A Comparison of Veterans with Problematic Opioid Use Identified through Natural Language Processing of Clinical Notes...

Area of Science:

• Health Informatics
• Clinical Informatics
• Natural Language Processing in Healthcare

Background:

• Opioid use disorder (OUD) is frequently under-coded in clinical practice.
• Problematic opioid use may be documented in unstructured clinical notes within electronic health records (EHRs).
• Identifying these cases is crucial for comprehensive patient care and epidemiological accuracy.

Purpose of the Study:

• To develop and apply a natural language processing (NLP) tool to identify problematic opioid use from clinical notes.
• To compare the demographic and clinical characteristics of patients identified by NLP versus those with ICD OUD diagnostic codes.
• To assess the prevalence and distinct profiles of patients with under-documented opioid use concerns.

Main Methods:

• A cohort of 222,371 patients from two Veteran Affairs service regions was analyzed.
• A hybrid NLP tool, combining rule-based analysis and a support vector machine, was developed and applied to clinical notes.
• Patients were also identified using International Classification of Diseases (ICD) opioid use disorder diagnostic codes.
• Performance metrics for the NLP tool included 96.6% specificity and 88.4% sensitivity on test data.

Main Results:

• The NLP tool identified 57,331 patients with problematic opioid use exclusively from clinical notes.
• Only 6997 patients were identified through ICD OUD codes.
• Patients identified exclusively by NLP were more likely to be women.
• Patients identified by ICD codes were more likely to be male, younger, have concurrent benzodiazepine prescriptions, more comorbidities, and more care encounters.
• Both groups showed substantially elevated comorbidity levels compared to patients not identified by either method.

Conclusions:

• Clinicians may under-code opioid use disorder, leading to underestimation of its prevalence.
• NLP tools can effectively identify problematic opioid use documented in clinical notes, capturing a different patient demographic.
• Healthcare teams should actively review clinical notes to uncover and address all instances of problematic opioid use.