Pragmatic Trial of User-centered Clinical Decision Support to Implement EMergency Department-initiated BuprenorphinE for Opioid Use Disorder

The study design is an 18-month pragmatic, parallel, cluster randomized, superiority trial using constrained randomization of clusters to arms. The unit of randomization (i.e. cluster) is the ED. EDs will be randomly allocated with an allocation ratio of 1:1. Adequate lead time will be allotted to install the intervention in the electronic health records (EHR) at all intervention sites--including a three month implementation and washout phase. The intervention will then begin at the same time across all sites with the CDS intervention fully implemented in the intervention sites' EHRs at the start of the trial. Clinicians at control sites will retain all control of their practice and practice as usual without the CDS intervention installed in their EHR.

Pragmatic trials study an intervention under the usual conditions in which it will be applied; as opposed to an explanatory trial which would test an intervention under ideal conditions. In cluster randomized trials, treatment intervention is allocated to clusters (i.e. groups of individuals) rather than individuals. This is done to manipulate the physical or social environment of the intervention when an individual intervention would likely result in contamination between intervention and control participants at the group level. The parallel cluster randomized design was chosen over a stepped wedge design due to the high likelihood of confounding by temporal trends from ongoing efforts to mitigate the opioid epidemic. A major challenge of the cluster randomized design is from potential confounding due to a limited number of heterogeneous groups. Constrained randomization offers a solution to this source of confounding by balancing key cluster-level prognostic factors across the study to avoid distorting estimates of treatment effect due to the confounding factors. This allocation technique more evenly distributes potential confounders between intervention arms by specifying the confounding factors, characterizing each cluster in terms of these factors, identifying a subset of randomization combinations of clusters that adequately balance confounding factors between intervention arms and randomly selecting one of these combinations as the allocation scheme. Potential confounders that will be used for this trial are: EHR vendor, ED annual volume, ED type (e.g., academic, community, urban, rural, etc), ratio of ED clinicians who have a waiver to prescribe BUP, current rate of ED BUP prescribing, resources in ED to facilitate management of patients with OUD, and willingness of staff to adopt the practice of ED-initiation of BUP.

Intervention:

The intervention for this study includes the user-centered CDS as well as education of ED clinicians practicing at all study sites.

The need for flexibility in the graphical user interface of the intervention resulted in the decision to develop the CDS as a web application. This provides the ability to access the tool both embedded within the EHR or directly over the Internet. The web application was developed as a single-page application (SPA) based on React JavaScript library. The CDS is a user-initiated, Substitutable Medical Applications and Reusable Technologies (SMART) on Fast Health Interoperability Resources (FHIR) application that streamlines a flow diagram of the clinical protocol for ED-initiated BUP.