Opioid Agonist Therapy Initiation: Validation of the Accuracy of Electronic Health Record Data Capture

Increasing opioid agonist therapy (OAT=buprenorphine and methadone) initiation rates is important, as OAT cuts opioid use disorder (OUD) mortality in half. Previous studies have attempted to use EHR data to capture cohorts of patients with OUD in order to define (and improve) OAT initiation rates. However, knowing the proportion of patients eligible to initiate OAT using EHR data alone may be difficult due to inconsistent EHR reporting of methadone maintenance. Our objective was to compare data obtained via the EHR with data that added manual chart review in order to validate a hybrid approach to understanding OAT eligibility and initiation rates.

We used previously described methods to extract EHR data based on ICD-10 codes for opioid overdose, withdrawal, and OUD for a 12 months period in our 4 ED system and excluded admissions and deaths. Based on EHR data, we divided the cohort into ‘new OAT initiation likely’ (defined as methadone administered of 10, 20, or 30mg or a prescription for buprenorphine), ‘likely already on OAT,’ (defined as methadone given at a higher dose or buprenorphine administered but not prescribed), or ‘missed opportunity’ (defined as no medications given or prescribed). We then manually reviewed 10% of charts to determine the accuracy of the categorizations (chart review cohort, CRC). Finally, we compared rates of the outcome of interest, OAT initiation, between the EHR-only cohort and the CRC.

From 10/23–9/24, 1976 ED visits met inclusion criteria. EHR categorization yielded 348 (18%) ‘new OAT,’ 332 (17%) ‘likely on OAT,’ and 1296 (66%) ‘missed opportunity.’ Removing those already on OAT, this indicates a 22% rate of OAT initiation (348/1644). Within the 1296 ‘missed opportunity’ group, chart review revealed that 51% either did not have OUD (17%) or were already on OAT (34%, 2/3 methadone and 1/3 buprenorphine), and 12% were offered OAT, (half accepted); therefore, 57% were misclassified. Comparing OAT initiation rates after excluding those already on OAT, the initiation rates were 22% for the EHR-only cohort and 46% for the CRC.

Although different EHRs may perform differently, methadone maintenance is inconsistently captured by EHR data. Manual chart review is likely necessary to accurately know OAT initiation rates.