Digital Phenotyping in Pediatric Mental Health: From Passive Sensing to Precision Psychiatry—A Narrative Review

The escalating crisis in pediatric mental health has rendered traditional, episodic care models insufficient. Reliance on infrequent clinical interviews often fails to capture the rapid, non-linear fluctuations of adolescent psychiatric disorders, leaving clinicians blind to prodromal signs of relapse. This narrative review synthesizes emerging evidence from 2020 to 2026 regarding digital phenotyping—the continuous, passive collection of behavioral data via smartphone sensors—as a paradigm shift toward precision psychiatry. We examine how raw sensor inputs, including accelerometry, GPS mobility patterns, and human-computer interaction (HCI) logs, are transformed by machine learning algorithms into objective ”digital biomarkers.” The reviewed literature demonstrates that these biomarkers—specifically sleep-wake cycle fragmentation, life-space contraction, and semantic shifts in typing—can successfully predict depressive relapse, differentiate bipolar mania from ADHD, and quantify suicide risk with high sensitivity. Crucially, these digital signals often precede subjective symptom reporting by 7–10 days, offering a vital window for Just-in-Time Adaptive Interventions (JITAIs). However, the translation of these computational insights into routine clinical practice faces significant ”last mile” challenges, including the integration of high-velocity data into Electronic Health Records (EHRs), the risk of clinician alarm fatigue, and ethical concerns regarding patient privacy and the socioeconomic ”digital divide.” We conclude that while digital phenotyping offers a revolutionary stethoscope for the mind, its successful adoption requires a hybrid framework where Artificial Intelligence augments, rather than replaces, the therapeutic alliance. This review provides a strategic roadmap for moving the field from passive sensing to proactive, preventative clinical care.