Wearable-based phenotyping of activity patterns during pregnancy

Area of Science:

• Obstetrics and Gynecology
• Digital Health
• Biomedical Engineering

Background:

• Wearable activity trackers offer objective behavioral data but their utility in detecting high-risk pregnancies needs assessment.
• Pregnancy complications like hypertensive disorders, fetal growth restriction, and preterm birth significantly impact maternal and infant outcomes.

Purpose of the Study:

• To evaluate conventional and novel digital biomarkers from wearable activity trackers for early detection of pregnancy complications.
• To assess the association between activity patterns and adverse pregnancy outcomes using longitudinal data.

Main Methods:

• Longitudinal Fitbit data from 336 participants in the Deep Phenotyping of Pregnancy Project (DP3) were analyzed.
• Novel metrics, cyclicality and time-to-drop, were derived alongside average weekly step counts.
• Activity patterns were compared between participants with and without common pregnancy complications.

Main Results:

• While average step counts did not differ, participants with complications showed significantly less cyclical activity (77.6% vs. 87.2%, p < 0.05).
• A shorter time-to-drop was observed in complicated pregnancies (28.4 weeks vs. 34.3 weeks, p < 0.05).
• Sleep duration and active minutes were not associated with pregnancy outcomes.

Conclusions:

• Pattern-based digital phenotypes, specifically activity cyclicality and time-to-drop, are more indicative of adverse pregnancy outcomes than aggregate step counts.
• These findings support the use of passive, scalable risk stratification for precision prenatal care.
• Digital biomarkers derived from wearable technology can enhance early detection and management of pregnancy complications.