Predicting Mood Disorders Through Wearable and Smartphone-Based Circadian Rhythm Analysis: Innovations in Digital Psychiatry

초록

Circadian rhythm disruptions are well-recognized as key features of major depressive disorder (MDD), bipolar disorder (BD), and related mood episodes. Recent advancements in wearable devices and smartphones have provided unprecedented opportunities for continuous, non-invasive monitoring of sleep-wake patterns and circadian rhythms. This lecture presents findings from our studies exploring the predictive potential of digital phenotyping in mood disorders. The first study developed a predictive model for mood episode recurrence using over 140 parameters derived from real-time data collected through wearable devices. This model demonstrated high accuracy (AUC: 0.937–0.963) in predicting depressive, manic, and hypomanic episodes within three days of onset. Circadian rhythm metrics, particularly phase delays and activity deviations, significantly contributed to the predictions, underscoring the critical role of circadian rhythm alignment in clinical management. The second study focused on predicting the recurrence of mood episodes using sleep-wake patterns and mathematically derived dim-light melatonin onset data collected through wearable devices and smartphones. Employing a machine learning approach, this study achieved robust accuracy (AUC: 0.80–0.98) in predicting depressive, manic, and hypomanic episodes the following day. Circadian rhythm phase shifts emerged as the most critical predictor. These study findings highlight the transformative potential of wearable and smartphone-based circadian rhythm monitoring in psychiatry. By integrating circadian rhythm metrics with digital phenotyping, these approaches enable precise and personalized prediction and management of mood episodes. This talk will discuss the clinical implications of these findings, focusing on enhanced prevention strategies and the future role of digital technologies in advancing mental health care.