Wearable-Derived Signals With the Strongest Evidence for Predicting Relapse in SMI

Some of the most useful wearable-technology-derived signals of relapse in serious mental illnesses (SMI), like depression, are not the sophisticated accelerometer or heart rate variability data a wearable generates, but something far simpler: the patient stops wearing or charging the device at all, according to Scott Fears, MD.^1,2 Fears, who is the chief medical officer at Amae Health and professor of psychiatry at UCLA, suggests that when this happens, clinicians can flag this update and check in with the patient directly, turning a data gap into an early outreach trigger.

That principle carries over into how Google built Sensor FM, the wearable-data foundation model Fears references. Rather than discarding gaps in heart rate or activity data as noise, the model's training treated missing data itself as informative, so the absence of a signal becomes part of what the model learns from rather than a limitation to work around.

Interpreting the more complex signals still requires a human layer, Fears shared with Psychiatric Times. Because Amae Health sees patients frequently in person, clinicians can use that face-to-face time to make sense of ambiguous patterns picked up through vocal and wearable data, cross-checking data streams against direct clinical observation to figure out what a signal is actually pointing to.

Fears framed this multimodal approach against psychiatry's long, largely unsuccessful search for a single defining biomarker comparable to fasting glucose in diabetes. That search has been frustrating for decades, he said, because mental health conditions do not organize themselves around one clean physiological marker. Instead, his team is pursuing biomarkers that cut across diagnostic categories, noting that patients with thought disorders such as schizophrenia and psychosis often share biomarker patterns with those who have depression or anxiety rather than presenting with a distinct signature.

The goal, Fears said, is to determine whether these cross-cutting biomarkers can predict which treatment is likely to work, indicate whether a patient is responding once treatment starts, and flag early evidence of relapse in someone who was previously stable.

"We've got some promising biomarkers that point to relapse before the individual can confidently say, 'Yeah, I'm depressed,' or 'I'm anxious,' or even the clinician can say, 'Yeah, I can say with confidence,'" Fears said, adding that these signals are emerging earlier than either subjective patient report or a clinician's own observations would typically catch.

He cautioned that pinning down exactly how early these biomarkers can predict relapse will require further research to refine, but said the field already has enough evidence to know the signal exists ahead of both patient and clinician awareness.

Dr Fears is the chief medical officer at Amae Health and a professor of psychiatry at UCLA.

References

1. Amae Health partners with Google Health Enterprise to bring wearable data integration to the treatment of complex mental health conditions. News release. August 4, 2026. Accessed August 4, 2026. https://www.globenewswire.com/news-release/2026/08/04/3338378/0/en/amae-health-partners-with-google-health-enterprise-to-bring-wearable-data-integration-to-the-treatment-of-complex-mental-health-conditions.html

2. Kuntz L. Bringing wearable data into psychiatric care: Amae Health partners with Google Health Enterprise to improve outcomes. Psychiatric Times. August 4, 2026. https://www.psychiatrictimes.com/view/bringing-wearable-data-into-psychiatric-care-amae-health-partners-with-google-health-enterprise-to-improve-outcomes