Artificial intelligence-powered documentation fails to significantly improve clinician efficiency
1. Liu and colleagues compared efficiency between two groups of clinicians, with one group using an artificial intelligence (AI)-powered documentation platform.
2. No statistically significant differences in efficiency and financial metrics between the two groups were found, but some improvements were seen in subgroups.
Evidence Rating Level: 1 (Excellent)
Study Rundown: AI-powered tools are increasingly used to automate clinical documentation and have the potential to alleviate administrative burdens. Liu and colleagues conducted a longitudinal study that compared the efficiency of clinicians who used the Dragon Ambient eXperience (DAX) documentation software to another group of clinicians from similar practices who did not use DAX. Primary outcome measures included electronic health record (EHR) metrics, such as documentation time (Note-Time), work time outside of work (WOW), completed appointment rate; and financial metrics, such as gross revenue per visit. Physicians were stratified by their specialty and DAX use status, with active users being those who transferred more than 25% of their DAX notes to the EHR and high users being those who transferred more than 60%. For primary outcomes, there were no statistically significant differences between the two groups after controlling age, gender, provider type, experience, and baseline outcome. However, subgroup analyses showed a minor decrease in Note-Time for low-volume clinicians and family medicine physicians. Overall, this study demonstrated that the AI-powered software did not improve clinician efficiency but has the potential to save time for selected physicians.
Click here to read the study in NEJM AI
Relevant Reading: AI-Powered Clinical Documentation and Clinicians’ Electronic Health Record Experience: A Nonrandomized Clinical Trial
In-Depth [randomized controlled trial]: 238 clinicians in family medicine, internal medicine, and general pediatrics in our patient clinical practices were recruited into the study. 112 clinicians were allocated to the DAX group, while 103 were included in the control group. Clinicians in the DAX group underwent a 1-hour training session in using DAX, a software that generates visit notes from audio recordings of clinical encounters. Clinician characteristics, such as age, gender, years of practice, and medical specialties were collected. Patient volume was calculated from the number of hourly visits and was categorized into low-, medium-, and high-volume clinicians. Outcomes included EHR use and financial metrics. The former included time in EHR (EHR-Time), documentation time (Note-Time), work time outside of work (WOW), and completed appointment rate. The latter included gross revenue per visit and work relative value units (wRVUs) per visit. The outcomes were assessed 180 days after DAX adoption. In the primary outcome analyses, no statistically significant differences between the two groups were found, after controlling baseline variables. However, exploratory subgroup analyses found that for low-volume clinicians, Note-Time was lower for DAX users (mean ratio 0.91, 95% CI 0.83-0.99). Improvement was also found for family medicine clinicians (mean ratio 0.91, 95% CI 0.85-0.98). Limitations of this study included lack of randomization, inability to obtain comprehensive editing time in DAX, and learning curve associated with using DAX. In summary, this study did not show that AI-powered clinical documentation software can statistically significantly improve clinician efficiency but may be useful to clinicians in some practice settings.
Image: PD
©2025 2 Minute Medicine, Inc. All rights reserved. No works may be reproduced without expressed written consent from 2 Minute Medicine, Inc. Inquire about licensing here. No article should be construed as medical advice and is not intended as such by the authors or by 2 Minute Medicine, Inc.
