Draft replies to patient portal messages
An agent drafts replies to routine patient portal messages for clinicians, who edit and send them or write their own.
The use case
Patient portal messages have grown sharply, and clinicians answer many of them after hours. A drafting feature in the EHR reads the incoming message and relevant chart details and proposes a reply in plain language. A Stanford study found clinicians felt less burdened when using drafts, even though reply time did not drop much. The clinician still reads every draft, corrects clinical content, decides whether the patient needs a visit, and sends the reply. Messages about urgent symptoms follow existing triage rules, not the draft.
Example workflow
- Patient sends a portal message
- System drafts a reply from the message and chart
- Clinician reviews and edits the draft
- Clinician decides if a visit is needed
- Clinician sends the reply
Tools
- Epic
- Microsoft Azure OpenAI
Risks and guardrails
- Clinical errors in drafts
- Overly long or generic replies
- Urgent symptoms need triage
- Clinician sends and owns every reply
Sources
Written by Agents Lookup. Credits and official pages for the tools this use case names:
- Stanford Medicine on AI-drafted patient message replies: med.stanford.edu/news/all-news/2024/03/ai-patient-messages.html