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

Area of Science:

• Medical Informatics
• Clinical Practice
• Artificial Intelligence in Healthcare

Background:

• The increasing adoption of AI-driven ambient scribes in medicine necessitates an understanding of their impact on clinical practice.
• Limited research exists on how ambient scribes influence the documentation and treatment of psychiatric symptoms in primary care.

Purpose of the Study:

• To compare the documentation and treatment of psychiatric symptoms in primary care outpatient notes generated with and without ambient scribes.
• To characterize differences in neuropsychiatric symptom documentation, prescription of antidepressants, and mental health referrals.

Main Methods:

• A case-control study analyzing electronic health records from primary care visits.
• Compared 1:1 matched notes (N=20,302) from visits using ambient scribes, human scribes, or no scribes, alongside pre-scribe deployment notes.
• Utilized a HIPAA-compliant large language model (GPT4o) to estimate Research Domain Criteria (RDoC) symptomatology and analyzed antidepressant prescriptions and referrals.

Main Results:

• Ambient scribe use was associated with significantly greater estimated levels of RDoC symptomatology across all domains (p <.001).
• The likelihood of psychiatric intervention (referral, new diagnosis, or antidepressant prescription) was significantly lower in ambient-scribed visits compared to unscribed visits (aOR 0.83, 95% CI 0.72-0.95).
• No significant difference in intervention likelihood was found between human-scribed and unscribed visits.

Conclusions:

• AI-driven ambient scribes in primary care are linked to enhanced documentation of neuropsychiatric symptoms.
• However, their use correlates with a reduced likelihood of initiating psychiatric interventions.
• Further investigation is required to ascertain the clinical outcomes associated with these observed changes.