Background: Prescribing of high-risk medications, such as benzodiazepines and sedative hypnotic (“Z-drugs”) to older adults is common. Many efforts to facilitate deprescribing have had limited success especially in chronic users, in part because of sub-optimal timing or poor design. While electronic health record (EHR)-based tools have potential, they have had limited application for deprescribing.
Objectives: We sought to directly compare the impact on deprescribing of 15 different EHR tools designed using behavioral science.
Methods: NUDGE-EHR is a two-stage 16-arm adaptive trial. In Stage 1, we cluster-randomized 201 primary care providers to receive either usual care or one of 15 new EHR tools for their patients ≥65 years chronically using benzodiazepines or Z-drugs across 26 clinics in Massachusetts (US). Each tool was classified whether it included one of 7 different behavioral components: 1) timing (at encounter opening vs. medication ordering), 2) boostering, 3) cold-state priming, 4) simplified language, 5) sign-off approval, 6) pre-commitment, and 7) risk framing. In interim analysis 9 months after randomization, we ranked the arms based on their impact on the primary outcome: patient-level deprescribing (i.e., discontinuing or tapering) measured using EHR data. We then re-randomized providers to receive usual care or one of the 5 most promising arms and followed for 11 months. Outcomes were evaluated using generalized linear mixed models with a logit link and binary errors, evaluating the association between the components and primary outcome, pooling across both stages and adjusting for clustering.
Results: 3063 unique patients met eligibility criteria. Across the 16 arms, 34.3% of patients had eligible medications deprescribed. Compared with messages timed at ordering a medication, messages at encounter opening were associated with a 25% higher odds of deprescribing (Odds Ratio [OR]: 1.25,95%CI:1.01-1.56). Boostering (OR: 1.15,95%CI:0.83-1.59), simplified language (OR: 1.20,95%CI:0.88-1.64), and pre-commitment (OR: 1.21,95%CI:0.89-1.64) were also associated with a non-significant greater odds of deprescribing versus usual care. Among arms using timing at encounter opening, pre-commitment was associated with greater deprescribing (OR: 1.67,95%CI:1.00-2.87).
Conclusions: Alerts that triggered when opening an encounter may be more effective than when refilling medication. Among alerts that were triggered during opening an encounter, using pre-commitment, leveraging a collaborative approach with patients, significantly improved deprescribing. As EHR tools are widely-used, these findings offer lessons for how to improve their effectiveness on prescribing.
