Personalized Depression Forecasting🔗

Article 5 is entitled “Personalised Depression Forecasting Using Mobile Sensor Data and Ecological Momentary Assessment”. It presents results of the MAIKI trial, in which patients with elevated depressive symptoms underwent treatment using a digital depression intervention. During the treatment phase, passive sensor features and EMA data were recorded with the newly developed “MAIKI” application. This information was harnessed within deep learning architectures that predict and forecast changes in depressive symptom severity during treatment. We found that depression forecasts were possible using this approach, and that performances improved when models were fine-tuned to each individual, leaving a common backbone layer across patients.

Article 5 has been published in “Frontiers in Digital Health” (Kathan, Harrer et al., 2022).