Daylight saving time affects European mortality patterns

Western Europe implements a biannual time change policy (daylight saving time) with clocks set forward by one hour on the last Sunday in March and back on the last Sunday in October. Although originally intended for energy conservation and time harmonisation, concerns have emerged about DST’s public health impacts, prompting calls by expert panels and scientific bodies to discontinue seasonal clock changes. Evidence suggests that even minimal shifts in circadian timing can disturb sleep and physiology and may influence the incidence of diseases and accidents. However, population-level effects on all-cause mortality remain unclear due to conflicting results, limited sample sizes, short study periods, heterogeneous control windows, and varied statistical approaches. This study aims to assess how DST transitions affect all-cause mortality across 16 European countries from 1998 to 2012 and to characterise short-term mortality patterns around the transition dates.

Prior research indicates DST-related circadian misalignment can influence conditions with temporal and circadian rhythmicity. Studies have linked spring transitions with a modestly increased risk of myocardial infarction, ischaemic stroke, motor vehicle accidents, atrial fibrillation, patient safety-related incidents, and suicides. Nonetheless, findings are often conflicting, with small and selective samples, short time horizons, non-uniform control periods around DST, and differing statistical methods that limit comparability. It has been hypothesised that these condition-specific effects could translate into detectable changes in general mortality, but robust, large-scale population evidence has been lacking.

Design and setting: Observational, multicountry time-series analysis assessing short-term changes in all-cause mortality around DST transitions across 16 European countries (AT, BE, HR, CZ, DK, UK [England and Wales], FR, DE, IT, LU, NL, PL, PT, SI, ES, CH) from 1998 to 2012. Exposure windows: data aligned around the DST transition dates (last Sunday in March and last Sunday in October). Control periods were set to two months before and after each transition. Data sources and population: Daily death counts and annual population numbers were obtained from national statistical offices for 145 NUTS2 regions, stratified by sex and one-year age, then aggregated to 10-year age groups for analysis. Meteorological data (daily mean temperature [K], relative humidity [%], wind speed [m/s]) and regional yearly GDP (in millions of US dollars) were incorporated. Some country-year-age strata were missing (e.g., mortality by sex/age missing for HR and UK before 2002; population by age missing for CH 1998; PL 1998, 2011–2012; FR 2012; DE 2011–2012) and excluded accordingly. Year 2000 served as the reference year. Outcome: Daily death counts and rates (daily death rates per 100,000 computed as daily deaths divided by daily population estimates). Statistical analysis: Multiple negative binomial regression models (log link) estimated incidence rate ratios (IRRs) for mortality around DST. Models adjusted for temporal factors (year, season, month, day of the week), meteorological variables (mean temperature, relative humidity, wind speed), geographic variables (latitude and longitude of NUTS2 centroid), a sine-cosine function capturing daylight seasonality, country fixed effects, regional and annual GDP, sex, and 10-year age group. Week indicators around DST transitions captured effects by week (week −1 as reference; weeks 0–8 pre/post). Interactions with sex, age, month, day of week, and country were explored; most were significant except sex-by-DST and age-by-DST spring (not shown). Sample encompassed 59,067,376 deaths and 13,714,704 observations (2 sexes × 9 age groups × 15 years × 145 regions × average 365.27 days/year; 4 leap years).

• Dataset included 59,067,376 deaths (population range ~397 million in 1998 to ~422 million in 2012).
• DST spring transition associated with a significant decrease in mortality during the first two weeks post-transition: week 1 IRR 0.965 (95% CI 0.951–0.979; p < 0.001); week 2 IRR 0.972 (95% CI 0.958–0.986; p < 0.001). Week 0 IRR 0.988 (0.978–0.998; p = 0.023). Later weeks not significant.
• DST fall transition associated with a significant increase in mortality during the first two weeks post-transition: week 1 IRR 1.018 (95% CI 1.003–1.033; p = 0.016); week 2 IRR 1.023 (95% CI 1.008–1.039; p = 0.002). Week 0 IRR 0.983 (0.972–0.993; p = 0.001). From week 4 onward, a quasi-constant significant decrease was observed (e.g., week 4 IRR 0.971 [0.957–0.986], p < 0.001; weeks 6–8 IRR ~0.982–0.983, p ~0.017–0.021).
• Septadian pattern: consistent weekly nadir on Sunday and peak on Monday, independent of season and DST. Monday vs Sunday increase IRR 1.029 (95% CI 1.024–1.033; p < 0.001), corresponding to ~15,102 excess deaths per year; Sunday IRR 0.972 vs Monday reference. Pattern present only among individuals aged ≥40 (both sexes). Young men aged 10–30 showed an opposite weekend pattern (peak on Sundays).
• Seasonality and trend: Significant overall decline in mortality from 1998 to 2012; mortality highest in winter and lowest in spring/summer, with month-wise declines from January to September then increases thereafter.
• Demographics: Men had higher mortality than women across all age groups; mortality increased steeply with age (e.g., age 80+ IRR 219.048 vs 0–9.9 reference).
• Geography/economy: Northern and eastern regions (higher latitude/longitude) modestly associated with higher mortality. Country effects: lowest mortality in Switzerland and Italy; highest in Croatia and Poland. Higher regional annual GDP associated with lower mortality (IRR 0.999 per unit).
• Meteorology: Higher mean temperature (IRR 1.004 per K) and wind speed (IRR ~1.000–1.001) associated with higher mortality; higher relative humidity associated with lower mortality (IRR 0.980 per %).

This continental-scale study demonstrates that DST transitions are associated with short-term changes in all-cause mortality: a decrease following the spring forward and an increase following the autumn return to standard time during the first two weeks post-transition. These findings counter the prevailing assumption that sleep loss from the spring transition necessarily elevates overall mortality and suggest that, at the population level, the net effect may be modestly protective in spring and adverse in fall. The analysis also reveals a robust, year-round septadian mortality rhythm—lowest on Sundays and highest on Mondays—limited to adults aged 40 and over and independent of DST, which could confound prior DST studies if not accounted for. Potential explanations for the Monday peak include detection/reporting biases, healthcare weekend effects, lower weekend air pollution, and protective social behaviors on weekends; the opposite pattern in young men likely reflects transport accidents and self-harm concentrated on weekends. The observed post-fall week 4+ mortality decrease remains unexplained and occurs despite seasonal adjustments. Overall, the results clarify population-level mortality dynamics around DST, highlight a pervasive weekly pattern, and underscore the importance of accounting for temporal, meteorological, geographic, and socioeconomic factors in such analyses.

DST influences all-cause mortality in Western Europe, with a significant short-term decrease in deaths during the first two weeks after the spring transition and a significant increase during the first two weeks after the fall transition. Independently, mortality exhibits a consistent weekly pattern with a Sunday nadir and Monday peak among adults aged 40+. These findings provide large-scale population evidence to inform policy discussions on seasonal time changes. Future research should leverage cause-specific and hourly mortality data, individual-level demographics and health status, and information on sleep parameters and chronotypes, and compare regions with and without DST to elucidate mechanisms and subgroup effects.

• Cause-specific daily mortality by age, sex, and region was unavailable at the European level, limiting subgroup and mechanistic insights; death certificate inaccuracies further constrain interpretation.
• Lack of individual-level demographic and clinical data beyond age and sex; no access to time-of-death precluded hourly analyses (e.g., distinguishing late Sunday night vs early Monday deaths).
• Potential DST-related effects confined to specific subgroups may be diluted at the population level; comparisons with Western countries without DST were not feasible.
• Observational design precludes causal inference.
• No data on sleep parameters or chronotypes, which could modulate responses to DST.