Multimodal Digital Indicators of PTSD and Symptom Cluster Variation

Posttraumatic stress disorder (PTSD) is characterized into four clusters within the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition: intrusion, avoidance, negative alterations in cognition and mood, and alterations in arousal and reactivity. Traditional assessment relies on self-report measures and clinician-administered interviews, which are subject to recall bias, limited accessibility, and infrequent administration. Digital biomarkers, defined as objective behavioral and physiological signals from audio and video data, offer an alternative for PTSD assessment. This study investigated whether multimodal digital biomarkers could predict PTSD symptom severity within individual symptom clusters. Participants (N = 52) meeting the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision (DSM-5-TR) Criterion A completed self-report PTSD assessments (PTSD Checklist for DSM-5 [PCL-5]) and provided unstructured video responses to the prompt "tell me about your day." A computational framework extracted over 200 digital biomarkers across visual, acoustic, physiological, and linguistic domains. Feature selection was conducted using the Boruta algorithm, and confirmatory analyses employed Bayesian multivariate multiple regression modeling the four PCL-5 subscales jointly. Multimodal digital biomarkers explained meaningful variance (arousal: 30.4%, intrusion: 28.7%, negative mood: 23.8%, and avoidance: 22.3%); similarly, prediction of total PCL-5 scores explained substantive variance (R² = .30). Variability in chin-raising muscle activity (AU17) emerged as a cross-cluster predictor in three subscales. These findings suggest that digital biomarkers may be informative at both the symptom cluster level and the level of global PTSD severity. Findings are preliminary given the small sample, cross-sectional design, and lack of external validation. Future research should replicate the findings in larger, more diverse samples and explore digital biomarkers on longitudinal symptom monitoring and treatment response tracking.