Using Smartphone-Tracked Behavioral Markers to Recognize Depression and Anxiety Symptoms: Cross-Sectional Digital...

Area of Science:

• Digital phenotyping
• Machine learning applications in mental health

Background:

• Depression and anxiety are common mental health disorders that are frequently missed or misdiagnosed.
• Delayed diagnosis and treatment of these conditions are linked to poorer patient outcomes.

Purpose of the Study:

• To investigate the utility of smartphone-derived behavioral markers for diagnosing depression and anxiety.
• To explore the potential of digital phenotyping in improving the recognition of these mental health conditions.

Main Methods:

• Utilized the Behapp platform for passive tracking of location and app usage in 217 individuals.
• Quantified 46 behavioral markers, including time spent at home and GPS trajectories.
• Applied machine learning to identify relevant markers and develop diagnostic prediction models.

Main Results:

• The total number of GPS-based trajectories emerged as a potential marker, with fewer trajectories associated with symptomatic individuals.
• Machine learning models incorporating GPS trajectories showed modest improvement over demographics-only models (AUC Mdn=0.60 vs 0.51).

Conclusions:

• Smartphone-tracked behavioral markers demonstrated limited discriminant ability in this study.
• These digital markers hold potential for future development in supporting depression and anxiety diagnostics.