However, clinically significant inaccuracies were noted “occasionally”
TUESDAY, Dec. 9, 2025 (HealthDay News) — Artificial intelligence (AI)-based ambient scribes may reduce documentation time and improve physician well-being, according to a study published online Nov. 26 in NEJM AI.
Paul J. Lukac, M.D., from the University of California, Los Angeles (UCLA), and colleagues examined the impact of ambient AI scribes. Outpatient physicians (238), representing 14 specialties, were randomly assigned (1:1:1) to either one of two AI scribe applications — Microsoft Dragon Ambient eXperience (DAX) Copilot or Nabla — or a usual-care control group. Change from baseline log writing time-in-note was the primary outcome. Secondary outcomes included several survey instruments: Mini-Z 2.0 to assess burnout, work environment, work pace, and electronic health record (EHR) stress; physician task load (PTL) to assess cognitive load related to stress from EHR documentation; and Professional Fulfillment Index-Work Exhaustion (PFI-WE).
The researchers found that Nabla users experienced a 9.5 percent (P = 0.02) decrease in time-in-note versus the control group. DAX users showed no significant change versus the control group (−1.7 percent). Compared with control, there were improvements in total Mini-Z (scale of 10 to 50; DAX, +2.83; Nabla, +2.69), PTL (scale of 0 to 400; DAX, −39.9; Nabla, −31.7), and PFI-WE (scale of 0 to 4; DAX, 0.32; Nabla, −0.23) for users of either scribe. There was one grade 1 (mild) adverse event reported, while clinically significant inaccuracies were noted “occasionally” on 5-point Likert questions (DAX, 2.7; Nabla, 2.8).
“This technology requires active physician oversight, not passive acceptance,” senior author John N. Mafi, M.D., M.P.H., also from UCLA, said in a statement. “The path forward requires embracing innovation while maintaining medicine”s fundamental commitment to patient safety through rigorous evaluation and ongoing monitoring.”
Abstract/Full Text (subscription or payment may be required)
Use of Emojis in Electronic Health Records Increased in 2024 to 2025
Methods for How Death, Dying Are Taught in Medical School Vary Considerably
About One-Third of Professionals Prescribe Subcutaneous Contraceptive for Self-Administration
AI Model Can Screen for Cognitive Impairment From Clinical Notes
BIND Screener Can ID Brain-Related Comorbidities in Duchenne Muscular Dystrophy
AI-Powered ECG Analysis Promising for Early Detection of COPD
Guidance Updated for Laboratory Testing for Drugs of Misuse in ED
HIV Viral Load Test With Next-Day Results Does Not Impact Linkage to Care