Digital Biomarkers from “video Selfies” Improve Prediction of Mental Health: Results from a Multimodal Ecological Momentary Assessment Study ​

Personal digital devices are transforming mental health assessment. Leveraging the ubiquity of smartphones and public familiarity with smartphone-recorded “selfies,” we developed a novel “video selfie” ecological momentary assessment, and tested the utility of digital biomarkers to improve prediction of mental health and well-being in a sample of non-clinical participants. Our video selfie paradigm yielding 71,902 videos (N = 1,634 participants) demonstrated high feasibility, indicated by high adherence (77% mean response rate) and high video quality with 92% of recordings meeting quality criteria. We show that video selfies demonstrate incremental validity beyond conventional self-report ecological momentary assessment, with video-based multimodal features collectively improving model performance for predicting concurrent NIH PROMIS scores for anxiety and depression and flourishing (Flourishing Index) (5.2%-7.0%). This research lays the groundwork for methodological innovation that advances mental health and well-being “in-the-wild” and at scale. Future applications include the integration of video-based data streams to improve time-sensitive delivery of digitally-mediated mental health interventions. ​