Revisiting Vinod Khosla’s vision for AI-powered healthcare

0:11Hello and welcome to today's episode of Fixing Healthcare. I'm one of your hosts, Jeremy Kor, also host of the Popular New Books and Medicine podcast and CEO of Executive Podcast Solutions. With me is Dr. Robert Pearl. For 18 years, Robert was the CEO of the Permanente Group, the nation's largest physician group. He currently serves as a professor at both of Stanford University School of Medicine and Business, and is author of the best-selling book ChatGBTMD, How AI-Empowered Patients and Doctors Can Take Back Control of American Medicine. All profits will go to Doctors Without Borders. If you want information on a broad range of healthcare topics, you can visit his website, robertparilmd.com.

0:44Today's interview is part of Season 10. This season focuses on technology and the role it will play in the future. Before we record the final couple of episodes, Robbie and I thought it would be enjoyable and educational to go back a couple of years for perspective. How likely are our guests to see the future clearly? Do they overhype or fail to fully anticipate how fast technology will advance? And who better to help us figure that out than Vinad Khosla, one of the world's most successful venture capitalists and an early investor in OpenAI. The company that created ChatGPT. This episode, number 40, was recorded in December 2021, a year before the release of ChatGPT.

1:18After you finish it, you should listen to his current perspectives recorded this season on episode 153, How Venture-Backed Tech is Transforming U.S. Healthcare, then compare his past predictions to today's reality. We know that you'll be impressed. Good afternoon, Bernard, and welcome to Fixing Healthcare. Great to be here. Always love talking about it. You're a passionate advocate for the positive role technology can play in healthcare. How do you believe it will be the solution for what ails the industry today? There's really two questions to answer. How does technology help in the United States?

2:02And then how does in healthcare, and how does it help the 7 billion people on the planet in healthcare? Those are somewhat different questions. I'll try and constrain it to the US healthcare system in terms of my answers. But I think it's globally applicable and the solutions are the same, but with a different tilt and different cultural context. In the US, we have relatively better care, but it's expensive. My personal view, the most expensive part of the system is expertise. And expertise can relatively be tamed with technology and AI. So not everybody in the world can get an oncologist. We can capture some of that expertise so each oncologist can do 10 times more patient care than they would on their own without that help.

3:23That applies broadly. We can talk about the major chronic conditions and the major expense conditions in health care. Happy to take them one by one or more broadly. But I do think there's huge leverage in replacing expertise. There's also testing and other things that can be dramatically better and cheaper. But I'll let you go in any direction you want here. So let's start with AI that you mentioned. AI can only be as good as the data that it has to work from. and in medicine today, outside of some of the visual examinations like radiology or dermatology or pathology. But in the majority of medical care, it's based upon the electronic medical record, which is not a very reliable source of data.

4:16How do you see this being solved going forward? So you are absolutely right. All attempts at using medical data, which was really generated mostly in the United States for billing purposes or systems of record, not for clinical care purposes, is not going to help AI. I think if you take a couple of different areas in primary care, unless you capture the exact conversation between the patient and the physician and translate like Google Translate does the language of the layman, maybe in many, many different languages, whether it's Spanish or English or others, into the language of SNOMAD in the healthcare system.

5:11That's sort of step one. Observing that conversation becomes a training ground for it in primary care. At the other end, something like psychiatric care. Unless you capture the conversation between the patient and the psychiatrist, you're not going to get AI to help guide that system. Now, in each system, each of these, and that same paradigm applies to oncologists, to cardiologists, to musculoskeletal care, you pick it. And these verticals are very, very important. But you're not going to get this through either what's called virtual care today, which is mostly telehealth and video is just a very bad medium for to capture this data.

6:09you have to set up systems that learn from each interaction from each text message back and forth each conversational fragment back and forth very few of the tele systems or the electronic systems are designed to do this and hence they're not really capable of ai learning there's a few rules-based systems, but I think those are inherently limited in what they can get to. Let me take psychiatric care. It's a really interesting example because you have to capture the conversation. And there's a few companies outside the US that have very good data sets of the actual conversation between a patient and their psychiatrist.

6:58That becomes a learning system. And you can learn from that, and the AI can learn CBT through conversation. There's also another large advantage. Part of it is how do you introduce these systems? It's not like you develop the system and start using it. My general view is these systems, whether they're in primary care or oncology or cardiology or psychiatric care improve about one, let's pick a nominal number, I'm making this up, 1 % a month for 60 months in a row and get substantially better and cheaper over that time. So we have to have tolerance for this kind of learning system. We also have to know how to introduce it.

7:54So in psychiatric care, for example, you wouldn't want to start doing CBT day one or talk therapy day one, but you could use it as a preliminary screening mechanism to see who needs the most urgent care and bring them to a psychiatrist. So screening, diagnosis, classification of urgency, those kinds of things will be done first, and the systems will learn from that. And then when they get more data, they get better and better in helping the psychiatrist. They may begin beyond just screening initially or initial determination of urgency of care needs, they may actually help in summarizing for the psychiatrist at the end of their session or guiding them on possible findings or diagnosis and using them to suggest directions for the psychiatrist to explore.

8:58So they may take an okay psychiatrist and make them much, much better in partnership. But in the end, if 7 billion people have to have a psychiatrist 24-7 available to them when they need it, whether it's at 9 a.m. or 3 a.m., this is the way we will have to go. And the same example applies to primary care. It will have to be done this way. the visual AI has made tremendous leaps forward. We now know that deep learning can do better than doctors when it comes to interpreting mammograms and difficult to read x-rays specific to pneumonia. And yet we're not using it in clinical medicine very much outside of research facilities.

9:54how are we going to be able to go from where we are now to making AI a powerful tool in the practice of medicine? Yeah. So previously I was just in psychiatry or in primary care or cardiology, I was talking about the interaction between the patient and the caregiver or the clinical care provider. In imaging, which is a very different beast, I would say today the technical problem or the technical question, it was 10 years ago, I said any radiologist not using such a system will in fact be killing people every day. unfortunately that is still happening through misdiagnosis and misdiagnosis or missed diagnosis rates are pretty high but it is starting to be used in orthogonal ways it's important keep in mind if physicians don't want to change their workload to leverage what they're doing, you're not going to be able to introduce these systems easily.

11:09But in this particular example of imaging, a company like Caption Health can, in fact, train anybody to take a cardiac echogram in a few hours in an FDA-approved manner. Why? Because the FDA, the AI, is doing the driving. Now it becomes much easier to do an echocardiogram in a primary care office without having a specialist present to do the echocardiogram. And you can do it with a handheld small ultrasound device. So using it in a way where training a person becomes easy and then And taking the echocardiogram becomes a non-specialist function, will reduce the cost of an echocardiogram done in a primary care office by 70 to 80 % quite easily.

12:06So it's the same AI imaging thing, but it's not being used to deliver the diagnosis. It's being used to train to essentially take the echocardiogram. So that's an example. The next step, and one of the problems in our system is the big providers, the GEs of the world, have been very slow to introduce these technologies in ways that are acceptable to radiologists or others. In an imaging, whether it's an ultrasound or an MRI, you could highlight the findings quite easily. but most of the systems being used today are older systems and these traditional movers are very slow at but there are small exceptions emerging so heart twister has a cmri a cardiac MRI that can be done in 15 minutes instead of an hour and 15 minutes.

13:18And it's a single button press on the MRI machine. Your typical CMRI exam is over 100 button presses by a specialist MRI technician. If you can reduce it to a single button press and no expertise needed in taking the CMRI and it gives you more information, it can be used as a routine screening exam for cardiac conditions instead of going to more invasive procedures, which is the alternative or a more expensive procedure that may cause four or five times more time in the machine itself. So it is being used at the edges. But I think we have to keep in mind anything that disrupts cost disrupts somebody's pocketbook.

14:14In it disrupts habits, the way workflow happens, the way people do their habits. And that changes slowly. My particular view of AI in clinical care will be it'll enter at the edges, like take an echocardiogram easier, faster, cheaper in the primary care office or do a CMRI instead of doing an ultrasound first and then go to a cath lab.

14:53sort of do more comprehensive stuff because it's both cheaper, more acceptable to the patient because it's 15 minutes in a MRI machine instead of an over in 15 minutes. So I think those kinds of things are happening at the edges, but the healthcare system is fairly slow to change their ways and adapt it. So the critical question becomes, other than these peripheral entries of AI into clinical care, how else will that happen? I am pretty optimistic that the most expensive conditions, take diabetes, cardiology, musculoskeletal, will be done in verticals outside of the healthcare system. Can you use a Livongo or a Hello Heart to reduce blood pressure by 20 millimeters without drugs?

15:54Could you do weight reduction by 8 % with a digital first offering? I think that's starting to happen. And these plans generally developed for enterprise self-insured employers are going into the plants and reducing their costs by handling 80 % of the conditions and in mental health, almost all the subclinical conditions for remotely from the home with vertically specialized solutions for the chronic condition or the acute condition like physical therapy. And that will start to reduce the cost. And then specialists will only be needed when the condition is severe or exceptional. And when it is, it'll become a learning opportunity.

16:55I think we see this happening in mental health some, obviously in diabetes and hypertension. S.O.O.D. Health is doing a great job of AI-driven muscular skeleton care. And we can go into the specifics of what exactly each of these are doing. AliveCore is doing a great job with remote cardiac care. But it is happening. It is happening slowly. And it will start outside the system. maybe take away the most profitable parts of these chronic conditions, which are the most expensive parts of these chronic conditions, away from traditional providers, and in some ways, hollow the system from outside in as payers and insurers go to these vertical solutions to handle deep conditions, do them mostly remotely, and over time, much more driven by AI.

17:53introducing AI in these verticals, whether it's primary care or musculoskeletal care or cardiac care, is much, much easier because the companies doing that are training new physicians or their physicians in the new workflow. and building a good partnership between the AI and the physician and letting the physician do much more accurately and with much higher standards of quality. And in the process, also training the AI. So imagine 10 years of learning by an AI, almost like a numerical intern would learn. I think that's happening. Let me focus on a different question related to the same AI functionality that bothers me greatly.

18:50You mentioned earlier an application that allows EKGs and rhythm strips to be created, and the application is very accurate. It creates the rhythm strips very well. But the last thing any physician wants is to receive hundreds of rhythm strips or EKGs from a patient, hiring nurses or someone else to interpret them is very expensive. We have the technology today to tell the patient you're fine or you're not fine. And yet none of the devices that I'm aware of has that level of medical presentation to the patient. Because if we had that, we could shift the entire paradigm, which says, I only need to see you when you're having a problem.

19:35If your diabetes is well managed, I don't need to see you often. But if it's not, I need to see you right away. And this machine will answer that question. And I can't find a company yet that has had the courage to be willing to step in there and provide that kind of information to the patient. Can it be done? Will it be done? What's it going to take? Yes, it can be done. It is being done. It is not widely known. But this is a great example. So let me dig a little bit deeper than the shallow explanations I've been given. Today, AliveCore can diagnose six conditions per FDA-approved diagnosis. That means they do not need a cardiologist to diagnose these six conditions that you can read off and EKG.

20:27Today, AliveCore has a two-lead and a six-lead ECG available for self-use at home. They will be introducing a 12-lead at-home use device, so you get all the data. But it really won't be needed because hopefully soon they'll have a dozen conditions. If you look on the Wikipedia page for an ECG, the dozen conditions that could be detected, they'll want to do them all, and they won't need a physician. Now, let me ask you, how many on average, so LiveCore has 110 ,000 subscribers that are self-paying for a monitoring service. Let me ask you your guess on how many ECDs a year are these patients doing on average?

21:24Take a guess. No idea, because some of them need to do one and some of them need to do 365. So I can't tell you. It depends upon the risk of the patients. So the average is closer to 365 a year than one. In fact, it's well over 20 a month. So almost every day, these patients are taking ECGs, some multiple times a day. And it does create the problem you mentioned. No physician's going to take a single patient and do multiple reads a day. So algorithms have to do it. And a live code does this very, very effectively. It's effective enough that these patients, almost all cardiac patients, post-diagnosis with cardiac disease are taking these very, very frequently and doing this very effectively without swamping the cardiologist.

22:34Let me ask you, Vinod, and this comes out of the self-driving car industry, and we can debate whether those cars are safer or not. My belief is they're going to be much safer than humans. And yet when a car crashes that's self-driven, people react as though it is a plane crashing from pilot error, ignoring all the mistakes that people make. Why won't the same phenomenon happen when AI does analysis of data better than humans? But why will patients forgive the AI when today they forgive the doctor but not the machine? This is a serious problem. problem and, in fact, a serious problem for Western AI.

23:20A friend of mine said to me, in China, if you have a self-driving car and it kills somebody, the government is very tolerant of it and will say, we want our technology to develop faster, so we'll take the cost of a few deaths. In the West, that's not how things are practiced. So almost certainly, almost all AI for the next five, 10 years will be delivered, mediated by a clinician licensed to practice at that level of medicine. So if some things are registered, nurse practitioner may do, but it'll be the highest quality that can be offered at the level of the license of the practitioner. And I think it will be overseen by a human who can then do more.

24:18So let me take the example. I went through some detail in the case of AliveCorp. Very, very unlikely that But, you know, clinician could do 20 ECGs per patient per month or 30 for that matter, because there's patients who do that. They can't afford it, so they can't monitor closely. But if it is reduced and delivered to a cardiologist to intervene when needed, that's how this will happen. Take musculoskeletal care. What SORD Health does, it's a beautiful example of using AI. Post-surgery, say a knee surgery, you want to do rehab and physical therapy is prescribed. Their AI does the therapy at home with the patient at any time of the day or night.

25:23And this is a stunning statistic. Last Christmas Day, 50 % of the people on physical therapy, prescribed physical therapy, it's usually a 6 to 12 week program post-surgery, did their therapy on Christmas Day. Think about it. Because of the accessibility, because anytime you're free, the AI is available. But here's the good part and the catch. The therapy plan post-surgery is done by a physical therapist in consultation with, say, the surgeon who did the surgery. The physical therapist then monitors the plan after each session of therapy done by the AI and modifies it. So a one-hour task of watching the patient and saying, put more weight on your left leg, that task can be done by an AI.

26:25But the therapist reviews progress and modifies to adjust the plan, which goes to the AI, which can guide the patient. So that's a classic example of a closed loop in which the human is spot spreading and the delivery of AI becomes safe and much more convenient, much more cheaper, and much higher quality. The patients are doing many more sessions when they can do it at nine o 'clock at night when they got free or when the football game ended, instead of saying, I need to schedule this physical therapy session in the middle of my workday three weeks from now when I don't know what emergencies I'll be dealing with.

27:11And that's why you get much more frequent therapy and much more frequent monitoring. And the results are very clear in the case of SORD Health. The progress and the efficacy is better than if a physical therapist was doing full-time therapy for these 12 weeks with this post-surgical patient. Having grown up in rural Iowa, rural health care and access to rural health care for Americans is something I'm very passionate about. Many people in rural areas still don't have access to true high speed Internet and often feel completely forgotten about. They struggle with food deserts, access to care, high poverty rates.

27:54As a venture capitalist, you probably watch innovation happening all across the world. I even recently heard of a company in Africa that delivers blood for blood transfusions over long distances and hard to reach areas via drone. Are there any innovations and technologies that you're seeing springing up in other parts of the world that could be used here to help improve rural health care? And if so, what? Yeah, frankly, all of these will apply to rural health care. I'm particularly excited about rural in India. Yeah. Wherever healthcare depends on deep expertise, which is generally expensive, think oncologist or a brain surgeon, and you can do it remotely, care quality will improve and care costs will decline.

28:41And accessibility will improve because a patient won't have to go from that with a small town in Iowa to a major medical center and travel for three days for an appointment. So I do think that will happen and it will be very, very beneficial. There's a beautiful health system in Alaska called the NUCA Health System. And because expertise is harder to get, they have broadened their primary care. And their definition of primary care is much broader than anybody else's. But what happens as a result, their primary care may be on per member per month, be three to four times more expensive than a typical health plan's primary care system.

29:35But what they have found is the total cost of care for those patients declines. and more is handled by the primary care physician. Now imagine that health system where the cost of primary care is made cheaper by a company like Cure Eye, one of my sub companies that Bob is familiar with. And it becomes$5 or$7 per member per month, much more usage, three to four times the usage. and broader than what NUCCA provides today for$100 a month per member. And it has the downstream implications of reducing cost of care. That's really good news for rural health care, remote health care, but also urban health care.

30:28The same will happen no matter whether you're rural or urban. Let me go back to the electronic health record that you mentioned earlier. It's always seemed to me that if the large manufacturers of the applications would open their APIs, application programming interfaces, that third-party developers could come in, could transport it off of the computer onto a tablet, could add applications. We're in an application world on mobile devices, and yet the EHR is still on large computers and computing systems. Do you agree that by opening the APIs, this could be done? And if so, what's it going to take to make it happen?

31:09Can it be done? Absolutely. You know, could a caption health algorithm run on a Philips ultrasound machine? Absolutely. It's trivia. Could be done in a month. Now, it may take a little longer to go through the regulatory process. But once they're regulatory approved, they're approved on the Tereson device. They're approved on other devices. it's pretty easy to add another device it works on. Now, what will happen and is happening is the major players like the Philips protect themselves because they don't want to disturb their business. And I don't even think it's a smart business decision, but they're being very, very parochial, saying we won't open up our interfaces, and that's a shame.

32:03but then their competitors are doing it butterfly is much more open than philips and so the caption software will run on a butterfly handheld device and it's much more accessible uh maybe not the you know it's two thousand dollars maybe it's not the same quality as a$50 ,000 Philips ultrasound machine, but that is happening. That will put pressure on the Philips of the world. Now, I can tell you that the Hotwista CMRI algorithm work on Siemens machines. And Siemens is helping them and has helped them open up their machines to run these algorithms. So that is happening. It's happening slowly. We have to keep in mind, seven years ago, when I wrote my piece called 20 % Doctor Included, nobody believed AI could do any imaging either.

33:11Maybe two or three years ago, people started to believe it. And then last year, I remember JAMA, one of the JAMA editors, quote something like, we are no longer accepting publications that claim AI algorithms are better than humans. That's sort of proven. No more innovation in that statement, whatever the condition, whether it's brain imaging, cancer imaging, x-rays of the lungs, you name it. but it takes time. After JAMA said that last year, or the JAMA editors said that, you're starting to see movement from the traditional players. Now there's a backlog of machines installed in offices, medical offices and hospitals that will take a long time to flow through the system.

34:11If you get a self-driving car today, It's only new cars that will have this, and the old cars will be around for 15 years. The same is true of equipment and hospitals. Now, I do think the more innovative systems will take the help of AI in these vertical approaches and will do it outside the system at much lower price points, especially in plans like ACOs and Medicare Advantage and others, where the cost burden relies on them, and they can get payback on these new investments very, very rapidly. I think the business implication is it'll be the newer players who adapt these technologies, disrupt the price points in economics, and will slowly eat into traditional systems, and that will go bankrupt.

35:06This is sort of the model that happened with bookstores and Amazon. The bookstores ignored Amazon for a long, long time as it slowly ate into sales with this new model. Later, they adapted this model, but mostly too late. Think of your favorite bookstores that have gone under. Barnes & Noble, whether they're big or they're small. Will the same thing happen to the electronic health record companies, the ones that refuse to open up their application programming interfaces? Well, there are clearly newer efforts in EHR companies, but EHRs are different because they are billing systems and they're systems of record.

35:56They're not clinical care systems mostly. And you, of all people, know the frustration of using those systems. But because they have so much legacy data, they have a lot of stickiness. So I don't think they go out of business or they get out of, say, UCSF or Stanford very easily. I think both use the Epic system. They may open up their interfaces, which will help a lot to add stuff on top. But you are starting to see systems that look like EHRs that are clinical practice systems in front of the clinician while they're with the patient during that interface. And I think those may be working set systems, not systems of record like an EPIC is.

36:52There won't be billing systems, but there'll be the clinical interaction systems. You're starting to see big things of that coming from the old model of scribes. Well, the need for documentation into these legacy systems has created demand for scribes. The scribes are much easier to augment or replace with AI systems. And a company like Robin Healthcare is doing that. And that then introduces a new kind of intermediate health record that then writes into your epic to the open interfaces. So I think something like that will happen. There's other startups developing what I call working set systems without trying to displace or throw away the old systems.

37:45Mostly because they're sick. they may have 20 years of patient history in a system. And you can't throw that away unless you can move it to a new system. I'm continually amazed that we pick medical students based upon their ability to take tests that require memorization of arcane facts. I understand that in the 20th century, we did this because you need a 50-pound backpack to carry all the medical information with you, and it would take an hour to research it all. Now we have iPhones and applications sitting in our pockets. Should we be choosing medical students differently? And how should we train them differently for the 21st century?

38:30Absolutely. And I said this about 10 years ago, and I wrote about it seven years ago in my paper called 20 % Doctor Included, which by the way is a 100-page document on how this transit, not only what will happen, but how the transition might happen in stages. And I say the following. Medical students are picked on IQ. If you want to get into Stanford Med School or Harvard Med School, you have to have IQ off the charts. And IQ often does not coincide with EQ. I think humans will be providing the human element of care. And that means we select people, students for their EQ, and we should adapt some of the techniques the USC film school might use in selecting students for high EQ, because the IQ part more and more over the next 10, 15, 20 years, and in my document, I painted a 25-year picture of the transition to seven or eight generations of machines and AI, the humans will be more and more left with the EQ element.

39:55And we should change what we select. Last year, Vinad,$89 billion of venture funding went into healthcare startups, most of it into care delivery, particularly care delivery systems using telemedicine. Is this exuberance justified in your opinion? Look, the venture business always over-invests in hard categories. The characteristic of venture startups is most will lose money, but more money will be made than lost, which says a small percentage of these will win, but the winners will make more money than the losers lose. So net-net, broadly, if somebody is broadly invested, they'll do well and make money on their investment.

40:46So, imagine after$89 billion,$80 billion gets lost. Companies, that money gets lost as investment. But the$9 billion turns into$900 billion, creates another Apple or Google. That's the model of venture capital. And so, more money will be made than lost, but most companies will lose money for investors. So what advice, given that, do you have for people who want to be entrepreneurs in the healthcare space? I think there's lots of room to be entrepreneurs. And every part of the system that is expansive can be disrupted, almost every part. and it doesn't matter whether it's care delivery as we mostly talked about testing whether it's cmris or cardiac mris or cardiac ultrasound or blood tests you know you can do 20 tests for the same price as one test much faster and cheaper with a machine like the Gen-like.

41:58Ord 20 should be available to a physician to use. He shouldn't say, hey, check your hemoglobin. Oh, your hemoglobin's okay. Three weeks later, after you do testing through Quest Labs, and then he says, oh, I should check your PSA or something else. why are you feeling so low energy? You should have all 20 because it costs the same delivered to you and a GenLite can do that today. In fact, a GenLite, 10 minutes in the physician's office over before the physician's visit with the patient can do 85 % of the tests the physician might order prospectively and they can do it going at risk and take the risk the physician may not order any test because the economics of it work because testing is so cheap.

42:57A company like Opentrons, which is doing COVID testing, is introducing a panel that for cheaper price than a Quest test, they will scan not only for COVID, but in the process, also the top 20 respiratory diseases. So just because you don't have COVID, you still want to know what you got. And that should be done as one panel for a cheaper price than you can get from a Quest today. And this is just being delivered, which is why OpenTrons is doing the vast majority of the testing in New York City, doing weekly testing in New York City schools, because the cost is so low. They're doing full PCR COVID testing at a higher quality than any of their traditional comparators at$28 a test for a PCR test.

43:53And in that kind of a price, they haven't set pricing, you can also do all 20 respiratory diseases that might be in the patient sample. And that's what will happen. I think the diagnostics will change. Same is happening and we don't need to go into things like drug discovery using AI, biologics through AI, antisense oligodesign through AI. Also in cell therapy and others, AI is starting to play a huge role. And I won't spend time on that today in the absence. We're running out of time. Bias and racism in healthcare have been well documented when it comes to people of color. How do you see technology being helpful?

44:44And are there pitfalls that we need to be careful to avoid? So first thing you just acknowledged for me is bias is built into healthcare. But it's built into healthcare delivers by humans because humans are biased. And we should acknowledge that. I think AI systems will capture these biases, but they will also surface these biases. Just because we have a company doing recruitment with AI, they discovered learning from human interviewers of kids on campus for recruiting for a large corporation, that humans were very, very biased. then because they discovered it in a quantifiable and measurable way they were able to eliminate the bias so i think the ai systems will surface explicitly the biases something that hasn't been surfaced there is no practitioner i know who will say hey i'm biased by against people of color but it does show up in the data.

45:56And hence, it gives you an opportunity to fix it. And it is being fixed in AI systems. And there's a lot of attention on it. I agree completely. In fact, I wrote a paper about that, that the bias supposedly in AI is just the bias in humans, and that actually AI has that potential to surface it and to be able to correct it. So let me ask you the following question. In December of 2019, before COVID came as sure, the federal government said that healthcare costs were gonna rise five to 6 % a year, every year for the next decade. That's$2.5 trillion. If we just forget about the question about inflation and supply chain and all the issues that are out there, but just focus on the intrinsic healthcare cost increases.

46:42Is your projection given technology that a decade from now, it will go up, the amount that was predicted go up more or somehow go up significantly less? That question is obviously a very complex question, this needing its own podcast. But I would say the following. If the incumbents apply this technology, the incumbents have no reason to disrupt their revenue. So costs will keep doing what they're doing. If new players disrupt the incumbents using new technology, as I've described, then in fact, I think we can change that cost trajectory. Cost delivered through SORD Health will be 10x cheaper than delivered with a traditional physical therapist for musculoskeletal.

47:46Care delivered by a live core for cardiac care will be 10x cheaper. And if it becomes the standard care for all patients, it will probably be that for self-insured employers, payers, Medicare Advantage programs. Then those costs, pieces of costs will be declined. So disruption of traditional players will be necessary for this to take hold. Otherwise, they have high cost structures and they have zero incentive to reduce their revenue, which is what reduced cost means to those players. Vinod, thank you so much for being our guest today. Your insights into technology, entrepreneurship and the future of health care are wonderful.

48:36Our listeners will have much to think about. And hopefully our nation can move forward in the ways that you describe and make America, once again, the best healthcare system in the world. Thank you. Thank you. Robbie, what do you think about what Vinod said? Jeremy, Vinod is a true visionary. He can see the power of technology long before others. His view that it can help make all clinicians as good as the best is inspiring. and his hope to bring the best healthcare to people across the globe is wonderful. As he said, the current healthcare system is predominantly Monday to Friday, 8 a.m. to 5 p.m., which unfortunately is true, but not what people want and need.

49:26Technology can expand access to 24 by 7 while minimizing the impact on care providers. A decade from now, I believe physicians will rely on technology tools to help them provide at least 50 % of the medical care they currently deliver. If there's one place I may see technology slightly different is that I do believe there will be a major role for telemedicine. It won't be just a doctor seeing a patient virtually rather than in the office. It will be the ability of video to bring groups of clinicians together, to offer expertise from national and global experts rather than locally, and to offer specialty expertise to primary care doctors while the patient is still in the exam room.

50:20Phrased differently, when combined with many of the other tools that Vinod described, it can eliminate the restrictions placed by distance and time, and as a result, drive quality up and costs dramatically down.

50:50Follow us on LinkedIn, Facebook, and Twitter at FixingHCPodcast. Thank you for listening to Fixing Healthcare, Breaking the Rules with Dr. Robert Pearl and Jeremy Corr. Have a great day.