Wales, across the whole study period (panel a), and during heatwaves and cold snaps therein (panels b and c).
Considering the whole study period of 30 January 2020 to 31 December 2022, cumulative temperature-related deaths exceeded cumulative COVID-19 deaths by 8% in South West England. While this exceedance did not occur in the other regions, temperature-related deaths amounted to 58% (East Midlands) to 75% (London) of COVID-19 deaths by the end of 2022. These results demonstrate that even during a global pandemic that dominated everyone’s lives, heat and cold were a comparable but arguably much more under-reported killer than COVID-19. Reducing temperature-related mortality would free up resources and capacity for the health services to respond to major pandemics when they occur.
Since extreme weather events are where we would expect the health effects to be largest, focusing on them provides important information on their interplay with other parallel health crises, including their compound health effects. Figure 2b shows that during the ten heatwave episodes (spanning a total of 70 days; see Table S2) in the study period, identified through UKHSA’s Heat Mortality Monitoring Reports 26,29,30 , temperature-related deaths outnumbered COVID-19 deaths in 9 of the 10 regions (except in North West England). This exceedance is particularly apparent in the southern regions where heat stress is more pronounced 31. The ratios of temperature-related deaths to COVID-19 deaths in the southern regions range from 1.7 in East of England and South East England to 2.7 in London. The ratios for the rest of the regions lie between 1.1 and 1.3, except for North West England which has a ratio of 0.8. These results highlight that even during the COVID-19 pandemic, heatwaves posed a serious threat to public health, which is often downplayed 32 or misrepresented as something enjoyable by the media in the UK 33.
Figure 2c shows the corresponding results during eight cold snaps in the study period, which are defined here as days on which a Level 3 Cold Health Alert was issued by UKHSA for any region in England (also spanning 70 days; see Table S2). A Level 3 (amber) Cold Health Alert represents a situation in which impacts are likely to be felt across the health and social care sectors, and potentially the whole population 34. During these cold snaps, temperature-related deaths were lower than COVID-19 deaths in all regions, with the ratios ranging from 0.4 in East of England to 0.8 in South West England. These results are likely to be driven by the large surges in COVID-19 mortality following the first emergence of the coronavirus and the domination of the Alpha variant, both of which occurred in winter (Figures 1a and S1). In this sense, our results should not be interpreted as low temperatures being less important than COVID-19 to health in winter, as we have already shown that cold-related mortality occurs throughout the year and dominated over COVID-19 in the second half of the study period (Figure 1a). Future outbreaks of COVID-19 or novel viruses could have a different seasonal pattern from the COVID-19 pandemic studied here. Therefore, they could have different health impacts relative to extreme cold in winter.
The co-occurrence of non-optimal temperatures and COVID-19 meant that all-cause mortality in England and Wales was, on average, higher in the study period than in the previous ten years (black line and bar in Figure 1a). During extreme events, the health system needed to deal with an unprecedented compound health impact from both extreme weather and COVID-19. Figure 3a shows the total number of deaths arising from high temperatures and COVID-19 during the 70 heatwaves days in the study period. Regional compound (heat-related and COVID-19) mortality ranged from 19 deaths per 100,000 people (95% confidence interval: 16–22) in North West England, to 24 deaths per 100,000 people (95% confidence interval: 20–29) in Wales.
These compound mortality levels are put into context by comparing Figure 3a with Figure 3c, which shows the reference levels of heat-related mortality from 70 heatwave days, calculated from the average of all identified heatwaves in the ten years that preceded COVID-19 (i.e., 2010-2019; see Methods). Substantially fewer heat-related deaths occurred during 70 heatwave days in 2010-2019, with the regional number ranging from 5 deaths per 100,000 people (95% confidence interval: 3–8) in North West England, to 13 deaths per 100,000 people (95% confidence interval: 12–15) in London. In other words, demand on regional health services was 1.7 (London) to 3.8-fold (North West England) when extreme heat coincided with COVID-19 in the study period, compared to the previous decade.
During the 70 cold snap days in the study period (30 January 2020 to 31 December 2022), regional compound (cold-related and COVID-19) mortality ranged from 80 per 100,000 people (95% confidence interval: 75–86) in Yorkshire and the Humber, to 127 deaths per 100,000 people (95% confidence interval: 123–132) in East of England (Figure 3b), highlighting the higher absolute demand on the health system during cold snaps that heatwaves in the study period. These compound mortality numbers are substantially higher than the reference numbers of cold-related deaths from the same number of cold snap days in 2010-2019 (Figure 3d), which ranged from 34 deaths per 100,000 people (95% confidence interval: 31–38) in London, to 48 deaths per 100,000 people (95% confidence interval: 34–50) in North East England. This means that depending on the region, demand on health services was 2.1 (South West England) to 3.5-fold (East of England) when extreme cold co-occurred with COVID-19, compared to extreme cold in the previous decade.
By adding temperature-related deaths (with COVID-19 deaths removed prior to calculation) and COVID-19 deaths together to estimate the compound mortality impact, we have assumed that they are independent of each other. Figure 4 shows the average number of temperature-related deaths per 100,000 population per day of individual heatwaves (panel a) and cold snaps (panel b) in all regions in 2016-2019 (non-COVID-19 years; grey markers) and 2020-2022 (COVID-19 years; coloured markers), versus the average temperatures of these events. While heat-related mortality generally increased with the average heatwave temperature for all events, and cold-related mortality generally decreased when cold snaps were milder, the extreme temperature events that co-occurred with COVID-19 have different distributions (grey dashed lines) from the events that occurred without COVID-19 co-occurrence (black solid lines). For heatwaves (Figure 4a), mean temperature and heat-related mortality shifted higher in the COVID-19 years, compared to events that were not affected by COVID-19. For cold snaps (Figure 4b), the variances in temperature and cold-related mortality were larger in events in the COVID-19 years than non-COVID-19 years. Two-sample Kolmogorov–Smirnov tests confirm that the COVID-19 event distributions are significantly different from the non-COVID-19 distributions at the 5% significance level. These results suggest that COVID-19 may have impacted temperature-related mortality during extreme weather events.
On the other hand, extreme heat may have exacerbated COVID-19 mortality in England and Wales too. This is evident on the hottest day ever recorded in the UK (19 July 2022), when the 91 more daily COVID-19 deaths occurred, compared to the average on days between 10 July and 25 July 2022 (Figure 1b). Modelling COVID-19 mortality is not within the scope of this study, but our results highlight the complex interplay between extreme temperatures and the COVID-19 pandemic, as well as its implications on population health and health services capacity.