Presentation (CTI)

OBJECTIVES: To characterize clinical identification and management patterns among patients with confirmed, likely, and potential AUD using linked structured claims, EHR, and human-in-the-loop large language model (LLM) extracted clinical notes.
METHODS: A retrospective study was conducted using NorstellaLinQ’s US real-world linked open claims, structured EHR, and clinical notes (June 2020-present). From 10,711,582 patients with AUD indicators in claims and EHR, 1,376,492 had at least one clinical note and 403,771 had meaningful AUD note signals, stratified into confirmed AUD (N=318,270; F10.xx codes or AUD-specific pharmacotherapy), likely AUD (N=12,724; structured clinical indicators without formal diagnosis), and potential AUD (N=72,777; LLM-extracted clinical note signals only). Care setting, specialty, severity, and note volumes were assessed. Treatment initiation was not assessed given pharmacotherapy was part of the confirmed AUD definition.
RESULTS: Care pathways diverged substantially by cohort: confirmed AUD patients (N=318,270) were concentrated in behavioral health and primary care, while likely AUD (N=12,724) and potential AUD (N=72,777) patients more frequently surfaced in specialty and acute settings such as gastroenterology and hepatology, consistent with incidental detection or later-stage complications. AUD was documented in clinical notes for 86% of confirmed, 82% of likely, and 38% of potential AUD patients. Among likely AUD patients, 82% had AUD discussed in clinical notes prior to formal diagnosis.
CONCLUSIONS: AUD care is fragmented across specialties with severity driving routing to acute settings. Physician notes document AUD-related activity in more patients than carry a formal diagnosis and capture it upstream of diagnosis in the likely AUD population. These findings highlight note-based surveillance as a mechanism for earlier AUD identification and underscore the limitations of structured data for estimating real-world AUD burden.