Published: 07 August 2025
Authors: Johannes Leimhofer, Milica Petrovic, Andreas Dominik, Dominik Heider, Ulrich Hegerl.
Abstract:
A popular trend in depression forecasting research is the development of machine learning models trained with smartphone sensor data and periodic self-ratings to detect early indications of changes in depression severity. This mixed-methods umbrella review identifies smartphone hardware and software sensor data streams used in mental health monitoring and assesses their cross-platform availability across Android and iOS devices. The study found that commonly used data streams include accelerometers, GPS, cameras, Bluetooth, Wi-Fi, step counters, battery levels, app usage, call and message logs, and screen status. The authors conclude that several sensor streams are increasingly available across platforms, supporting their potential use in depression forecasting systems, but further research is needed to ensure reliable outputs across different mobile hardware and software configurations.