Integrating Active and Passive Digital Phenotyping in Bipolar Disorders

Abstract

Bipolar disorders (BDs) are complex psychiatric conditions marked by mood shifts, resulting in substantial individual and societal burdens. Despite their impact, therapeutic advances are limited, largely due to the heterogeneity of symptoms. Recent research emphasizes the potential of ambulatory assessment data, leveraging technology like smartphones and wearables, to capture the nuanced interplay between mood, physical activity, and sleep in BDs. Research and clinical practice would benefit from a better understanding of behavioral phenotypes and how they relate to everyday symptoms in individuals with BDs. Examine relationships between behavioral phenotypes and symptoms using active (EMA of daily mood) and passive (Fitbit sleep and steps) ambulatory assessment data. Hypotheses: 1) Within-person increases in total sleep time and steps taken per day will both be associated with an improvement in mood. 2) Steps per day and mood, as well as total sleep time and mood, will be concurrently related at the between-person level. 3) Variability in total sleep time will predict mood outcomes. Participants with BDs (n = 177) were selected from the Providing Mental Health Precision Treatment study (follow-up of 12 months; total N = 1,919). Participants completed daily mental health symptom assessments via phone app and wore a wearable (i.e., Fitbit). regularly for passive data collection on sleep and steps. Mixed effects models were used to analyze within- and between-person variation. Dynamic structural equation modeling was used to test hypothesized relationships across variables over time. Mood, steps per day, and total sleep time all vary significantly over time. No between-person associations were found between steps per day and mood on the same day. Total sleep time was associated with better mood the next day (a_ = .33, 95% CI [.08, .53]). More variability in total sleep time is associated with worse mood (a_ = -.26, 95% CI [-.49, -.00]). How might these relationships be affected by potential moderating variables, such as age, medication, therapy, or specific BD type? Could consistency of sleep be more important than quantity? Does better mood lead to improved engagement in activities that are not captured by step count, i.e., more sedentary or relaxing activities?