Skip to main content
CMAJ : Canadian Medical Association Journal logoLink to CMAJ : Canadian Medical Association Journal
. 2024 Sep 16;196(30):E1042. doi: 10.1503/cmaj.240363

Artificial intelligence scribes in primary care

Payal Agarwal 1,, Rosemarie Lall 1, Rajesh Girdhari 1
PMCID: PMC11412733  PMID: 39284604

Artificial intelligence (AI) scribes can mimic human scribes

These tools use speech recognition, natural language processing, and AI technologies to listen during a clinical encounter and generate clinical documentation.1 Clinicians can then review, edit, and sign the generated note in the medical record.

Early evidence suggests that AI scribes lessen administrative burden and improve quality of time spent with patients2,3

A 10-week, California-based pilot found that primary care providers spent less time documenting during appointments and using electronic medical records outside office hours.3 Patients felt comfortable with AI scribes and reported clinicians spent less time looking at a computer.3 The use of AI scribes may also improve the quality of medical notes by generating more timely and complete documentation.3,4

Clinical documentation generated by AI scribes may contain errors and must always be reviewed3,5

Atrifical Intelligence tools can introduce mistakes, including hallucinations (i.e., documenting things that did not occur) or omission of key information.3,5 They may struggle with different languages and documenting physical examinations.2,3 Clinicians are ultimately responsible for the quality of their documentation and must review all AI-generated notes to ensure accuracy and completeness.4,5

Users must ensure software is compliant with local privacy regulations4,5

To date, AI scribes remain unregulated. Clinicians must understand how clinical information captured by the software is stored, retained, accessed, and subsequently used.4,5 Data stored outside of Canada may be subject to foreign laws. The onus is on the clinician or their institution to understand privacy implications and potential for patient harm, and to ensure the software performs as intended.

Users must obtain consent before using an AI scribe4,5

Clinicians must explain, receive, and document informed consent from patients, including reasons for use and potential risks such as privacy implications and subsequent ownership, storage, and use of health data (e.g., further training of the AI scribe).5

Footnotes

Competing interests: Payal Agarwal is leading an evaluation of AI scribes that is funded by the Ontario Ministry of Health, in partnership with OntarioMD and Women’s College Hospital. She reports meeting support from Canada Health Infoway and Digital Health Canada. She is Chief Medical Information Officer at Grand River Hospital and St. Mary’s General Hospital. Rosemarie Lall reports funding from the CONCOR1 project, honoraria from the Ontario College of Family Physicians, and meeting support from the Ontario Medical Association and the Scarborough Ontario Health Team (SOHT). She is co-chair and member of the Scarborough Family Physicians Network, peer leader with OntarioMD, and member of the SOHT Collaboration Council, the Scarborough General Branch Society Executive, the Department of Family and Community Medicine at Scarborough Health Network (SHN) Executive, and the SHN research ethics board. Rajesh Girdhari reports funding from the Department of Family and Community Medicine at the University of Toronto and Ontario Health. No other competing interests were declared.

This article has been peer reviewed.

References


Articles from CMAJ : Canadian Medical Association Journal are provided here courtesy of Canadian Medical Association

RESOURCES