Interrupted Time Series of User‐centered Clinical Decision Support Implementation for Emergency Department–initiated Buprenorphine for Opioid Use Disorder

Abstract

Objective Adoption of emergency department (ED)‐initiation of buprenorphine (BUP) for opioid use disorder (OUD) into routine emergency care has been slow, partly due to clinicians’ unfamiliarity with this practice and perceptions that it is complicated and time‐consuming. To address these barriers and guide emergency clinicians through the process of BUP initiation, we implemented a user‐centered computerized clinical decision support system (CDS). This study was conducted to assess the feasibility of implementation and to evaluate the preliminary efficacy of the intervention to increase the rate of ED‐initiated BUP. Methods An interrupted time series study was conducted in an urban, academic ED from April 2018‐February 2019 (pre‐implementation phase), March 2019‐August 2019 (implementation phase), and September 2019‐December 2019 (maintenance phase) to study the effect of the intervention on adult ED patients identified by a validated EHR‐based computable phenotype consisting of structured data consistent with potential cases of OUD who would benefit from BUP treatment. The intervention offers flexible CDS for identification of OUD, assessment of opioid withdrawal, motivation of readiness to start treatment and automates EHR activities related to ED‐initiation of BUP (including documentation, orders, prescribing, and referral). The primary outcome was the rate of ED‐initiated BUP. Secondary outcomes were launch of the intervention, prescription for naloxone at ED discharge, and referral for ongoing addiction treatment. Results Of the 141,041 unique patients presenting to the ED over the pre‐implementation and implementation phases (i.e. the phases used in primary analysis), 906 (574 pre‐implementation and 332 implementation) met OUD phenotype and inclusion criteria. The rate of BUP initiation increased from 3.5% (20/574) in the pre‐implementation phase to 6.6% (22/332) in the implementation phase (p=0.03). After adjusting for the temporal trend of the number of physician’s with X‐waiver training and other covariates, the relative risk of BUP initiation rate was 2.73 (95% CI 0.62, 12.0; p=0.18). Similarly, the number of unique attendings who initiated BUP increased modestly 7/53 (13.0%) to 13/57 (22.8%, p=0.10) after offering just‐in‐time training during the implementation period. The rate of naloxone prescribed at discharge also increased (6.5%, pre‐implementation; 11.5%, implementation; p<0.01). The intervention received a System Usability Scale score of 82.0 (95% CI 76.7‐87.2). Conclusion Implementation of user‐centered CDS at a single ED was feasible, acceptable, and associated with increased rates of ED‐initiated BUP and naloxone prescribing in patients with OUD and a doubling of the number of unique physicians adopting the practice. We have implemented this intervention across several health systems in an ongoing trial to assess its effectiveness, scalability, and generalizability.