AI authors promise to reduce the burden of documentation, but new real-world data reveal a more complex reality: modest efficiency gains, unchanged after-hours work, and important questions about how clinicians actually use the time they save.
Study: Changes in physician time spend and visit volume with the adoption of AI-powered scribes. Image credit: Andrey_Popov/Shutterstock.com
Artificial intelligence (AI) scribes can modestly benefit clinicians by reducing time spent recording and documenting electronic health records and increasing weekly visit volume, according to a new study published in GLASS.
AI scribes are emerging to deal with the growing documentation burden
Electronic health record (EHR) documentation is a time-consuming task for clinicians, averaging 2.3 hours per 8 hours of patient care. This documentation time is associated with physician burnout, often limiting clinical competence, patient access, and quality of care.
Artificial intelligence (AI) powered documentation tools, also known as AI scribes, have been developed to reduce EHR documentation burden and improve clinician satisfaction. However, studies investigating the effectiveness of AI scribes have produced mixed findings, and evidence of their impact on productivity remains limited.
Given this gap in the literature, the current study investigated whether the adoption of AI scribes in health systems can produce changes in EHR time expenditure and weekly visit volume, and whether these changes vary by clinician characteristics.
Want to see how AI works in real clinics? Read this expert interview
US multi-site study examines real-world adoption of AI scribes
The study involved five academic health care institutions in several regions of the United States that introduced AI scribes to their clinicians. The study population included ambulatory clinicians, advanced practice clinicians, and resident physicians who had the option of using an AI scribe.
Across the entire study population, clinicians who were given access to AI scribes were considered adopters of AI scribes, regardless of whether they actively used the tool (intention-to-treat definition). Conversely, clinicians who did not have access to AI scribes were considered non-adopters.
The main parameters analyzed in the study were time spent on documentation, time spent in the EHR outside of scheduled working hours, and weekly visit volume.
AI scribes modestly reduce EHR and documentation time
The study involved 8581 clinicians, including 1809 adopting AI scribes. Participants came from primary care, medical and surgical specialties, and the majority were general practitioners, followed by advanced clinicians and medical specialists.
Analysis of pre- and post-adoption trends showed that AI scribe adoption was associated with 13 fewer minutes of EHR time and 16 fewer minutes of documentation time per 8 hours of planned patient care, representing 3.0% and 10.0% relative reductions in time spent, respectively.
Additionally, adopting AI scribes was associated with 0.49 more weekly visits made, representing a 1.7% increase in weekly visit volume. However, no significant changes in EHR time outside of work hours were observed after the adoption of the artificial device.
Among clinician groups, the greatest improvements after adopting AI scribes were seen in primary care specialists, advanced care clinicians, female clinicians, physicians, and clinicians who used AI scribes in 50% or more of visits.
The revenue analysis estimated an additional $167.37 in monthly evaluation and management (E/M) revenue per clinician associated with adoption of the prosthesis.
Time savings can be shifted to other clinical responsibilities
The study reveals that adoption of AI scribes by clinicians in academic healthcare institutions is associated with a modest reduction in total EHR and documentation time and a modest increase in weekly visit volume.
Specifically, the study finds that the adoption of AI scribes does not significantly change out-of-hours work, despite modest reductions in total EHR time and documentation time. These findings suggest that although the adoption of AI scribes has saved clinicians time on documentation, some of these time savings may be reallocated to other patient care activities, such as reviewing documentation for accuracy, monitoring electronic inbox messages from patients, dealing with test results, or conducting a medical record review.
As observed in the study, the time saved by adopting AI scribes was highest among primary care specialists, advanced practice clinicians, female clinicians, residents, and clinicians who used AI scribes in 50% or more of their visits. Among these groups of clinicians, residents are a critical population for evaluating the utility of AI scribe adoption, given how critical documentation is to learning and the unknown implications for resident learning.
Since the number of residents who used AI scribes was limited in the study, the researchers highlight the need for future studies to assess the impact of AI scribes on residents’ time expenditure and influence on learning.
Despite the highest benefits seen for clinicians who used AI scribes in 50% or more visits, only 32% of users used AI scribes that often. This finding highlights the need for robust education and support for adopters.
Due to the non-randomized study design, the observed changes may not be solely related to the adoption of the technical design. Some unmeasured differences between adopters and non-adopters may also influence the findings. Future studies should investigate the reproducibility of these findings and identify factors that may enhance the benefits of this technology.
In addition, the study included only academic health care institutions, with an average weekly volume of visits of approximately 20. The observed benefits may differ in non-academic settings, where clinicians have a much higher volume of visits.
