31.4 the Feasibility and Clinical Utility of Digital Phenotyping to Assess Sleep Duration and Quality in Youth

People with bipolar disorder (BD) often experience significant sleep dysregulation, the nature of which can vary widely depending on their current mood state. Changes in sleep can predate significant mood changes, offering an “early alert” of an impending episode. Related, improving sleep stability can indicate a positive treatment response and possible remission of the mood episode. Digital phenotyping, which measures behavior and mental status using data collected from smartphones, may be a practical way by which to conduct long-term sleep monitoring in youth. We evaluated the feasibility and clinical utility of this approach in adolescents with BD and typically developing (TD) peers. Participants (aged 14-19 years) and their caregiver were interviewed monthly about the adolescent’s mood and behavior for 18 months. During this time, the adolescent used the Beiwe app, which collects surveys, location/mobility (global positioning system [GPS]), and phone locked/unlocked status, among other data. We assessed associations between phone sensor–derived sleep duration/quality and both participant self-report about sleep and clinician-rated changes in mood. Passive and survey data were obtained with minimal missingness among BD (n = 28) and TD (n = 23) participants. Phone sensor–derived sleep duration data correspond with participant self-reported sleep and wake times. Inspection of behavioral features and screen lock/unlock data indicate unique sleep and circadian patterns between BD and TD participants. Related, changes in objectively assessed sleep are observed prior to clinician-rated mood changes. Digital phenotyping is a feasible approach to collect objective sleep data from adolescents over a long period of time. This population, which is often reluctant to engage in mental health services and among whom smartphone use is ubiquitous, may be an ideal population for this low-burden approach. Data patterns differ across BD and TD adolescents and across mood state, suggesting clinical utility for diagnosis and relapse prevention.