Imagine you are in kidney failure and need a transplant. You likely have a lot of questions, maybe more than you can cover in your next appointment with your doctor. Or maybe you don’t know what to ask.
Soon, these initial conversations could be done with an AI-powered avatar or chatbot that looks and sounds like a human.
Two studies with faculty leaders of the Vanderbilt Center for Transplant Science show the efficacy of having a virtual conversation with an avatar or chatbot. In the latest version of one of the interfaces, patients can even choose the race and sex of the avatar.
“In the future, if you or I make a doctor’s appointment, we’re not going to meet the doctor first,” said Peter Reese, MD, Director of the Vanderbilt Center for Transplant Science and first author of the study “An Interactive Virtual Nephrologist to Engage Patients with Advanced Chronic Kidney Disease About Transplantation,” recently published in the American Journal of Kidney Diseases.
“Before we meet the doctor, we’re going to meet the virtual clinician,” Reese said. “The job of that virtual clinician will be to do an intake and figure out what our priorities are.”
In Reese’s study, researchers developed a virtual transplant nephrologist, or kidney doctor — an interactive, video-based platform designed to personally educate patients. Meeting the virtual doctor would be as easy as clicking a link on a smartphone or computer.
Reese said that when patients find out they need a transplant, it is a very emotionally charged time. They may freeze up or not understand what questions to ask. But interacting with an AI-powered avatar gives them an opportunity to explore topics in an unhurried, private environment. They may say things to an avatar that they wouldn’t in a doctor’s visit.
“When you meet with the virtual nephrologist, it looks like you’re on Zoom,” Reese said. “It looks like you’re talking to someone who’s very photorealistic, with gestures, eyebrows moving, the whole thing.”
About 150 participants completed the study. Participants were between 18 and 75 years old; the median age was 66. After choosing the preferred sex and race of the virtual doctor, patients could choose from among 11 topics and could pause, change topics and request more information. Participants spent a median of 12 minutes interacting with the virtual clinician.
Female participants were much more likely to choose a female avatar (79%), which is consistent with earlier studies that show women more often select female providers in practices where both male and female providers are available. Only 46% of male participants selected a male virtual nephrologist. Most Black participants (71%) chose a dark-skinned nephrologist, whereas 55% of non-Black participants chose a light-skinned nephrologist.
“These data suggest that customizable virtual encounters may help satisfy preferences that are often unmet in usual care but are feasible to deploy efficiently with AI-based software,” the authors said.
The study found participants were significantly more likely to speak to a human clinician about kidney transplantation after the virtual encounter.
The study was conducted with a virtual clinician that was limited to specific topics. But Reese and his team have now developed a 2.0 version that is integrated with a large language model, allowing the avatars to answer kidney-related questions that patients ask. That will be the focus of a future study.
Reese is not alone in studying this area at Vanderbilt Health. In a study published in 2024 in the Journal of Community Genetics, first author Elisa Gordon, PhD, MPH, Professor of Surgery, led a research team that developed a culturally appropriate chatbot for people of African ancestry undergoing testing for live kidney donation to consider Apolipoprotein L1 (APOL1) genetic testing.
Her study was based on her earlier research finding that transplant nephrologists order genetic tests but were not trained to provide the genetic counseling ethically required before ordering the test and again after test results arrived. Ethically, Gordon explains, people undergoing genetic testing need to receive counseling to make informed treatment decisions. The chatbot could help bridge that gap and help integrate genetic testing into clinical practice.
“To help with that integration process and to offload some of the work from the transplant nephrologist, I created a conversational chatbot,” said Gordon, Associate Director of the Vanderbilt Center for Transplant Science.
Gordon, her collaborators, and her community advisory board developed the chatbot’s scripted responses and conducted 10 focus groups to obtain input on the content and design from 54 participating living kidney donors and kidney transplant recipients in Chicago and Washington.
Participants interacted with a text-based interface like a text conversation on a smartphone. They found the chatbot “informative about APOL1 genetic testing, neutral and unbiased, trustworthy, and culturally targeted to individuals of African ancestry,” the study found. The interface allows patients to email a copy of their chatbot conversation for future reference.
Gordon has finalized her study examining the impact of the chatbot on potential donors’ donation decisions as part of her broader clinical trial published recently in the Clinical Journal of the American Society of Nephrology. She has since developed, with co-investigator Zhijun Yin, PhD, MS, at Vanderbilt Health, a new beta interactive (unscripted) chatbot on lifestyle behaviors to protect kidney health that taps into a large language model. Gordon and Yin plan to refine the chatbot to ensure that ethical guidelines govern the AI interaction, as an ethical model for other chatbots and AI broadly.
She emphasized that chatbots are not a replacement for talking with your provider; rather, they are a supplement.
“That said, people do trust their chatbots,” she said. “It’s remarkable.”
