How Clinicians Are Using AI to Communicate With PatientsUTMB CIO George Gaddie and CMIO Carlos Clark Describe AI Assistant Program
The University of Texas Medical Branch recently began piloting an AI assistant integrated into the organization's Epic MyChart patient portal platform, and already clinicians are seeing efficiencies in responding to the large volume of electronic messages they receive from patients, said George Gaddie, UTMB CIO, and Carlos Clark, UTMB's chief medical information officer.
The AI tool drafts responses to patient messages, which clinicians can review, edit, modify -or even reject - before sending, Clark said.
"It allows us to be more efficient in our workflows responding back to patient messages," he said. "As you can imagine, after COVID, a lot of in-basket messages have dramatically increased because a lot of our patients have become very accustomed to utilizing the patient portal to help answer their questions or address any other medical needs," he said.
The AI assistant helps clinicians turn around responses to patients in a quicker fashion. "That helps reduce our physician burnout but also be more timely in getting back to our patients, which is, of course, our goal, to have better patient outcomes."
Some clinicians receive up to 100 messages weekly from patients, and the time saved with the help of the AI assistant can add up to hours, Clark said. "Our goal with this tool is to help our patients get responses quicker, as well as help the providers not have to type long messages back to patients."
The AI model suggests a response to a patient message, Gaddie said. The AI assistant is currently used only when clinicians are responding to initial messages from patients - not for carrying on longer conversations or to answer follow-up questions in the same thread, he said.
The AI tool within Epic's MyChart is embedded in the system and teaches itself as it processes patient messages. It learns from whether clinicians reject or edit responses generated by the assistant, Gaddie said.
"That clinician feedback goes back to the engine and learns whether a response that was generated was not acceptable," he said. "And it will help continue to tune itself. So, the more people, the more physicians that start using the tool and start providing the feedback, the better the responses will become."
UTMB is an academic medical center with four major campuses, six hospitals and about 100 clinics across southeast Texas. The organization also provides care for about 80% of the prisoners in the state of Texas, Gaddie said.
In the joint interview (see audio link below photo), Gaddie and Clark also discuss:
- More details about how the AI assistant, from Microsoft and embedded in Epic MyChart, works;
- Privacy and security considerations with the AI assistant;
- Promising uses of AI in other areas of healthcare, including in pathology and in helping clinicians keep up with evidence-based medical research.
Gaddie is vice president of information services and CIO of UTMB. He has extensive experience in clinical information systems and in leading large teams to ensure quality service and a reliable IT infrastructure. Gaddie is a certified healthcare CIO and a member of several industry professional associations.
Clark joined UTMB in 2008 after completing his residency at UTMB in Galveston and his medical degree from the University of North Texas Health Science Center, Texas College of Osteopathic Medicine. He was appointed the medical director of the internal medicine residency clinic and has been engaged in clinic process improvement to provide quality care for the patients and the best learning environment for residents.