DeMMO Models Digital Mobility Outcomes Across Diseases for Progression Monitoring.

Digital mobility outcomes (DMOs), derived from wearable sensors, offer a promising way to track disease progression in daily life. However, most studies typically focus on a single disease at a single point in time, failing to capture how multivariate DMO relationships evolve across multiple diseases and clinical outcomes. Existing temporal multi-task frameworks can model progression within one disease but cannot jointly model multiple prediction outcomes across diseases, especially when patient cohorts do not overlap. To address these gaps, researchers propose DeMMO, an interpretable framework for longitudinal, multi-disease, and multi-outcome learning. DeMMO represents each disease-outcome objective using a longitudinal DMO coefficient matrix and combines temporal regularization with stable and visit-specific feature selection. Its core innovation is an automatic mechanism that learns signed relationships directly from these longitudinal mappings, allowing for selective information sharing between diseases and outcomes even without shared participants. Evaluated on the large-scale Mobilise-D dataset, which includes 24 harmonized DMOs over five visits across multiple mobility-limiting conditions, DeMMO achieved superior overall and outcome-specific prediction performance compared to nine strong baselines. The framework also identifies reliable longitudinal DMO patterns, which are valuable for clinical validation and continuous disease monitoring.