Electronic Health Record Intervention and Deprescribing for Older Adults

Importance: Potentially inappropriate medications are commonly overprescribed to older adults. Although electronic health record (EHR)-based tools can increase use of evidence-based medications, their ability to reduce prescription of potentially inappropriate medications remains unclear. Objective: To test the effects of 2 EHR interventions, designed using behavioral science techniques, on the deprescribing of potentially inappropriate medications compared with usual care in older patients. Design, Setting, and Participants: In this 3-group parallel randomized clinical trial, 201 primary care physicians (PCPs) in an academic center in Massachusetts were cluster-randomized in November 2022. Follow-up ended March 15, 2024. The intervention focused on patients of randomized PCPs who were 65 years or older, had a PCP visit between November 10, 2022, and March 15, 2024, and were prescribed at least 90 pills of benzodiazepines, nonbenzodiazepine sedative hypnotics, or at least 2 anticholinergic medications in the past 180 days. Interventions: PCPs were randomized to usual care (no intervention) or to 1 of 2 sequential EHR interventions: a precommitment intervention, in which an EHR message was sent to the physician during the first patient visit asking the PCP to initiate deprescribing discussions with a second reminder EHR message at the patient's second visit encouraging deprescribing; and a boostering intervention, in which PCPs received a notification encouraging deprescribing at the first patient visit and an in-basket reminder 4 weeks later. Main Outcomes and Measures: The primary outcome was deprescribing at least 1 medication on or after the first patient visit though the end of follow-up. Deprescribing was defined as physician-directed discontinuation or medication tapering assessed at the patient level using EHR data. Generalized estimating equations with a log link and binary-distributed errors were used for analyses, adjusting for clustering and multiple testing using Holm-Bonferroni corrections. Results: Of 1146 participants (mean age, 73.6 years [SD, 6.4]; 69.7% female, mean follow-up, 289.9 days), 373 (32.5%) had at least 1 medication deprescribed: 145 (36.8%) in the precommitment group, 122 (34.3%) in the boostering group, and 106 (26.8%) in usual care. Compared with usual care, deprescribing was 40% more likely (relative risk [RR], 1.40; 95% CI, 1.14-1.73; absolute difference, 10.4%) in the precommitment group and 26% more likely (RR, 1.26; 95% CI, 1.01-1.57; absolute difference, 6.5%) in the boostering group. No serious adverse events were reported through the adverse event reporting system. Death rates based on manual chart review were 1.4% in the precommitment group, 3.9% in the boostering group, and 1.8% in the usual care group. Conclusions and Relevance: These results support use of EHR tools designed using behavioral science principles to significantly increase rates of deprescribing potentially inappropriate medications used by older adults. Trial Registration: ClinicalTrials.gov Identifier: NCT005538065.