A Joint Survival Modeling and Therapy Knowledge Graph Framework to Characterize Opioid Use Disorder Trajectories.

We developed a multi-stage computational framework to model transitions across OUD onset, treatment-related remission, and relapse, and further link risk factors to treatment and intervention. Using linked EHR and survey data from All of Us, we constructed three time-to-event phenotypes: (i) OUD onset, defined from opioid prescriptions and diagnostic codes; (ii) transition to remission; and (iii) transition to relapse following remission. For each participant, we derived longitudinal predictors from clinical conditions and survey concepts, including recent (1-, 3-, 12-month) event counts, cumulative exposures, and time since last event occurrence. We fit regularized survival models for each transition and aggregated feature selection frequencies and hazard ratios to identify a compact set of high-confidence predictors. Across all stages, we observed consistent contributions from pain, mental health, and polysubstance use. Chronic pain syndromes, tobacco and nicotine dependence, anxiety and depressive disorders, and cannabis dependence were frequent and prominent predictors of OUD onset and relapse, whereas tobacco dependence during remission and other remission-coded conditions were strongly associated with the transition to remission. To support computational prioritization of therapeutic strategies, we constructed a therapy knowledge graph integrating genetic targets, biological pathways, and published evidence to link identified risk factors with candidate treatments reported in recent OUD studies and clinical guidelines.