The short answer
In this UK study, people whose sleep and wake times were most regular from day to day were less likely to die during follow-up than people whose times were least regular. Regular timing fit the death records better than how long people slept did.
This describes groups of people over several years. It does not prove that keeping a steady schedule lowers anyone’s risk of death.
What this study cannot tell us
It cannot show that a steady schedule lowers the risk of death. The authors call their findings correlational, and say sleep regularity may be both a cause and a marker of earlier death. An irregular schedule could be a sign of poor health; this study cannot establish causation.
It cannot say what happens to an individual who changes their routine. Nobody in the study was asked to change their sleep times.
It cannot establish the exact strength of the association. The authors suggest the true strength likely lies between their minimally adjusted result and their fully adjusted result.
Who was studied
The study used UK Biobank, a large research group of volunteers who joined at ages 40 to 69 years. It analysed 60,977 people who wore a wrist movement sensor.
When the sensors were worn, ages ranged from about 43.5 to 79 years. The average age was 62.8, about 55.0% were women, and about 97.2% were White.
The authors say the findings come from an older age group of mostly White participants and should be repeated in other groups of people.
What was compared
People were sorted into five equal-sized groups by how regular their sleep was. Each higher regularity group was compared with the least regular group.
Regularity was scored on the Sleep Regularity Index, which runs up to 100, where 100 means sleeping and waking at exactly the same times every day. In plain terms, the most regular group went to sleep and woke within about 1-hour windows on most days, while the least regular group had about 3-hour windows.
The authors ran the same comparison for sleep duration, so the results could be put side by side.
What was measured
Everyone wore a movement sensor on the wrist for 7 days. The sensor data were used to estimate when each person slept, and from that the regularity score and the hours slept.
Deaths were then followed in national records, and on average people were followed for 6.30 years. There were 1,859 deaths from any cause.
Results use a minimal and a full model. The minimal version adjusts for age, sex and ethnicity. The full version also adjusts for physical activity, socioeconomic status, social visits, smoking, urban or rural home, shift work and some medicines. The full version covered fewer people (46,721) than the minimal version (60,780).
The main finding
Comparing the most regular group with the least regular group, the hazard ratio for death from any cause was 0.52 (0.45–0.60) in the minimal model and 0.70 (0.59–0.83) in the full model. A hazard ratio below the neutral value means a lower rate of death in that group during follow-up. The numbers in brackets are confidence intervals, which show uncertainty around each estimate.
For sleep duration, the reported estimates were 0.69 (0.60–0.80) in the minimal model and 0.76 (0.65–0.89) in the full model. These use different duration groups; the claim sheet identifies each comparison. The authors compared how well each approach fitted the death records and found the regularity models fitted better.
In both models, each of the more regular groups had a lower risk of death from any cause than the least regular group.
Limits of the study
Sleep was measured for 7 days only. That recorded week is a snapshot of a person’s sleep, and the authors list this as a limitation.
The study is observational, so differences in health, work and life circumstances between people may explain part of the result.
The full model adjusts for variables such as smoking status that the authors say may have both confounding and mediating roles. Their suggestion that the relationship lies between the minimal and full estimates is an interpretation, not an established bound.
The more regular groups were also different in other ways, such as more physical activity and fewer shift workers, as the participant table shows.
If you are 70 or older
The sensors were worn at ages from about 43.5 to 79 years. That reaches into older age, but in the text we could read the paper does not give separate results for its oldest participants.
The authors suggest that falling asleep and waking within about 1-hour windows each day may be a practical way to make sleep timing more regular. The study did not test this, or whether it changes anyone’s risk. If sleep problems or medicines affect your schedule, your own clinician can help.
How it fits the other evidence
A separate analysis of the same UK Biobank sensor data, using a different method, reached a similar conclusion. Because it draws largely on the same participants, its agreement is not independent confirmation.
A separate study of Japanese adults, which asked people about their sleep habits rather than using sensors, also found higher mortality in people with irregular sleep. A recent review of the topic found many studies pointing the same way, but those studies used different measures, so the review could not combine them into a pooled estimate.
What it means for your marker page
The sleep regularity page on MarkerScore explains how this marker is used and lists its sources. This article reads a source in plain English.
The paper
Windred et al. (2024). Sleep regularity is a stronger predictor of mortality risk than sleep duration: A prospective cohort study. Sleep. Read the paper PMID 37738616
Other papers mentioned
- Cribb et al. (2023). Sleep regularity and mortality: a prospective analysis in the UK Biobank. eLife. Read the paper
- Omichi et al. (2022). Irregular sleep and all-cause mortality: A large prospective cohort study. Sleep Health. Read the paper
- Kalkanis et al. (2025). Sleep regularity as an important component of sleep hygiene: a systematic review. Sleep Medicine Reviews. Read the paper
Where every number comes from
Each number above links here. The words in quotation marks are copied from the paper (PMC10782501 (open-access full text from NCBI PMC)).
60,977 (people)
The analysed sample was 60,977 UK Biobank participants.
Found in: Abstract; Introduction; Discussion (article zsad253)
“We calculated Sleep Regularity Index (SRI) scores from > 10 million hours of accelerometer data in 60 977 UK Biobank participants (62.8 ± 7.8 years, 55.0% female, median[IQR] SRI: 81.0[73.8–86.3]).”
62.8 (years)
The average age of participants was 62.8 years (standard deviation 7.8).
Found in: Abstract (article zsad253)
“60 977 UK Biobank participants (62.8 ± 7.8 years, 55.0% female, median[IQR] SRI: 81.0[73.8–86.3])”
40 to 69 years (years)
UK Biobank enrolled people aged 40 to 69 between 2006 and 2010.
Found in: Methods; Overview (article zsad253)
“A baseline cohort of approximately 502 000 participants aged between 40 and 69 years were recruited to the UK Biobank between 2006 and 2010”
7 days (days)
Participants wore a wrist accelerometer for 7 days, between 2013 and 2016.
Found in: Methods; Overview (article zsad253)
“From this baseline cohort, 103 669 participants wore Axivity AX3 devices (Axivity, Newcastle upon Tyne, UK) on their dominant wrist for 7 days under free-living conditions between 2013 and 2016.”
43.5 to 79 years (years)
Table 1 age-range row: across the ten percentile columns, the smallest value is 43.50 and the largest is 79.00 years.
Found in: Results; Table 1, age row (article zsad253)
“43.89–78.72 43.59–79.00 43.69–78.84 43.50–78.20 44.10–78.88 43.92–79.00 43.50–78.88 43.69–78.84 43.90–78.61 44.10–78.20”
100 (index points)
The Sleep Regularity Index runs to 100, which means perfectly regular sleep and wake times.
Found in: Methods; Sleep regularity and sleep duration (article zsad253)
“An SRI of 100 represents perfectly regular sleep–wake patterns, and zero represents random patterns.”
1,859 (deaths)
There were 1,859 deaths from any cause.
Found in: Results; Table 1 (article zsad253)
“All-cause mortality rate was 4.84 deaths per 1000 person-years, with 1859 all-cause deaths, 377 by cardiometabolic causes, and 1092 by cancer.”
6.30 years (years)
Average follow-up was 6.30 years.
Found in: Results; Participant characteristics (article zsad253)
“The mean (±SD) follow-up period between accelerometer recording and study endpoint (March 21, 2021) was 6.30 ± 0.83 years.”
0.52 (0.45–0.60) (hazard ratio)
Most regular fifth vs least regular fifth, all-cause mortality, minimally adjusted model (age, sex, ethnicity).
Found in: Results; Table 2 (article zsad253)
“Individuals in the 80–100th sleep regularity percentiles had the lowest risk compared to the 0–20th percentiles (minimal model: HR = 0.52 [0.45–0.60], p < .001, full model: HR = 0.70 [0.59–0.83], p < .001).”
0.70 (0.59–0.83) (hazard ratio)
Most regular fifth vs least regular fifth, all-cause mortality, fully adjusted model.
Found in: Results; Table 2 (article zsad253)
“Individuals in the 80–100th sleep regularity percentiles had the lowest risk compared to the 0–20th percentiles (minimal model: HR = 0.52 [0.45–0.60], p < .001, full model: HR = 0.70 [0.59–0.83], p < .001).”
0.69 (0.60–0.80) (hazard ratio)
Sleep duration 60–80th percentiles versus 0–20th percentiles, minimally adjusted model.
Found in: Results; Table 3 (article zsad253)
“Longer sleep duration predicted lower risk of mortality up to the 60–80th percentiles in the minimal model (HR = 0.69 [0.60–0.80], p < .001), and up to the 80–100th percentiles in the fully adjusted model (HR = 0.76 [0.65–0.89], p < .001).”
0.76 (0.65–0.89) (hazard ratio)
Sleep duration 80–100th percentiles versus 0–20th percentiles, fully adjusted model.
Found in: Results; Table 3 (article zsad253)
“Longer sleep duration predicted lower risk of mortality up to the 60–80th percentiles in the minimal model (HR = 0.69 [0.60–0.80], p < .001), and up to the 80–100th percentiles in the fully adjusted model (HR = 0.76 [0.65–0.89], p < .001).”
about 1-hour windows (hours)
The most regular fifth went to sleep and woke within about 1-hour windows on most days.
Found in: Discussion (article zsad253)
“people with the top 20% of SRI scores went to sleep and awoke within approximately 1-hour windows on the majority of their study days”
about 3-hour windows (hours)
The least regular fifth went to sleep and woke within about 3-hour windows.
Found in: Discussion (article zsad253)
“people with the bottom 20% of SRI scores went to sleep and awoke within approximately 3-hour windows”
five equal-sized groups (groups)
People were split into quintiles (five equal-sized groups) of sleep regularity, and each of the top four groups was compared with the lowest.
Found in: Methods; Statistical analyses; Table 2 (article zsad253)
“Hazards of mortality were estimated for each of the top four SRI and sleep duration quintiles compared to their respective lowest quintiles, which were hypothesized to have the highest mortality risk.”
Average physical activity rises across the regularity groups (no unit)
Table 1 mean physical activity (milli-g) by regularity group, least to most regular: 25.93, 27.68, 28.70, 29.42, 30.11.
Found in: Results; Table 1, physical activity row (article zsad253)
“25.93 ± 7.97 27.68 ± 7.85 28.70 ± 7.80 29.42 ± 7.98 30.11 ± 8.08”
Share of shift workers falls across the regularity groups (no unit)
Table 1 share of shift workers by regularity group, least to most regular: 10.37%, 8.46%, 7.00%, 5.80%, 5.12%.
Found in: Results; Table 1, shift work row (article zsad253)
“10.37 (1252) 8.46 (1024) 7.00 (847) 5.80 (702) 5.12 (622)”
Full model includes variables that may be both confounders and mediators (no unit)
The authors say the fully adjusted models contain variables (for example smoking status) that may be both confounders and mediators.
Found in: Discussion, limitations (article zsad253)
“Fourth, our fully adjusted models contain variables that potentially have both confounding and mediating effects (e.g. smoking status).”
Deaths from NHS Digital (England) and NHS Central Register (Scotland) (no unit)
Mortality data came from NHS Digital (England) and NHS Central Register (Scotland).
Found in: Methods; Mortality records (article zsad253)
“Mortality data were received from NHS Digital (England) and NHS Central Register (Scotland).”
Minimal: age, sex, ethnicity. Full: adds physical activity, socioeconomic status, social visits, smoking, urban or rural home, shift work, cholesterol or blood pressure medicine (no unit)
The minimal and full models as defined in the note to Table 2.
Found in: Results; Table 2 (article zsad253)
“Minimal model was adjusted for age, sex, and ethnicity. Full model was adjusted for age, sex, ethnicity, physical activity, socioeconomic status, social visits, smoking status, urbanicity, shift work, use of medication for cholesterol or hypertension.”
Correlational findings; regularity may be a cause and a marker (no unit)
The authors say the findings are correlational and sleep regularity may be both a cause and a marker of earlier death.
Found in: Discussion (article zsad253)
“Finally, we acknowledge the correlational nature of our findings. Sleep regularity may be both a cause and marker of premature mortality risk.”
A recorded week of sleep data per person (no unit)
The authors say a single week gives only a snapshot of sleep patterns.
Found in: Discussion (article zsad253)
“First, the single week of data collected for each individual provides only a snapshot of their sleep–wake patterns”
True strength lies between the minimal and fully adjusted results (no unit)
The authors say the true strength of the relationship likely lies between their minimally and fully adjusted models.
Found in: Discussion, limitations (article zsad253)
“The true strength of the sleep regularity-mortality relationship therefore likely lies between our minimally adjusted models, which contain only non-mediating covariates, and our fully adjusted models.”
Older age group of mostly homogeneous ethnicity (no unit)
The authors say the findings come from an older age group of mostly homogeneous ethnicity and should be repeated in other groups of people.
Found in: Discussion, limitations (article zsad253)
“Third, our findings are within an older age group of mostly homogeneous ethnicity, and should be replicated across other cohorts, including cross-culturally.”
Regularity models fit mortality better than duration models (no unit)
The authors compared model fit (Akaike Information Criteria) and regularity models fit all-cause mortality better than equivalent sleep duration models.
Found in: Results; model fit comparison (article zsad253)
“Formal model comparisons (model 1 vs. model 2) revealed that sleep regularity models provided a better fit to all-cause mortality data than equivalent sleep duration models.”
60,780 (people)
Number of participants in the minimally adjusted models.
Found in: Results; Table 2 (article zsad253)
“N = 60 780”
46,721 (people)
Number of participants in the fully adjusted models.
Found in: Results; Table 2 (article zsad253)
“N = 46 721”
Corrections
None recorded since publication.
Educational content only — not medical advice, diagnosis, or treatment. Not a medical device; not FDA evaluated. Consult a qualified healthcare professional about your own health, and call emergency services for urgent symptoms. Full medical disclaimer →