Policy Maker by Day, Doctor by Night | Dr. Thomas Keane

Speaker A: We, as doctors, we don't get trained to think about anything other than the clinical outcome that the patient is going to have. But unfortunately, getting that, uh, fortunately or unfortunately, getting that optimal clinical outcome means understanding the entire context of the patient's care. So that means, will the patient actually be able to afford the drug that I prescribe, or are they going to abandon it at the pharmacy counter? Currently, they estimate that about 125,000 deaths a year are attributable to non compliance with medication. Um, and that non compliance compliance is oftentimes informed by patients not being able to afford the medication. One of the procedures that I do as a radiologist is plug up blood vessels in patients who are, uh, bleeding to death. And, uh, you can plug them up with a variety of different materials. And some of these materials are very expensive, and some of them are very cheap. So after I get access to the blood vessel and I'm ready to tie it off to plug it up, I will actually ask my, uh, nursing colleagues, which of the various devices does this hospital get for the least amount of money, what is the cheapest? And I keep track of how much plugging up this blood vessel is going to cost me because I want to lower the price, both for the patient and for the hospital. It's conceivable I could do the procedure and use so many materials that the hospital would actually lose money on it.

Speaker B: Wow.

Speaker A: Yeah. So you become really aware of the bigger picture of healthcare, just beyond, um, the clinical choices you make.

Speaker B: Ladies and gentlemen, welcome back to another very special episode of the Healthcare Trailblazers podcast back at Health and Human Services headquarters in the supercool studios. Um, and honored to be here with Dr. Tom Keene, who is the National Coordinator for Health Information Technology.

Speaker A: My official title is the National Coordinator for Health Information Technology. I prefer to say just national coordinator because it sounds like I have a remit to coordinate everything in the world, but we coordinate health information technology. And, uh, thank you for having me, Mendel.

Speaker B: It's a pleasure. And with the title done so many things in the last couple years, I think, um, you know, they keep piling on, uh, responsibility on you. So I've heard really tremendous things. So you are a, uh, actually still practicing radiologist?

Speaker A: I am, yeah.

Speaker B: Um, for, I'm guessing, a few years. How do you go from radiology to government?

Speaker A: Uh, that's a great question. When I was in, um, medical training, I, uh, had a roommate who, uh, at the time, his uncle was the vice, uh, president of Kenya.

Speaker B: Okay.

Speaker A: And he got involved in uh, health care policy, obviously, because he came from a family that was involved in healthcare policy, international healthcare policy and the like. But he was training here in the United States.

Speaker B: Okay.

Speaker A: And uh, he got the opportunity through our institution to do a fellowship here at hhs and he asked me, uh, would you like to do a fellowship at HHS as well? So I did. Uh, way back in the day I had the opportunity to work with great people like uh, Alex Azar and Demetrius Kazukis and Tevy Troy and Secretary Mike Levitt. Uh, kind of caught the bug for doing policy work but naturally, uh, wanted to practice medicine, so got out and uh, began practicing full time, but kept a foot in the policy, uh, world. And then in 2018 had the opportunity to come back, uh, into full time government work. I still practice medicine part, uh, time on the side, but did full time government work in the Deputy Secretary's office and then at the national coordinator for health IT. And then in Trump 47, I was asked to uh, join as the national, uh, coordinator leading the agency that I previously worked at. So that's sort of my background. But the initial introduction was really, uh, the training institution facilitated, uh, getting involved in policy work. And having had a roommate who was really interested in it, I took uh, the opportunity and really caught the bug.

Speaker B: What happened to the roommate?

Speaker A: Uh, he is the head of heart and lung transplant surgery at the University of West Virginia. And uh, I believe he is the president of the, the African American Society of Thoracic Surgeons. Very cool. So, yeah, he's doing really well.

Speaker B: Uh, so you kind of were doing policy and practicing basically from day one, it sounds like.

Speaker A: Yes, yeah, we're doing both and I think that's really important. Um, so I continue to practice at the University of Maryland at Jefferson Medical College at the va. And so when I'm at work here, I get to make policy and see other people make policy and then on the weekends or in the evenings or on my days off I go and I live with the consequences of those policies and I see what policies would be most impactful, what policies are working, what policies are not. I think it's good to have the lived experience of a practitioner, uh, who has to live with the choices that I make when I'm in this job.

Speaker B: So, so I have an interesting question. It is, it is pretty, um, it is somewhat obvious to me how important it is the fact that you are a doctor and a radiologist in your informing policy. But how has informing policy changed or informed the way that you practice?

Speaker A: Uh, that is a Very good question. Um, I can tell you what prompted me to get back into government full time, uh, in 2018, is that I had been a member of two very strong radiology practices where the doctors were happy, we enjoyed what we were doing, we enjoyed each other. And both of these practices blew up or failed or unraveled as a result of changes in the health care market. It wasn't that we weren't satisfying the path patients or satisfying the hospitals that we covered. It had to do with mergers, acquisitions, divestitures, and hospitals which were being driven by larger market forces. And I thought to myself, you know, I've had not one, but two practices where I was really happy unravel as a result of larger market forces. I'd like to understand those larger market forces, and the best place to do that might be in government. So when I go back into, uh, practice, I am cognizant of what the intention of government programs are and what the intention of health care leaders, uh, from the president on down is. And I try and reflect that in my practice. So, for example, a, um, big, uh, focus of government is quality. They want to make sure that, uh, patients are safe and that there's high quality health care and that health care is affordable. And so when I make choices in, uh, my day to day practice, I try and reflect that even though there's no, um, direct incentive for me to do so. So to give a concrete example, um, one of the procedures that I do as a radiologist is plug up blood vessels in patients who are, uh, bleeding to death. And, uh, you can plug them up with a variety of different materials. And some of these materials are very expensive and some of them are very cheap. So after I get access to the blood vessel and I'm ready to tie it off to plug it up, I will actually ask my, uh, nursing colleagues, which of the various devices does this hospital get for the least amount of what is the cheapest? And I keep track of how much plugging up this blood vessel is going to cost me because I want to lower the price both for the patient and for the hospital. It's conceivable I could do the procedure and use so many materials that the hospital would actually lose money on it.

Speaker B: Wow.

Speaker A: Yeah. So you become really aware of the bigger picture of healthcare, just beyond, um, the clinical choices you make. You try and put it in a larger context. Another example that ties to work that we've done is when I'm prescribing a drug for a patient, I make sure to check what those Drug prices are for the patient to make sure that the patient can afford that drug, that it's not going to bankrupt them. Because usually there's multiple therapeutic alternatives, um, that are, uh, equivalent. And I say what's the best for the patient from a cost perspective. So that's how it's informed my day to day practice as well.

Speaker B: I love that, I love that so much. We recently, uh, we had a great time. We put together a conference right at the Willard next to the White House, um, called Care Conference. And it was focused on care coordination. And we had about 15 million patients represented in the crowd of about 150 people. So very kind of high concentration audience. Uh, and we had a bunch of wonderful folks, including like Jacob Schiff and Abe Sutton and Geeta and Emily, a bunch of good people out there. And so while talking to Abe, um, they were really trying to peel back the layer of the intent like you're talking about. And so by the end of it, I kind of suggested to the crowd a takeaway. And the takeaway was almost, almost word for word what you're saying you do in practice, which is why this resonates so much. My suggestion was, I think we all try. When you practice medicine, the thing that you want to do is provide the type of care that your family, that you would want your family to have 100%. And my suggestion was to add another layer on top of that, which was when you go and you start planning out how you're going to operationalize some of these new care management programs. Um, think of it as how do I provide care that a, I would give to my whoever, grandmother, aunt, mother, whatever, and that I would pay for out of my pocket. Um, and I said, I think if you can do those things, you'll find yourself well in line with the kind of direction where the puck is going

Speaker A: as far as programs, 100%. And we as doctors, we don't get trained to think about, um, anything other than the clinical outcome that the patient is going to have, which obviously we want to optimize and which obviously is what our main focus should be on. But unfortunately, getting that, uh, fortunately or unfortunately, getting that optimal clinical outcome means understanding the entire context of the patient's care. So, um, that means will the patient actually be able to afford the drug that I prescribe or are they going to abandon it at the pharmacy counter? Currently, they estimate that about 125 deaths a year are attributable to non compliance with medication. Um, and that non compliance is oftentimes informed by Patients not being able to afford the medication. So you have to take that into account. The other thing is transitions of care. And this is where I'd be interested in hearing about careco, your company. Because, uh, oftentimes when you discharge a patient from a hospital or from an acute care episode, you don't think about them transitioning back into the community, transitioning away from the acute care episode back into their normal life. And frequently patients bounce back frequently, uh, uh, their disease process gets worse in the community because they haven't been given, um, the right tools. So, um, you know, I'm interested in hearing. I can tell you how we're trying to support care coordination, but I'm interested in hearing a little bit about care co, because I know you've done care coordination not only in this company, but in a previous company. I'd love to hear about it.

Speaker B: Well, tables turned. Um, okay. Um, sure. So I actually think transitions of care is such a great example of the value that we're trying to drive at careco. But what we're doing at careco is something that I think should be done across the board. And the basic idea is we live in a time now where we have technology that can create this layer of intelligence that collects all the communications that happen with the patient and turn that into. Think of it as like a centralized brain, um, that can then direct care. And so we're all familiar with the AI use cases which we have obviously incorporated in careco as well, for the administrative burden of workflows like documentation creation, task generation, communication generation, things like that, which we do as well. So we collapse the administrative burden around care coordination. But what I think the interesting thing is, is the collection of data and turning that into intelligence that then does. Drives care to the best outcome as quickly as possible. And so, um, in fact, a couple months ago, or maybe about a year ago, I sat here, um, uh, with the, uh, ptac, uh, committee, and we spoke about this. And, um, I think there's a lot of interest around it where everyone's using these ambient AI tools, and yet the intelligence dies the moment that interaction dies. And so you have this conversation with a patient, and you, as a doctor, know this better than most. There's so much gold in that conversation. And yet all of that basically goes to the garbage. A tiny sliver in the form of a SOAP note or some form of documentation gets left behind in the ehr, and that's it. And what we think is not only should all of that be retained somewhere, but it should compound and it should collect and compound based on the other engagements with the patient. And so we're just trying to like, do what we do as humans. So the first time you talk to somebody, your patient, the first time you talk to your patient, you're, you're finding out about them, you're asking them questions. The second time you talk to that person, hopefully you are then like stacking that data in your head on top of the first data and then the third time and you never have like the same conversation over and over again, Right? Unfortunately, in healthcare it's not like that. You do have the same conversation over and over again. How many times have you told the doctor that you don't smoke? You said it at the front desk, then you said it to the nurse that brought you in, then you said it to the attendant that you said to the doctor. And so our thought is, hey, if we can capture every text message, every in person conversation, every phone call and every video visit, so every form of communication gets collected and then turns into this layer of intelligence that then tells whoever is talking to the patient exactly what to talk about to get to the, to the, to the goal. How cool could that be? And then there's other things around quality control and assurance. But I want to come back to the transitions of care example, because it's such a good one. Typically transitions of care involve many people, uh, probably five or six teams. So there's someone bedside at the patient. Then when the patient is actually discharged, they're supposed to get a phone call. Then there's the follow up, then you're trying to get them back with the pcp. Then there's following up on that. Right? And that can be many, many different people. But imagine if they were all sharing the same brain. So imagine if there was one system that was collecting the thing that was talked about at the bedside and that informed the next automatic conversation, said, okay, so here are the pain points and here's what you need to talk about. And then both of those, uh, communicated to the third one. So it collects everything. So the vision of careco is a layer of intelligence using the communication stack, which is, I think, a pretty interesting angle. And then uses that to direct care, um, track it and, uh, score it for quality and then collapse the administrative workflow.

Speaker A: Well, that's an extraordinarily powerful tool and process that you're building, uh, in two ways. The first is patients throw off tremendous amounts of information and data. And we now have the capacity not only to collect that data, that's part of What I do with electronic health records and what other companies do with wearables and with ambient scribes. But we have the ability to make sense of that with LLMs. And the ability to collect all that data and turn it into something meaningful and sensible is a very powerful idea. And I'm really glad you're building it. Let me talk about the, um, care coordination case, um, you know, for people like to do math in, uh, their head. I was an intern way back in 2003 and I was an intern in Maryland. And uh, I worked at a large, uh, I worked at Johns Hopkins, a large hospital. And anybody who wanted could dial a number and refer a patient in. I still remember the number because this number creates so much trouble for you when you're an intern. It's 410-955-4444. And anybody anywhere in the country could dial that number and say, I'd like to refer a patient to you. And many of those referrals are appropriate. Uh, oftentimes in the community they don't have the resources to properly take care of a patient. Many of them were, they were, uh, due to staffing issues perhaps at a hospital or due to a practitioner who wasn't uh, really sure what was going on with their patient. And at that time all we could do is really talk to the doctor, maybe get a couple of faxes to get information about that patient, and then generally we would accept the patient. And this created a lot of friction. Um, we were accepting patients who didn't necessarily need quaternary care. Uh, we were filling up our beds and we were preventing people from being admitted who needed to be admitted. This has all changed because of the new, uh, information technology layer that has been deployed over the last 25 years. Now I still practice in Maryland, this time at the University of Maryland. Now when we get those same calls for a referral into a state run medical hospital, um, I can actually look at what's going on with the patient at their actual medical record at the outlying hospital that they're at. I can look at their imaging, I can look at their notes, I can look at their vitals in real time if they're in the icu and I can talk to the doctor and say, yes, this patient is appropriate for transfer or no, they're not. And additionally, if the doctor says, well, I need some help managing this patient in place, we will actually assemble a care team that provides remote consults to that patient. Being able to see the same data oftentimes, even being able to see the patient because, uh, they now have cameras in the patient's room where we can actually. We can't do a physical exam from 200 miles away, but we can actually see the patient. That is what technology has enabled. And it's been absolutely transformative. I love the fact that I can, um, really provide care for patients outside of the scope that I just see in the hospital and better manage the patients, um, for other doctors and assemble care teams on the spot. So it's really been transformative. If you talk about care coordination.

Speaker B: Yeah. And I know you're doing really powerful work there. And I think one of the focuses that you have is around interoperability. Um, let's talk about that for a second. Because that's what you're really hitting on and being able to now all of a sudden see things in different areas. And we've come a long way over the last. Whatever it's been 20 years, ah, since we've gone digital, at least most of us. Um, and now you guys are taking that just to the next level. I mean, you're focused on AI and collaborating with private industry. Every time I talk to you and the teams, uh, here at hhs, at the onc, at, um, cmmi, it's, like, mind blowing, um, to see that a big focus is on interoperability and getting all these different systems to communicate to each other. We live in this weird healthcare world that the EHR systems essentially own the data. And even though officially the patient owns the data, we all know that that's just not the practical case. And that's what you guys, I think, are working on. How do you approach that and what's kind of the plan to conquer it, when it would seem to me that a lot of, like, the very value proposition, kind of the business case of these EHRs is their data. They seem to be. Even though the public posturing from. I won't name anybody, but the public posturing from many of the large EHRs is like, oh, we're here, we're collaborative. But I think their actual business model is they definitely are incentivized to keep a, uh, lock and key on that data, because otherwise, what are you. At some point, um, in this day and age, with the ease of building essentially, CRMs and billing systems, your insight

Speaker A: is, uh, 100% spot on. And for some context and also for your viewers and listeners, let me give you a little bit of the history of ONC and tell you how we got to where we are. Um, the agency was created by presidential executive order in 2004 with the understanding that health data and health electronics needed to be regulated in some form or fashion by government. In 2009, President Obama passed the Hitech act, which, uh, did two things. Number one, it mandated the adoption of electronic health records. It said that if you are a provider provider, you have to adopt an electronic health record. And it provided, depending on how you count, between 30 and $50 billion to facilitate that adoption. So in 2009 only about 15% of providers were using electronic health records. By 2015 that was over 90%.

Speaker B: Wow.

Speaker A: And now, yeah, in a five year time there was rapid adoption. And now in 2026 it's basically 99 to 100%. Everybody uses electronic health records. The rapid adoption of health records created two problems. The first is that, uh, there were a lot of frictions with providers. You're really changing workflows, you're changing the way providers have to interact with patients. Patients noticed at the time that the doctor was no longer looking at them, but was rather managing a computer. So the first thing was the burden that it created on providers. And the second thing is that these health tech companies all of a sudden found themselves sitting on top of this gold mine of data about the patient. So in 2016, also under President Obama, the Senate and House unanimously passed the Cures Act. And what the Cures act said was two things. Number one, the government has to focus on reducing the burden that's associated with electronic health record use in general health tech. And secondly, this data belongs to the patient. It has to move at the direction of the patient for the benefit of the patient, privately and securely. Now, as you say, there's a business case for holding onto this data. Um, you know, hospitals, uh, these are their customers. And hospitals are reluctant to share customer data because it could put them at a competitive disadvantage or it's what they perceive to put themselves at a competitive disadvantage. Uh, similarly, the electronic health record companies and other holders of health data would, I think, prefer to sell the data than to share it. Um, in many instances, or they might feel, rightly, we do elect to curate this data to clean this data. It's not cost free to do all of that and we should be able to make some margin on moving the data. So what we have in place is essentially, uh, three tools to make sure that data moves for the benefit of the patient. The first is we've stood up a nationwide interoperability network called tefca. And uh, like all government things, it's an acronym. It stands for the Trust Exchange Framework and Common Agreement I'm impressed that you

Speaker B: know that, that the actual.

Speaker A: I'm really bad at acronyms too, but this one, since I run the agency, I have to know. Um, I'm proud to say that when Secretary Kennedy took office, only about 10 million documents had been exchanged on this. And these are documents exchanged, ah, related to patient care. There's two use cases on it. Uh, first is treatment. One doctor needs to pull the records for a patient from another doctor and individual access. The patient wants access to that data themselves so that they can look at. And in those two use cases, only about 10 million documents have been shared. Now, fast forward one year later, it's been over a billion. We've really focused in laser like fashion on making sure the interoperability network, which is voluntary, works for everybody who's participating in it. And now we're looking at 100 to 150 million document shares per month. It's really taking off. So to get that data to move, we've stood up the infrastructure to make it happen. Uh, secondly, we've adopted standards. And standards are a very dry area. People say, you know, they say, why do I want to talk about standards for data representation and data transport? But it's basically the language that allows one computer to talk to another computer. And we've worked very hard on standing up these standards so that computers can talk to each other. But the third and probably most important plank is our information blocking regulation. Under the information blocking regulation, it says that if you hoard or silo data when it should be moving for the benefit of the patient, you can be fined up to a million dollars per instance. And in fact, the regulation doesn't say per instance of what, it just says per instance of information blocking. Does that mean blocking the data on one patient? Does that mean blocking one transaction? So, as you can see, the fines and fees that people could pay for information blocking, uh, can be pretty severe. Um, up until now, the information blocking regulation has not actually been enforced. But Secretary Kennedy and President Trump had said that this is going to be a priority of this administration. So my office has already issued a number of notices of potential nonconformity which says, hey, guess what? Your information blocking, you have to stop. And we've already gotten responses from that and we will be making announcements about this in the coming weeks. But additionally, the Office of Inspector General is looking to actually apply the civil monetary penalties to people who, um, our information blocking, and it's our hope that data will flow, and we believe that it will. It's going to be a Little bit of an uphill battle, but we do think we'll be able to reset market expectations to have data flow.

Speaker B: What about functionality blocking?

Speaker A: Mhm.

Speaker B: So it's kind of like, okay, officially we'll give you whatever data, but again, I think uh, some of these have monopolized and I don't even say that in a bad way. They've done a really good job at least like gathering the ability to be like the heart and brain of the kind of health system and operation. And they definitely want to keep as much functionality in their thing. And so they'll use certain things like making their, whether it's API documentation extremely complicated or not letting you actually set up a sandbox until you have a health system partner, but then the health system partner won't let you actually set up a sandbox unless you have the sandbox. So there's a lot of different creative ways. I just heard of one, uh, recently again I won't name anybody, but a different large EHR that I think just switched up all their APIs. So all the point solutions that were kind of dependent on them.

Speaker A: And what's an API?

Speaker B: Oh, I don't know.

Speaker A: Asking you for the benefit of the API.

Speaker B: Asking API stands for, um, I don't know what it stands for.

Speaker A: Application Programming Interface. There we go. It's the way computers talk to each other. Yes, but I'm sorry, I just wanted to make sure that the listeners were. Yeah, go ahead. I'm sorry to interrupt your cadence.

Speaker B: No, no, your point spot on. So it's the thing that allows them to talk to each other. And so if you had a point solution that was relying on a specific communication point at the EHR to then transmit the data and then that change, all of a sudden your entire application is gone. And being that so many of these applications are really mission critical for patient care, you need to be up and running. So it causes a lot of um, challenges. So it would seem to me that there's ways to kind of be officially good with. Oh, we don't do any data blocking, but there's still a lot of efforts around functionality blocking 100% and I'm glad

Speaker A: that you brought that up. So we under, you know, we actually administer and uh, require that all electronic health records support an application programming interface that we call our G10 API. That's because it's um, code of federal regulation, ah, 170.315 G10 that says what the functionality of that API must be. And we've of course heard from the market that there's all kinds of ways to make the API non functional. There's latency, maybe you can only make a call every 10 seconds. There's excessive fees, um, there's bad documentation, there's incomplete information, there's information that's not conformant to a standard. And so it is our, uh, intention in future rulemaking to really specify what the functionality of these APIs is so that they actually work and so that the ability to essentially deprecate the technology as a means of information blocking is no longer a thing. Additionally, we know that many, uh, Companies will produce APIs for business partners or for people who pay for them that work very robustly. And what we want to do is make sure that these APIs are available to everybody. If you're producing an API for somebody, we would like that API to be accessible to other actors in the market who are similarly situated under similar terms and circumstances. Now, there's a lot of thinking that has to go into how you write a regulation that does that. But we are acutely aware of the problem and we are, uh, addressing it, uh, in future rulemaking.

Speaker B: And I will say you feel that, you feel that in the market, you feel that you guys are on it, because you can tell the public posture has changed from these folks dramatically. Um, and so, uh, I hope and presume with the people that are leading this that that'll also be effective.

Speaker A: Well, let me tell you what happens when incentives are aligned. Um, you can actually get pretty good interoperability. So, uh, last summer, uh, under President Trump and Secretary Kennedy's direction, we finalized what we called our HTI 4 rule. By the way, I'm going to get away from numbering, uh, our rules. We're going to start giving them names so that people know what they, what they.

Speaker B: You come with citations. It's amazing.

Speaker A: Exactly. So, uh, I was thinking of maybe skipping a number or something like that just to throw the market off. But, uh, our HCI4 rule, we did two things. We said that, that, um, electronic health records have to support functionality to allow for the, uh, display of drug prices at the time of prescribing. We call that the real time prescription benefit functionality. We also required them to support electronic prior authorization, meaning that they had support functionality to allow the electronic health record to talk to computers at the insurance companies so that prior authorization for procedures could be negotiated between the computers rather than negotiated between the doctor and the insurance company representative by facts. So let's talk about the real time prescription benefit functionality. There are aligned, uh, interests there. The insurance companies Would like people to use drugs that are unformulary, would like people to choose the lowest cost drug and would like to avoid prescription abandonment at the counter. They don't like it when their patients, who they're trying to keep the burden of disease low on, are actually abandoning therapy because it's expensive. They don't like getting angry calls from the patient from the pharmacy parking lot saying I can't afford my drug. What the heck are you guys doing? So we mandated uh, that this functionality had to be supported. And even though we did it last summer, Right now about 90% of electronic health records actually support this functionality. And what it means is that when a doctor chooses a drug at the time of prescribing, uh, he or she can see what the cost of that drug is and what the cost of therapeutically equivalent alternatives are. Hey, you chose atenolol as the blood pressure medicine, but maybe Metoprolol will be $10 cheaper for the patient. That information allows the doctor and the patient in real time at the point of care to choose what is not only the most therapeutically effective therapy, uh, but also what's the most therapeutically effective therapy that the patient can afford. Interestingly, the reason this will works so well and has been adopted so quickly is because everybody is aligned in this. The insurance companies will actually not only provide this information to the electronic health record companies for free, they will actually pay the electronic health record companies to carry this information because it facilitates the efficient use of insurance. So when incentives are aligned data flows, the standard, the uh, NCPDPRTB standard, uh, as it is to just.

Speaker B: I will not ask you for that one.

Speaker A: Acronyms, uh, you know, is actually very robust, it's actually widely deployed and it's being used now what we're doing is asking the uh, electronic health record companies to carry even additional information. So in many instances, as you know, the insurance price for a drug, even if it's very low, might actually be higher than the price that the patient would have to pay for cash, particularly for generic drugs. So um, for example there are vendors of cash pay, uh, pricing information like Trumprx, like Mark Cuban Cost plus drugs like Amazon, uh, prescribing, I forget exactly what theirs is called. GoodRx. These provide cash prices so the patient might say, you know what, I'm not going to use my insurance. My insurance copay is $15, but I can actually get it on Mark Cuban cost plus drug for $8 or I can get it on TrumpRx for $7. Uh, we've actually looked at TrumpRx, I think has been out for about four or five months. And I think early research shows that it saved patients about $400 million because it shows these, uh, cash pay prices. And so myself, Dr. Oz and Secretary Kennedy sent out a letter to, um, the electronic health record companies, asking them to subscribe to these data feeds. The data feeds are free, asking them to subscribe to them. So that in addition to the information that comes from the pharmacy benefit managers and insurance companies, you can also see these cash pay alternatives. And um, we wanted to make this happen quickly, which is why we haven't just adopted it in regulation. We might in the subsequent regulation, but for now we've asked them to adopt it voluntarily. And the, uh, response that we've gotten from the market has been overwhelmingly positive because they understand that it's a benefit for the patient and nobody really loses.

Speaker B: Wow. Um, by the way, super in the weeds, just a side comment, but, um, secondary insurance information is a very like low visibility data point that there's really like not even structured in the EHR. Um, and that would definitely add to this, so 100%.

Speaker A: And uh, as we are writing our forthcoming rule, which my lawyers will tell me, uh, every time I mention it, I violate the Administration Administrative Procedure Act. So I don't want to mention it too much.

Speaker B: This is pre decisional, iterative, and confidential.

Speaker A: Yeah, all of that, exactly. Um, so we are considering all of that. And what can I ask in careco, uh, what sort of, uh, frictions do you find, um, from not having access to the secondary insurance information?

Speaker B: So careco in particular, um, we're not really doing much eligibility, but my experience on that comes from the company where I started from. For the last eight years we were doing the actual service of care coordination. Um, and that is still a very human manual process. So you can pull a list pretty easily out of an EHR that has the primary insurance, but very, very few will have cohesive secondary insurance information. And for a lot of these programs that makes a massive difference to the patient's coverage. So the secondary will pick up 80% of the 20% that's maybe not covered by whether it's Medicare or whatever. Um, and that could mean the patient participating in the program or not. And for, in fact, I'd say about 80% of patients don't have that secondary and they 80% of patients that are eligible for a lot of these care management programs, and that is the reason they do not participate in the care management program is because of that $8 co pay that is simply too much for them. Um, and so getting rid of the co pay is a whole different discussion, but that's with Congress. So we'll leave that for the folks across the street.

Speaker A: Um, but we will have a discussion after this about, um, your insights from your prior company and uh, what you learned about secondary insurance. Because it's something that we're thinking about and something that I think we can do something about in the electronic health record.

Speaker B: Cool. Well, that's exciting. Um, where do you think government does not as we kind of go down this path? Right. Where does government not belong? And I'll give some context to where I'm thinking. The whole premise, I think, of this entire discussion. Why do you care about efficiency and quality? Is because you pay for it.

Speaker A: Right?

Speaker B: The government is the biggest. Biggest or after commercial payers?

Speaker A: No, no, government is the biggest.

Speaker B: Okay?

Speaker A: Single biggest payer.

Speaker B: Single biggest payer in the country. Um, and it's health care costs are ballooning out of hand and at some point it's going to go bankruptcy. So we got it. We got to get this under control. And that's why it would seem to me that that's the underlying reason for government to even be in this conversation. Be like, hey guys, we got to make this simpler. We got to make it more efficient, we've got to make it more cost reductive. Um, but do you have any opinions on where that line should be? Because at some point the furthest extent of this is why doesn't government just have a national ehr and say, you know what, guys, we're sick and tired of telling you, like, update this, update that. Here's a platform, just do this.

Speaker A: Well, that's a great question. I think the last half century has shown that, uh, markets tend to produce the most efficient and innovative outcomes. They drive the highest quality and the lowest price. We see this in the computer and the technology industry. It's relatively unregulated. And, um, government did play a role. It set up the standards for the Internet, the tcpip. Uh, I think Tim Berners Lee came out with, uh, HTTP. Um, but much of this was developed in the private sector. And technology has absolutely transformed our lives. So we understand that the private sector is where the innovation happens. It's where the cost lowering and quality improvement actually occurs. Um, one area that comes, uh, to mind is an artificial intelligence. Now there is a need for good governance and regulation around artificial intelligence. Currently, as a, uh, radiologist, I have for 20 years used artificial intelligence tools in our practice. People, uh, Talk about it nowadays, like, oh, AI is coming online. And I say, welcome to 2005 in radiology. We've been using our artificial intelligence to help us interpret images since then. And the tools are getting better and better. These all go through FDA device clearance. They actually have to show that they're safe, that they're effective. And then of course, they go through clearance with the doctors. Doctors say, yes, this works or no, it doesn't. There have been sepsis tools that have been released that doctors have found, at least in their initial incarnations, didn't really work and so they turned them off. By contrast, I routinely use in my practice stroke detection tools that don't work 100% of the time, but on balance help me get better at my job and help me be more efficient at my job. So there's also that human in the loop layer. What I would not want to see and what I think the administration doesn't want to see is a patchwork of regulations in 50 states or in, ah, thousands of localities in which people are, um, putting constraints, uh, on how artificial intelligence is used or moved or, or how the ecosystem around artificial intelligence is, uh, regulated. It's really something that should be done nationally and coordinated nationally. Um, the federal government has the resources to do it. We have the insights to see what's going on in all 50 states and in all different locales, and we hear from all stakeholders. So it would be a tragedy if we saw the potential revolution that artificial intelligence can create in healthcare, strangled by having a patchwork of 50 state regulations that are inconsistent with each other or poorly thought out. Um, so that is one area where, you know, this administration would really like to see the state step back and leave it to the federal government. Now we are taking that very seriously. But crafting meaningful regulation that protects patients and also allows the field to advance is a heavy lift. But we are doing. The FDA has already done that with devices and they've done a great job. I mean, uh, as I said, it's transformed the practice of radiology. I'm sorry, I stepped on what you were going to say.

Speaker B: No, yeah, I couldn't agree more. I know that's a pretty heated, uh, conversation in this town. And, um, I can see the argument for things like gun laws and abortion and things that are really hot topics that people feel really personal about. But when it comes to technology and specifically access and medical care, you know, uh, I don't. I think you'd be hard pressed to find someone who's super excited about the fact that There's a patchwork of regulation around, like, where doctors can operate and that you have to register in each state in order to have a license. So, like, what's the upside there? I'm not really sure.

Speaker A: Right. Well, one of the things that's being thought through internally is whether if you have autonomous AI, if it should be trained in the same way and held to the same standards as a physician. And in that case, yeah, there's 50 state licensing laws in states. I have licenses, I'm embarrassed to say, in 20 different states because radiology can be done remotely across state lines. And so.

Speaker B: I'm sorry, why? Like, is there a good reason for you to have to have done that 20 times and that you can't operate in the other states?

Speaker A: I would prefer that not to be the case. And the VA only requires you to have a, uh, license in one state to operate in any of the VA hospitals. That's a separate discussion, and I will leave that to, uh, to the states. But to the extent that the states regulate the practice of medicine and that they have their own equities in how they want medicine practiced in their particular state, the idea that an autonomous AI would be licensed under current licensing laws is something that we've considered and something where the states could step in if we go in that direction. Um, but we really think that in the same way that, uh, drug and device policy and payment policy is really made at the national level, for the most part, uh, we really feel that, um, that's something that should be happening at the federal level. And, uh, to do it otherwise would probably not, um, improve quality or safety, but would slow advancement.

Speaker B: Yeah.

Speaker A: Um, uh, you brought up the patchwork of state laws or different states. Let me tell you something where we are trying to coordinate between hospitals and, uh, the community, because I know you're in the care coordination business, um, we recently released what we call our behavioral health information technology grants. Behavioral health data is particularly sensitive substance use data. Behavioral health data enjoys special protections under a regulation called 42 CFR Part 2. Right. Yes. And we know that the way of, uh, taking care of patients who have behavioral health or substance use disorders, a small part of it is the medical system. System, um, unlike, say, getting a liver transplant, where the medical system is really key to that, um, you can have an acute crisis in behavioral health or substance use disorder, but really the key is getting you to transition back into your life, getting you to transition back into the community. So what we've done is we've set up pilots in nine states with 45 exchange partners where as patients are discharged from the hospital after an acute care episode, they're able to communicate the data from their care episode to occupational service providers, vocational service providers, housing providers, uh, community, uh, engagement and community based organizations to allow that patient to transition back into the community to not relapse.

Speaker B: Who m communicates the data, the behavioral health facility or the patient?

Speaker A: Uh, the data gets communicated directly by the behavioral health facility to, um, these trusted exchange partners. And it's all subject to the regulations, it's all subject to patient consent. But we know that if a patient has an acute substance use disorder, um, episode, ah, where they almost die and they're treated in the hospital and now they have to be transitioned to a residential treatment facility, and then they have to leave that residential treatment facility to return to the community. Uh, all of this care has to be integrated. And we've released these behavioral health information technology grants to pilot the communication of this data so that the data is communicated seamlessly. And so the vision that careco has of collecting all of this data from patients and then using it, uh, leveraging it to optimize patient care so that this sort of data can actually flow to optimize patient care. And so far it's been a big win. Um, we're excited about it. Uh, I had the opportunity to go to Huntington, West Virginia recently and see a program called Project Hope. And it's for women with small children who have substance use disorder. And they get admitted for six months to a residential treatment facility with their children, where they learn not only how to overcome their addiction and to manage their addiction, they also learn basic life skills like how to, um, uh, you know, take care of your children, how to manage that with a job, how to find a job and get situated in a job, how to live within a community and how to keep a home. And the, uh, average tenure in the program is six months. And they've had 170 graduates who have successfully completed the program, are in recovery, and have mainlined back into the community with jobs and with custody of their children, it's been a fantastic success. Um, what we'd like to see is the data that, uh, is necessary to make this happen to be able to happen in all 50 states through the programs that we administer.

Speaker B: Wow. Going all the way back to the hitech act, which then created the necessity to have EHRs, which then created this, uh, I think if I followed the story, TEFCA was maybe what I'm referring to. But then there was a whole nother thing. That had to be created to kind of roll back the issues that were created by the introduction of the ehr.

Speaker A: That's exactly right.

Speaker B: You are leading version two of this, which is the AI revolution. And you already alluded to, um, you know, the Trump administration just put out, um, just put out, uh, the framework for AI regulation, I think a couple days ago or a week ago. And so, I know, not necessarily healthcare specific, but definitely would probably tie in. Um, but what have you learned maybe to do or not to do from the history and the experience of that first push and watching the things that went well and watching obviously the things that went not well to the point where there had to be kind of a rollback, uh, um, um, thing put together. Um, how is that informing your, um, introduction into this kind of next phase of digital? Which is.

Speaker A: I think what we've learned is that the best thing to do is to facilitate markets and to see where there might be market failures and step in to plug those gaps. So obviously patient privacy, security and safety are tantamount. Uh, effectiveness of, uh, therapies is tantamount. But if there was a footfall with the HITECH act, it did allow for rapid adoption of electronic health records. That rapid adoption was done by mandate and subsidy. It wasn't something that was done, um, by market forces. This stands, by the way, in contrast to the adoption of PAC systems. PAC systems are the picture archiving systems, uh, by which radiologic images are displayed to doctors. Uh, the first PAC System debuted in 1989 at the University of Maryland. And basically by 11 years later, by the year 2000, they were completely universal. The old days of having an actual piece of film with an X ray on it were long gone. And the reason that it was adopted so quickly is in the long run it was cheaper. It's cheaper to hold this stuff electronically than it is to hold it in actual giant vaults of films. But it also increased access. Physicians were able to all over the hospital or in the ambulatory care clinic to see the imaging that was relevant to their patients. So adoption was rapid and it was worth paying the money for. Cost came down, quality went up, and this all happened by market force. By contrast, the EHR adoption was mandated. You have to adopt EHRs and here's subsidy to do it. And so the normal market mechanisms of lowering cost and improving quality and making sure that there's a fit to purpose didn't happen. So as the AI revolution happens, what we want to do is facilitate the market while assuring patient privacy, security and safety, making sure that there's not scams in the market too. We want to make sure that there's some effectiveness, but we don't want to be directive to the market and say, oh, this is what you have to do. So if you think about how the market has adopted it, some of the earliest use cases have been these administrative use cases, the AI scribes. Right, and what do AI scribes do? They're not making uh, critical clinical decisions about the patient. And um, there's always a human in the loop. You actually read your own report to make sure that it says the patient had an intracranial hemorrhage, not an intracranial hemorrhoid. You read it to make sure that uh, it says that there's no new tumor rather than just new tumor. You want to make sure that um, you read your own report. So for these early use cases, the market has chosen to adopt these and they're widely loved by doctors because now I can talk to you as a patient and have the AI scribe capture all of that information and structure it into a meaningful note, saving me documentation time. So the market has determined that similarly with radiology, there's all kinds of opportunities for um, AI to help us interpret images. But what's one of the critical use cases is stroke. Um, time is brain. When a patient has a stroke, they have about a four hour window in order to have that stroke treated before they have permanent disability. So when a patient presents to the emergency room, we have stroke, uh, detection, uh, software that colors the images of the patient's brain where the computer thinks there might be a stroke and it allows me to triage the cases faster and say, oh, I have eight stroke patients presenting across five different hospitals. But this one the computer is telling me might be positive. I'm going to look at that one for, and if I see that in fact there is a stroke, I can call the radiologist who's actually going to treat the stroke. We actually treat the strokes by pulling blood clots out of patients brains. I can actually, yeah, it's really cool. I call uh, the radiologist and I say you need to drive in and treat this patient right now while I carefully look for all of the different places the stroke are. But the point is that the technology, uh, is determined by the mark market. We're not forcing it, we're facilitating it. The fda, uh, makes sure that it's safe and effective through uh, its programs. And um, there's a human in the loop. So they're able to determine is this working or Is this not? So basically we're going to take a market based approach and there are rules for government in that.

Speaker B: Yeah, certainly. All right, well, moving on to like the future. What does. This is always a fun question. What does the health system look, if you had control over the direction that this goes in and the things that you and your team and teams here are working on all kind of happen. Um, what does the ideal, um, healthcare system look like in 10 years from an AI perspective? Is it. And I'll give you just because that's maybe too broad. There's a sentiment that I've heard where we think anything that is not hands on care can be done agentically. Uh, then there's probably the flip side of that. So are we looking at like, yeah, maybe that's enough color?

Speaker A: Well that's a great question. And um, I think that uh, the National Coordinator Agency, the office of the National Coordinator, is actually facilitating this because we're facilitating the movement of data. So what does this mean? That all of the administrative waste that happens in the system is eliminated? I referred to um, a real time prescription benefit that allows doctors to choose the lowest cost alternative. We're also automating electronic prior authorization. We estimate that it costs about $2 billion a year in administrative time, uh, getting prior authorization for care that a, uh, doctor has prescribed for a patient. There is no reason that the, this cannot be adjudicated between the electronic health record and the um, uh, insurance company computers. And both us and my colleagues at CMS, uh, Dr. Oz, Chris Clomp, John Brooks, these guys have adopted standards to allow these computers to talk to each other. And to their credit, America's Health Insurance Plans, the Electronic Health Record association and the AMA are all collaborating with us to actually make this an implementable reality. The fact that we've defined standards and defined deadlines isn't enough. We actually have to collaborate with the market to make electronic prior authorization happen. So we estimate that within the next year about 80% of prior authorizations will be adjudicated in real time electronically in the background. So uh, all of this administrative waste in the system, getting this squeezed out by technology, whether it's agentic AI or just traditional, uh, technology, uh, this means, um, not just prior authorization, but quality reporting, uh, it means uh, transitions of care. All of these things that normally required human interaction, even the billing and coding and all of these things can be done electronically. So a crushing of the administrative costs in the system by AI then the second use case is the augmentation of the clinician, um, by Clinical use cases. If I can read 150 mammograms in a shift, because I know that the AI can really triage the mammograms for me and tell me which ones I really need to take a look at. And I can rely, based on the science and the underlying evidence, on the AI triaging these mammograms, my productivity goes up. And the time that a patient has to wait for their mammographic result or the time they have to wait to get an appointment goes down. Scheduling and referrals is another, uh, AI, uh, driven use case. Right now it's very hard to get an appointment with a doctor. Right. It could take a month, or for a specialist, three or four months. And yet it turns out that about 30% of a physician's time is actually slack time, not used because the algorithms for actually filling the physician's time slots are inefficient. We can actually make physicians and general care physicians, nurse practitioners, physicians, assistants, all more accessible through AI. So there's the administrative use cases and the clinical use cases. And what I would like to see is quality go up, uh, costs go down, uh, access improve all through the use of AI. And I actually believe I don't have to do anything to make that happen other than support the standards and, uh, continue directionally to make sure that government, uh, plays its proper role in facilitating the market. And I think we'll actually see that. I mean, look at how the iPhone. The iPhone was 2008. I mean, our world has completely changed as a result of that.

Speaker B: Yeah. So last time we were down here, had the honor of having Dr. Oz and Amy Gleason on the podcast. And one of the things that's, obviously that Amy in particular is so passionate about is, uh, and I think all of you in general is patient empowerment and using data for patients. And I know that that's something that you've been working on. Uh, so what are some of those efforts around giving patients, and we talked about earlier in this conversation, um, patients, uh, not really owning their data, even though they do own their data. What are you working to help transform that and actually give that data back and the power back to the patients?

Speaker A: Well, we really understand that agentic AI can empower patients if they have their data. Uh, so, for example, uh, on our TEFCA network, we have a, uh, use case called Individual Access Services, in which I can go on my phone through an individual access service provider, and I can download my data from all of the sites of care that I've ever received care at and download it into an agentic AI that can then help me manage my health care. So, for example, I, uh, am a runner, and, uh, obviously like to optimize my running time and my diet and everything else so that I don't suffer an injury, but so that I get the maximum cardiovascular. Cardiovascular benefit out of it. I have been able to download all of my personal information and download it into apps that allow me to optimize my training regimen.

Speaker B: And you choose the agentic AI?

Speaker A: I choose the agentic AI. That's exactly right. In fact, I demonstrated this, uh, functionality to Secretary Kennedy, and I showed it to him, and after I handed him my phone, I realized I just handed him my personal medical records. And to his credit, he looked at it, he kind of nodded, and he said, uh, you, resting heart rate is 42. That's very good. You must be a runner. And I said, thank you, and I took the phone back. He was very. He was really a gentleman about it. He didn't snoop too much. But, yes, this is. And, um, you know, I'm doing this for wellness and health, but for a patient who's had a transplant or a patient who has diabetes or a patient who has chronic disease, we already see that these personal assistants are really, uh, transforming the way that they are cared for and are improving outcomes.

Speaker B: Yeah, it's just the beginning of, um, Step one is giving patients the, uh, data. Step two, which is what I think is we're just at the beginning of, is giving the patients a reason to use the data. And that's, I think, the key missing ingredient that the agentic AI and AI brings. Because up until now, the data was there. I officially own it, but I had nothing. It's all just doctor language I don't understand. It doesn't, uh. It doesn't speak to me. Now, with AI and with nlp, you can just, like, you can talk to it, and it can make sense to you, and it can help you with your goals. And so, um, really an exciting time.

Speaker A: It really is. And that's one of the great promises of AI Patient empowerment, which is what we're all about here and in this administration.

Speaker B: Yeah. Well, uh, Dr. Keen, um, I'm always astounded when I come down here and talk to you folks, not just the work that you're doing. Um, and there's so much that you do, they. That obviously, you didn't even mention whether it was, like, the, you know, the partnerships with clear and ID me and, like, you're constantly there's constantly things coming out in partnership with private sector and innovation and um, even the revamping of this studio I think is a very small reflection. Yeah, I think it's a very small reflection of like the general revamp um that's been going on here. So it's always astounding to me. Um and the most impressive thing to me that I always say over and over again is the past is the passion that every single person from like I said hhs, cmmi, M uh ONC that I uh have the pleasure of speaking with it's clear that this stuff like they live this stuff, they do this on their free time. This is really clearly a passion and it's abundantly clear with you as well.

Speaker A: Same as you with Careco.

Speaker B: Thank you. Um, so thank you so much for the time.

Speaker A: Um, thank you for having me. I've enjoyed the conversation. I'm looking forward to continuing it.

Speaker B: Awesome. Thanks.