Impact of Ambient Artificial Intelligence Scribes on Emergency Department Documentation Burden: A Retrospective Cohort Study (Preprint)

Abstract

<sec> <title>BACKGROUND</title> Clinician burnout has reached crisis levels in emergency medicine, with clinical documentation burden identified as a central contributing factor. Ambient artificial intelligence (AI) scribes offer a promising approach to reduce documentation burden, but scant quantitative evidence exists on objective time savings in the emergency department (ED) setting. </sec> <sec> <title>OBJECTIVE</title> To evaluate the effect of ambient AI scribes on on-shift documentation time in a busy ED during a staged rollout, accounting for physician- and patient-level factors. </sec> <sec> <title>METHODS</title> We conducted a 6-month retrospective cohort study at a tertiary academic ED from February to August 2025. The analytic cohort included 7,640 encounters managed by 87 attending physicians across four ED care settings. We limited our analysis to encounters managed individually by a single physician and excluded cases with human scribes. The primary outcome was on-shift documentation time derived from electronic health record audit logs. We used mixed-effect linear models with physician random intercepts to adjust for patient and encounter characteristics. </sec> <sec> <title>RESULTS</title> Ambient AI scribes reduced on-shift documentation time by 68.2 seconds per encounter (95% CI [82.9, 53.5]; P&lt;.001). High-utilizing physicians (those with usage rates ≥9.9%, the cohort mean) saved 73.9 seconds per encounter compared to 27.5 seconds for low/moderate utilizers (difference: 46.4 seconds, P=.013). Note character count decreased by 958 characters (95% CI [1757, 158]; P=.019); after-shift documentation time increased modestly by 12.3 seconds (95% CI [1.3, 23.3]; P=.029). </sec> <sec> <title>CONCLUSIONS</title> In this single-center study, ambient AI scribes were associated with a statistically significant reduction in on-shift documentation time, translating to approximately 23 minutes saved over a typical 8-hour shift with 20 patient encounters. These findings support their clinical utility in mitigating documentation burden; however, the effect varies based on physician, patient, and workflow factors. </sec>

Ähnliche Arbeiten