Sleep Regularity Beats Sleep Duration as Predictor of Death Risk

Sleep regularity — how consistently you fall asleep and wake up at the same times each day — is a stronger predictor of mortality risk than total nightly sleep hours, according to a landmark study published in the journal Sleep (Oxford University Press). This finding challenges one of the most widely cited health recommendations of the last two decades, suggesting that public health messaging built almost entirely around the "eight hours" target has been optimizing for the wrong variable. For anyone who already tracks biometrics closely, this is both a vindication of wearable data and a reason to rethink which numbers actually matter.

What the 2024 Study Actually Found

The research — Windred DP, Burns AC, Lane JM, Saxena R, Rutter MK, Cain SW, Phillips AJK, Sleep, 2024;47(1):zsad253, DOI 10.1093/sleep/zsad253 — appeared online in September 2023 and in the January 2024 print issue (volume 47, issue 1), with an official publication date of January 11, 2024. It analyzed a large prospective cohort to quantify how much sleep regularity and sleep duration each independently contribute to all-cause and cause-specific mortality. Rather than leaning on sleep diaries or one-off questionnaire snapshots, the investigators used accelerometer-derived data — wrist-worn activity trackers worn continuously — to capture objective, longitudinal sleep-wake patterns across multiple days per participant. The cohort came from the UK Biobank: 60,977 participants (mean age approximately 62.8 years; about 55% female), drawn from more than 10 million hours of accelerometer recordings. Over a follow-up period of up to 7.8 years, 1,859 participants died — the mortality events on which the analysis is based.

The central finding was direct: after adjusting for sleep duration and a comprehensive set of confounders — including age, sex, ethnicity, and a range of sociodemographic, lifestyle, and health factors — sleep regularity remained a robust, statistically significant predictor of mortality. Participants with highly irregular sleep schedules faced meaningfully higher risks of all-cause mortality and elevated risks from cardiometabolic disease and cancer, compared with those who kept consistent schedules. Crucially, the authors reported that adding sleep duration to a mortality model already containing the Sleep Regularity Index did not significantly improve prediction (p ≈ 0.14–0.20) — the concrete statistical basis for the claim that regularity, not duration, carries the dominant signal.

The magnitude of the effect was substantial. Relative to the least-regular quintile of sleepers, participants in the four more-regular quintiles showed roughly 20% to 48% lower all-cause mortality risk, 16% to 39% lower cancer mortality risk, and 22% to 57% lower cardiometabolic mortality risk. The study's authors computed the Sleep Regularity Index (SRI) for each participant, then stratified the cohort from most irregular to most regular and compared survival outcomes. The gradient was clear and dose-responsive: the more irregular the sleep, the higher the mortality hazard — and that relationship held after extensive covariate adjustment.

The Sleep Regularity Index: What It Measures and How It Works

The Sleep Regularity Index (SRI) did not originate with the 2024 mortality study. Andrew J.K. Phillips and colleagues introduced it in a 2017 paper in Scientific Reports ("Irregular sleep/wake patterns are associated with poorer academic performance and delayed circadian and sleep/wake timing") as a quantitative, actigraphy-derived measure of day-to-day sleep-wake consistency, validated in that work on 61 undergraduates tracked over 30 days. Notably, Andrew J.K. Phillips is also a co-author of the 2024 mortality paper, giving a clear methodological through-line from the metric's introduction to its large-scale epidemiological application. Daniel Windred, lead author of the 2024 mortality paper, worked with this research group to extend and apply the SRI to population-scale questions. The mortality study brought the SRI to mainstream attention, but the metric had already been used and validated in smaller cohorts before this work pushed it to the forefront of sleep epidemiology.

Understanding the sleep regularity index formula matters because it defines precisely what "regular" means in a scientific context — and the answer is more specific than most people assume.

The SRI Formula Explained

At its core, the SRI quantifies the probability that a person is in the same sleep-wake state (asleep or awake) at any two time points exactly 24 hours apart. The original Phillips et al. (2017) formulation is defined so that it can be expressed on a percentage-style scale:

SRI = -100 + (200 / (M − 1)) × Σ δ(s(t), s(t + 24h))

Where s(t) is the binary sleep-wake state (0 = awake, 1 = asleep) at epoch t; δ equals 1 if the states match across the 24-hour lag and 0 if they differ; and M is the number of valid consecutive 24-hour epoch pairs across the recording period. In this original formulation the index runs from −100 (perfectly irregular — always in a different state at the same clock time day over day) to +100 (perfectly regular — always in the same state at every clock time). Some downstream implementations and summaries re-express or rescale the metric onto a 0–100 range for readability. Because both conventions appear in the literature, when you see a reported SRI value it is worth confirming which scale was used.

In practical terms, more-regular sleepers score toward the high end of whichever scale is used, and shift-working or heavily socially-jet-lagged individuals score substantially lower. Exact population averages depend on the cohort, the scale convention, and the sleep-classification algorithm, so specific numeric bands should be treated as illustrative rather than universal — the safest comparison is always within a single study using a single, stated scale.

Critically, this formula demands longitudinal, high-resolution data — at minimum several consecutive days of continuous actigraphy sampled at short (e.g., one-minute) epochs. That's exactly why the SRI was impractical to compute at population scale before wrist-worn accelerometers became cheap and ubiquitous, and why this study's use of large-biobank accelerometer data — more than 10 million recorded hours — was methodologically significant.

The Sleep Regularity Questionnaire: A Self-Report Alternative

For clinical and research settings without wearable hardware, a sleep regularity questionnaire approach offers a lower-cost proxy. Participants self-report their typical bedtimes and wake times across weekdays and weekends; the variance between those anchors — sometimes called social jet lag magnitude — serves as a rough surrogate for irregular sleep timing. A representative sleep regularity questionnaire instrument might ask:

• What time do you typically fall asleep on workdays versus free days?
• How often does your sleep schedule shift by more than one hour between days?
• Do you use an alarm clock on weekends?
• How often do you nap, and at what times?

Validated self-report tools that capture regularity-relevant dimensions include the Munich Chronotype Questionnaire (MCTQ), which explicitly computes social jet lag from workday versus free-day sleep timing, and certain items of the Pittsburgh Sleep Quality Index (PSQI). Questionnaire instruments based on MCTQ principles are freely available through Till Roenneberg's chronobiology group and have been used in many epidemiological studies. While cruder than the formula-derived SRI, questionnaire-based regularity measures capture regularity-relevant timing information. Importantly, the 2024 mortality paper used objective accelerometer data — which gives its findings considerably stronger methodological standing than any self-report instrument could provide.

Sleep Regularity vs. Sleep Duration: Why the Distinction Is Consequential

Conventional wisdom — and the framing of most public health guidelines worldwide — has centered on sleep duration: adults need seven to nine hours per night, per recommendations from bodies including the American Academy of Sleep Medicine (AASM) and the Centers for Disease Control and Prevention (CDC). That's not wrong, but the 2024 study's findings suggest it's importantly incomplete.

Metric	What it measures	How it is typically captured	Mortality predictive power (per this study)
Sleep Duration	Total hours of sleep per night	Self-report, actigraphy, polysomnography	Did not significantly improve prediction once regularity was in the model (p ≈ 0.14–0.20)
Sleep Regularity (SRI)	Day-to-day consistency of sleep-wake timing	Multi-day actigraphy; questionnaire proxy	Stronger independent predictor: 20–48% lower all-cause mortality across more-regular quintiles
Sleep Quality	Subjective restfulness, fragmentation, efficiency	Pittsburgh Sleep Quality Index, EEG polysomnography	Not the primary focus of this study
Sleep Timing (Chronotype)	Preferred phase of sleep (early vs. late)	Munich Chronotype Questionnaire, actigraphy	Related to but conceptually distinct from regularity

The distinction matters mechanistically, not just statistically. Sleep duration tells you how much sleep someone accumulates; sleep regularity tells you how well their circadian system is entrained to a 24-hour rhythm. These are separable biological phenomena. You can consistently sleep six hours at the same clock time every night (high regularity, short duration) or erratically sleep nine hours at unpredictable times (high duration, low regularity). The 2024 study's key contribution is demonstrating that the circadian entrainment signal — captured by the SRI — dominates the total-hours signal in predicting who dies sooner, even when both are included in the same model.

This aligns with a broader shift in chronobiology. Decades of research — from the 2017 Nobel Prize in Physiology or Medicine awarded to Jeffrey C. Hall, Michael Rosbash, and Michael W. Young for their work on the molecular mechanisms controlling the circadian clock, to large-scale epidemiological studies of night-shift workers — have established that misalignment between behavioral timing and internal biological clocks carries serious, measurable physiological costs. Irregular sleep is, in effect, self-administered chronic circadian disruption. The immune system, glucose metabolism, cardiovascular regulation, and DNA repair mechanisms all operate on circadian schedules; when sleep timing is chaotic, those systems can receive conflicting or poorly timed cues, with plausible downstream consequences for disease risk.

What Sleep Regularity Means in Everyday Terms

For readers not steeped in sleep science, the practical meaning of sleep regularity comes down to one intuitive question: If someone observed your sleep-wake state at 11:30 p.m. on a Monday, how reliably could they predict your state at 11:30 p.m. on Tuesday? On Wednesday? On a Saturday? High regularity means that prediction is almost always correct. Low regularity means it's close to a coin flip.

Social and occupational forces systematically undermine regularity in modern life. Shift work is the most extreme case, but "social jet lag" — the gap between biological sleep timing during the work week and the later, longer sleep of weekends — affects a large share of the working population in industrialized countries, according to analyses by Roenneberg and colleagues. Late-night screen use, variable work schedules, cross-timezone travel, and the cultural normalization of "catching up on sleep" over weekends all erode the SRI without necessarily reducing average sleep duration. Someone who sleeps exactly eight hours every night but at wildly different clock times would score poorly on the SRI — and, per the 2024 findings, would face elevated mortality risk despite meeting the duration guideline.

Why this matters beyond the fitness tracker: If sleep regularity is a stronger predictor of mortality than duration, then the most important sleep intervention may not be "go to bed earlier" or "sleep longer" — it may be "go to bed at the same time every day, including weekends." That is a behavioral prescription, and it is considerably harder to sell than a supplement or a sleep-stage score.

Methodology: Why Accelerometers Make This Study More Credible

A persistent weakness of sleep epidemiology has been its reliance on self-reported data. Survey instruments — however carefully designed and validated — are vulnerable to recall bias, social desirability effects, and the genuine difficulty humans have accurately estimating their own sleep timing across multiple days. The 2024 Sleep study addressed this directly by deriving sleep-wake states from wrist accelerometry: the 60,977 participants wore devices recording wrist movement at high resolution across an extended wear period, and validated algorithms translated those motion signals into minute-by-minute sleep-wake classifications, yielding more than 10 million hours of usable data.

This approach has known limitations — actigraphy cannot distinguish sleep stages, and wrist movement is an indirect proxy for sleep state — but for the specific question of timing consistency, it is far more reliable than self-report. Whether someone fell asleep at 10:47 p.m. versus 1:13 a.m. on consecutive nights is precisely the kind of variation that questionnaire recall systematically flattens and that continuous accelerometry faithfully captures.

Using a large prospective cohort — with participants enrolled before mortality outcomes occurred, then followed for up to 7.8 years — also substantially strengthens causal inference relative to cross-sectional designs. Confounding remains possible in observational data, but the study's extensive covariate adjustment (age, sex, ethnicity, and a broad set of sociodemographic, lifestyle, and health factors, including physical activity levels derived from accelerometry rather than self-report) removes the most plausible alternative explanations more rigorously than most prior sleep studies could manage.

Implications for Wearables, Health Tech, and Quantified Self Culture

The health-tech world's relationship with sleep tracking is both its strength and its limitation here. Devices from Oura, Garmin, Apple, Fitbit, and WHOOP all generate continuous, longitudinal sleep-wake data — exactly the kind the SRI requires. In principle, any of these platforms could compute a real-time Sleep Regularity Index for their users and surface it as a primary health metric. Currently, most of them foreground total sleep time, sleep-stage distributions (light, deep, REM), and proprietary "readiness" or "recovery" scores. Regularity, when it appears at all, tends to be a secondary feature or buried in weekly summary views.

The 2024 study provides a compelling evidence base for a product and UX shift: if regularity predicts mortality more strongly than duration, then the notification reading "You got 6h 42m last night — try for 8 hours" may be less health-relevant than one reading "Your sleep timing varied by 2 hours 15 minutes this week — that inconsistency is your primary risk factor." That's not a trivial change. It requires product teams to communicate a more abstract concept — phase consistency rather than hours accumulated — to users who have spent years being trained to optimize a single nightly number.

There is also a critical data-quality caveat. Consumer wearables use proprietary, often unpublished algorithms to classify sleep from accelerometry. The SRI, as defined in research, requires validated sleep-wake classifications at high temporal resolution using algorithms benchmarked against polysomnography. Whether consumer device outputs can substitute for research-grade actigraphy when computing SRI has not been systematically established — a gap that represents both a genuine research opportunity and an important limitation for anyone hoping their consumer ring or watch is already computing a publication-quality regularity score.

The broader theme — that passively collected, longitudinal behavioral data can reveal health signals invisible to periodic clinical snapshots — is one that AI and machine learning researchers in digital health have been pursuing aggressively. The machine learning pipelines now being deployed across health tech are precisely what would be needed to scale SRI computation, integrate it into personalized risk models, and flag at-risk individuals before mortality outcomes materialize.

Key Takeaways

• Sleep regularity is a stronger predictor of mortality risk than sleep duration, according to Windred et al. (2024), published in Sleep (Oxford University Press, 47(1):zsad253), based on 60,977 UK Biobank participants and 1,859 deaths over up to 7.8 years of follow-up.
• The Sleep Regularity Index (SRI) — originally developed by Phillips et al. (2017) in Scientific Reports and applied at scale by Windred and colleagues (Phillips is a co-author of both) — quantifies day-to-day consistency of sleep-wake timing by comparing sleep-wake states 24 hours apart. It is defined on a −100 to +100 scale in the original formulation, and is sometimes rescaled to 0–100; confirm the convention before comparing scores across studies.
• More-regular sleepers showed roughly 20–48% lower all-cause mortality, 16–39% lower cancer mortality, and 22–57% lower cardiometabolic mortality across the four more-regular quintiles versus the least-regular quintile, after controlling for sleep duration and a comprehensive set of standard confounders.
• Sleep duration remains relevant, but adding it to an SRI-based model did not significantly improve mortality prediction (p ≈ 0.14–0.20) — a meaningful reversal of conventional epidemiological emphasis.
• The likely mechanism involves circadian misalignment: irregular sleep disrupts the timed coordination of immune, metabolic, and cardiovascular systems governed by molecular biological clocks.
• Sleep regularity questionnaire-based tools — including the Munich Chronotype Questionnaire (MCTQ) and social jet lag calculations — offer lower-cost proxies for the SRI in settings without wearable hardware, but objective accelerometry produces substantially more reliable estimates.
• Consumer wearable platforms already collect exactly the data type needed to compute SRI at scale, but most currently surface duration and sleep-stage metrics rather than regularity as a primary output — a meaningful gap between the science and the product.
• Social jet lag — irregular sleep driven by differing weekday versus weekend schedules — is one of the most common and most underappreciated drivers of low SRI in industrialized populations.

Research Gaps and the Road to Clinical Translation

The 2024 Sleep study is unlikely to be the last word on this question. Several important gaps remain. First, causal direction requires more rigorous investigation: irregular sleep may partly reflect underlying illness rather than cause it, and teasing those pathways apart in observational data — even prospective data with extensive covariate adjustment — is genuinely difficult. Randomized intervention trials that experimentally improve sleep regularity and then measure downstream biomarkers (inflammatory markers, glucose tolerance, cardiovascular endpoints) would substantially strengthen the causal case.

Second, the optimal SRI threshold — the score below which clinical intervention is warranted — has not been established. As with blood pressure or LDL cholesterol, translating a continuous risk metric into an actionable cut-point requires large-scale calibration studies across diverse populations with varying chronotypes, occupational demands, and cultural sleep norms. Third, the interplay between regularity, duration, timing (chronotype), and sleep quality as independent and interacting risk factors is still being mapped. The 2024 study is a powerful piece of that puzzle — not the complete picture.

For public health bodies, the practical implication is that guidelines may need to evolve beyond the duration-centric "seven to nine hours" framing to incorporate regularity as an equally important — or more important — behavioral target. For clinicians, the study suggests that asking patients not just "how long do you sleep?" but "how consistent is your sleep schedule, including weekends?" is a meaningful addition to routine health screening. And for the sleep technology industry, the study is a quiet but clear signal: the most mortality-predictive sleep metric is one that most devices are not yet prominently reporting. That gap between what the science demonstrates and what the product surfaces is, sooner or later, a competitive and public-health opportunity that someone will move to close.

---

Primary source: Windred DP, Burns AC, Lane JM, Saxena R, Rutter MK, Cain SW, Phillips AJK. "Sleep regularity is a stronger predictor of mortality risk than sleep duration: A prospective cohort study." Sleep. 2024;47(1):zsad253. DOI: 10.1093/sleep/zsad253. Oxford University Press. Published online September 2023; print January 2024.

SRI methodology: Phillips AJK, et al. "Irregular sleep/wake patterns are associated with poorer academic performance and delayed circadian and sleep/wake timing." Scientific Reports. 2017;7(1):3216. (Original introduction of the Sleep Regularity Index.)