Adaptive Decision Support for Addiction Treatment in the Emergency Department: ADAPT Factorial Trial Results

Buprenorphine (BUP) is a life-saving treatment for opioid use disorder (OUD) that can be safely initiated in the ED; in our prior work, clinical decision support (CDS) to support BUP initiation was underutilized, however. This study, part of a Multiphase Optimization Strategy project, evaluates methods to increase engagement with CDS to promote initiation of BUP for ED patients with OUD.

Between Mar-Sep 2025, we conducted a 2×2×2 randomized factorial trial at 3 EDs in a single health system, evaluating encounter-level CDS strategies including prompts to (1) nurses at triage, recommending Clinical Opiate Withdrawal Scale (COWS) assessment, (2) clinicians during order entry, recommending patient assessment, and (3) clinicians at discharge, facilitating customized discharge materials. Adult ED patients with suspected OUD were identified using a machine learning–based OUD phenotype; triage chief complaints; pre-hospital receipt of naloxone; or COWS score. Eligible encounters were randomized to one of 8 arms. The primary outcome was encounter-level CDS engagement, defined as any interaction with the CDS tool, which facilitates OUD diagnosis, order entry, and documentation. BUP initiation rate was a secondary outcome. For the primary main-effects analysis, we compared primary and secondary outcomes between encounters exposed versus not exposed to each intervention (nurse prompt, provider order prompt, discharge prompt), with p-values estimated from generalized estimating equation model.

A total of 2761 encounters involving 1130 unique patients were included. CDS engagement was significantly increased in encounters randomized to receive the clinician assessment prompt (5.4% with vs 3.1% without; p<0.001) and clinician discharge instruction prompt (6.2% with vs 2.3% without; p<0.001) but not in encounters randomized to receive the nursing prompt for COWS assessment (4.7% with vs 3.8% without; p=0.58). No significant differences in BUP initiation rates were observed across arms (p=0.78).

Although the evaluated implementation strategies improved CDS use, this increase did not translate into higher rates of buprenorphine initiation. These results suggest that CDS engagement is necessary but not sufficient to change care delivery in this clinical scenario. Further optimization should consider additional workflow, behavioral, and system-level factors.