Abnormal rest-activity patterns are well established during acute mood episodes and can precede clinical recurrence. Evidence for persistent disturbances during euthymic or inter-episode periods is more limited, derives mainly from small or short actigraphy studies, and has not previously compared major depressive disorder (MDD) and bipolar disorder (BD) phenotypes within the same large cohort. This study aimed to characterise rest-activity phenotypes in inter-episode mood disorder participants using consumer wearable data, focusing on both mean-level features and variability metrics.
The analysis used Fitbit-derived data from the All of Us Research Program (AoURP). Clinical groups were restricted to inter-episode windows. The final sample comprised 24,019 healthy controls (HC), 3,590 people with MDD, and 533 people with BD. Data were de-identified and accessed as existing research data via the All of Us Researcher Workbench; no additional institutional review board approval was required for this secondary analysis.
Four actigraphy-derived features were analysed: daily step count, wakefulness after sleep onset (WASO), total sleep time (TST), and sleep midpoint. For each feature the authors modelled three complementary dimensions: (i) average level and sensitivity to photoperiod, (ii) within-person fortnight-to-fortnight variability, and (iii) between-person heterogeneity in baseline level. This approach distinguishes mean-level differences from instability at the individual level and heterogeneity across individuals.
Daily step count was lower in both clinical groups compared with healthy controls. Reported effect sizes were d = -0.16 for MDD and d = -0.22 for BD. Both MDD and BD groups exhibited blunted photoperiod sensitivity, indicating a reduced change in activity across varying daylengths. Within-person variability in step counts was reduced in MDD and BD relative to controls, with within-person variability 4–7% lower in the clinical groups. Between-person heterogeneity in baseline step count was also reduced in MDD (14% lower), suggesting a more uniform, damped activity profile across individuals with MDD.
Mean-level differences in sleep features were relatively sparse. In contrast to the damped and rigid activity patterns, measures of sleep timing and duration showed increased instability. Both within-person and between-person variability for sleep features followed a graded pattern: HC < MDD < BD. Within-person variability increased by approximately 20–33% in BD relative to HC, and between-person heterogeneity rose up to 62% in BD relative to HC. The graded increase implies BD intensifies the variability pattern observed in MDD rather than representing a distinct categorical departure.
The results indicate a clear dissociation between physical activity and sleep across inter-episode mood disorders. Physical activity is characterised by lower average levels and greater rigidity — lower within- and between-individual variability — whereas sleep timing and duration are markedly more unstable, both within individuals over time and across individuals. Because these patterns were graded rather than strictly diagnosis-specific, the authors interpret MDD and BD as lying along a shared continuum of rest-activity disturbances in inter-episode periods.
The study highlights that mean-level measures alone can miss important residual disturbances present between episodes. Variability metrics derived from consumer wearables captured inter-episode abnormalities in both activity and sleep that are not apparent from averages. The authors suggest these wearable-derived variability measures warrant further evaluation as research phenotypes in prospective mood-state studies, where they may help track or predict changes in mood state over time.
The research used openly available de-identified data from the All of Us Research Program; all participants provided informed consent at enrollment in AoURP. The analysis was conducted as a secondary analysis by authorised researchers within the secure All of Us Researcher Workbench and did not require additional institutional review board approval. Data access protocols and eligibility criteria are available through the All of Us Research Program support pages. The authors declared no competing interests and indicated adherence to relevant reporting guidelines.