Digital biomarkers for sleep
The most widely captured digital health domain, and the one where the distance between consumer marketing and research evidence is widest.
%20(1).png)
Sleep is measured every night by tens of millions of consumer devices, which makes it the most widely captured digital health domain and also the one where the distance between consumer marketing and research grade evidence is widest. The measures collected here describe how long someone slept, how continuous that sleep was, and how long it took them to get there.
Polysomnography remains the reference standard. It is also expensive, requires a sleep laboratory, and changes the thing it measures by placing the participant in an unfamiliar bed connected to equipment. Wearables trade single night precision for something polysomnography cannot offer: many consecutive nights in the participant's own bed, which is where night to night variability becomes visible.
That trade shapes how these measures should be used. A wrist device is not a substitute for a diagnostic sleep study. It is a way to observe patterns across weeks, to detect change within a person, and to reduce reliance on recall, which diverges from objective measurement in predictable ways. People with insomnia in particular tend to underestimate how much they sleep, so a questionnaire and a sensor can disagree while both are informative.
Sleep measures in this library
How these measures are used
Sleep measures serve three purposes in research. They act as endpoints in insomnia and circadian trials, where change in continuity or duration is the outcome of interest. They act as safety and tolerability signals in drug trials where sedation, activation or disrupted sleep is a known or suspected effect. And they act as covariates almost everywhere else, because sleep affects mood, cognition, activity and glucose regulation, and an unmeasured shift in sleep can confound any of them.
Protocols typically combine an objective measure with a validated questionnaire such as the Insomnia Severity Index or the Pittsburgh Sleep Quality Index. This is deliberate. The questionnaire captures how the person experiences their sleep, the sensor captures what the device can detect, and treatment decisions in insomnia are driven by the experience as much as the number. Reporting change in both is now standard practice in this field.
What the evidence supports today
Consumer sleep trackers have been tested against polysomnography repeatedly, and the pattern of results is consistent. They estimate total sleep time reasonably well in healthy sleepers, they are considerably weaker at detecting wakefulness within the night, and their staging output is the least reliable part of what they report. Accuracy degrades in exactly the populations clinical studies care about most, including older adults and people with disrupted sleep.
The American Academy of Sleep Medicine's 2018 clinical practice guideline supports actigraphy for assessing sleep in specific clinical situations, which is the clearest professional endorsement this domain has. No consumer sleep tracking output currently holds a regulatory qualification as a trial endpoint.
The practical consequence is that these measures are strongest for within person change over time and weakest for absolute values or cross device comparison. A study designed around change from baseline on one device is on much safer ground than one comparing values between devices.
Common questions
Are consumer sleep trackers accurate enough for clinical research?
For total sleep time in healthy adults, generally yes. For detecting wake after sleep onset, less so, and for sleep staging, considerably less. Accuracy also drops in older adults and in people with disturbed sleep. They are best used for tracking change within a person over many nights rather than for establishing an absolute value on any single night.
Do these measures replace a sleep study?
No. Diagnosis of sleep disorders still requires polysomnography or a validated home sleep test. Wearable measures complement that by describing what happens across the weeks between clinic visits, which a single overnight study cannot show.
Why do wearable results disagree with what participants report?
Because they measure different things. Self report captures perceived sleep, which is influenced by how distressing the night felt. Sensors estimate physiological sleep from movement and heart rate. In insomnia the two diverge systematically, with participants typically underestimating how long they slept. Both signals are useful and neither is simply wrong.
How many nights should a study collect?
More than most people expect. Sleep varies substantially from night to night, so a handful of nights gives an unstable estimate. Two weeks of recording is a common minimum for characterising a person's typical sleep, and studies interested in variability itself need longer.
Run a study on these measures
WeGuide captures wearable data and patient reported outcomes in one platform, from screening through to analysis.
