Simple EHR nudges cut use of high-risk medications

By Priyanjana Pramanik, MSc.Reviewed by Lauren HardakerFeb 4 2026

A randomized JAMA trial shows that simple, behaviorally designed electronic health record prompts can shift prescribing habits in primary care without new staff, added time, or complex programs.

Study: Electronic Health Record Intervention and Deprescribing for Older Adults. Image credit: Yelena Temirgaliyeva/Shutterstock.com

In a recent study published in JAMA, researchers evaluated whether two interventions, precommitment and boostering, could increase the rates at which physicians deprescribe medications that should be limited in older individuals, such as benzodiazepines or sedative hypnotics.

Compared to usual care, both interventions, which were informed by behavioral science, increased deprescribing rates, with the precommitment strategy producing a larger impact, though overall deprescribing rates remained modest.

Why risky medications persist in older adults

Potentially inappropriate medications, such as nonbenzodiazepine sedative hypnotics, benzodiazepines, and strongly anticholinergic drugs, are commonly prescribed to older adults despite well-documented risks, including increased falls and hospitalizations. Although clinical guidelines state that the long-term consumption of these medications should be limited, deprescribing remains uncommon in routine practice. Barriers include limited clinician time, patient resistance, inertia, and insufficient practical tools.

Previous deprescribing interventions, such as pharmacist-led reviews, specialist involvement, or patient education, have shown mixed effectiveness and are often resource-intensive. In contrast, electronic health record (EHR)-based clinical decision support tools have successfully improved evidence-based prescribing, yet their effectiveness for reducing inappropriate long-term medication use in primary care is less clear.

Behavioral science offers strategies to overcome decision-making barriers, such as precommitment and reinforcement, which may be particularly well-suited for integration into EHR workflows by shifting decisions away from time-pressured clinical encounters.

Physician-level randomization targets high-risk prescribing patterns

Researchers tested whether two EHR interventions grounded in behavioral science principles could increase the rate at which physicians deprescribe medications that can have adverse outcomes for older adults, compared with usual care.

They conducted a three-group cluster-randomized clinical trial at a medical center within a large Massachusetts health system. A total of 201 eligible primary care physicians (PCPs) were randomized in a 1:1:1 ratio to usual care, a precommitment intervention, or a boostering intervention.

Patients were eligible if they were aged 65 years or older, had an appointment with an included PCP during the study period, and had received one or more prescriptions for high quantities of benzodiazepines, nonbenzodiazepine sedative hypnotics, or at least two strongly anticholinergic medications.

Randomization occurred at the physician level to reduce contamination. The precommitment intervention involved sequential EHR notifications prompting physicians first to discuss medication risks with patients and later to deprescribe it. The boostering intervention delivered an initial deprescribing prompt, followed by an optional EHR inbox reminder approximately four weeks later, if requested by the clinician rather than being sent automatically. Usual care physicians received no visible prompts.

The primary outcome was deprescription of at least one targeted medication, defined using EHR data as discontinuation, dose tapering, or nonrenewal, including both active and passive discontinuation. Researchers accounted for clustering in the analysis and applied statistical corrections for multiple comparisons.

Medication discontinuation increases without dose reductions

Among 1,146 eligible patients with an average age of 73.6 years, approximately 70 % of whom were female, 32.5 % experienced deprescribing of at least one potentially inappropriate medication during a mean follow-up of about 290 days.

Deprescribing occurred most frequently in the precommitment group (36.8 %), followed by the boostering group (34.3 %), and least often in usual care (26.8 %), meaning roughly one in three patients experienced deprescribing even with intervention.

Compared with usual care, patients whose physicians received the precommitment intervention were 40 % more likely to have a medication deprescribed, corresponding to an absolute increase of 10.4 %.

The boostering intervention also significantly increased deprescribing, with a 26 % relative increase and a 6.5 % absolute difference compared with usual care. These effects were consistent after adjustment for patient demographics and across most prespecified subgroups.

The interventions were particularly effective among patients taking only one class of targeted medication, while effects were smaller and not statistically significant among patients using multiple medication classes.

No significant differences were observed between groups for secondary outcomes measuring pill quantity or cumulative medication dose prescribed, indicating that overall medication exposure did not decline despite higher discontinuation rates. No serious adverse incidents related to deprescribing were reported, and mortality rates were low across all groups, although deaths were numerically higher in the boostering group.

Behavioral EHR nudges change clinician prescribing behavior

This trial demonstrated that EHR interventions informed by behavioral science can significantly increase discontinuations or tapering off of potentially inappropriate medications in older adults within primary care, without evidence of short-term safety concerns.

Both precommitment and boostering strategies were effective, with precommitment yielding the largest improvement. These findings extend prior work by showing that simple, EHR-embedded behavioral nudges, without additional staff or patient-facing programs, can change clinician behavior at scale, though they do not eliminate inappropriate medication use.

Key strengths include the randomized design, integration into routine clinical workflows, broad inclusion criteria, and use of objective EHR-based outcomes. The pragmatic nature of the trial enhances generalizability to real-world primary care settings.

Several limitations should be noted. Deprescribing that occurred outside the health system may not have been captured, and passive discontinuation may not always reflect intentional clinician action. The study did not report downstream clinical outcomes such as falls or hospitalizations, despite these being prespecified, because claims-based data linkage was not available at the time of publication, and secondary dose-based outcomes showed wide confidence intervals. Additionally, findings reflect a single academic health system.

Overall, the study supports the use of behaviorally informed EHR tools as a scalable and effective approach to reducing potentially inappropriate medication use among older adults, while highlighting the need for complementary strategies to achieve larger clinical impact.