Effect of a Data‐driven Intervention on Opioid Prescribing Intensity Among Emergency Department Providers: A Randomized Controlled Trial

Abstract

Little is known about accuracy of provider self-perception of opioid prescribing. We hypothesized that an intervention asking emergency department (ED) providers to self-identify their opioid prescribing practices compared to group norms—and subsequently providing them with their actual prescribing data—would alter future prescribing compared to controls. This was a prospective, multi-center randomized trial in which all attending physicians, residents, and advanced practice providers at four EDs were randomly assigned to either no intervention or a brief data-driven intervention during which providers were: (1) asked to self-identify and explicitly report to research staff their perceived opioid prescribing in comparison to their peers, and then (2) given their actual data with peer group norms for comparison. Our primary outcome was the change in each provider's proportion of patients discharged with an opioid prescription at six and twelve months. Secondary outcomes were opioid prescriptions per hundred total prescriptions and normalized morphine milligram equivalents prescribed. Our primary comparison stratified intervention providers by those who underestimated their prescribing and those who did not underestimate their prescribing, both compared to controls. Among 109 total participants, 51 were randomized to the intervention, 65% of whom underestimated their opioid prescribing. Intervention participants who underestimated their baseline prescribing had larger-magnitude decreases than controls (Hodges-Lehmann difference -2.1 prescriptions per hundred patients at 6 months [95% CI -3.9 to -0.5] and -2.2 per hundred at 12 months [95% CI -4.8 to -0.01]). Intervention participants who did not underestimate their prescribing had similar changes to controls. Self-perception of prescribing was frequently inaccurate. Providing clinicians with their actual opioid prescribing data after querying their self-perception reduced future prescribing among providers who underestimated their baseline prescribing. Our findings suggest that guideline and policy interventions should directly address the potential barrier of inaccurate provider self-awareness.