Sam Chang: Hi, my name is Sam Chang. I'm a urologist in Nashville, Tennessee at Vanderbilt University Medical Center, and I'm truly delighted to have a future superstar in urologic oncology, Dr. Avi Baskin. Dr. Baskin did his residency at UCSF and completed his urologic oncology fellowship at Vanderbilt and is now an assistant professor at University of California, Irvine. We've asked Dr. Baskin to give us some insights on a program that he helped develop in Nashville and that he's bringing to the West Coast, looking at patient counseling and the utilization of AI. So Avi, thanks so much for spending some time with us.

Avi S. Baskin: Thanks, Dr. Chang. I really appreciate being here. One of the things I think we all face as clinicians is there's such a drain on our time with the amount of patients that we have to see, especially at these very high-volume centers. I think one of the things that we really wish is we could give more patients more time, and sometimes visits are limited to only a few minutes just because of the volume of patients we have to see. But the problem with that, especially in urology, is the amount of information and counseling that we have to deliver takes a lot longer than the time that we have.

A lot of the time we spend giving the same information, the same spiel over and over again to each patient about prostate cancer or vasectomy or any sort of thing that we commonly treat. We were thinking of ways and solutions that we can work with to try to alleviate the situation. And I think using artificial intelligence the way we have offers a solution to provide patient-specific education to each person that walks through the door and increase our efficiency at the same time.

Sam Chang: Well, there's no question, our templated notes sound like we have spent an hour or so with each of our patients as we go through the different variations of therapy, side effects, pros and cons, et cetera. But you're right, the quantity of information that we try to deliver in a focused period of time can be overwhelming sometimes, can be confusing many times. And so tell us a little bit about how you came about the nitty-gritty of this project, and then maybe you can show an example.

Avi S. Baskin: Yeah, so here's what we did at Vanderbilt. We ran a pilot study with 10 patients to test out this technology. And what the technology is—it's basically called an AI avatar, and it uses this artificial intelligence technology to basically make a version of you as Dr. Chang or Dr. Baskin, so that version looks like you're having a video call with someone.

The idea is that the patients can interact with this AI avatar version of Dr. Chang prior to the clinic visit. And what they can do is, let's say they're coming in to talk about prostate cancer, for instance. They can say, "Hey, what is prostate cancer? What does the prostate do? What are the treatment options?"

So when they get to the actual visit, they've had a chance to ask all of these questions, get informed by an avatar that's trained on prostate cancer-specific knowledge that is chosen by the clinician. And then that way the patients come to the visit with a much higher level of education. That's the goal at least. When we ran this study, we found the patients loved interacting with the avatar, but there were definitely some improvements that need to be made, and I can get into those.

Sam Chang: And the information that initially was formulated with this avatar was information that we vetted in terms of guidelines, et cetera, nothing straight from a Google search or anything. I mean it may have included some, but obviously data, evidence that we thought was accurate, up-to-date and allowed hopefully some nuanced discussion as well.

Avi S. Baskin: Correct. And when we think of—people may have used things like ChatGPT or whatnot—those resources are essentially trained on the entire internet. I don't want to oversimplify it, but the entire internet. When you ask a question about prostate cancer into one of those, you'll likely get a very good answer, but sometimes as a layperson, it's hard to distinguish what applies to you and what doesn't. What we've done with this is the avatar is only trained on—it's basically a long document of curated information in kind of a question-and-answer format about what prostate cancer is, common questions patients would ask—so that when patients are interacting with this avatar, it's not drawing from anything else; it's drawing from a specific set of information so you know it can only go in those directions.

And furthermore, it has specific instructions. So we basically told the avatar a few things. We said if patients ask to be prescribed Viagra or to prescribe any medication, it says, "Hey, I can't prescribe anything."

And then the avatar is also trained specifically for Vanderbilt. So a lot of times patients will ask, "Hey, where do I park? What can I eat around campus?" And the avatar is trained to know all these things to hopefully assist them and decrease the call burden on all of the staff.

Sam Chang: Well, I mean, I think just like a picture is worth a thousand words, let's see an example. Avi, do you have one that we could share with those on UroToday to see what we've got set up?

AI-generated Avi S. Baskin: Hi, UroToday, I'm Dr. Baskin. Well, I look and sound like Dr. Baskin, but I'm actually an AI-generated avatar. I've been trained with accurate, standardized information about prostate cancer. My role is to answer common questions before a clinic visit so patients arrive prepared and ready to focus on what matters most: their goals and preferences. For example, many men ask me, "What are the main treatment options for prostate cancer?" I explain that options can include surgery, radiation, or sometimes focal therapy. Each approach has its own risks and benefits, especially related to urinary and sexual function.

My job is to give patients this foundation so when they meet the real Dr. Baskin in person, the conversation can go deeper into what matters most to them. Our early work showed patients were curious and engaged, but the biggest opportunity is what comes next. Imagine if a patient could log back in after their visit to ask follow-up questions they didn't fully understand or review the visit in their preferred language to increase access. Imagine if the avatar could take a detailed medical history before the appointment, so more of the clinic time is spent on patient values and shared decision-making. That's the future vision: expanding this from one moment in the healthcare journey to something that supports patients across every phase of care.

Sam Chang: So Avi, that was not you.

Avi S. Baskin: It was not me. So it looks like me, and I think for someone who hasn't met me, they might think it was me. There are subtleties, but I think the technology is quite good. And I think one of the questions that we toyed with was, should this avatar look exactly like the clinician? Is that good or bad? And some of the data and research that we reviewed show that it provides a familiar face and makes them seem more human, and the literature would support that.

But one of the questions I actually asked each of the 10 patients in the pilot study was—because you could do this in a variety of formats. You could do it as a text message or email or you could do voice like a phone call. But the patients actually really liked the video portion of it, so I think it's important to listen to their feedback. There's still obviously a long way to go in development on multiple fronts, but I definitely think there's some really exciting technology here that can help our patients.

Sam Chang: That ability to have both a pre-visit discussion with the avatar, the ability to have a post-visit discussion with the avatar—you can see how you can really raise efficiency. You always worry about possible errors. And so tell me the next step. You completed the initial pilot, got patient information, feedback, et cetera. Where are you going to go next?

Avi S. Baskin: There are a few areas that I think are really interesting. The first is I want the avatars to basically be able to take a medical history successfully. And I don't think that's going to be incredibly challenging, but there are already intake forms that we have, things that we already do. And if it can take a medical history, so by the time you get to the point of the visit where it's answered the patient's questions about prostate cancer, it's taken a medical history and it can transcribe those things for the clinician to see, I feel like you're starting your visit at a much later point with that patient where they have a lot of good knowledge and information that they've gathered. You already have most of the medical history that you need from them, and it's written right there. You can confirm some things and then you really get to spend that time on patient preferences, what matters to them most, connecting with patients, doing the things that are enjoyable and also fun to do as a physician, which is connect with the individual patients. That's the next step.

But there are a variety of other things that I'm looking at too. As the avatar mentioned, the AI Dr. Baskin mentioned, there's so much potential in the post-visit space, meaning patients go home, they get a new diagnosis of cancer—a lot of times they don't exactly remember what happened at the appointment. They have a spouse at home who wants to know what happened. What if you could log back in and you could link it to the medical record and the patients could say, "Hey, you said I needed surgery. What was that process like?" Or, "Could you explain to my wife what a prostate biopsy is?" That's one thing that I think would be really good.

The other area for this that I think would be excellent is if you're getting all this medical information prior to clinic appointments—and one of the big issues, especially at big tertiary centers that we face, is urologists, and this goes across the medical disciplines, but urologists are so specialized, even in urologic oncology. You may have subspecialization, especially in prostate or bladder cancer. And from the outside when you have these referrals coming in, it's not always clear which doctor the patient should go to from an institutional standpoint. But let's say you have these patients interacting with avatars and getting this medical information—you can create a system upfront where the avatar is able to screen and say, "Hey, this is a patient with a history of prostate cancer," but their actual problem is kidney stones, and those inefficiencies in the system could be addressed in this way. So there's a lot of potential here and there's a lot to develop, but there are so many directions we can go with it.

Sam Chang: That leads then to my last question here, Avi: all right, where do we go next? I mean, all the ideas that you mentioned, just right away, it's like, "Oh, that's so appealing. Oh, that would be very helpful."

Not only for patients, because obviously patients get somewhat disheartened when they see someone and that someone—being me—basically says, "I really don't treat large kidney stones," or "These recurrent urinary tract infections are better treated by one of my partners," et cetera. So what's the next step?

How can we actually start integrating this into clinical practice? Because I think that's going to be one of the questions that everybody has. It's like, "Oh, this is a great idea. This is what my car service center asked me to do—fill out all these things, ask me questions." How do we actually now integrate this into our clinical practice?

Avi S. Baskin: Yeah, it's an excellent question, and it's challenging. I mean, some of the things that we ran into just getting this pilot study off the ground with existing technology—there are a lot of issues around security, patient safety, data safety, and then trust. As we know and as is out there, these AI systems are not 100% accurate. So we have to decide where do we draw the line on what the threshold is. But I think practically, testing this out in academic centers, we're working to build the next piece, which is the part that does the medical history as well as the patient education, and then working to build that and test it in patients, iterating, getting feedback, and then slowly rolling it out. I think one of the nice things about being in an academic setting is you can test these things without commercial pressures, and that's the next step. But there are so many avenues—I think we're trying to bite off one piece at a time.

Sam Chang: Yeah, it's very exciting, Avi, because you can see, just like anything, it's accelerating. There's no question that many more people feel more and more comfortable with this over time or at least have exposure, and so having this go beyond the handouts that we give to actually true interaction and being able to assimilate questions and then answer them accurately—just fantastic. I'll be very excited to see your next study as you integrate and try to bring this into fruition in clinical practice, and after the next iteration, reach out to us and give us an update regarding how much closer we are to using this every day in clinical practice.

Avi S. Baskin: Thank you, Dr. Chang. I really appreciate it. Thanks for having me on.