Abstract TU264: Smartphone-based Digital Phenotyping as a Scalable Approach to Measure Sleep Health in Emerging Adults

Background. Sleep is a key lifestyle factor for cardiometabolic health, but most epidemiologic studies rely on self-reported measures of sleep that are limited by recall bias and low temporal resolution. Smartphone-based digital phenotyping—using passive sensor data to capture real-world behavior—offers a scalable, low-burden alternative for collecting objective, high-frequency sleep data. However, its feasibility, acceptability, and concordance with self-reported sleep data in emerging adults remain underexplored. Methods: We used a customized smartphone app to collect passive sensor data (GPS, accelerometer, screen state) and nightly self-reported sleep data over eight days from 548 participants enrolled in the smartphone substudy of the Economic and Educational Contributors to Emerging Adults’ Oral and Cardiometabolic Health Study (The 3E Study). Sleep measures, including bedtime and wake time, were assessed using both smartphone passive sensor data and self-reported survey data. Data completeness was defined as the proportion of participants with valid sleep estimates over the study period, and concordance was assessed using Spearman rank correlation. On day nine, participants completed a survey assessing comfort, behavioral changes, and perceived burden of smartphone data collection. Results: Across eight days, we collected over 2.7 billion raw sensor data points (~5 million observations per participant), including 22.1 million GPS, 1.4 billion accelerometer, and 2.1 million screen state observations. Preliminary smartphone-derived sleep data were 67% more complete than self-reported data and showed moderate concordance with self-reported sleep (ρ=0.4, p<0.001). The majority of participants identified as Latine (39%) or Asian (35%), and 58% as female. Over 78% of participants reported being comfortable with data collection and found it non-intrusive; 72% reported little to no changes in daily behavior. Conclusions: Smartphone-based digital phenotyping appears to be a feasible and acceptable approach for capturing objective sleep data in emerging adults, with moderate concordance to self-reported measures. By enabling high-resolution, continuous assessment of real-world behaviors, this approach may advance understanding of how sleep and other lifestyle factors contribute to cardiometabolic health.