Methods Protocol: Identifying Objective Mobile Sensing Markers of Depressive Symptoms Using Individualized and Federated Learning

This proof‑of‑concept study aims to examine how mobile sensing (MS)–derived features relate to self-reported depressive symptom severity. The protocol primarily focuses on estimating within-person and partially pooled associations between MS features and symptoms, thereby contributing directly to the MONDY study’s [1] Research Question 1 (RQ1): whether MS data can function as objective markers of daily depressive symptom severity. While predictive modeling is applied as an operational tool within this framework, its function is secondary—supporting the evaluation of association strength, model-based symptom inference, and the stability of estimated relationships across individuals. Prediction performance is therefore interpreted as evidence of how well MS-based models capture the underlying associations of interest. The data analyzed originates from the exploratory MONDY observational trial described by Reich et al. [1], based on the clinical trial registered under DRKS00032618. Reich, H., Schreynemackers, S., Amin, R., Ludwig, S., Zippelius, J., Leimhofer, J., Dunker, T., Schriewer, E., Carell, A., Weber, Y., & Hegerl, U. (2025). Links between self-monitoring data collected through smartphones and smartwatches and the individual disease trajectories of adult patients with depressive disorders: Study protocol of a one-year observational trial. Contemporary Clinical Trials Communications, 45, 101492. https://doi.org/10.1016/j.conctc.2025.101492