Getting seven hours of sleep a night is widely recommended, but how you measure those seven hours matters. A large new study using UK Biobank data shows that device-measured sleep period time — the time a person actually spends in bed intending to sleep — is more strongly linked to lower risks of death, cardiovascular disease, diabetes, and dementia than either total sleep time recorded by a device or self-reported sleep duration.
The study, published in Sleep by researchers at the Children’s Hospital of Eastern Ontario Research Institute, the University of Ottawa, and the University of Sydney, compared three ways of measuring sleep duration and their associations with five major health outcomes in 69,000 to 77,000 adults aged 40 to 69 years.
What they found
Participants wore wrist accelerometers for seven days, yielding two device-based metrics: total sleep time (TST), which captures actual minutes of sleep, and sleep period time (SPT), which captures the total time spent in bed intending to sleep. Participants also reported their typical nightly sleep duration by questionnaire. The researchers tracked health outcomes over an average of 8 years.
The associations between sleep duration and most outcomes followed an inverse J-shaped pattern — meaning that very short sleep and very long sleep both carried higher risk, with the lowest risk at an intermediate duration. This pattern held across all measurement methods but was most pronounced for device-based measures. For depression, the relationship was U-shaped.
The clearest differences emerged when comparing mortality risk at 7 hours of sleep versus a 5-hour reference point:
- People whose sleep period time (device SPT) was 7 hours had a 30% lower risk of dying during follow-up (hazard ratio 0.70, 95% CI 0.61-0.79).
- Those with 7 hours of device-measured total sleep time had a 17% lower risk (HR 0.83, 95% CI 0.78-0.89).
- Those who self-reported 7 hours of sleep had a 14% lower risk (HR 0.86, 95% CI 0.74-1.00).
The nadir — the sleep duration associated with the lowest risk across outcomes — differed by measurement method. For self-reported sleep, the lowest-risk point was around 7.2 hours. For device-measured sleep period time, it was around 7.7 hours. For device-measured total sleep time, the lowest-risk range was wider, spanning 6.8 to 9.3 hours.
These patterns were broadly consistent across outcomes: inverse J-shaped for all-cause mortality, cardiovascular disease, type 2 diabetes, and dementia, with the strongest associations consistently seen for sleep period time. The U-shaped curve for depression meant that both short and long sleepers faced elevated risk relative to those in the middle range.
Why it matters
Current sleep guidelines recommend 7 to 9 hours per night for adults, a range based largely on studies using self-reported sleep duration. The new findings support those guidelines but add a critical layer of precision: the measurement method matters when quantifying the link between sleep and health.
The study suggests that sleep period time — how long a person spends in bed intending to sleep — may be a more relevant target for future device-based sleep recommendations than total sleep time. This distinction has practical implications. Many wearable devices and smartphone apps report total sleep time, but the time a person allots for sleep in their daily schedule (going to bed early enough, staying in bed long enough) may matter more for health outcomes than the exact minutes of actual sleep captured by movement-based algorithms.
The stronger associations for device-based measures also highlight a persistent challenge in sleep research: self-reported sleep is subject to recall bias, rounding, and social desirability effects. People commonly round to the nearest hour or overestimate their sleep duration, which can attenuate statistical associations. The dose-response curves in this study were similar in shape across methods, but the magnitude of the associations was substantially larger for device-measured sleep period time.
For clinicians, the findings reinforce that asking a patient “how many hours do you sleep” may not capture the full picture. Device-based tracking, even if imperfect, can reveal patterns — particularly sleep period time — that self-report alone misses.
Limits
The study is observational, meaning it can identify associations but cannot prove that changing sleep duration directly causes changes in health outcomes. The UK Biobank cohort is healthier and less diverse than the general UK population, which may limit generalizability. Accelerometer-based sleep measures have known limitations: they cannot distinguish between quiet wakefulness and light sleep as accurately as polysomnography. Additionally, the single 7-day measurement window may not capture habitual sleep patterns over longer periods.
Bottom line
Both self-reported and device-measured sleep duration show broadly similar dose-response relationships with major health outcomes, but device-measured sleep period time — how long a person stays in bed intending to sleep — shows substantially stronger associations with lower mortality and disease risk. The findings support current guidelines of 7 to 9 hours of sleep per night regardless of measurement method, while suggesting that future device-based recommendations may benefit from focusing on sleep period time rather than total sleep time alone.
Source
Chaput JP, Biswas RK, Ahmadi M, Cistulli PA, Bian W, Stamatakis E. Dose-Response Associations Between Sleep Duration and Health Outcomes in Adults: Comparison Between Self-Reported and Device-Based Measures. Sleep. 2026; zsag193. DOI: 10.1093/sleep/zsag193. PMID: 42454954.