Digital Phenotyping in Health Using Machine Learning Approaches: Scoping Review

Area of Science:

• Digital Health
• Computational Psychiatry
• Behavioral Informatics

Background:

• Digital phenotyping involves real-time data collection from personal devices in naturalistic settings.
• Existing research shows heterogeneity in clinical applications, data types, collection methods, analysis, and outcomes.

Purpose of the Study:

• To conduct a scoping review mapping published research on digital phenotyping.
• To outline study characteristics, data collection and analysis methods, machine learning approaches, and future implications.

Main Methods:

• A systematic literature search was performed for studies published in 2020-2022 on PubMed and Google Scholar.
• PRISMA-ScR guidelines were followed for literature screening, data extraction, and charting.
• Included studies were analyzed for origin, design, clinical focus, data types, collection modes, analysis, and limitations.

Main Results:

• 46 studies were included, with most research originating from North America and focusing on wearable data.
• Observational studies were most common, primarily investigating psychiatric, mental health, and neurological disorders.
• Machine learning was used in 7 studies, with random forest, logistic regression, and support vector machines being prevalent.

Conclusions:

• The review offers foundational and application-oriented insights into digital phenotyping in healthcare.
• Future research should prioritize prospective, longitudinal studies with diverse populations and larger datasets.
• Addressing privacy, ethical concerns, and developing personalized digital health interventions are crucial next steps.