Supporting Healthcare Initiatives to Facilitate Treatment of Opioid Use Disorder in the Intensive Care Unit

Mainehealth, Portland ME

Investigators

Linked publications & trials

Abstract

This proposal aims to improve the identification and management of patients with Opioid Use Disorder (OUD) in MaineHealth intensive care units (ICUs) through the use of a clinician-facing dashboard, tailored educational programming, and natural language processing (NLP). Our overarching goal is to reduce practice variations, increase adherence to evidence-based standards for managing OUD, and strengthen linkages to outpatient OUD treatment services. The lack of standardized guidelines for managing critically ill OUD patients nationally underscores the urgency and significance of this research. Aim 1: We will implement a validated clinician-facing dashboard at three MaineHealth ICUs to monitor key quality indicators related to OUD care, including medications for OUD (MOUD) administration, infectious disease screening, and naloxone prescribing. Clinician champions will review the dashboard and collaborate with care teams to ensure recommended interventions are consistently delivered. We will evaluate whether dashboard implementation is associated with improvements OUD related outcomes including MOUD utilization, hospital readmissions and emergency department visits. Aim 2: We will design and implement an educational intervention for critical care providers across the three participating ICUs to standardize OUD management practices. The training will incorporate case-based modules, site-specific order sets, and provider feedback, with support from the Community Engagement, Bioethics, and Outreach (CEBO) Core and the MaineHealth Addiction Medicine Council. Surveys and interviews will assess changes in provider knowledge, attitudes, and confidence pre- and post-intervention, and examine whether increased engagement leads to improved clinical outcomes and order set utilization. Aim 3: We will use NLP to identify patients with OUD by analyzing unstructured clinical notes within the electronic health record. Using supervised learning methods, we will extract clinical phrases and tags associated with OUD. We will evaluate the performance of our NLP algorithm compared to structured data sources. Outcomes: We will assess whether our combined intervention improves provider knowledge and adherence to OUD care standards, increases MOUD utilization, and reduces readmissions and ED visits. The use of NLP will help identify patients with OUD that can be under-recognized using structured data alone.