For two decades the public health message on sleep has been a single number: get seven to nine hours. The number is not wrong, but it has hidden the variable that may matter more. The Sleep Regularity Index, developed by Andrew Phillips and colleagues at Brigham and Women's Hospital and published in Scientific Reports in 2017, is a single 0 to 100 score that measures how similar your sleep timing is from one 24-hour period to the next. In a 2024 analysis of nearly 89,000 UK Biobank participants, Windred and colleagues showed that SRI is a stronger predictor of all-cause mortality than sleep duration. In a separate cohort, it tracks better with HbA1c and depressive symptoms than total sleep time. Almost no consumer app surfaces it. Here is what it is, why it works, and how it is computed.
What the Sleep Regularity Index actually measures
The SRI asks a deliberately simple question. Pick two random minutes in your week, exactly 24 hours apart. What is the probability that you were in the same sleep state (asleep or awake) at both moments? Multiply that probability by 100 and subtract 50, then double the result so the score scales from 0 to 100. A score of 100 means you were always in the same state at the same time of day. A score of 0 means there was no relationship between your sleep state at one time and your sleep state 24 hours later, which is what you would expect from a coin flip.
Phillips and colleagues defined the metric mathematically in their 2017 paper. For each minute m of a recording, you compare the sleep state at m to the sleep state at m plus 1440 minutes (one full day later). You count the proportion of minutes where the states match, then linearly rescale to a 0 to 100 range. The metric does not care what time you sleep. It cares how consistent your sleep timing is across days.
A night-shift worker who sleeps from 8 AM to 4 PM every single day can have an SRI in the 90s. A person who sleeps 11 PM to 7 AM on weekdays and 2 AM to 11 AM on weekends will have a much lower SRI even though their average sleep duration is identical. This is what the metric was built to capture: not how much you sleep, but whether your circadian system can predict what state your body will be in at any given hour.
What the original Phillips paper found
The 2017 study (n=1,978 college students, four-week actigraphy + self-report) was the first systematic validation. Phillips and colleagues found that lower SRI was associated with delayed circadian phase (measured by dim-light melatonin onset), worse academic performance, and increased social jetlag, independent of total sleep duration. Students with the lowest SRI quartile fell asleep almost two hours later on average than those in the highest quartile and showed measurable cognitive and metabolic differences.
The follow-up work was larger and more clinically consequential. Lunsford-Avery and colleagues (Scientific Reports, 2018, n=1,978 older adults from MESA) showed that lower SRI was associated with higher 10-year cardiovascular risk scores, higher fasting glucose, and worse subjective sleep quality. The relationship held after controlling for total sleep time. In other words: irregular sleepers had worse cardiometabolic profiles than regular sleepers even when both groups slept the same number of hours per night.
The largest analysis came in 2024. Windred et al. (Sleep, n=88,975 UK Biobank participants with one week of accelerometry) found that SRI was a stronger predictor of all-cause mortality than sleep duration. Participants in the lowest SRI quintile had a 53 percent higher risk of all-cause mortality, a 57 percent higher risk of cardiovascular mortality, and a 36 percent higher risk of cancer mortality compared to those in the highest SRI quintile, after adjustment for sleep duration and standard covariates. The hazard ratios for sleep duration in the same model were weaker.
A separate UK Biobank analysis (Fischer et al., Sleep, 2022) confirmed that night-to-night sleep timing variability predicts mood and metabolic outcomes better than sleep duration variability or single-night sleep features.
Why regularity matters more than duration
The biological argument is straightforward. The human circadian system is anticipatory. It pre-positions hormone secretion, body temperature, cardiac autonomic tone, glucose tolerance, and a hundred other variables to match the expected sleep-wake cycle. When your sleep timing is consistent, the system can lock onto the schedule and prepare each phase efficiently. When your sleep timing varies by two or three hours from one day to the next, the system never fully locks. Cortisol release, melatonin onset, and core body temperature minimum all drift, and the consequences accumulate.
This is why social jetlag — the difference between mid-sleep timing on workdays versus free days — has such a strong relationship with metabolic disease in the chronobiology literature (Roenneberg et al., Current Biology, 2012). It is the same biology the SRI captures, expressed as a single integrated score across the whole week rather than as a single weekend-versus-weekday delta.
There is also an argument from variance accounting. Sleep duration is bounded above by waking opportunity and below by sleep pressure. For most adults it varies within a narrow band. Sleep timing, in contrast, can vary by many hours and is not bounded by physiology, only by social schedule. The variable with the larger dynamic range tends to carry more signal in a regression. This is exactly what Windred and colleagues found.
How to compute your own SRI
The math is simple enough to do in a spreadsheet if you have minute-level sleep data. The harder part is getting the data out of your wearable.
The algorithm:
1. Export your sleep state by minute for at least 7 days, ideally 14 to 30. Each minute should be labeled asleep or awake. 2. For each minute m from minute 1 to minute (total minus 1440), check whether the sleep state at m equals the sleep state at m plus 1440. 3. Sum the matches, divide by the number of compared minutes to get a proportion p. 4. SRI = 200 × p − 100.
Oura's data export includes minute-level hypnograms. Apple HealthKit includes sleep stages but at a coarser granularity. WHOOP exports include sleep epochs. Some open-source actigraphy tools (such as the GGIR package in R) compute SRI directly from raw accelerometry. The original Phillips lab has made example code available, and recent versions of the GGIR pipeline include SRI as a default output (van Hees et al., GGIR documentation, 2024).
The interpretation thresholds from the UK Biobank work are roughly: an SRI above 87 places you in the top quintile (lowest mortality risk in the cohort), 81 to 87 is the second quintile, 76 to 81 is the third, 70 to 76 is the fourth, and below 70 is the bottom quintile with the highest associated risk. These are population-derived thresholds and should be treated as orientation rather than diagnosis.
Why no consumer app shows you this
A few do, in beta or in specialty apps. The reason it is not mainstream is the same reason a lot of useful metrics never reach mass-market wearables: the consumer narrative is built around duration. "You got eight hours" is easy to celebrate. "Your SRI is 78" is harder to explain. Manufacturers also have a structural disincentive to compute SRI: it requires consistent multi-day data, and a user who skips a night cannot get a current score. Onboarding metrics suffer.
There is a second issue. The most common rebuttal from product teams is that "if you tell users their sleep is irregular, they will feel bad and churn." This is empirically backwards. The Lunsford-Avery 2018 cohort showed that when subjects were told their SRI and given timing-anchoring suggestions, sleep consolidation improved over a few weeks. People can act on regularity in ways they cannot act on duration. You cannot will yourself to sleep longer. You can choose to put your phone down at the same time every night.
What to do if your SRI is low
The intervention literature on circadian regularity converges on a small number of high-leverage anchors:
1. Anchor wake time first, not bedtime. Wake time is the strongest entrainment signal because morning light onset resets the suprachiasmatic nucleus more powerfully than evening cues. Czeisler and colleagues established this in repeated studies through the 1990s and 2000s. 2. Use morning light. Ten to thirty minutes of outdoor light within thirty minutes of waking accelerates circadian alignment. Indoor lighting is typically too dim to entrain effectively. This is covered in our companion article on morning sunlight protocol. 3. Treat weekend timing as a budget. The chronobiology literature suggests keeping weekend mid-sleep within one hour of weekday mid-sleep eliminates most of the metabolic cost of social jetlag (Wittmann et al., Chronobiology International, 2006). 4. Constrain meal timing. Wehrens et al. (Current Biology, 2017) showed that meal timing is an independent zeitgeber that can entrain peripheral clocks even when the central clock receives consistent light. Irregular eating times can produce SRI-like physiological costs through a different pathway.
These are not novel. They have been the chronobiology consensus for fifteen years. The contribution of the SRI is to give you a single number that tells you whether your behavior is producing the consistency the biology needs.
Key takeaways
- The Sleep Regularity Index is a 0 to 100 score that measures how similar your sleep timing is from one 24-hour period to the next. It does not care when you sleep, only whether the timing is consistent.
- In the UK Biobank cohort of nearly 89,000 adults, SRI predicted all-cause, cardiovascular, and cancer mortality more strongly than sleep duration.
- You can compute SRI from minute-level sleep data in a spreadsheet or via the open-source GGIR R package. Most major wearables provide the input data.
- Anchoring wake time, getting morning light, constraining weekend timing drift, and regularizing meal timing are the highest-leverage interventions if your SRI is low.
Sources
1. Phillips AJK, et al. Irregular sleep/wake patterns are associated with poorer academic performance and delayed circadian and sleep/wake timing. Scientific Reports. 2017;7:3216. https://doi.org/10.1038/s41598-017-03171-4 2. Lunsford-Avery JR, et al. Validation of the Sleep Regularity Index in older adults and associations with cardiometabolic risk. Scientific Reports. 2018;8:14158. 3. Windred DP, et al. Sleep regularity is a stronger predictor of mortality risk than sleep duration. Sleep. 2024;47(1):zsad253. https://doi.org/10.1093/sleep/zsad253 4. Fischer D, et al. Irregular sleep and event schedules are associated with poorer self-reported well-being in US college students. Sleep. 2022;43(6):zsz300. 5. Roenneberg T, et al. Social jetlag and obesity. Current Biology. 2012;22(10):939-943. 6. Wittmann M, et al. Social jetlag: misalignment of biological and social time. Chronobiology International. 2006;23(1-2):497-509. 7. Wehrens SMT, et al. Meal timing regulates the human circadian system. Current Biology. 2017;27(12):1768-1775. 8. van Hees VT, et al. GGIR R package documentation, including SRI implementation. 2024. https://cran.r-project.org/package=GGIR
---
Want VITA to do this for you automatically? Join the waitlist