Neil Thornbury, MD (Host): Welcome to Pulse and Perspectives, the show that brings you the heartbeat of healthcare innovation in the vantage point of healthcare industries. I'm your host, Neil Thornbury, CEO of T.J. Regional Health. In today's episode, we're venturing into the realm of artificial intelligence, AI, a transformative force in healthcare.

We have the perfect guide for this journey, Dr. William Thornbury, a local primary care physician with a robust background in AI and technology. Join us as we unravel how AI is reshaping healthcare and the profound impact it promises for our future.

Now, the term artificial intelligence has been very mainstream here, especially in the last 12 to 18 months. And certainly there's a lot of conversation around that. So Dr. Thornbury, if you will, just take a few minutes and tell us a little bit about yourself and tell us what the term artificial intelligence means just in general.

William Thornbury, MD (Guest): Well, thank you Neil, for having me. I do admit to having the same parents, but I'm sorry to say that you and I rarely, rarely get to see each other anymore. But this is a good venue for it. My background was initially in pharmacy and then in medicine, and uh, I'm very fortunate that one point you asked me to look into the Lean Systems program at UK.

That course, is run by Toyota, kind of the grandparent of Lean Systems and Lean Thinking. And that brought my appreciation of how health systems work and drove me into telemedicine which is a place I thought I would never be. And just trying to lean up the efficiencies of our primary care office.

That work that we did there was kind of groundbreaking work. We in Kentucky built the first mobile to mobile telemedicine system in the world and tested all that in rural Kentucky in 2009, 2010, just after the iPhone came out. That has kind of kept me in the neighborhood of Ongoing Technology, and I do get the opportunity to serve a few companies as an advisor in technology.

And about 20 months ago, I think a lot of us that were paying attention to the field, began to understand that there was an emerging work in artificial intelligence that has really blossomed in the last, I think, 20 months. And for our guests that aren't really familiar with artificial intelligence or AI, you could think of that maybe as kind of a machine generated automation of tasks, learning, conceptualization and expression.

And a good way to think about might be, what will be the next step of the internet? You know, for us, we're all in, in health informatics, we're moving information from one place to another, assessing information, and these engines, these computers help us do our work. The artificial intelligence now serves as, this has been going on since about 1950.

Again, up until about 2012 or 2015, very, very poor progress has been made, but as we've come into these, what is termed generative artificial intelligence, or machines that are predictive, that is to say, when they're trying to predict the next word, the next thought, the next paragraph, if I say left, it will think right, if I say up, it'll think down, and in doing that, it has become more helpful for many of us in the real world to do our work, and what's really been dramatic about artificial intelligence is what is termed emergent properties, where these engines now, as they're to develop more and more computational power; they're developing these new traits that weren't built in to the system, and I think that's caught us, caught all of us in this industry off guard. And I'll give you an example. You might build a system that's what we call a large language model, which helps maybe do your homework. It'll write Shakespeare for you.

It can write your reports, but you may ask it, well, can you do mathematics? And it says, well, I can't do mathematics. You build up the computational power a few thousand times, and all of a sudden it starts doing mathematics. And it learned it in a way that we don't learn. It learned it differently. And that's one fine example again, of how these machines and how these systems are going to become online and help us think and understand our world better.

Host: Yeah, it's a great example. And I tell you one that out of healthcare that really where I started my interest like yourself into artificial intelligence, and although at the time we weren't really necessarily coining it artificial intelligence like we are today, but if you remember back in 2016, when Google DeepMind developed AlphaGo and played a Go player, which is a board game called Lease It All and at that point, what they did is they tried to develop this, AlphaGo AI machine learning program, or even project. And again, they used it on board game Go, which is a game that's 4,000 years old with a lot of complexities. And they were just trying to develop and see how it would learn. And that's really where I started taking a lot more interest in artificial intelligence.

So it's pretty interesting how it's evolved and then really taken off in the last two years. So I understand January of this year, Dr. YThornbury, that you attended a symposium in Washington regarding AI. Can you try and offer a few of the insights of the conference just for the listeners and what are you hearing out there?

William Thornbury, MD (Guest): Well, my role was serving as president of the Kentucky Medical Licensure Board, and so as you can imagine, with something this powerful is what we're seeing today. The government was going to have an interest in this, and so they asked all the states and territories to come together in D.C. to have a symposium to try to understand where we are on this and maybe where we're going. My insight immediately was there's a gentleman that led the symposium and said the executive order 14110 came out and is now fully implemented. And it said the government is not only going to oversee this, but there are going to be very substantial guardrails.

I think my take home from the medical side is, you know, number one, how's AI going to work for us and for us, I think it is more of a consultant. It's going to render an opinion, and I think presently in the near term, the physician's going to have to review that opinion to decide, well, with that piece of information, how am I going to move the case forward?

I think we'll be working with this as a part of informed consent. I think when we work with the families, we'll say, well, just like we want to do a procedure on our patients to try to improve their health, we're going to say, well, we're going to work with this type of consultant. It's an artificial intelligence. We want your permission. We want you to fully, be fully transparent how we're doing this. I think one of the other concepts was what we call explainable AI, which is a black box saying, well, when we use algorithms or systems, we might put in, say three or four or five or six metrics and come out with a decision.

And, of course, you can understand the metrics because you're placing them in. Well, when AI renders an opinion, sometimes it renders an opinion and it doesn't tell you why it rendered the opinion. And it may be dancing with a number of variables, but those are not always clear. And I think what we're going to have to do as clinicians is understand what the variables it's using, what the weight of the data is.

For example, if it's looking up all the information on some type of breast cancer, well, is it using all studies? Is it using just studies with a p value less than .05? Are these studies in the last 5 years or 3 years or 25 years? So what we need to understand is what's the black box that we can share with the family when this information is being generated to us, what's it going to mean to us?

I think as medical boards, what we want to understand is presently, we're under statute in Kentucky to regulate only physicians and clinicians. We don't really regulate artificial intelligence. And right now, the FDA is not regulating artificial intelligence. It's not said that it's a medical device.

So is there some type of shared responsibility? They brought in an attorney that really I found very intriguing. And the attorney kind of said, well, we've polled potential juries, and if the artificial Intelligence recommended something that was within the standard of care and the clinician disagreed, there might be a jury problem for the clinician.

But on the other hand, if the artificial Intelligence says, well, I disagree with you and the clinician is recommending the standard of care, potential juries might support that. And again, these cases have not been brought forward. They will come up. But, it opens just a whole bag of questions like, what is the standard of care?

Right now, our standard of care in medicine is changing every 42 days. But these intelligences I hesitate to say they're logarithmic, I think they're more hyperbolic. These things are going to move so far, so fast, so quickly, at some point I think a lot of us believe that this will become the standard of care.

It's certainly not there now, but when it does, what's that going to mean for the work that we do? And for us, we need to not only know how to use it clinically as a consultant, but what are the shortcomings of the artificial intelligence? For us to weigh that position that it's rendering, we need to know, again, what's inside the black box and where are the weak points of the argument.

The last thing that kind of caught my attention was that can you imagine a person, for example, say a primary care clinician that started rendering for example, opinions in say cardiology or infectious disease based on artificial intelligence? And this is outside their scope of practice. We don't really, at this point, the boards of medical licensure have not seen these cases, so we've not adjudicated them.

But I think that we're watching them very closely. The last point that I will make before I close the conference idea was just the importance of moving medical informatics and the use of artificial intelligence into not only the medical education, but not only the education that we're receiving clinically.

And I think that what you're going to see is for us, how are we going to use this in a way where we can be helpful to our patients and competitive in the market? And you can imagine that if we didn't have the internet, what would a hospital today look like without the internet?

What would a health clinician or health system look like in three to five years without the internet? And it's going to move a lot faster than the internet has moved in 20 years. It's going to move in a matter of probably years and two years ago, we would never have a conversation like this.

And now you can imagine we're talking about these systems. We can see the future where they are going to be smarter than we are and how to deal with that.

Host: So let me ask you just on utilization, and again, when you talk healthcare and you talk, and from my perspective, hospitals and hospital systems, I think a lot of what's out there right now seems to be very vague on where do we start? And I always say, if you want to know where to start, ask yourself, where are the biggest opportunities and what are the biggest issues you have, and then try to start using and leveraging resources to solve those problems.

And I feel like AI is going to be a solution that's going to be really strong moving forward. But, you know, again, where is AI today? You know, I see it in radiology. I see it in administrative tasks, scheduling, phones. I see everybody kind of dipping their toe into different parts of healthcare. What do you see is really kind of a utilization that is, that you see coming up here in the near term, or what's going to be a bigger impact maybe in the longterm.

William Thornbury, MD (Guest): Well, in the near term, I'll give you an example, in Washington, there was a clinician from Tufts University in Boston, and what they do is they use artificial intelligence to basically clear their ER. So they have a lot of trauma, they have a lot of cases that come in that require computed demography imaging.

And the AI very easily, because this is data, it can look at the image, it can say, with very high certainty, this looks clear, there doesn't look to be a problem, it's going to be over read eventually, but these cases that are non problematic and don't need to be put in front of the radiologist can be told to the ER physician within a 99 or 99.9 percent probability that there's not a problem and let the ER clinician decide how to handle that.

So, they're moving those cases out to lean up their ER so they can pull the sick people in. Now, eventually, all these cases are over read. In our practice, what we do is we use it as a supportive consultant piece. Again, we want to remain constant with the standard of care. We want to remain constant with the current literature.

And so, we might use this as an opinion to say, well, we want to look this information up. We want to look it up effectively. We'll use systems that bring us the standard of care, that bring us the guidelines, that tell us what literature supports these guidelines. We use it for something as complicated as that.

On the other hand, when I worked with patients just a few minutes ago, we brought up a piece of technology that's free, it costs no money, and this is a patient that was a new diabetic, well educated person, and we used to refer this maybe to a diabetic educator. We would try to have them buy a book or two, maybe look, watch a video, and what we can do now is we can use these technologies to, to not only explain diabetes in a very simple, like, third grade, fourth grade, fifth grade manner, but we can also get these technologies to build a one week diet plan, to offer recipes for certain days, to offer different snacks because diabetes tends to fail commonly in the supermarket.

We can get it to give us a grocery list and this costs no money for these patients. And it's really just about like having a dietician with you. If I can't have one with me, it'll be the next best thing. You asked me a direct question, I'll give you a direct answer. How would I use this as a clinician or as a health system?

The first thing I'd want to do is, I want to look at every day, where am I doing repetitive work? What's my simple repetitive work that maybe just might take a few minutes a day? But you might do that 10 or 20 or 30 times a day. And so, say if you were in a private practice and you saw a patient and wrote a super bill and you gave the super bill in and that was faxed to the coder.

Well, you can imagine how AI might take that super bill and be able to look at it, visually interpret it. It might pull in an ICD-9 code and code that for you, and then have a human oversight just to review that code, and that may end up say in any given week, many, many, many hours. So just like lean systems, we look at what's in front of us, what we do every day, what's something that we can simplify so we can make our humans more efficient to do higher complicated tasks.

Host: So 60 seconds, any last thoughts on what you want to leave the audience and the listeners with?

William Thornbury, MD (Guest): Well, my take home is that in healthcare, clinicians that are using artificial intelligence and use it properly, health systems that are using artificial intelligence and using it properly, for better patient care and line systems are going to outperform systems and clinicians that don't. And they're going to have better outcomes.

It's not even going to be close. And I think that what we want to try to do is we want to try to engage these systems early. And if you don't know anything about it, like, there's probably 50 percent of the country that doesn't know really anything about AI. And we want to begin our education about what is artificial intelligence?

What's it going to mean to us? What are examples that are simple systems that are AI and how can we use this to become more competitive in the marketplace, but also how can we be better clinicians and serve our patients and our community better?

Host: Great thoughts, truly. And I, as I say, today's the best day to be in healthcare and AI really is bringing a lot of excitement and energy. So really take home. We've got, AI first and foremost, it has to be used safe. It's got to stick to the standard of care and it's got to be helpful to the patient. There is a lot of different clinical uses and administrative uses that are out there. And I think we just have to keep watching those and see what the studies produce. So we are applying it in a very safe way. And I think we both can agree that it's going to have a major impact moving forward. And those that leverage AI, are going to really exponentially outperform those that do not.

So thank you for tuning in to Pulse and Perspectives. It was a pleasure to have Dr. Thornbury with us today sharing his insights on the exciting intersection of AI and healthcare. We hope you found our discussion enlightening and that it sparked some ideas on these innovations that might shape the future of your healthcare landscape.

For more discussions like this, make sure you subscribe to our podcast on your favorite platform. Please share on your social media channels. I'm Neil Thornbury and until next time, keep your pulse on the future and keep your perspectives wide open. Thank you very much.