A prognostic enrichment strategy with latent class analysis (LCA) has not been examined in hospitalized patients with opioid misuse. In a cohort of 6,224 patient encounters, LCA identified four subtypes with differing patient characteristics in demographics, substance use patterns, and hospital utilization. Face validity was provided with natural language processing and topic modeling from a data corpus of 422,147 clinical notes. Identifying subtypes with latent class analysis and examination with topic modeling may inform health systems on targeted therapeutic approaches.

Learning Objective: 1. Define the success of latent class analysis to subtype opioid misuse in hospitalized patients.
2, Examine the role of topic modeling with latent dirichlet allocation to provide supportive evidence from the clinical notes for the subtypes
3. Perform prognostic enrichments strategies for targeted interventions


Brihat Sharma (Presenter)
Loyola University Chicago

Majid Afshar, Loyola University Chicago
Dmitriy Dligach, Loyola University Chicago
Robert Kanie, Loyola University Chicago
Elizabeth Salisbury-Afshar, Rush University
Niranjan Karnik, Rush University
Cara Joyce, Loyola University Chicago

Presentation Materials: