Could Generative AI Redefine American Medicine?

Welcome, everyone. I'm Dr. John Whyte. I'm the Chief Medical Officer at WebMD. Everyone's talking about artificial intelligence. Is artificial intelligence going to make the doctor, patient relationship better? Is it going to make it worse? People are talking about ChatGPT. But only about 20% of people have actually tried it. Is that going to help? Or is that going to hurt?

My guess today says artificial intelligence can actually improve the physician patient relationship, but we have to utilize it in the right way. There's a new book out. It's entitled ChatGPT MD: How AI empowered Patients and Doctors Can Take Back Control of American Medicine. Dr. Robert Pearl, thanks for joining me today.

John, it's a pleasure to be with you once again.

I want to start off with, and taking a step back, what's wrong with the physician patient relationship today that you say has to change? And you've chronicled it in your new book.

I'm glad you're starting there, John, because as physicians, we always want to make the right diagnosis before we recommend the treatment. And in this particular situation, I believe that the biggest problem is that the amount of work that physicians are being asked to do today is simply more than they can usually accomplish. And if that's the challenging you have, what do you have to do?

You have to shorten the visits. You have to keep it narrowly focused on whatever the acute problem is. And you don't have the time to develop the kind of personal relationship that we know existed in the past before you and I were practicing medicine. It's been estimated it would take a primary care doctor 27 hours a day to fulfill all of the requirements, which is asking too much of clinicians today.

But Robbie, I have to say, you take a very positive view in your book of the potential role of AI. But let's be honest, there's a lot of doctors out there. There's a lot of patients out there that are fearful about it. Is it going to make mistakes? Is it going to make the relationship worse? How can a chat bot actually do a better job in terms of answering questions about medical issues? Those are the things that people are saying. So why do you have this very optimistic view, and many other people simply do not?

I can't speak why other people do not. But if I had to guess, and this is part four of the book, ChatGPT MD, they have concerns about privacy and security and bias. And of course, the medical errors, the hallucinations that exist today. And they are right to be concerned about each of those pieces.

I think two perspectives I'd like to give. The first one is, in some of those areas, like privacy and security, will ChatGPT make what exists today-- going to a Google, clicking on a link-- will it make the privacy, the security, the bias worse or better? With misinformation, intentional misinformation worse or better? I think it's going to make it a lot better for all the reasons I explained in the book, because it's not an advertising based model. It's a subscription model.

And a subscription model is not a question of getting people to go to-- people paying for the advertising, it's providing the service optimally to the user, who in this case, is the patient. ChatGPT is doubling in power every year. Five years from now, it'll be 30 times more powerful. I tell people, patients, don't

use ChatGPT today to make a diagnosis of managing chronic disease. Learn from it. Figure out how to use it. Think about what it's going to be like in five years. I'm looking five years out, and they're probably looking at what exists right now.

I'm looking one year out, saying, how's it going to change in one year? Because everything is moving so quickly. Can you give us two examples from your book in how you think AI is actually going to improve that physician patient relationship?

I think the conclusion is that it will take 20% of what physicians do today away from them, freeing up their time. I'm receiving emails from clinicians across the United States who know the book's coming out. And they say, Doctor Pearl, you know, we heard how good it is if they can diagnose you. I put information in on my patients. I get great answers.

I was on a podcast last week with a woman who was talking about her husband who had a skiing injury. And I said, you know, why don't you go into ChatGPT, put all the systems in place and see what the diagnosis is going to be? And it came up with the exact diagnosis of a rotator cuff injury, the recommended diagnostic maneuvers and the probable surgical treatment that he'd be requiring for him.

So I think you have to understand the specifics that are there taking the work away. So where can that work be taken away? The first thing to me is chronic disease. You know, we see patients in the office three or four times a year for a problem that exists every single day. We have wearable monitors that you had on your show before, and no one's really using them. And why aren't they using them? Because right now, the data coming off of them would have to go to your clinician. I don't know any clinician who wants all that data.

But if it could be analyzed and looked at by ChatGPT. And every day, or frequently, patients told how they are doing. Imagine how much better that care could be. 400,000 people die annually from misdiagnoses. I'm not even counting the medical errors or the failures to control chronic disease. Pure misdiagnoses. Will this improve it? And I think that is the mindset where I look at that. I believe that we will be able to do much better by general AI with the clinician.

So what's the biggest obstacle today from integrating generative AI into the health care process? Is it money? Is it resources? Is it lack of acceptance? There's a component of that, as well. So what's that biggest obstacle? What would you change with if I gave you a magic wand? What would you use it for?

I can answer that question very straightforward. And you said it a little while ago. It's the reimbursement system of American medicine. Because as long as we're paying transactionally on a fee for service basis, the more you do, the more you get paid. If the patient does it, you don't earn any money. This is not going to happen. But we have to shift to a system that rewards clinicians for keeping patients healthier. And we shouldn't see taking someone with a myocardial infarction, unblocking the coronary arteries as success, when it could have prevented them from blocking in the first place.

And the data says-- I talk about the Kaiser Permanente experience-- you could lower the incidence of myocardial infarction 30% to 40%. Strokes, 30 to 40%,. Certain cancers, 30% to 40%. And if we can eliminate those, think about the dollars it frees up to invest in a variety of ways. So you're absolutely right. As long as we have a transactional service system, doctors are not going to be that interested in applying it. The day we shift to capitation, this technology now becomes more positive and powerful.

I want to push on that, because you ran a physician group, as you mentioned, in your role at Kaiser Permanente and Permanente medicine. Do you think most physicians are accepting of generative AI?

The physicians that I talked to are excited about it. It's not what exists today. I want to keep going back to that. But I also want to watch how fast it's moving, what's called plug-ins. So plug ins, you know, you can think of it as like a plug-in, literal one. It's really not that, it's a Bluetooth connection, a variety of ways of getting information to the system. When I wrote the book ChatGPT MD, I wrote it with ChatGPT intentionally. And the chapter that I had ChatGPT write by itself, I labeled it as such, so the reader would know what I wrote and what ChatGPT--

That's important. Transparency.

Well, I started the process by downloading the 1.2 million words that I had published on a variety of locations, and that was laborious. Six months later, that's how long it took me to actually complete this process. And I did it so fast because I knew that by two years from now, anything I wrote would be out of date. That process now is facilitated by plugins.

When I talk about wearables, wearables can connect into the application using plugins, they're called GPTS. But they're basically, for listeners and viewers, think of it as a plug-in that you're able to accomplish. And now, what has come out? It's called memory, which means it now can remember those things. And I also want to point out the NVIDIA-- which is the major chip, it drives all of the chips that drive all of generative AI-- has just announced that it has a nurse AI bot that is able to make diagnoses and be able to manage chronic disease.

This is moving so fast. And by that, I don't mean moving fast in a way of expanded applications with the old technology. The technology is moving faster than we're able to move. Getting more reliable, continually. I think there still going to be problems that need to be observed and figured out. But I think that this is the opportunity to be able to now shift and improve quality, making care more convenient and care more affordable.

And you point out, these are iterative tools and technologies, and we need to be engaged to continue to improve upon them. Full disclosure, I've known you for many years. I won't say how many. You've written a lot of books over that time. And what I love about your books are you always have a specific goal in mind, which is about impacting the healthcare system. I want you to tell our audience, what's your goal with this book? What do you want to accomplish?

I'd like to accomplish three things, John. The first one, is I'd like the process of transformation to be led by clinicians, not by corporate entities. That's why I have the subtext there. And I'm worried it's not gonna happen, for the reasons that you point out. Because there are things that have to happen transforming systems of care that physicians are going to say, I don't have the time.

And we also know that the cognitive bias is against loss. They're going to say, I'm afraid of my income. I'm afraid of my work date. Important concerns. But as a result of that, someone else will take the lead. And I don't think that's going to be as good for the providers of care or the recipients of care. So that's the first thing that I'd like to do.

I think the second thing is that I want people to be able to understand where it's going. I would like clinicians, and I'd like patients to understand this too, and not to fear it. Not to see what we have today, to see what's coming in the very near future. So that's the second part that is there. And the third goal that I have is I feel really bad right now for both doctors and patients. We know that burnout is over 60% for clinicians. We know that 70% of patients say they don't like the system of health care. And I think that this is a tool that could work.

But what I'd love to see, John, as I think I'd like to see every society starting right now, having a group of people experimenting with it, not applying it yet, clinically. It's like everything else we do in research. Testing it. Seeing the answers that come out of it. Trying to figure out the prompts. And the more information you provide, the better it's going to be. And why should every clinician have to figure it out on his or her self? Each specialty could be different, but the tool is the same. So those are the three things, John, I would love to see happen coming out of this. Because if we don't do it, I do not see a solution anywhere on the horizon.

Your book How AI Empowered Patients and Doctors Can Take Back Control of American Medicine. Doctor Pearl, it is always good to see you. Thank you for taking time and really leading this thoughtful discussion about what can transform medicine. So thank you.

Thank you, John. Anyone wants more information, my website is Robertpearlmd.com. And you can go there and get information about generative AI or any of the other topics that we've talked about. And John, I really appreciate your leadership of American medicine. And I'm confident you'll be part of the solution. And I look forward to working together with you once again. Thank you so much for having me this morning.