[Paper Review] A Joint Survival Modeling and Therapy Knowledge Graph Framework to Characterize Opioid Use Disorder Trajectories

Mengman Wei, Stanislav Listopad+1 more|arXiv (Cornell University)|Jan 19, 2026

Opioid Use Disorder Treatment0 citations

TL;DR

The paper presents a multi-stage survival modeling framework for OUD trajectories (onset, remission, relapse) using All of Us data, and links risk factors to a therapy knowledge graph to support treatment prioritization.

ABSTRACT

Motivation: Opioid use disorder (OUD) often arises after prescription opioid exposure and follows transitions among onset, remission, and relapse. Linked EHR-survey resources such as the All of Us Research Program enable stage-specific risk modeling and connection to intervention options. Results: We built a multi-stage framework to model time-to-onset, time-to-remission, and time-to-relapse after remission using All of Us EHR and survey data. 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. We fit regularized Cox models for each transition and aggregated selection frequencies and hazard ratios to identify a compact set of high-confidence predictors. Pain, mental health, and polysubstance use contributed across stages: chronic pain syndromes, tobacco/nicotine dependence, anxiety and depressive disorders, and cannabis dependence prominently predicted onset and relapse, whereas tobacco dependence during remission and other remission-coded conditions were strongly associated with transition to remission. To support therapeutic prioritization, we constructed a therapy knowledge graph integrating genetic targets, biological pathways, and published evidence to map identified risk factors to candidate treatments in recent OUD studies and clinical guidelines.

Motivation & Objective

• Model three clinically meaningful OUD transitions: onset, remission, and relapse.
• Identify compact, high-confidence predictors from high-dimensional EHR and survey data.
• Link identified risk factors to evidence-based treatments via a knowledge graph for decision support.

Proposed method

• Construct three time-to-event outcomes for OUD onset, remission, and relapse from linked EHR and survey data.
• Engineer time-indexed features (recent counts, cumulative counts, days since last) across pain, mental health, and polysubstance use concepts.
• Fit L1-penalized Cox models (lasso) for each transition with start–stop formulation and cross-validated tuning.
• Aggregate feature selection frequencies and hazard ratios across data partitions to identify robust predictors.
• Build a therapy knowledge graph integrating genes, pathways, and drugs to connect risk factors with MOUD and other treatments.
• Use a Personalized PageRank approach on the graph to rank candidate drugs and provide explainable outputs.

Figure 1: Predictors of OUD onset. Hazard ratios (HRs) from time-to-event models for OUD onset. HR $>1$ indicates increased hazard (earlier onset) and HR $<1$ indicates decreased hazard. Predictors shown are the 24 prioritized features selected by the modeling/feature-selection pipeline; Selected in

Experimental results

Research questions

• RQ1 What are the stage-specific risk factors for OUD onset, remission, and relapse in a large, diverse cohort?
• RQ2 Can high-dimensional, time-varying predictors from EHRs and surveys be distilled into a compact, interpretable set of risk factors for each OUD transition?
• RQ3 How can a knowledge graph be used to connect identified risk factors to evidence-based treatments and prioritize therapeutic options?

Key findings

• Pain-related conditions, tobacco/nicotine use, and mental health disorders are prominent predictors of OUD onset.
• Remission is strongly associated with MAT/MOUD exposure and recent clinical engagement, with other substance-use and anxiety-related predictors also important.
• Relapse risk is linked to recent clinical activity and polysubstance indicators, with some extreme hazard ratios likely reflecting sparse data or complexity.
• The knowledge graph prioritizes established MOUD drugs and identifies mechanistic clusters for non-MAT candidates, offering hypothesis-generating treatment options.

Figure 2: Top predictors of OUD remission. Hazard ratios (HRs) summarize associations with time to OUD remission; HR $>1$ indicates a higher hazard of remission (i.e., faster remission), and HR $<1$ indicates a lower hazard. To improve readability, the x-axis is shown on a log scale; values above 10