Full-day sleep pattern analysis in common mental disorders: Leveraging highly discrepant recordings from two consumer tracking devices.

Sleep and circadian rhythm disruptions are increasingly studied through consumer sleep-tracking devices, both in research and in real-world contexts. These devices offer a unique perspective on mental health, given the strong connection between sleep disturbances and Common Mental Disorders (CMD). In this study, we sought to identify and characterize abnormal sleep behaviors by examining discrepancies between two complementary sleep-tracking devices. Rather than treating inter-device disagreement as measurement noise, we interpreted it as a potential behavioral signal. This approach uncovered six statistically robust outlier patterns in sleep health that were interpretable and clinically relevant. These patterns span a full 24-hour window-including nocturnal, diurnal, and peri-sleep activities-thus providing a holistic view of sleep-related behavior. We analyzed data from 149 patients (72% woman), ranging from 18 to 71 years old, and diagnosed with non-severe CMD over a period of three months. At the end, 4,824 days of sleep recordings were collected from two devices: a less accurate wristband tracker (W) and a more precise sleep-tracking mat (M). Using k-means clustering on high-discrepancy recordings (>5 hours), we identified six robust patterns of full-day sleep behavior that exhibited consistency at the individual user level, suggesting an origin in the patient's behavior rather than random noise. To further validate these clusters, we integrated additional behavioral metrics in the analysis such as daily step distribution or smartphone usage as indicators of physical or social activity. By leveraging device discrepancies, we revealed several sleep patterns of potential clinical relevance-indicative of oversleeping, unintended sleep onset outside the bed, or atypical sleep-wake cycles. These findings highlight the potential of passive sleep monitoring to support early detection of pathological changes (e.g., depressive episodes) and to inform clinical decisions by identifying behavioral side effects of treatment.