Wearable-based phenotyping of activity patterns during pregnancy.

Wearable activity trackers provide scalable, objective measures of behavior, but their potential use for early detection of high-risk conditions during pregnancy remains to be evaluated. Using longitudinal Fitbit data from 336 participants in the Deep Phenotyping of Pregnancy Project (DP3), we evaluated both conventional and novel, pattern-level, digital biomarkers of activity in relation to common pregnancy complications, including hypertensive disorders of pregnancy (HDP), fetal growth restriction (FGR), and spontaneous preterm birth (sPTB). Beyond average weekly step counts, we derived two novel metrics: cyclicality, reflecting the persistence of weekly rhythmic activity patterns, and time-to-drop, the gestational length interval until activity levels declined below group norms. While mean step counts did not differ among the groups, the groups of participants with complications were significantly less likely to exhibit cyclical activity patterns (87.2% vs. 77.6%, p < 0.05) and showed shorter time-to-drop compared to controls (34.3 vs. 28.4 weeks, p < 0.05). Neither sleep duration nor active minutes were associated with the outcomes. These findings suggest that pattern-based digital phenotypes, rather than aggregate step counts, capture activity deviations associated with adverse pregnancy outcomes. Our data support scalable, passive risk stratification and inform precision prenatal care.