About the Article
Article Files
[This article belongs to Volume - 70, Issue - 12]
Published on : 2025-12-16 23:15:17
Article Code: AMJ-16-12-2025-12360
Title : Digital Phenotyping of Sleep Disorders in College Students Using Wearable Sensors and Ecological Momentary Assessment
Author(s) : Jennifer L. Walsh, MD, PhD, Kenji Nakamura, PhD, Ahmed R. Hassan, PhD, Maria C. Rodríguez, PsyD
Abstract :
Background: Sleep disorders affect up to 60% of college students, yet traditional assessment methods lack ecological
validity. Digital phenotyping through wearable sensors and ecological momentary assessment (EMA) offers
continuous, real-time monitoring but requires robust analytical frameworks.
Objectives: To develop a digital phenotype for insomnia in college students using multimodal data from wearables
and EMA, and to evaluate its predictive validity against gold-standard clinical measures.
Methods: We conducted a 90-day prospective cohort study of 782 college students (mean age 20.4±2.1 years) across
four universities. Participants wore Oura Rings for continuous sleep monitoring and completed 6 daily EMA surveys
via smartphone. Daily features (sleep efficiency, heart rate variability, circadian alignment) were integrated with
mood and fatigue ratings. Insomnia severity was assessed monthly using the Insomnia Severity Index (ISI). Mixed
effects models and gradient boosting identified predictive features.
Results: Digital phenotype achieved 84.2% accuracy (AUC 0.89) in discriminating moderate-severe insomnia (ISI≥15)
from no-mild insomnia. Top predictors were sleep efficiency variability (OR=2.34, p<0.001), heart rate variability
during sleep (OR=0.71, p<0.001), and circadian misalignment (multi-model inference R²=0.42). EMA compliance was
78.3%. Students with insomnia showed 2.3-fold higher night-to-night sleep variability (p<0.001).
Conclusions: Digital phenotyping reliably identifies insomnia in college students, capturing dynamic patterns invisible
to cross-sectional assessment. This approach enables early detection and personalized intervention delivery.