Integrating human-smartphone interactions and actigraphy in digital phenotyping of obesity and depression: Social-physical coherence algorithm development and validation study.

Abstract

Circadian misalignment is associated with obesity and depression, yet few tools directly quantify the alignment between physical and social activity. We developed social-physical coherence, a digital-phenotyping measure defined as the magnitude-squared coherence between wrist-actigraphy physical activity and smartphone-derived social activity at the circadian (~24-hour) frequency, indexing how closely the two rhythms are aligned. In 135 adults (mean age 43.8 ± 12.3 years; 65.2% female; 42.2% with a history of major depression; 18.5% with obesity), physical and social activity were recorded for at least four weeks (3,973 person-days), and rest-activity rhythm indices - including interdaily stability (IS) - were computed from both data streams. Social-physical coherence correlated positively with app-derived IS (ISapp; r = 0.51-0.64 across 1-6 weeks). Higher coherence and ISapp were associated with lower body mass index during weeks 2-4 (β =  -0.22 to -0.25), remaining significant after Benjamini-Hochberg correction and after adjustment for sleep duration and chronotype; actigraphy-derived IS was not, and among all rhythm indices only ISapp paralleled this relationship. Coherence was uncorrelated with social jetlag, and no measure was associated with depressive symptoms. Social-physical coherence, derived from digital phenotyping of physical and social activity, is inversely associated with obesity but not depression.