News · Sleep
Your sleep tracker and your own guess tell different stories
A study in the journal SLEEP tracked up to 76,811 UK adults with wrist monitors and found seven hours in bed lined up with 30% lower death risk, more than double the signal from self-reported sleep.
Based on a peer-reviewed study in SLEEP
- Researchers led from the University of Sydney and the University of Ottawa used UK Biobank data, published in the journal SLEEP.
- Participants aged 40 to 69 wore wrist accelerometers for 7 days and also reported their own sleep, giving a direct head-to-head comparison.
- Up to 76,811 adults were analyzed for death from any cause and 69,153 for cardiovascular disease, over an average 8.0 years of follow-up.
- Compared with 5 hours, 7 hours of 'sleep period time' was associated with 30% lower mortality risk. The same 7 hours measured as actual sleep gave 17%, and self-reported gave 14%.
- Lowest risk sat at roughly 7.2 hours for self-reported sleep and 7.7 hours for sleep period time.
- The shape held across outcomes: an inverse J-curve for death, heart disease, diabetes and dementia, and a U-shape for depression.
- This is observational, and the wrist monitors ran for only one week, standing in for years of habit. The paper is paywalled beyond its abstract.
Ask someone how long they slept and they will tell you seven hours. Strap a monitor to their wrist and the number moves. A study in the journal SLEEP has worked out how much that gap matters, and the answer is: quite a lot.
Researchers used UK Biobank participants aged 40-69 who wore wrist-worn accelerometers for 7 days and self-reported their daily sleep duration. Having both measures in the same people let them ask a question the field had mostly ducked: does the health advice change depending on how you count?
The question underneath the guidance
Every sleep recommendation you have read rests on studies that asked people to estimate. Wearables changed what is measurable, but not the advice, and until now nobody had checked whether the two methods point the same way.
That is what this study set out to do: assess whether associations between sleep duration and health outcomes differ by measurement method, and compare dose-response curves to evaluate whether sleep duration recommendations should vary based on the method of assessment.
How it was done
The analysis covered 69,153 adults for cardiovascular disease and up to 76,811 adults for mortality, with an average follow-up of 8.0 years. Alongside death and heart disease, the researchers tracked incident type 2 diabetes, dementia and depression.
Three versions of sleep duration went into the models: the week of wrist data split into two different device measures, and each participant’s own estimate.
Same shape, different sharpness
Across all sleep measurement methods, adjusted dose-response curves showed a similar inverse J-shaped pattern, and a U-shaped association for incident depression. Put simply, too little sleep looked worse than too much, but both ends carried risk.
Where the methods parted company was strength. A sleep period time of 7 hours was associated with a 30% lower risk of mortality relative to the 5-hour reference value. In comparison, a total sleep time and a self-reported sleep duration of 7 hours were associated with 17% and 14% lower mortality risk relative to the 5-hour reference value.
Three numbers, one relationship, seen through progressively blurrier lenses. When people estimate their own sleep, the error flattens the curve.
The measure nobody talks about
The study’s quieter finding concerns which device number matters. Sleep period time, the whole stretch from dropping off to final waking, outperformed total sleep time, which counts only the minutes genuinely asleep.
The researchers suggest sleep period time may be more relevant than total sleep time for future device-based recommendations. For anyone reading a tracker each morning, that is a hint that the time you spend in the sleep window may say more than the app’s efficiency score.
The optimum landed close to familiar territory. The lowest estimated risk across outcomes occurred at approximately 7.2 hours for self-reported sleep and 7.7 hours for sleep period time.
What it does not settle
This is observational research, so it maps associations rather than proving that changing your sleep changes your risk. The direction of causation is genuinely uncertain at the long end: illness causes long sleep at least as readily as the reverse.
The wrist data also covers a single week, standing in for years of habit, and UK Biobank volunteers are healthier and more affluent than the general population. The paper sits behind a paywall, so the authors’ own list of limitations is not available here and these caveats are the conservative ones.
What survives is a reassuring conclusion and a useful correction. The findings support existing evidence that 7-9 hours of sleep per night is optimal for adults, regardless of how sleep is measured.
The advice was right. It was just being measured with a ruler that shook, and the sharper instrument shows the effect was bigger than anyone was crediting.
People also ask
What is the difference between 'sleep period time' and 'total sleep time'?
Sleep period time is the whole window from falling asleep to finally waking, including the brief awakenings almost everyone has and mostly forgets. Total sleep time counts only the minutes actually asleep inside that window. This study found sleep period time carried the strongest associations with health outcomes, which is why the researchers suggest it may be more relevant than total sleep time for future device-based recommendations.
Should I trust my tracker over my own sense of how I slept?
For estimating duration, the device data behaved more consistently here. Associations were substantially stronger for device-measured sleep period time than for self-report, which is what you would expect if devices measure duration with less error. That said, consumer wearables are not the research-grade accelerometers used in this study, and no tracker diagnoses a sleep disorder. Persistent poor sleep is a conversation for a doctor, not a dashboard.
So how many hours should I actually aim for?
The study supports the existing advice: 7 to 9 hours of sleep per night is optimal for adults, regardless of how sleep is measured. The lowest estimated risk sat near 7.2 hours by self-report and 7.7 hours for sleep period time. If you are using a tracker, remember it may be reporting time in bed asleep rather than pure sleep, so the target shifts slightly depending on which number your device shows.
Why was the death-risk figure so different between methods?
Because measurement error blurs associations. When sleep is estimated from memory, the noise flattens the curve, making the link look weaker than it is. The 30% figure for device-measured sleep period time against 17% for measured total sleep and 14% for self-report is best read as the same underlying relationship seen through three lenses of differing sharpness, not three different truths.
Does sleeping more than 9 hours harm you?
The curves in this study turned upward at the long end as well as the short end, which is the usual finding. But long sleep is often a marker of existing illness rather than a cause of it, since people who are unwell sleep more. That is a well-known limitation of this kind of research, and it means the long-sleep side of the curve should be read cautiously.