Health Affairs This Week places listeners at the center of health policy’s proverbial water cooler. Join host Jeff Byers, editors from Health Affairs Publishing, and guests as they discuss health policy’s most pressing news and trends.
Hello and welcome to Health Affairs This Week. I'm your host, Byers. We are recording on 11/09/2026. This is a last call for the insider event on November 17 next week with Jetson Leader Lewis of Boston University and the former deputy executive director of the White House Anti Fraud Task Force. He will join Chris Fleming from Health Affairs to get insiders up to speed on fraud policies and what to watch in the health care industry moving forward.
Jeff Byers:Today on the pod, we're talking AI, and to do that, we have David Brailer, founder and chairman of Health Evolution. David, welcome to the podcast. Thanks for coming on.
David Brailer:Good to see you, Jeff. How's everything?
Jeff Byers:Things are going well. Things are going well. You know, we're moving to the the cooler months here on the East Coast, so I'm excited about that.
David Brailer:Well, I'm not gonna tell you it's 80 degrees in California.
Jeff Byers:Well, yeah, well, I didn't hear it. So, you know, going back to even warmer months, in June of this year, you wrote in Health Affairs Forefront, a piece titled Why AI Will Accelerate Healthcare Inflation. So, we're gonna assume, for the purposes of this podcast, that listeners haven't read the piece. You know, we'll put a link in the show notes, of course, But briefly, can you explain the main thrust of your argument?
David Brailer:Sure. Yeah. And I wrote that piece because I'm kind of surrounded by AI tech utopians who believe very positive things about AI, which I am inclined to. But one of the places where I've had multiple arguments is this belief that it would create paradise in healthcare. And I took particular issue with this issue around cost.
David Brailer:And so, what I basically said was, AI will be inflationary, perhaps quite inflationary in healthcare, really for three reasons: economic theory, early evidence, and what we learned from electronic records. Let me just hit each of those. Economic theory basically says you get what you pay for. And we know what we pay for in healthcare is volume and intensity. And AI is likely to help us do more billing, more code finding, more capture of things that we already do in the industry that drive billing, drive revenue.
David Brailer:Secondly, there's early evidence. When I wrote that piece, there were two reports. There are now at least five or six major reports that show significant early evidence of AI driven billing costing billions of dollars more than baseline. And the forecasts for it are quite significant. The third was around electronic records.
David Brailer:Back when I was the first national coordinator for health information technology in the Bush administration, and we did the EHR push, many respected economists and scientists, including the RAND Corporation, had forecasts that showed that electronic records would save potentially 15 or 18% of health care costs. And that would have been at the time $300,000,000,000 give or take. 400 maybe. And the evidence that was accumulated over the next decade showed that it actually increased healthcare spending by about 2% per year because of billing intensity and other things. So my point was simply, AI is a magical technology, but it doesn't have the kind of magic that escapes the economic forces that surround healthcare, and therefore we put to use with more billing, more intensity, more code capture, etc.
Jeff Byers:Yeah. I wanted to ask a little bit more about that, whether AI will increase, you know, health care spending much like the same way EHRs did. But going... You know, really quickly going to EHRs, you know, you just mentioned you lent a big hand in helping, I would say, digitize the the doctor's office in in many ways. You sort of compared that experience to AI adoption in your piece.
Jeff Byers:You know, I do kinda wonder when you look at it, you know, from this spot in San Francisco of, like, you know, tech enthusiasts, you know, how do you see adoption between EHRs and AI comparing with each other, you know, decade later or two?
David Brailer:Yeah. Well, look, I think there's... They're really part of a spectrum because AI would be useless in healthcare or any other economic sector without the post Internet era conversion of paper records to digital. Healthcare, we called that electronic records, and manufacturing, it had different names. But, you know, it's a stream, because what did EHRs do?
David Brailer:They didn't create medical records. They changed them from written ones that are stored in files, or images that are stored in sleeves in a file room, into digital structured information that can therefore be shared, communicated, reviewed, compiled, etcetera. And what does AI do? Well, it takes digital data and turns facts into insights. It's, If you think about it as nothing else, it's a massive distillation engine that can find things, put it together, assimilate it, and draw conclusions, or at least assimilate it in a way you can draw conclusions.
David Brailer:So they're on a spectrum. They share that in common. Both of them fundamentally change how people work. You know, the nature of the electronic record meant sitting at a keyboard, which has become the torture for many doctors and nurses now, getting in the way of the patient interaction. But, you know, it changed how they worked, and it changed how they looked for information, how they made decisions.
David Brailer:And AI clearly will do the same by being able to take that work and going further. The best example would be when Google came to be, we were able to search for information in ways that we never thought we could. And now with AI, we don't even have to search. We just have to ask a question, and it will go find the answers and assimilate it. So they're quite similar in that.
David Brailer:I think there's one huge difference, which is that electronic records were and are highly regulated by HIPAA, by state law, by other reasons. And the reason for that is it is the source of truth in health care. The electronic record is the record of your health care, and there's significant liability or consequences if an organization gets that wrong, or loses it, or can't find it, or shares it with the wrong people. AI is the Wild West. It isn't healthcare what it is in the rest of our economy.
David Brailer:There are no rules. There are no regulations. There are massive data access things that are happening that would really go outside of a regulatory structure. So, they're quite different in how they're being deployed, and I think that will have huge impacts on how they end up being kind of used in the long run.
Jeff Byers:Yeah. Do you get a sense of how perception might be for doctors using AI to make diagnosis and treatment decisions? And the reason I ask this is because, you know, I'll ask, you know, ChatGPT or Copilot something about my, you know, blood count records or something along those lines, and I'm like, yeah, okay, I can I can ask, but, like, if I I wonder how I would feel if a doctor asked the same questions I'm asking the same program? I don't know. Any thoughts?
David Brailer:Well, first of all, I think you are evidencing what is happening with AI, which is many, many consumers are using it to ask, why does this hurt? What is this lab report? You know, what could be wrong that explains something going on with me or my child or my spouse or my parent? And that, I think, there are some people that believe that this will fundamentally alter the consumer role in the healthcare industry by people becoming more assertive. I remember when I was in clinical practice right after the internet era began, I was practicing.
David Brailer:I was an HIV doctor, so we treated AIDS patients, which were very, very intensively sick. And I remember when they started coming in with reams of paper printed off the internet about studies, or about treatments, or about there are various ideas for therapies that were being widely circulated, and this is kind of that. But your question about clinicians, I think last I just saw a statistic that seventy five percent of clinicians consult AI at least once a week with respect to patient care. And the challenge there is that we don't know what AI they're consulting, how they're consulting it. Are they asking the right question?
David Brailer:Is the system hallucinating, or is it giving them the right answer? But, you know, that's no different than doctors not looking up something in a journal or writing down something wrong. The doctor is still responsible for the treatment decision. They still bear the consequences of getting it wrong. And so, you know, until the tech utopians achieve the world where AI is able to practice medicine, that's the way it will continue to be.
David Brailer:And I think that's a good thing because it does help doctors get access to information more quickly and can use it in their treatment.
Jeff Byers:Yeah. Thanks for that. And then kind of going back to this idea of EHR AI comparisons, you know, we know that EHRs weren't perfect from an interoperability standpoint. Some of the tech vendors, I believe, had the idea of, like, that was their data, not the patient's data, or were, like, to share such valuable patient data. You know, as AI further entrenches into health care, do you like, just from a forecasting, you know, putting on your predictions hat, do you see any of like similar potential challenges arising when it comes to like sharing data?
David Brailer:Well, let's be honest, interoperability has been my career bane. You know, that was what I went to Washington to do, and we still have not accomplished it. And you're laying out some of the reasons why there's a reluctance to share information, but also it's not just the hospitals and doctors. The vendors don't want to lose the information out of their system because it makes it easier for doctors and hospitals to shop around to another electronic record. And so there's a lot of friction that's built in, and we still have yet to get to the point where it flows freely and easily and at low cost.
David Brailer:AI, I think, first of all, will cause people to see a reason to share data they hadn't seen before. And you just said it yourself. A lot of people are trying to put their medical record together and give it to an AI system to say, tell me what my general health status is, or what preventative activities I should follow. In fact, the CMS, the new Medicare Health Advisory Committee that's underway, has a committee that's looking at data liquidity with a specific use case of having patients be able to compile their data for AI to use it. And that means deep, long, longitudinal data pools.
David Brailer:I think AI can also make data sharing easier, because one of the challenges we've had is it's such complexly structured data to be able to move it around. And AI can kind of sort out how to move data around. That's how it assimilates so much information and structures that it doesn't matter if it's in a PDF or if it's in a structured file. It can pull the data and assimilate it in a way that makes it way easier. So I expect AI to advance interoperability quite a bit.
Jeff Byers:Okay. Okay. Interesting. And I want to get back to your piece. You wrote a couple of different pieces for Forefront.
Jeff Byers:But before moving on to that, you know, I wanted to quickly ask you, you noted that EHRs did cause a lot of administrative time for physicians to increase, potentially leading to burnout. We know that burnout is an issue within the health care workforce and the workforce in general, but, also big big narrative in the health care workforce. You know, might AI suffer the same fate of increasing workloads and stress under growth at all cost mindsets, which can happen in a, you know, for profit sector?
David Brailer:Well, yeah. So look, I think think there's no question that electronic records had a last mile problem, which is the systems weren't set up well. The data that was coded from paper to digital wasn't right. And doctors for an entire generation, day, and nurses were seen as the human fix it bots that, like the hospital or the group would say, you need to figure out how to get these records correct. You need to figure out how to make them workable.
David Brailer:That's not our job. And that did lead to a tremendous amount of burnout, frustration, and antithesis from providers who just that was not the job that they trained to do. And, you know, yet to this day, we're now in a place where nobody would go back to paper records. Like, they would never do that. It's so much easier to pull the information.
David Brailer:You can work on it at home, or when you're traveling, and you know, all the reasons that we're a digital society now, they're amplified in healthcare. I think AI will have a different impact. And that's because it will help us assimilate information very quickly, so that makes it easier. But it also means there's a lot more information that the clinician is going to be expected to assimilate to be able to make treatment decisions. And that's going to mean they're going to have to process through much more complex findings, much more complex options for treatment.
David Brailer:And, you know, I think that's probably good for care, but it also is going to be frustrating for them. And I think the real thing is, let's bear in mind that if the tech utopians are correct, society will try to find a pathway to autonomous AI tools that can practice medicine. I'm personally skeptical that that will happen. But even if the tools advance just to help clinicians better and better, it's the clinicians who are training those tools. And I think we've got to bear in mind that AI can go look up textbooks or find things and put it together.
David Brailer:But the AI that we see that's becoming much more effective, look at like coding replacement, like replacing software engineers, is happening because software engineers trained it how to actually do those jobs. And so I think most AI companies are expecting clinicians to train their AI systems to do most of that work, and that's going to be incredibly frustrating for clinicians. So it could go both ways, but I see it being a little bit of bad, a little bit of good.
Jeff Byers:Okay. So going back to your Forefront piece, Jonathan Ketchum and Ken Terry both disagreed with your observations and wrote separate pieces in Forefront. Again, we'll put links to the show notes. You wrote a follow-up piece. You know, they're also not here to defend themselves or their, you know, perspective, so I don't want to put words in their mouth.
Jeff Byers:But your Forefront piece, you know, in response to theirs was called, AI will accelerate inflation because it cannot deliver value in today's healthcare market. You know, briefly, can you describe your case in response to that?
David Brailer:So, first of all, I really appreciated Ketchum and Terry's commentary. I also got really dozens, hundreds of other personal responses from people that read the pieces and either liked it or disagreed with it. Ketchum and Terry and I agreed on a lot of things, but the two principal points of disagreement were their assertion that AI was going to lower costs because it would improve labor productivity. And that's a completely true statement. AI, if it does anything, will improve labor productivity.
David Brailer:People should be able to get a lot more things done per hour worked, because it's an amplifier of human data search, human decision process. Or you can get a lot more done with less labor invested. And then the other was that AI would improve the quality of care. So let speak to both of those. On AI lowering labor productivity costs, I have no doubt about that.
David Brailer:It will do that. And administrative offices, potentially clinically, it's going to take a lot of low value labor and to replace it. However, the question is, will those costs get passed on to the people who buy care? Employers, governments, people. And I asserted no, because it's not a competitive marketplace.
David Brailer:Over 70% of healthcare markets have a dominant monopolistic healthcare system. The payer market is not highly competitive if you look at it by region and by kind of product. And so there's no mechanism by which if a hospital or a doctor's group makes more money because it costs them less, somehow someone's going to compete against them, and they have to pass the cost back to the purchasers. It just doesn't work that way. Everybody knows this.
David Brailer:If you're flying from San Francisco to New York, there are a lot of airlines that fly that route, and it's very price competitive. But if you're flying from San Francisco to Portland, Oregon, there's like one airline that does that, and the prices are outrageously high. So, you know, we all live around competitive markets and cost. The other one, I think, was more substantive, and it was about AI improving quality. And I titled the piece with, I think, a controversial title that it can't deliver value in today's market, because we don't reward value in today's healthcare market.
David Brailer:We reward volume and intensity. And this is not to say that AI will not help doctors make better diagnoses, or pick better treatments, or reduce complications of therapy, or whatever it might be. That happened with electronic records. In fact, when electronic records came out, there was another forecast that said they would save one hundred thousand lives per year through reducing medical errors and reducing drug prescribing errors and other kinds of mistakes. It turned out that that piece was retracted about a decade later because there was no evidence at all.
David Brailer:And to date, there has been no evidence electronic records have actually saved lives. And when I look at electronic records, after twenty years of detailed, thorough investigation, there's no evidence that it's improved lifespan, it's improved health span, longevity. It's had no real benefit to society with respect to what we would call quality of care. I don't see how AI will. I think it will be harnessed as an economic technology to drive the competitive nature of some healthcare markets where it can, but the rest of it is going to be trapped.
David Brailer:So that is where we had the disagreement. And, again, I'm not trying to say AI won't do a bunch of cool things in terms of helping this patient, or that patient, or this doctor. My point is it will not scale up to a market where it delivers societal value. And so I think that is very much where we're going to end up unless we change the incentives of healthcare. So actually, we do what we've been trying to do for my whole generation, which is to pay for value, to pay for quality, not to pay for volume.
Jeff Byers:Yeah. So you mentioned in your piece, this notion of an AI modifier code. Can you quickly describe what that is and how might it be implemented? You know, I think that kind of gets to what you're talking about.
David Brailer:Yeah. So the challenge here is, let's fast forward. It's five years from now.
Jeff Byers:Okay.
David Brailer:And we want to know whether or not AI had any impact. Okay. Well, how do you do that? You want to look at care where the doctor used AI to help them find an answer. That's kind of AI augmented care.
David Brailer:Or you want to look for where the doctor had an AI tool that actually helped them do the specific thing. Like maybe it guided them with a robot during a procedure. Or it was completely AI making the treatment decision. You can't do that because there's no mechanism. There's no place that says in Mrs.
David Brailer:Jones' claims file or medical record from last year how AI was used in any of the treatment decisions. So I argue that we should have a modifier, not a reimbursement modifier. There are two kinds of modifiers. There are reimbursable ones, like, let's say, this patient was sicker than this patient. I need to be paid more.
David Brailer:Or this was done in a surgery center versus a hospital room, you're going to be paid less. Those are reimbursement modifiers. I'm talking about informational modifiers like are used to say this was a telehealth, this was a preventative care, it's quality of care. Social determinants of care are informational modifiers that say this was AI supplemented care, this procedure was done with AI augmentation, it was AI guided. Just to have the ability to have those so researchers can go back and say, ah, look, when this procedure is done without AI versus when it's used with AI, it costs this difference.
David Brailer:It has this quality of outcome. If we don't develop that, we will never know the impact of AI on care. And so I think it's an essential policy and research need for the industry.
Jeff Byers:Well, Brailer, thank you for that. You know, as we kinda wrap up, I wanted to make sure I asked you, you know, whether we're looking beyond AI in health care adoption and just investments in health care in general, I guess I wanted to quickly ask you since I have you here, I know you invest a lot in health care. What are you bullish on? What are you bearish on?
David Brailer:Yeah. Well, I tell you what I'm bearish on is AI companies. You know, mostly because they're enormously overvalued. I think you have to see, this is kind of like what we're watching with Anthropic and its valuations. These companies would have to outperform in a way that has never happened to live up to their valuation.
David Brailer:So I'm cautious there. I believe the companies will be successful. I just don't think they'll ever generate their value, you know, at the current kind of trade. What I'm really excited about is cell and gene therapy. Just the transformative nature of those treatments, you know, and they're very difficult, they're very expensive.
David Brailer:But that market, I think, will transform healthcare over the next generation. And a lot of that is being AI accelerated. I just was reading about an AI guided evolutionary process that was just incredible to change how drugs were made. And so I think that's just an exciting transformational area right now that I'm paying a lot of attention to.
Jeff Byers:Well, thanks for that. And I know you have another piece on Forefront coming up in early December. Do you want to tease anything out from that, or you want to keep it close to the vest for now?
David Brailer:No, I'm happy to. This third piece came about as the result of the dialogues that we've had in the pages of Health Affairs. And it was this tantalizing question, Well, okay, AI is trapped by our current payment system. Can't AI help us change our payment system? And so, I went back and kind of looked at how do you think through a question like that.
David Brailer:There's no evidence on it today, but trying to lay out how would you think about AI's potential to change the way we reward value in the healthcare market. And that's what this third piece is about.
Jeff Byers:Alright. Well, we aren't able to put links of that into the show notes for today, but we will make sure we amplify it on Health Affairs channels. You can sign up for our free newsletter, Health Affairs Today, and you'll you'll get an update when that publishes in early December. But with that, that's all the time we have for today. David Brailer, thanks again for joining us today on Health Affairs This Week.
Jeff Byers:And if you, the listener, enjoyed this episode, please send it to a friend. Thanks, and we will see you next week.