Using wearable sleep and daily rhythm data to predict mood episodes in mood disorder patients

168 patients were monitored, yielding accurate predictions for mood episodes based on sleep-wake data.

• Models developed predict future mood episodes using only sleep-wake data collected through smartphones and wearables.
• An average of 587 days of clinical follow-up and 267 days of wearable data were analyzed.
• The derived 36 sleep and circadian rhythm features enabled next-day predictions for depressive, manic, and hypomanic episodes with AUCs of 0.80, 0.98, and 0.95, respectively.
• Daily circadian phase shifts were identified as significant predictors: delays in sleep onset were associated with depressive episodes, while advances were linked to manic episodes.
• This prospective observational cohort study indicates the potential of sleep-wake data and prior mood episode history for mood episode prediction.

Simplified