Integrating active and passive digital phenotyping to study the temporal dynamics between physical activity and mood in bipolar disorders

Abstract Bipolar disorders (BDs) are characterized by pronounced mood fluctuations that can impair daily function and treatment. While long-term benefits of physical activity for mood are well-established, less is known about how day-to-day changes in activity impact mood in individuals with BDs. This study leverages 12 months of digital phenotyping and ecological momentary assessment data from the University of Michigan’s PROviding Mental Health Precision Treatment study to examine day-level associations between mood, physical activity (steps, very active minutes), and resting heart rate (RHR) in 252 adults with BDs. Participants completed daily mood ratings via smartphone (prompted at 6:00PM) and wore Fitbits that passively recorded steps, very active minutes, and RHR (available on days with ≥30 min of still periods). Using Dynamic Structural Equation Modeling, we observed a small but credible bidirectional relationship between daily steps and mood: individuals who reported better mood than their own average took more steps the following day, and vice versa. Very active minutes predicted next-day mood but not the reverse, which—while possibly due to their low frequency in the sample—may suggest that higher-intensity activity does not produce a feedback loop similar to total steps. No credible cross-lag associations emerged between daily RHR and mood. Conditional models found that sex and age influenced several activity dynamics, and baseline anxiety was positively associated with daily RHR variability. These findings highlight that increases in physical activity, especially in daily steps, may support mood regulation in BDs on the day level.