Heat- And Cold-Associated Mortality in Germany, 2000–2023

Background: The number of deaths in Germany attributed to heat has risen in recent years. At the same time, winters have become milder, which implies that there may be fewer deaths due to extreme cold. In this study, we estimated the mortality associated with both heat and cold in Germany for the years 2000–2023.

Methods: We used daily death counts and mean temperature records to estimate the percentage of heat- and cold-associated deaths for the years 2000–2023 by means of a generalized additive model (GAM). We also carried out a meta-analysis to estimate the effect of temperature on the relative mortality risk (exposure–
response curve).

Results: We confirmed the recent increase in heat-associated mortality already reported elsewhere, while also documenting a decline in cold-associated mortality from approximately 2015 onward. As a result, overall temperature-associated mortality has fallen slightly over the past 10 years. We estimate that 4–5% of deaths are temperature-associated; typically, 3–4% are linked to cold, and just under 1% to heat.

Conclusion: The slight decline in temperature-associated mortality does not imply that there is no cause for concern in the future, nor can it be simply extrapolated to future developments. Temperature-associated mortality reflects both the relative mortality risk at a given temperature and the number of days with that temperature. If the number of days with high temperatures continues to rise, adaptation strategies will have to be developed rapidly to prevent a marked rise in heat-related deaths.

Cite this as: Rau R, Dietrich C, Köppen B: Heat- and cold-associated mortality in Germany, 2000–2023. Dtsch Arztebl Int 2026; 123: 357–61. DOI: 10.3238/arztebl.m2026.0059

The increased frequency of extreme temperature events associated with climate change raises questions about health consequences for the human population—particularly the danger of death. In the research presented here, we explored the trends in heat- and cold-associated mortality in Germany during the period 2000–2023. Winklmayr et al. (1) analyzed data from the years 2018–2020 and concluded that, despite possible adaptation to higher temperatures, heatwaves represent “a significant threat to human health in Germany.” It can therefore be assumed that the increase in ambient temperature will, in the long term, lead to increased mortality (2, 3, 4). At the same time, milder winters may bring about a reduction in cold-related mortality. However, the evidence does not all point the same way (e.g., (2, 3, 4)). Currently, the overall effect tends to be determined by cold-associated mortality, because far more fatalities are attributed to cold than to hot weather (3, 4, 5, 6).

Analysis of temperature-related mortality necessitates differentiation between direct and indirect causes of death. While direct heat- or cold-associated diagnoses account for only a small fraction of total mortality—in Germany, 14 persons died of heatstroke or sunstroke and 270 of hypothermia in 2023 (7)—assessment of the overall effect requires the inclusion of indirect causes. Temperature-associated excess mortality is documented in the literature and covers a spectrum all the way from cardiovascular and respiratory disease to automobile accidents and mental illness. When in this article we speak of heat- and cold-associated mortality, this is comparable with cause of death statistics, where typically the condition specified as underlying disease is the first link in the causal chain and not necessarily the immediate cause of death.

Heat and cold differ not only with regard to causes of death, but also in their dynamics. Models have to take time lags into account: In extreme heat events, mortality rises immediately, returning to baseline around a week later (8, 9, 10, 11, 12). This is attributable to acute cardiovasculatory disease, e.g., stroke. In conditions of extreme heat, death can be caused by impaired autogenous heat regulation, dehydration, and elevated blood viscosity and cholesterol levels. Moreover, study findings point to significant risks of death in persons with respiratory disease, diabetes, neurological disorders such as dementia, and mental illness, as well as external causes such as drowning or automobile accidents as a direct or indirect consequence of heatwaves (10, 11, 12, 13, 14, 15).

In contrast to the immediate effect of heat, the consequences of low temperatures emerge only after 3–4 days, with the danger of death staying high over the next 2–3 weeks (8, 9, 10, 11, 12). This time lag is determined essentially by respiratory disease and by cold-associated cardiocirculatory stress (hypertension, vascular stenosis). More deaths in persons with dementia or diabetes and from external causes such as falls and automobile accidents can be observed in the days and weeks after cold exposure (10, 11, 12, 16, 17). Interestingly, according to Breitner et al. cold often exerts an initial protective effect, i.e., for a short time the risk of death drops below baseline mortality, before this “benefit” is outweighed by an elevated risk of death in the days thereafter (8).

Analyses have already been undertaken to quantify the temperature-related mortality in Germany. The data documented so far come from research conducted in individual cities (8, 18) or are focusing on selected causes of death (19, 20, 21). Furthermore, most studies have concentrated on heat-associated mortality in the summer (1, 13, 22). In contrast, this original research investigates both heat-associated and cold-associated mortality.

Data

We evaluated the daily numbers of deaths in the period from 1 January 2000 to 31 December 2023, as compiled by the German Federal Statistical Office and provided at state level as a special analysis (23, 24). A total of 21 468 666 German residents died during that period, an average of around 2450 per day. As shown in the eTable, the data of the daily special analysis deviate only minimally from the official annual figures. More importantly, however, the number of deaths rose over the observation period, from 830 000–840 000 annually at the beginning to more than 1 million each year towards the end. To account for any confounders in this increase in deaths, the model incorporated the Federal Statistical Office’s annual population figures. Estimates for the daily population size were derived by linear interpolation of the data published on 31 December each year.

eTable

Comparison of the Federal Statistical Office’s official annual number of deaths (2000–2023) with this study’s annual figures aggregated from daily data

The daily temperatures were based on the daily mean of the air temperature at 2 m above the ground, published by the German Weather Service (25). The daily average temperature used in our analyses was the arithmetic mean of all available weather stations per federal state.

Methods

The temperature-associated deaths were estimated in a two-stage process. In the first step, the exposure–effect curves for each of the 16 German federal states were determined using the model developed by Gasparrini et al. (26). This model has many desirable properties. For instance, it assumes a quasi-Poisson distribution of the deaths. This permits greater variance than the commonly utilized Poisson models (27) featuring the restrictive assumption that mean and variance must be identical. The influence of temperature on mortality is estimated using a generalized additive model (GAM) (28, 29). This class of model enables a data-driven estimation of the association that is more flexible than the classic (generalized) linear models. Moreover, the model permits capture of a time lag between exposure and mortality: deaths due to a heatwave or cold spell do not necessarily occur on the same day, but may, as already mentioned, be delayed. The model we used takes this into account by permitting a delay in effect of up to 25 days depending on the temperature.

The population size was added to the original model (26) as an offset. This does not represent an additional variable with its own parameter, but takes account of changes in population size. A detailed explanation of all model components used can be found in the eMethods.

The illustration in Figure 1a, centered on the minimum mortality temperature (MMT; 17.9 °C), visualizes the positive properties of the exposure–effect association in the federal state of Bavaria. Because no linear, quadratic, or other specific form of association is assumed, the model permits flexible estimation of the risk of death, which can vary depending on temperature and time lag. The figure also shows why daily data—if available—are preferable to the frequently used weekly data. The change in mortality is heavily dependent on time, and summarizing 7 days of exposure to heat or cold neglects this complex structure.

The curves in Figure 1b and Figure 1c each show a cross section of Figure 1a for a specific temperature. They make it apparent that for heat the relative risk increases sharply on the same day and then decreases to something resembling a neutral, inconspicuous level over a roughly 2-week period. (Heat is defined as 25 °C in Figure 1b.) In contrast, a protective effect can be observed immediately after the onset of a cold spell, i.e., the relative risk drops below the reference value of 1. (Cold is defined as −10 °C in Figure 1c.) This observation may seem surprising at first but duplicates the findings of previous studies in, for instance, southern Bavaria and London (8, 30). The ensuing increase in mortality risk for cold is lower than that for heat but persists for a longer time.

Figure 1

Estimated exposure–effect curve for Bavaria, representative of the 16 German federal states, in the period 2000–2023 a) Centered on the MMT (17.9 °C) b) Estimated effect of temperature on mortality with a lag time of up to 25 days at 25 °C c) Estimated effect of temperature on mortality with a lag time of up to 25 days at –10 °C

In the second step, a meta-analytic model (31) was used to generate an overall estimated exposure curve for the whole of Germany from the individual federal state data (Figure 2). The relative mortality risk rises continuously with increasing cold, but not as rapidly as with heat. The lighter red areas on either side of the curve show the 95% confidence intervals. The curve is narrow because of the large size of the study population, becoming wider only at extremely low or high temperatures. The numbers of days with these temperatures were relatively low, however, as shown by the histogram behind the exposure–effect curve. This frequency distribution reveals the number of days with the given temperature (rounded to whole °C) during the study period.

As shown in eFigure 1, this model allowed for relatively accurate estimation of the trend in daily mortality. While the observed daily deaths are entered as black dots, the red line shows the deaths estimated by the model. With 2448 deaths per day, the mean absolute estimation error was 58 deaths, or 2.36%.

eFigure 1

Observed (black dots) and estimated (red line) number of daily deaths in Germany in the period 2000–2023

By means of the estimated exposure–effect model, the heat- and cold-associated deaths in the period 2000–2023 were estimated for each given exposure in each federal state and year, and cumulated over all 16 states (30, 32). They were expressed as a percentage of all deaths that occurred during the year. Percentages are in our view preferable to absolute numbers, because a rising population alone would lead to more deaths, even if there were no change in either the exposure–effect association or the numbers of hot and cold days. A trend line and a non-parametric LOESS smoothing function (33) were added to help visualize the trend. The confidence intervals are based on empirical Monte Carlo simulations (30). All statistical calculations were performed using R, version 4.5.3 (34).

Results

The red line in Figure 3 shows the trend in heat-associated mortality in Germany for the period 2000–2023 (with 95% confidence intervals). The proportion of all deaths associated with heat varies between 0.5% and 1.7%. As already observed in Germany, e.g., in (1), our research revealed a higher proportion of heat-associated mortality in recent years than at the beginning of the time series. This can be seen from the dashed trend line. However, the continuous red line, a non-parametric estimate of the trend, shows that the increase did not occur at the same rate from 2000 to 2023. Cold-associated mortality is shown in the same way in blue. As described in the literature, e.g., (2), cold-associated deaths represent a higher proportion of all fatalities, ranging from 2.7% to 4.6%. Therefore even in the year with the lowest figure for such deaths, the proportion was considerably greater than the highest percentage for heat. Furthermore, this analysis revealed a decreasing tendency in the proportion of deaths represented by cold-associated mortality since 2014. The overall temperature-associated mortality is shown in green in Figure 3. Owing to the higher proportion of deaths on cold days than on hot days, overall temperature-associated mortality in Germany has been decreasing slightly for almost a decade. Whether this will turn out to be a long-term trend or is just a short-term phenomenon cannot be determined owing to the short length of the observation period.

Discussion

The aim of the analysis presented here was to determine the trends in heat- and cold-associated mortality in Germany since 2000. To this end, we used a model with flexible estimation of the influence of temperature on mortality. The model takes account of temperature effects not only on the day of exposure but also those emerging several days later, and also that this time lag may vary depending on the prevailing temperature (see especially Figure 1). We were able to show (eFigure 1) that this model captures mortality with a high degree of accuracy.

When interpreting the results, it must be recalled that this was an observational study, permitting conclusions not about causality but merely about associations. Three general trends were observed (Figure 3):

Figure 3

The proportion of deaths associated with heat (red) or cold (blue) and the total thereof (green) in Germany in the period 2000–2023

• Heat-associated mortality increased somewhat over the study period.
• The mortality associated with low temperatures has, in contrast, been sinking slightly for almost a decade.
• Estimated overall temperature-associated mortality, i.e., deaths related to heat or cold considered together, is also decreasing. This is because a higher proportion of deaths are associated with low than with high temperatures. Typically, just under 5% of all deaths can be attributed to low and high temperatures.

Sensitivity analyses showed that the findings are relatively robust (see the eMethods for details). One potential factor is the COVID-19 pandemic, because the estimated exposure–effect curve covered the period 2000–2023 and thus included the excess mortality due to COVID-19. However, exclusion of the years 2020–2023 did not change the association (eFigure 2). Neither did the general pattern of temperature-associated mortality differ among different observation periods, with the exception of 2000–2018 (eFigure 3).

eFigure 2

Exposure–effect curve for Germany over the whole period (2000–2023, in blue) and without COVID-19 (2000–2019, in red)

eFigure 3

The proportion of deaths associated with heat (red) or cold (blue) and the total thereof (green) in Germany

As often described in the literature, cold-associated mortality begins at temperatures below the MMT. In the model for Germany the MMT would be 17.8 °C. However, inclusion of relatively moderate temperatures makes the number of “cold deaths” seem artificially high. A more restrictive cold threshold of 5 °C thus inevitably leads to lower rates of cold-associated mortality, but even then there is no qualitative change in the overall result (eFigure 4). With cold defined as the 2.5% percentile and heat as the 97.5% percentile (eFigure 5), higher heat-associated than cold-associated mortality was observed, which in view of the rapid rise in relative mortality during heatwaves (Figure 2) is not surprising. Nevertheless, in this scenario too there was a tendency towards decreasing temperature-associated mortality in the last 10 years of the study period.

Figure 2

Estimated cumulative exposure–effect curve with 95% confidence interval

eFigure 4

The proportion of deaths associated with heat (red) or cold (blue) and the total thereof (green) in Germany in the period 2000–2023

eFigure 5

The proportion of deaths associated with heat (red) or cold (blue) and the total thereof (green) in Germany in the period 2000–2023.

The results of this analysis emphatically do not permit the conclusion that temperature-associated mortality will be unproblematic in Germany in the future. The effects of climate change cannot be balanced against each, so, for example, our findings do not make protection plans (35) obsolete. For instance, there is already a tendency towards increased mortality that can be linked to higher temperatures, as also found in other studies (1).

Temperature-associated mortality depends on two factors: (1) the (relative) mortality risk at a given temperature and (2) the number of days for which this temperature prevails. Figure 2 visualizes this association: The relative mortality risk increases much more rapidly at high than at low temperatures. If average temperatures were to continue increasing at the same rate as in recent decades (36), the frequency distribution in the figure would shift to the right, meaning more days of elevated mortality risk and thus more “heat deaths.” Our analysis revealed no signs of any adaptation to rising temperatures (eFigure 6), i.e., no evidence that comparably high temperatures would result in lower mortality (1, 37). Initiatives to improve adaptation are advisable to counter the risk of increasing heat mortality.

eFigure 6

The estimated cumulative exposure–effect curve with 95% confidence interval for Germany, based on the metaregression of the pooled results for the 16 federal states

The goal of this research was to determine the trends in temperature-associated mortality over the whole of Germany. Future work on this topic should include systematic local analysis of regional climates. Even evaluation at federal state level, which forms the basis of our findings, does not always offer the necessary small-scale resolution with regard to local risk assessment. Apart from regional climates (e.g., in northwestern or southeastern Germany), attention needs to be paid to towns and cities, which are hotter, or heat up more quickly and to a greater extent, than the surrounding less urbanized areas with less soil sealing, e.g., the suburbs, and rural areas (e.g., 8, 10, 11, 18, 37). This would enable more accurate localization of any regional adaptation deficits. In combination with age-group analysis, which was not possible with our source data, targeted measures to benefit vulnerable segments of the population could be developed.

Acknowledgments

We thank the two anonymous reviewers of the manuscript and are grateful to Prof. Antonio Gasparrini and Dr. med. habil. Ulrich Hammer.

Conflict of interest statement
The authors declare that no conflict of interest exists.
Manuscript submitted on 21 November 2024, revised version accepted on 7 April 2026
Translated from the original German by David Roseveare

Corresponding author
Prof. Dr. Roland Rau

References

1. Winklmayr C, Muthers S, Niemann H, Mücke HG, An der Heiden M: Heat-related mortality in Germany from 1992 to 2021. Dtsch Arztebl Int 2022; 119: 451–7

2. Rai M, Breitner S, Wolf K, Peters A, Schneider A, Chen K: Impact of climate and population change on temperature-related mortality burden in Bavaria, Germany. Environ Res Lett 2019; 14: 1–11

3. Huber V, Krummenauer L, Pena-Ortiz C, et al.: Temperature-related excess mortality in German cities at 2°C and higher degrees of global warming. Environ Res 2020; 186: 1–10

4. Martinez-Solanas E, Quijal-Zamorano M, Achebak H, et al.: Projections of temperature-attributable mortality in Europe: A time series analysis of 147 contiguous regions in 16 countries. Lancet Planet Health 2021; 5: 446–54

5. Ingole V, Sheridan SC, Juvekar S, Achebak H, Moraga P: Mortality risk attributable to high and low ambient temperature in Pune city, India: A time series analysis from 2004 to 2012. Environ Res 2022; 204: 112304

6. García-León D, Masselot P, Mistry MN, et al.: Temperature-related mortality burden and projected change in 1368 European regions: A modelling study. Lancet Public Health 2024; 9: e644–53

7. Statistisches Bundesamt: Sterbefälle, Sterbeziffern (ab 1998). GBE – Gesundheitsberichterstattung des Bundes. 2026. www.gbe-bund.de:443/gbe/isgbe.archiv?p_indnr=6&p_archiv_id=7333299&p_sprache=D&p_action=A (last accessed on 15 May 2026).

8. Breitner S, Wolf K, Devlin RB, Diaz-Sanchez D, Peters A, Schneider A: Short-term effects of air temperature on mortality and effect modification by air pollution in three cities of Bavaria, Germany: A time-series analysis. Sci Total Environ 2014; 485: 49–61

9. Huynen MM, Martens P, Schram D, Weijenberg MP, Kunst AE: The impact of heat waves and cold spells on mortality rates in the Dutch population. Environ Health Perspect 2001; 109: 463–70

10. Demoury C, Aerts R, Vandeninden B, Van Schaeybroeck B, De Clercq EM: Impact of short-term exposure to extreme temperatures on mortality: A multi-city study in Belgium. Int J Environ Res Public Health 2022; 19: 3763

11. Parliari D, Cheristanidis S, Giannaros C, et al.: Short-term effects of apparent temperature on cause-specific mortality in the urban area of Thessaloniki, Greece. Atmosphere 2022; 13: 852

12. Ma Y, Zhou L, Chen K: Burden of cause-specific mortality attributable to heat and cold: A multicity time-series study in Jiangsu Province, China. Environ Int 2020; 144: 10599

13. Leyk D, Hoitz J, Becker C, Glitz KJ, Nestler K, Piekarski C: Health risks and interventions in exertional heat stress. Dtsch Arztebl Int. 2019; 116: 537–44

14. Walinski A, Sander J, Gerlinger G, Clemens V, Meyer-Lindenberg A, Heinz A: The effects of climate change on mental health. Dtsch Arztebl Int 2023; 120: 117–24

15. Basagaña X, Sartini C, Barrera-Gómez J, et al.: Heat waves and cause-specific mortality at all ages. Epidemiology 2011; 22: 765

16. Ryti NRI, Guo Y, Jaakkola JJK: Global association of cold spells and adverse health effects: A systematic review and meta-analysis. Environ Health Perspect 2016; 124: 12–22

17. Moriyama M, Hugentobler WJ, Iwasaki A: Seasonality of respiratory viral infections. Annu Rev Virol 2020; 7: 83–101

18. Rai M, Breitner S, Huber V, Zhang S, Peters A, Schneider A: Temporal variation in the association between temperature and cause-specific mortality in 15 German cities. Environ Res. 2023; 229: 115668

19. Zafeiratou S, Samoli E, Analitis A, et al: Assessing heat effects on respiratory mortality and location characteristics as modifiers of heat effects at a small area scale in Central-Northern Europe. Environ Epidemiol 2023; 7: e269

20. Zhang S, Breitner S, Rai M, et al.: Assessment of short-term heat effects on cardiovascular mortality and vulnerability factors using small area data in Europe. Environ Int 2023; 179: 108154

21. Zhang S, Breitner S, de’Donato F, et al.: Heat and cause-specific cardiopulmonary mortality in Germany: A case-crossover study using small-area assessment. Lancet Reg Health Eur 2024; 46: 101049

22. Huber V, Breitner-Busch S, He C, Matthies-Wiesler F, Peters A, Schneider A: Heat-related mortality in the extreme summer of 2022—an analysis based on daily data. Dtsch Arztebl Int 2024; 121: 79–85

23. Statistisches Bundesamt (Destatis): Sterbefälle nach Tagen, Wochen und Monaten – endgültige Daten. 2000–2019; 2024. www.destatis.de/DE/Themen/Gesellschaft-Umwelt/Bevoelkerung/Sterbefaelle-Lebenserwartung/Publikationen/Downloads-Sterbefaelle/statistischer-bericht-sterbefaelle-tage-wochen-monate-aktuell-5126109.html (last accessed on 25 February 2026).

24. Statistisches Bundesamt (Destatis): Sterbefälle nach Tagen, Wochen und Monaten. 2020–2024; 2024. www.destatis.de/DE/Themen/Gesellschaft-Umwelt/Bevoelkerung/Sterbefaelle-Lebenserwartung/Publikationen/Downloads-Sterbefaelle/statistischer-bericht-sterbefaelle-tage-wochen-monate-aktuell-5126109.html (last accessed on 25 February 2026).

25. Deutscher Wetterdienst (DWD): Tagesmittel der Lufttemperatur in 2 m Höhe; 2024. https://opendata.dwd.de/ (last accessed on 24 February 2026) beziehungsweise die Zeitreihen historical und recent: https://opendata.dwd.de/climate_environment/CDC/observations_germany/climate/daily/kl/ (last accessed on...). Die Daten werden unter der Creative Commons BY 4.0 “CC BY 4.0” Lizenz zur Verfügung gestellt.

26. Gasparrini A, Scheipl F, Armstrong B, Kenward MG: A penalized framework for distributed lag non-linear models. Biometrics 2017; 73: 938–48

27. Brillinger DR: The natural variability of vital rates and associated statistics. Biometrics 1986; 42: 693–734

28. Hastie TJ, Tibshirani RJ: Generalized additive models. London, UK: Chapman & Hall 1990.

29. Wood S: Generalized additive models: An introduction with R. Boca Raton, FL: Chapman & Hall/CRC 2006.

30. Gasparrini A, Leone M: Attributable risk from distributed lag models. BMC Med Res Methodol 2014;14: 55

31. Sera F, Gasparrini A: Extended two-stage designs for environmental research. Environ Health 2022; 21: 4

32. Gasparrini, A, Masselot P, Scortichini M, et al.: Small-area assessment of temperature-related mortality risks in England and Wales: A case time series analysis. Lancet Planet Health 2022; 7: e557–e64

33. Cleveland WS, Grosse E, Shyu WM: Local regression models. In: Chambers JM, Hastie TJ (eds.): Statistical models in S. Boca-Raton, FL: Chapman & Hall/CRC 1992: 309–76

34. R Core Team. R: A language and environment for statistical computing. Vienna, Austria 2026. www.R-project.org/ (last accessed on 26 May 2026).

35. Bundesministerium für Gesundheit 2025. hwww.bundesgesundheitsministerium.de/themen/praevention/hitze.html. Stand: 19. März 2025 (last accessed on 21 May 2025).

36. Deutscher Wetterdienst (DWD): Zeitreihen für Gebietsmittel für Bundesländer und Kombinationen von Bundesländer, erstellt am: 2. Mai 2025. https://opendata.dwd.de/climate_environment/CDC/regional_averages_DE/annual/air_temperature_mean/regional_averages_tm_year.txt (last accessed on 25 May 2025).

37. An der Heiden M, Muthers S, Niemann H, Buchholz U, Grabenhenrich L, Matzarakis A: Heat-related mortality. Dtsch Arztebl Int 2020; 117: 603–9.

38. Westermann: Diercke Weltatlas. Braunschweig, Westermann Schulbuchverlag 2023.

e1. Gasparrini A, Guo Y, Hashizume M, et al.: Mortality risk attributable to high and low ambient temperature: A multicountry observational study. Lancet 2015; 386: 369–75