Changes in psychiatric documentation and treatment in primary care with artificial intelligence scribe use

Despite increasingly widespread use of artificial intelligence-driven ambient scribes in medicine, the extent to which they may impact clinician practice is not well-studied.

To characterize differences in documentation and treatment of psychiatric symptoms in primary care outpatient notes generated using ambient scribes.

Case-control electronic health records.

Primary care annual visit notes from the Massachusetts General and Brigham and Women's Hospital systems between February 2023 and February 2023.

Random sample of 20,302 notes from 4 types of visits, matching 1:1 using sociodemographic and clinical features: those using an ambient scribe, those using a human scribe, those occurring during the same period without a scribe, and those occurring prior to scribe deployment.

Use of an artificial intelligence-driven ambient scribe.

Neuropsychiatric symptom documentation, in terms of estimated Research Domain Criteria, using a HIPAA-compliant large language model (GPT4o; gpt-4o-11-20); incident antidepressant prescriptions and diagnostic codes; referral for mental health follow-up.

In the ambient scribe group, mean age was 48 (SD 14) years; 59% of notes reflected individuals of female sex, and 5.0% met criteria for moderate or greater depressive symptoms by PHQ-9. Estimated levels of RDoC symptomatology in all 6 domains were significantly greater in the ambient-scribed notes (p <.001 for all contrasts). In a logistic regression model, likelihood of a psychiatric intervention (referral, new diagnosis, or antidepressant prescription) was significantly lower among ambient-scribed visits compared to unscribed (aOR 0.83, 95% CI 0.72-0.95), but not for human-scribed compared to unscribed (aOR 1.01, 95% CI 0.87-1.17).

In this case-control design examining outpatient primary care notes, we found that incorporation of artificial intelligence-driven ambient scribes in primary care was associated with greater levels of neuropsychiatric symptom documentation but lesser likelihood of acting on psychiatric symptoms. Further study will be required to determine whether these changes are associated with differential outcomes.