Integrated into the patient portal, the tool asks the person about their symptoms and then writes a more complete message, which they are free to modify or ignore before sending.
Vanderbilt Health has deployed its first AI tool aimed directly at patients at scale, making it one of the first major U.S. teaching hospitals to go beyond the simple chatbot backed by the portal. The rollout began on April 1 at One Hundred Oaks Primary Care Center, in a phased approach.
Asking a better question, not getting an answer
The assistant is activated in the My Health at Vanderbilt portal when the patient selects "medical question". It then asks follow-up questions adapted to the reason — symptoms, questions about a treatment, postoperative follow-up — and then writes a more detailed message that the patient reads before sending. The stated objective is therefore not to answer for the caregiver, but to improve the quality of the information that reaches him. The patient retains the possibility of sending his message as he wrote it at any time.
Technically, the tool is based on a large OpenAI language model, which runs in the Microsoft Azure environment already used by the institution for its clinical applications. Vanderbilt's triage protocols, referral guidelines, and nurse triage repositories are injected through augmented generation of recovery. The rollout is to be scaled up to other consultations, until the assistant is the default messaging experience.
Read the full article on the original source
Go to the source →