Digital biomarkers for mental health and stress
The domain that promises the most and delivers the least reliably. Arousal is measurable, and it is not the same thing as stress.
.png)
This is the domain where digital measurement promises the most and delivers the least reliably, and the directory treats it accordingly. Passive sensing can produce plausible looking indicators of stress, mood and behaviour from heart rate, electrodermal activity, sleep, movement and phone use. What it cannot yet do is establish that those indicators correspond to the psychological states they are named after.
The measures collected here are behavioural and physiological signals with a defensible relationship to mental health outcomes: sleep regularity, activity patterns, physiological arousal, and the circadian structure of daily behaviour. They are presented as what they are, correlates and context, rather than as measurements of mood.
The distinction matters commercially as well as scientifically. Consumer products routinely label a physiological arousal index as a stress score, which implies a level of construct validity that the underlying signal does not support. Arousal rises with exercise, caffeine, cold, illness and excitement as readily as with distress. A measure that cannot distinguish a difficult meeting from a brisk walk is not measuring stress, and describing it as though it were is the failure mode this domain most needs to avoid.
Mental health and stress measures in this library
How these measures are used
In research these measures are used as objective context around validated psychological instruments rather than in place of them. A trial in depression measures depression with a validated scale, and uses passive sensing to describe what changed in sleep, activity and daily routine alongside it. That combination can show whether a symptomatic improvement was accompanied by a return to normal behaviour, which self report alone cannot establish.
The second use is detecting change in routine over time. Disruption to sleep timing, a contraction in the range of places a person goes, or a sustained fall in activity are behavioural changes that precede or accompany deterioration in several conditions, and continuous data makes them visible between appointments.
Studies in this area carry heavier consent and governance obligations than most. The data is intimate, inference about mental state is sensitive, and participants should understand what will be inferred, by whom, and what happens if a signal suggests risk.
What the evidence supports today
The evidence supports association rather than measurement, and that distinction should govern how these measures are used. Sleep disruption, reduced activity and irregular daily routine are reliably associated with mood and anxiety disorders across many studies. What has not been established is that a passive index computed from those signals measures the psychological construct it is named after with enough validity to serve as a trial endpoint.
Physiological arousal measures such as electrodermal activity and heart rate variability respond to a wide range of stimuli, only some of which are psychological distress. Their specificity is the unresolved problem, not their sensitivity.
No passive mental health measure in common use currently holds a regulatory qualification as a clinical trial endpoint, and validated questionnaires remain the accepted outcome measures in this field. The realistic near term contribution of passive sensing is to describe behavioural change objectively and frequently, which is genuinely useful, and to stop short of claiming to measure how someone feels.
Common questions
Can a wearable measure stress?
It can measure physiological arousal, which is not the same thing. Arousal rises with exercise, caffeine, cold, illness and excitement as well as with psychological distress, and the sensor cannot distinguish between them. Consumer stress scores are built on this signal, and they should be read as arousal indices rather than as measurements of how someone feels.
What is passive sensing?
Collecting behavioural and physiological data continuously in the background, without asking the participant to do anything. Typical sources include movement, sleep, heart rate and, in some studies, phone use patterns. Its advantage is that it does not depend on the participant remembering or being willing to report; its cost is that it is intrusive and requires careful consent.
Do these measures replace depression or anxiety questionnaires?
No. Validated questionnaires remain the accepted outcome measures in mental health trials. Passive measures are used alongside them to describe what happened to sleep, activity and routine, which adds objective context that self report cannot supply.
What consent issues are specific to this domain?
Participants should be told explicitly what will be inferred from their data, not only what will be collected. Inference about mental state is more sensitive than the raw signals it comes from. Protocols should also state in advance what happens if data suggests someone is at risk, since discovering that mid study without a plan is both an ethical and a practical failure.
Run a study on these measures
WeGuide captures wearable data and patient reported outcomes in one platform, from screening through to analysis.
