Identifying opioid relapse during COVID-19 using natural language processing of nationwide Veterans Health Administration electronic medical record data.

• DOI: 10.1037/abn0000984
• Publication Date: 2025-03-20T13:19:15Z
• Authors: Nicholas A. Livin...; Amar D. Mandavia; Anne N. Banducci; Rebecca Sistad Hall; Lauren B. Loeffel; Michael Davenport; Brittany Mathes-W...; Maria Ting; Clara E. Roth; Alexis Sarpong; Noam Newberger; Zig Hinds; Jennifer R. Fonda; Daniel Chen; Frank Meng
• Keywords: Health records; Electronic health record; 2019-20 coronavirus outbreak; Medical record; Veterans Affairs

Novel and automated means of opioid use relapse risk detection are needed. Unstructured electronic medical record data, including written progress notes, can be mined for clinically relevant information, the presence substance relapse-critical markers recovery from disorder (OUD). In this study, we used natural language processing (NLP) to automate extraction relapses, timing these occurrences, veteran patients' record. We then demonstrated utility our NLP tool via analysis pre-/post-COVID-19 trends among veterans with OUD. For demonstration, analyzed data 107,606 OUD enrolled in Veterans Health Administration, comparing a pandemic-exposed cohort (n = 53,803; January 2019-March 2021) matched prepandemic October 2017-December 2019). The recall was 75% precision 94%, demonstrating moderate sensitivity excellent specificity. Using tool, found that odds postpandemic onset were proportionally higher compared trends, despite patients having fewer mental health encounters which derive instances onset. research application as hypothesized, elevated postpandemic. methods identify monitor holds promise future surveillance, prevention, clinical outcome research. (PsycInfo Database Record (c) 2025 APA, all rights reserved).