(287) EHR-based pragmatic trial evaluating the impact of simulation intervention on high-risk prescribing to older adults by junior physicians

Background: High-risk medications like benzodiazepines, sedative hypnotics, and antipsychotics are commonly prescribed for hospitalized older adults, despite guidelines recommending avoidance. Prior interventions have not fully addressed how junior physicians make such prescribing decisions, particularly when experiencing stress.

Objectives: Provider-facing interventions like simulation training could affect prescribing of high-risk medications but have not been evaluated for this purpose nor in a pragmatic trial using databases.

Methods: In this 2-arm pragmatic trial, we randomized 40 first-year general medical resident physicians (i.e., interns) assigned to evening inpatient general medicine services at an academic medical center. Interns were block-randomized 1:1 to either intervention (a 40-minute immersive simulation training) or control (online educational training) arms about high-risk medications, which they received at the beginning of their general medicine rotation. The primary outcome was the number of new benzodiazepine, sedative hypnotic, or antipsychotic orders for treatment-naïve older adults during hospitalization. Patients were identified using EHR data and were not otherwise recruited in any way. Secondary outcomes included prescribing by all providers, being discharged on a target medication, and prescribing of related or control medications. Outcomes were measured using EHR data over each intern’s service period and evaluated using generalized estimating equations, adjusting for clustering.

Results: In total, 522 treatment-naïve older adult patients were included in analyses, including 263 and 259 patients in the control and intervention arms, respectively. Over follow-up, interns prescribed ≥1 of these medications for 13 (4.9%) intervention and 13 (5.0%) control patients, respectively. The intervention led to no difference in the number of new prescriptions (Rate Ratio [RR]: 0.85, 95%CI: 0.31-2.35) versus control and no difference in secondary outcomes. In secondary analyses, the intervention-group interns wrote significantly fewer as needed (“PRN”) orders for the target drugs (RR: 0.29, 95%CI: 0.08-0.99) and instead tended to write more “one-time” orders than control interns, though not statistically significant (RR: 2.20, 95%CI: 0.60-7.99).

Conclusions: While this intervention did not impact total high-risk prescribing for hospitalized older adults, it did influence how the interns prescribed, resulting in fewer PRN orders, which suggests possibly greater ownership of care as PRN orders require discretion by others on the care team. Future evaluations using EHR data should similarly explore nuances in prescribing outcomes that may not be reflected in total prescribing alone.