373 Challenges in using real-world data to study opioid use disorder treatment in the hospital

Objectives/Goals: Our research group is focused on care of hospitalized persons with opioid use disorder (OUD) in the era of high-potency synthetic opioids (HPSO). In this work, we describe trends in patient-directed discharge (PDD) and inpatient treatment with medications for opioid use disorder (MOUD). We hypothesized that PDD is associated with MOUD dose and timing. Methods/Study Population: Patient data generated in the routine care of patients was automatically abstracted using a SQL query on Epic Clarity tables in the electronic health record (EHR). We included adult patients admitted to Johns Hopkins Hospital between July 1, 2019 and June 30, 2022, with an ICD-10-CM code for a list of opioid-related disorders (F11.X) consistent with OUD. Demographics, prior medication list, clinical care including hospital service, consultation services, COWS scores, length of time in emergency department, time of triage, time until receipt of methadone or buprenorphine, dosage and timing of MOUD, opioid medications other than methadone or buprenorphine, adjuvant medications; prior methadone or buprenorphine treatment and disposition. Query results were validated by manual abstraction of EHR. Results/Anticipated Results: The SQL identification of the cohort of patients with OUD was found to be accurate. Time of triage, discrete orders completed during hospitalization were well represented in the query. The query was able to identify individual opioid medications but unable to summarize total dose in Morphine Milligram Equivalents. The query did not extract accurate information from patient-controlled analgesia pumps due to the continuous nature of the medication rather than discrete doses reflected in the medication administration record. Finally, the query characterized prior treatment with methadone or buprenorphine as a binary variable – dosage and timing of that prior treatment could not be accurately represented. Finally, stimulant use is not reliably collected in the EHR and was unavailable. Discussion/Significance of Impact: Given the rise of HPSO, patients may not tolerate delay of MOUD. Improving the granularity of data collected will offer more insight into the inpatient treatment for OUD. Real-world data have promise but requires extensive technical expertise. Future work is needed to improve capture of derived variables such as total dosage of opioids in MME.