Dynamic bidirectional associations between GPS mobility and ecological momentary assessment of mood symptoms in mood disorders (Preprint)

Abstract

BACKGROUND Extensive research has been conducted on the digital phenotype in relation to mood disorders, but the impact of the daily time-lagging and bidirectional association on mood and global positioning system (GPS) mobility remains relatively unexplored. By leveraging the widespread use of smartphones, we examined the potential correlations between mood and behavioral changes, with implications for future scalability. OBJECTIVE This study aimed to investigate the bidirectional time-lag relationships between passive GPS data and active ecological momentary assessment (EMA) data using smartphone app technology. METHODS Between March 2020 and May 2022, we recruited 45 subjects (mean age=42.3, SD=12.1 years) followed for 6 months, including 35 individuals diagnosed with mood disorders referred by psychiatrists and 10 healthy control subjects. This resulted in a total of 5248 person-days of data. Over six months, we collected two types of smartphone data: passive data on movement patterns with nearly 100,000 GPS data points per individual and active data through EMA capturing daily mood levels, including fatigue, irritability, depression, and mania. Our study was limited to Android users due to operating system constraints. RESULTS Our findings revealed a significant negative correlation between normalized entropy (r = −0.353, P = 0.035) and weekly depression scores, as well as between location variance (r = −0.364, P = 0.029) and depression. In participants with mood disorders, we observed bidirectional time-lagged associations. Specifically, changes in homestay were positively correlated with fatigue (β = 0.047, P = 0.016), depression (β = 0.235, P = 0.013), and irritability (β = 0.149, P = 0.030), and negatively correlated with location variance (β = −0.869, P < 0.001) and depression (β = −0.015, P = 0.009). These findings suggest dynamic, bidirectional relationships between mobility patterns and mood variations in individuals with mood disorders. CONCLUSIONS This study demonstrates the potential of utilizing active EMA data to assess mood levels and passive GPS data to analyze mobility behaviors, with implications for managing disease progression in patients. Monitoring location variance and homestay can provide valuable insights into this process. The daily use of smartphones has proven to be a convenient method for monitoring patients’ conditions. Interventions should prioritize promoting physical movement while discouraging prolonged periods of staying at home.