Research Christmas 2025: Shifting perspectives
Acute effects of daylight saving time clock changes on mental and physical health in England: population based retrospective cohort study
Melanie A de Lange, Kate Birnie, Rebecca C Richmond, et al
BMJ 2025; 391 doi: https://doi.org/10.1136/bmj-2025-085962 (Published 17 December 2025)Cite this as: BMJ 2025;391:e085962
Abstract
Objective To explore the acute effects of daylight saving time clock changes on mental and physical health events in primary and secondary care in England.
Design Population based retrospective cohort study.
Setting English primary care practices contributing to the Clinical Practice Research Datalink GOLD database, linked to hospital admissions and accident and emergency data.
Participants 683 809 people (road traffic injuries: all ages; cardiovascular disease: aged ≥40 years; all other conditions: ≥10 years) registered with a participating English general practice, with a health event for one of the health conditions of interest in their primary or secondary care record in the eight weeks surrounding the spring or autumn clock changes between 2008 and 2019.
Main outcome measures Health events were defined as a diagnosis code (or symptom code and prescription for mental health conditions in primary care) of anxiety, major acute cardiovascular disease, depression, eating disorder, road traffic injury, self-harm, or sleep disorder in primary or secondary care or a psychiatric condition in accident and emergency. Negative binomial regression models, adjusted for day of the week and region (and Easter weekend in spring), compared mean event rates per day in the week after the clock changes and the control period (four weeks before the changes and weeks 2-4 after).
Results In the week after the autumn clock change, five health conditions had fewer events: anxiety (from 17.3 events per day (per year, per region) to 16.7; incidence rate ratio 0.97, 95% confidence interval 0.95 to 0.98), acute cardiovascular disease (from 50.0 to 48.9; 0.98, 0.96 to 0.999), depression (from 44.6 to 42.7; 0.96, 0.95 to 0.97), psychiatric conditions (from 3.5 to 3.3; 0.94, 0.90 to 0.98), and sleep disorders (from 5.4 to 4.9; 0.92, 0.87 to 0.97). Little evidence was found of reductions in eating disorder diagnoses, road traffic injuries, or self-harm or of changes after the spring clock change.
Table 1
Sample characteristics at each patient’s start of follow-up*. Values are numbers (percentages)
| Characteristics | Total sample (n=683 809)† | Spring clock change (n=410 763) | Autumn clock change (n=424 264) |
|---|---|---|---|
| Sex: | |||
| Male | 294 482 (43.1) | 174 981 (42.6) | 180 635 (42.6) |
| Female | 389 318 (56.9) | 235 775 (57.4) | 243 622 (57.4) |
| Missing | 9 (0.0) | 7 (0.0) | 7 (0.0) |
| Age category, years: | |||
| <10 | 14 323 (2.1) | 7148 (1.7) | 8260 (1.9) |
| 10-19 | 51 783 (7.6) | 28 783 (7.0) | 31 263 (7.4) |
| 20-29 | 86 196 (12.6) | 49 808 (12.1) | 52 911 (12.5) |
| 30-39 | 95 629 (14.0) | 56 299 (13.7) | 59 456 (14.0) |
| 40-49 | 105 432 (15.4) | 63 531 (15.5) | 65 852 (15.5) |
| 50-59 | 88 482 (12.9) | 53 497 (13.0) | 54 900 (12.9) |
| 60-69 | 88 393 (12.9) | 54 169 (13.2) | 55 425 (13.1) |
| 70-79 | 88 569 (13.0) | 56 383 (13.7) | 56 369 (13.3) |
| 80-89 | 56 143 (8.2) | 35 696 (8.7) | 34 671 (8.2) |
| ≥90 | 8859 (1.3) | 5449 (1.3) | 5157 (1.2) |
| Index of Multiple Deprivation: | |||
| Most deprived | 175 621 (25.7) | 107 208 (26.1) | 110 765 (26.1) |
| Rest | 507 650 (74.2) | 303 198 (73.8) | 313 167 (73.8) |
| Missing | 538 (0.1) | 357 (0.1) | 332 (0.1) |
† Patients can appear in both spring and autumn analyses, so total sample is less than spring and autumn samples combined.
Conclusion The week after the autumn clock change was associated with a reduction in events for cardiovascular disease, sleep disorders, and mental health disorders, but little evidence suggested that the spring clock change was associated with a change in the number of health events. Electronic health records contain the date that a health event is recorded by a clinician, which is not necessarily the date of symptom onset.
