Digital Phenotyping

Relapse in depression, bipolar disorder, and schizophrenia contributes substantially to disability, hospitalization, and health care costs. Early detection is therefore critical for enabling timely intervention. Digital phenotyping—the continuous passive and active collection of behavioral and physiological data through smartphones, wearables, and other personal devices—offers a promising strategy for identifying early relapse signals. Yet, current evidence remains fragmented, with studies varying in design, populations, data modalities, and analytic approaches.

To address this gap, we will conduct a systematic review and meta-analysis to quantify the predictive accuracy of digital phenotyping tools (e.g., sensitivity, specificity, area under the curve), assess their clinical utility and feasibility in real-world care, and identify methodological and reporting gaps. Following PRISMA 2020 guidelines, we will search major databases for studies of adults (≥18 years) with these disorders, where relapse is defined by clinician diagnosis, hospitalization, or symptom thresholds. Data will be synthesized using random-effects meta-analysis, with subgroup analyses by diagnosis, data modality, and model type.

Findings will provide an evidence base to guide clinical adoption and future research. They are particularly relevant in Saudi Arabia, where integration of digital health solutions is a national priority. This work directly supports Saudi Vision 2030 by advancing the use of AI and digital health to improve early intervention and reduce the burden of mental illness.