Heat mortality monitoring report: 2023
Updated 22 July 2024
© Crown copyright 2024
This publication is licensed under the terms of the Open Government Licence v3.0 except where otherwise stated. To view this licence, visit nationalarchives.gov.uk/doc/open-government-licence/version/3 or write to the Information Policy Team, The National Archives, Kew, London TW9 4DU, or email: psi@nationalarchives.gov.uk.
Where we have identified any third party copyright information you will need to obtain permission from the copyright holders concerned.
This publication is available at https://www.gov.uk/government/publications/heat-mortality-monitoring-reports/heat-mortality-monitoring-report-2023
1. Summary
During summer 2023, there were an estimated 2,295 (95% confidence interval (CI): 1,681 to 2,910) deaths associated with the 5 periods of heat across the summer. UKHSA has published annual statistics on the impact of adverse heat episodes since 2016. Previous reports are available at Heat mortality monitoring reports. This UKHSA report is the final assessment of summer 2023, following interim analysis published in February 2024.
The summer of 2023 was the eighth warmest on record. June saw the majority of the hot weather, with July particularly wet and unsettled. A late season surge in temperatures was also observed with high temperatures occurring in September. The year 2023 was also the first season in which the new impact-based Heat Health Alerts (HHA) system was in operation. A full seasonal summary of the weather conditions for summer 2023 is available from the Met Office.
2. Methods
In June 2023, UKHSA launched the new HHA system, which moved away from a simple threshold-based approach to an impact-based approach. Details of the new alerting approach can be found on the Weather-Health Alert system user guide.
For the purposes of this report, a heat episode for analysis of heat-associated mortality was defined as an amber HHA in at least one region or the mean Central England Temperature (CET) being at least 20°C. One day either side was also included. There were 5 heat periods which met this definition as outlined below, comprising a total of 28 days. This is the third highest number of heat episode days since UKHSA statistics began in 2016.
2.1 Episode 1 (E1)
The first heat episode lasted 7 days (8 June to 14 June), during which an amber HHA was issued in the West Midlands and East Midlands, East of England, Southeast and Southwest regions. The mean CET also reached 20°C during this period.
2.2 Episode 2 (E2)
The second heat episode lasted 4 days (23 June to 26 June), during which no amber HHAs were issued. However, a yellow HHA was issued for Yorkshire and the Humber, East Midlands, West Midlands, East of England, London, Southeast and Southwest, with mean CET reaching 20°C.
2.3 Episode 3 (E3)
The third heat episode lasted 3 days (7 July to 9 July), during which no amber HHAs were issued. However, a yellow HHA was issued for Yorkshire and the Humber, East Midlands, West Midlands, East of England, London and the Southeast with mean CET reaching 20°C.
2.4 Episode 4 (E4)
The fourth heat episode lasted 3 days (10 August to 12 August), during which no amber HHAs were issued. However, a yellow HHA was issued for Yorkshire and the Humber, East Midlands, West Midlands, East of England, London, Southeast and Southwest England. Mean CET also reached 20°C during this period.
2.5 Episode 5 (E5)
The fifth heat episode lasted 11 days (3 September to 13 September), during which a yellow HHA was issued for the Northwest, Yorkshire and the Humber, West Midlands and East Midlands, East of England, Southeast, Southwest and London, with the East of England, Southeast and London all upgraded to an amber warning on the 9 September. The mean CET in this period also reached 20°C.
As in previous years, the estimate of all-cause heat-associated mortality was calculated by taking the difference between the observed number of all-cause deaths obtained from General Registry Office (GRO) (corrected for delays in registration) on the episode days with the average from the combination of the 14 non-episode days preceding and subsequent to the heat episode (forming a baseline level), having subtracted the estimated number of deaths attributed to coronavirus (COVID-19) on those days (defined as deaths with COVID-19 recorded on the death certificate). Estimates are broken down by region, age group and gender.
For the first time, estimates of the years of life lost (YLL) from heat-associated mortality are also included in the report. These were calculated from the estimates of all-cause heat-associated mortality broken down by age group, with remaining life expectancy values for each age group obtained from the latest available Office for National Statistics (ONS) national life tables for England averaged across sexes and ages.
In addition, as in the report for summer 2022, this year’s assessment also includes an estimate of modelled heat-associated mortality based on the observed regional temperature-mortality relationships in recent summers. For this analysis, mortality data for deaths occurring 2018 to 2022 was obtained from the ONS (using final annual registrations data for deaths registered 2018 to 2022, and provisional monthly registrations data for deaths registered since the end of 2022), with deaths with a mention of COVID-19 removed. Temperature data was obtained from the Met Office, using a latitude-longitude grid at 0.1 degree resolution to derive daily temperature data at regional level. This allows comment on how the observed heat-associated mortality in 2023 compares to estimates from models which account for confounding factors. Further detail on the methodology is available.
3. Observed mortality
During summer 2023, there were an estimated 2,295 (1,681 to 2,910) deaths associated with the 5 heat periods that occurred during the heat-health alerting season. Total heat-associated deaths observed during summer 2023 were significant in all regions, apart from the Northeast of England and East of England. Overall, the highest heat-associated deaths per one million population were seen in the Southeast region. Heat-associated deaths were significant in all age groups above 65. Across genders, heat-associated deaths were relatively equal. See tables 1, 2 and 3 for full breakdown of estimates across the whole season and by episode.
Across each period, spikes in daily mortality values can be observed in Figure 1. There are additional spikes within the daily series that are not associated with heat episodes and an increasing trend in daily mortality towards the end of the season. E1 observed the second highest heat-associated mortality of 635 (317 to 953), with significant mortality in the 85-and-over and 65-and-over age groups and in males and females. Over the 7 days, there were on average 91 heat-associated deaths per day. The lowest total mortality was observed in E2, with 119 (-105 to 343) where on average there were 30 heat-associated deaths per day. A total of 221 (29 to 412) heat-associated deaths occurred during E3, with an average of 74 heat-associated deaths occurring per day over the 3-day period, where no significant deaths were observed in any age group or region. E4 observed a total of 387 (189 to 585) heat-associated deaths, with an average of 129 heat-associated deaths per day over the 3-day period. Significant heat-associated deaths were observed in the 65-and-over age groups and in males and females. E5 was the longest heat period with 11 days of temperatures steadily rising, peaking on the 7 September, with a total of 934 (522 to 1346) heat-associated deaths over the period, which was an average of 85 heat-associated deaths per day over the period. This was the highest number of heat-associated deaths in a single episode in 2023. Significant heat-associated deaths were observed in the West Midlands, Southeast, London and Southwest regions, for the 65-and-over age groups and for males and females.
Table 1. Total estimated heat-associated mortality by heat episode and region
| Heat Period | 1 | 2 | 3 | 4 | 5 | All | All |
|---|---|---|---|---|---|---|---|
| Date | 8-14 June | 23-26 June | 7-9 July | 10-12 August | 3-13 September | All | Per million population |
| Whole population | 635 (317-953) (note 2) | 119 (-105-343) | 221 (29-412) (note 2) | 387 (189-585) (note 2) | 934 (522-1,346) (note 2) | 2,295 (1,681-2,910) (note 2) | 40 (29-51) (note 2) |
| Average heat-associated deaths per day | 91 | 30 | 74 | 129 | 85 | 82 | |
| North East | 53 (-24-130) | 32 (-23-88) | 9 (-37-54) | 49 (0-98) | -21 (-117-75) | 122 (-28-272) | 45 (-10-101) |
| North West | 107 (-13-227) | 36 (-50-122) | 11 (-61-83) | 49 (-26-125) | 87 (-69-243) | 291 (57-524) (note 2) | 39 (8-70) (note 2) |
| Yorkshire and Humber | 67 (-36-170) | 16 (-57-89) | 22 (-40-83) | 61 (-4-127) | 96 (-38-230) | 262 (62-462) (note 2) | 47 (11-83) (note 2) |
| West Midlands | 62 (-45-169) | -20 (-95-55) | 27 (-39-93) | 37 (-30-103) | 147 (7-287) (note 2) | 252 (45-459) (note 2) | 42 (7-76) (note 2) |
| East Midlands | 66 (-28-160) | -4 (-69-61) | 47 (-11-105) | 24 (-35-82) | 77 (-43-198) | 209 (28-391) (note 2) | 42 (6-79) (note 2) |
| East | 29 (-64-121) | 4 (-61-70) | 48 (-10-105) | 9 (-47-65) | 53 (-66-171) | 143 (-35-322) | 22 (-5-50) |
| South East | 126 (-14-266) | 50 (-48-149) | 48 (-36-132) | 63 (-23-148) | 186 (4-367) (note 2) | 473 (204-742) (note 2) | 50 (22-79) (note 2) |
| London | 65 (-37-166) | 7 (-64-79) | 15 (-44-75) | 47 (-16-110) | 173 (40-305) (note 2) | 307 (111-502) (note 2) | 35 (13-57) (note 2) |
| South West | 63 (-44-169) | -6 (-81-69) | -7 (-70-56) | 47 (-21-115) | 144 (2-286) (note 2) | 240 (32-448) (note 2) | 42 (6-78) (note 2) |
Note 1: Estimated all-cause heat-associated mortality by region for each individual heat period and total cumulative heat-associated mortality across all heat periods.
Note 2: Statistically significant values.
Note 3: 95% confidence intervals are in brackets.
Note 4: ONS Mid-Year Population Estimates for 2022 are used to calculate deaths per million population.
Table 2. Total estimated heat-associated mortality by heat episode and age group
| Heat Period | 1 | 2 | 3 | 4 | 5 | All |
|---|---|---|---|---|---|---|
| Date | 8-14 June | 23-26 June | 7-9 July | 10-12 August | 3-13 September | All |
| 0 to 24 | 19 (-14-53) | 9 (-15-32) | -1 (-19-18) | 7 (-13-27) | -3 (-43-37) | 31 (-31-94) |
| 25 to 44 | -25 (-71-22) | -19 (-52-14) | 1 (-28-31) | -2 (-30-27) | 11 (-50-73) | -33 (-124-58) |
| 45 to 64 | 43 (-69-154) | -57 (-133-20) | 64 (-6-135) | 5 (-63-72) | 125 (-19-269) | 180 (-34-394) |
| 65 to 74 | 120 (-6-247) | 34 (-56-124) | 10 (-65-85) | 93 (12-174) (note 6) | 60 (-103-223) | 317 (70-564) (note 6) |
| 75 to 84 | 162 (-12-336) | 48 (-74-171) | 57 (-48-161) | 139 (30-247) (note 6) | 402 (176-628) (note 6) | 808 (472-1,144) (note 6) |
| 85+ | 315 (118-513) (note 6) | 104 (-36-244) | 89 (-29-207) | 145 (22-268) (note 6) | 339 (82-596) (note 6) | 992 (611-1,374) (note 6) |
| 65+ | 598 (306-890) (note 6) | 186 (-20-393) | 156 (-19-330) | 377 (194-559) (note 6) | 801 (422-1,180) (note 6) | 2,117 (1,552-2,683) (note 6) |
Note 5: Estimated all-cause heat-associated mortality by age group for each individual heat period and total cumulative heat-associated mortality across all heat periods.
Note 6: Statistically significant values.
Note 7: 95% confidence intervals are in brackets.
Table 3. Total estimated heat-associated mortality by heat episode and gender
| Heat Period | 1 | 2 | 3 | 4 | 5 | All |
|---|---|---|---|---|---|---|
| Date | 8-14 June | 23-26 June | 7-9 July | 10-12 August | 3-13 September | All |
| Male | 323 (96-549) (note 9) | -18 (-175-140) | 132 (-4-269) | 192 (52-332) (note 9) | 488 (197-779) (note 9) | 1,117 (682-1,552) (note 9) |
| Female | 312 (89-535) (note 9) | 137 (-22-296) | 88 (-46-223) | 195 (55-335) (note 9) | 446 (154-738) (note 9) | 1,178 (744-1,613) (note 9) |
Note 8: Estimated all-cause heat-associated mortality by gender for each individual heat period and total cumulative heat-associated mortality across all heat periods.
Note 9: Statistically significant values.
Note 10: 95% confidence intervals are in brackets.
Please note, the overall column and row totals in tables 1 to 3 may not align due to rounding of mortality estimates.
Figure 1. Heat impact on mortality: aged 65 and over
Shaded areas on Figure 1 highlight periods that meet UKHSA criteria for estimating heat-associated mortality and were analysed by age, gender and region. All other aspects of Figure 1 are:
- daily all-cause deaths adjusted for registration delay (solid black line)
- daily all-cause deaths adjusted for registration delay and with COVID-19 deaths removed (dashed black line)
- pre-pandemic 5-year mean daily deaths baseline (grey dotted line)
- plus or minus 2-week baseline (solid green line)
- COVID-19 deaths (dotted black line)
- mean central England temperature (solid red line)
4. Years of life lost
Years of life lost (YLL) is a measure of premature mortality, which takes into account remaining life expectancy at age of death. Heat-associated YLL for summer 2023 has been calculated based on the distribution of heat-associated deaths across the age groups 65 to 74, 75 to 84, and 85-and-over in Table 2. (It is not calculated for younger age groups as they did not see statistically significant heat-associated mortality in summer 2023).
While the 85-and-over age group had the highest number of heat-associated deaths across summer 2023, the highest heat-associated YLL was found in the 75-to-84 age group, and the 65-to-74 age group also had a higher central estimate for heat-associated YLL than the 85-and-over age group. None of the differences were statistically significant. See Table 4 for a breakdown of YLL by age group and heat episode.
Table 4. Total estimated years of life lost from heat-associated mortality by heat episode and age group
| Heat Period | 1 | 2 | 3 | 4 | 5 | All |
|---|---|---|---|---|---|---|
| Date | 8-14 June | 23-26 June | 7-9 July | 10-12 August | 3-13 September | All |
| 65 to 74 | 1,941 (-105- 3,987) | 550 (-906- 2,006) | 162 (-1,051- 1,375) | 1,504 (194- 2,814) (note 12) | 970 (-1,666- 3,607) | 5,127 (1,132- 9,122) (note 12) |
| 75 to 84 | 1,504 (-111- 3,118) | 445 (-691- 1,582) | 529 (-441- 1,499) | 1,290 (283- 2,297) (note 12) | 3,731 (1,633- 5,828) (note 12) | 7,499 (4,381-10,617) (note 12) |
| 85+ | 1,147 (428- 1,866) (note 12) | 379 (-131- 889) | 324 (-106- 754) | 528 (80- 976) (note 12) | 1,235 (299- 2,171) (note 12) | 3,613 (2,223- 5,002) (note 12) |
Note 11: Estimated all-cause heat-associated YLL by age group for each individual heat period and total cumulative heat-associated YLL across all heat periods.
Note 12: Statistically significant values.
Note 13: 95% confidence intervals are in brackets.
5. Modelled mortality
Temperature-mortality associations were established using regional daily all-cause deaths (excluding COVID-19 deaths) and regional daily average temperatures occurring from May to September in the preceding 5 years (2018 to 2022). The technique of distributed lag non-linear modelling was used to consider the effects of temperature on mortality with a lag of up to 3 days. These associations were then applied to the observed temperatures during the heat episodes in summer 2023 to model the expected number of heat-associated deaths in each heat episode. This approach allows us to compare the deaths observed in summer 2023 with what might have been expected based on temperatures alone.
Across the summer, the modelled estimate for England of 2,278 (1,996 to 2,489) heat-associated deaths is very similar to the estimate using observed mortality of 2,295 (1,681 to 2,910). This indicates that overall, the number of heat-associated deaths observed was similar to what would be expected based on summer mortality patterns in 2018 to 2022.
When considering the best estimates for the individual heat episodes, there is some misalignment. For example, higher observed mortality than modelled was seen in E1 (by 263 deaths) and E4 (by 202 deaths), indicating a bigger impact than would have been predicted by the model at the observed temperatures. Conversely, E2 saw 183 fewer and E5 306 fewer deaths observed than predicted by the model. None of these differences are statistically significant considering the confidence intervals. The confidence intervals for the observed estimates for all episodes are very large, indicating large uncertainty. In comparison, the modelled estimates based on modelling techniques and recent temperature-mortality are more constrained.
Figure 2. Comparison of total heat-associated mortality estimates using observed mortality and modelled mortality by heat episode
Black dots indicate estimate values, and black lines indicate the 95% confidence intervals. Dashed lines are for visual separation between heat episodes.
6. Discussion
Meteorologically, summer 2023 was the eighth warmest on record. Only 2 of the heat episodes were associated with an amber HHA in at least one region of England, however 3 other periods of heat resulted in yellow HHAs being issued. Despite the relatively mild and cool summer, 2023 observed the third highest heat-associated mortality since the heat-mortality reports were first published in 2016. While E5 was the longest and had the highest total heat-associated mortality, the highest intensity of heat impact appears to have been in E4, in which an average of 129 heat-associated deaths per day occurred.
For the first time in UKHSA heat-mortality monitoring reports, YLL has been calculated in addition to mortality. This allows greater understanding of the impact of heat episodes beyond the number of heat-associated deaths, particularly for relatively younger age groups: the 75-to-84 and 65-to-74 age groups were found to have higher YLL from heat-associated mortality than the 85-and-over age group. In this report, YLL figures were calculated using the observed mortality estimates grouped into fairly wide ranges (65-to-74, 75-to-84, 85-and-over) and not disaggregated by sex, and should therefore be treated with caution. Future reports will conduct this analysis at a more granular level.
The recent trend of increasing number of HHAs being issued and heat-associated deaths has continued in 2023, as have other trends observed since the start of the COVID-19 pandemic, such as the shifting mortality baselines and underlying patterns in mortality (for example increasing trend in COVID-19 deaths in summer months, general increased trend in cardiovascular disease deaths and more deaths occurring at home). This means that this analysis and the interpretation of heat-associated mortality estimates are more complex than pre-pandemic reports. Therefore, it’s important that caution is taken when interpreting these results.
As in the 2022 report, UKHSA has employed 2 methodologies to understand heat-associated mortality in heat episodes, allowing comparison between the predicted mortality based on recent summer mortality trends and the observed mortality. No statistically significant difference was observed overall between the estimates produced by the 2 methods, indicating that the overall heat-associated mortality in 2023 was in line with trends from the last 5 years.
Misalignment was seen in the point estimates in 4 out of the 5 individual heat episodes (with E1 and E4 having higher mortality than predicted by modelling, and E2 and E5 having lower mortality than predicted by modelling). Aside from methodological differences in the 2 approaches, a possible reason for the differences between observed and modelled estimates, particularly for E1, could be worse impact of heat episodes earlier in the summer season when the population is less prepared for extreme heat. The prolonged warmer weather in June and September and clear public messaging could plausibly have contributed to behaviour changes meaning that the public were more prepared for the extreme temperatures in E2 and E5. Factors such as high night-time temperatures and humidity could also play a role.
More work is required to assess the contribution of these different factors to heat-associated mortality. Future analysis at UKHSA will seek to apply the modelling approach used in the 2022 and 2023 reports to previous years to understand the range of factors that modify the impact of heat episodes on mortality. This will include disaggregation of mortality data by additional factors such as place of death and cause of death. In addition, ongoing evaluation of the impact-based heat-health alerting system introduced in summer 2023 will support a better understanding of how alerts may have affected behaviour and response.
Despite the relatively cooler weather in 2023 than other recent summers, significant heat-associated mortality was still observed in 4 out of 5 of the heat episodes that occurred, including in short spells of heat lasting only 3 days and in heat episodes where only yellow HHAs were issued. This highlights the need for coordinated responses even for yellow HHAs to protect those most vulnerable to the impacts of heat, and suggests that the current decision-making aid temperature thresholds that underpin the HHA system are appropriate. With strong evidence that heat episodes in England are becoming more intense, longer and more frequent, it is important that organisations ensure they have heat response plans in place, aligned with the recently updated Adverse Weather and Health Plan as well as longer-term adaptation strategies to reduce avoidable deaths during adverse heat events in summer 2024 and beyond.