A Second Opinion with Dr. Michael Abramoff | A Second Opinion Podcast

A Second Opinion with Dr. Michael Abramoff

Dr. Michael Abramoff is the founder and executive chairman of IDX Technologies, the artificial intelligence diagnostic company that became the very first company in any field of medicine to get FDA approval for completely autonomous artificial intelligence. His remarkable instruments can independently analyze images of the retina for early signs of diabetic eye disease that thousands lose their vision to each year. This is the first in a three-part series on innovation in medicine and health.

Dr. Michael Abramoff:   Very patients have an eye exam. We put the AI in, and now they can see that I can provide it, if they need them the same day. So we went from months of wait time to same day for patients today.

Bill Frist:                       You’re listening to A Second Opinion, your trusted source in gaging at the intersection of policy, medicine, and innovation, in rethinking American health. Dr. Michael Abramoff is the founder and executive chairman of IDX Technologies, the artificial intelligence diagnostic company that became the very first company in any field of medicine to get FDA approval for completely autonomous artificial intelligence. His remarkable instruments can independently analyze images of the retina for early signs of diabetic eye disease that thousands lose their vision to each year. This is the first in a three-part series on innovation in medicine and health. I’m your host, Senator Bill Frist. Welcome to A Second Opinion.

Bill Frist:                       Michael, artificial intelligence, machine learning, the eye being the window to the body, autonomous… What does all of that mean today? And start at the basics and we’ll come back and talk about what IDX does, but just in a general overview, how does it all fit together and why is the eye so important?

Dr. Michael Abramoff:   Yeah, great question. So artificial intelligence or augmented intelligence as many like to call it, is something where… something, a highly cognitive task, a very demanding task is now done by a computer, but it’s always a narrow task. We’re not talking here about general intelligence, anything a doctor does in this case, because we’re talking about healthcare, is specifically about narrow diagnostic or therapeutic tasks. And machine learning is a technique to create artificial intelligence. You can make artificial intelligence just by having the rules and prescribing what the computer should do in certain cases, but it gets very wieldy and out of control.

Dr. Michael Abramoff:   And actually, it doesn’t work really well. So we learned over the last decades, building… trying to build artificial intelligence systems outside healthcare, and later inside healthcare, that the best way to do it is machine learning, where essentially, you teach the computer by examples where to look for, what it should make a decision on and what not. Then, you mentioned autonomous. Autonomous is a challenging term, but if you think of autonomous cars, this may be more obvious. It’s really a car that drives itself without someone holding hands on the wheel. And so in healthcare, it means a medical system, a diagnostic system or a therapeutic system that makes medical decisions by itself. It doesn’t mean that it is not… that the patient doesn’t see the doctor anymore, but for this specific task, rather than sending maybe the patient somewhere else for a specialist, is now done by the AI and autonomous AI.

Dr. Michael Abramoff:   Of course, people worry about this. Is there risk of harm? Are these systems not biased? Are they fair? Do they actually improve patient outcomes? So I think it’s really important that we do it right, so AI needs to be done the right way. I think you’ve seen with some examples that if you don’t apply AI the right way and starting from maybe ethical principals, people get concerned, there’s a lot of worry right now about some users of AI, so we don’t want that because why are we trying to make the system… Why did we build these systems? It’s because if you have an autonomous AI making a decision that previously only a human doctor, especially if a very expensive specialist could do, you can save costs, you can improve quality, and you can improve most important, access. Because now, I’m a retina specialist myself. There’s maybe 3,500 like me in the US, but now I can make a diagnosis and have an AI make a diagnosis that’s more accurate than me as a doctor anywhere.

Dr. Michael Abramoff:   And I like to give the example of a grocery store, where you can now go to the grocery store in Delaware, and in the back, is there autonomous AI that’s going to make a diabetic eye exam in this case? And so, it allows you to have access to very high quality care at a very low cost anywhere the patient is.

Bill Frist:                       And so, let’s go a little bit deeper in that and let’s take the example of the eye.

Dr. Michael Abramoff:   Yes. Yes.

Bill Frist:                       Which is what we’re talking about. What makes the eye such an attractive organ in terms of making a diagnosis and then prescribing a therapy, which is what the word autonomous, the way you explained it to me, will actually do? Why the eye?

Dr. Michael Abramoff:   Well, I’m biased because I’m an ophthalmologist. I love the eye, that’s why I chose becoming a doctor and doing my residency and my fellowships. And so, it’s such a beautiful organ because it’s part of the brain. As an embryo when we develop, it’s part of the brain. It forms as part of the brain, and… but what is different from the brain, the brain is really well-protected in the skull. They say you need scans and MRI or other radiation equipment to look into the brain. The eye, because it’s made to see, allows light to be put into it and you actually image it.

Dr. Michael Abramoff:   So, now you can look at brain tissue without using radiation or injections or something expensive and maybe dangerous to the patient. And so, it allows you to look at the neural tissue, which is the brain, but it’s also in the vascular tissue because so many diseases of the brain and the neural system are related to how the blood vessels behave and the circulation. But we know from many, many studies that cardiovascular disease shows in the blood vessels in the eye. Also in the brain, but that’s hard to see. But in the eye, you can have a relatively local camera, no radiation, just a flat image and you can see neural tissue, brain tissue, and vascular tissue. And so that’s why Alzheimer, cardiovascular disease, diabetes, multiple sclerosis. So many diseases that show up in the neural system in the brain also show up in the retina in the eye.

Bill Frist:                       So, IDX, and we’ll come back to the company, the entity that you have founded, I want you to explain what it actually does, but typically I need an eye exam, especially at my age and diabetics, we’ll come back and talk about the subpopulations, and I would go to my primary care physician and say, “Oh, remember that you need to have an ophthalmologist look in your eye,” and I’d say, “Okay,” and she would say, “Do you want me to make the appointment for you?” And she’ll make the appointment. I’d hear about it a week later, and it would show up at some time, and then I’d go sit in that office and then I would talk to an intake person and I would talk to the nurse, and I would be taken back to a room and I’d have drops put in my eye and then I’d wait 30 minutes, and then I would have the ophthalmologist to come in, look in the eye and look into the machine.

Bill Frist:                       And then at the end of it, they would basically say, “Things are okay,” and then I would go home, get in the car to go home or maybe some drops to reverse and the sunglasses and the dark glasses, all of which is very important, because if the ophthalmologist picks up something to me, it could show what’s going on in my heart or metabolically by diabetes, just by looking in the eye. So, what does IDX do? Take me to the world of IDX and how it’s different?

Dr. Michael Abramoff:   And by the way, your experience is very common. What you just described is. So, but IDX was founded on very broad principles of lowering the cost of healthcare, improving quality, and improving access through autonomous artificial intelligence. More importantly, we do that through taking specialist diagnostic and therapeutic expertise from a few places where you can go for an experience like you, to everywhere where the patient already is, which is typically at home, in retail clinics, other places like that. Because like I said in the beginning, it allows you to do this very narrow task at any place where you can put a computer and a camera. And so, that is what IDX is about. In the eye, in the skin, in the ear, any place where they’re especially scared that we can put into primary care, retail clinics where the patient is, that’s what we do.

Dr. Michael Abramoff:   It’s interesting that we have a special way of creating these AIs, because we build it being as close as we can to the way clinicians, as far as we know in terms of neuroscience, as the brain of a clinician works. So we try to mimic it as much as we can. There’s a number of advantages that we could talk about, and I can explain.

Bill Frist:                       Yeah, yeah. Well, let’s go to that. But then, so the experience I just described, and you said a camera and a computer, so I would be able to go and just say to a retail clinic of some sort, and how long does it take? What does the machine actually do? Am I seeing a nurse? And who’s coordinating this? Just walk me through the process real quick.

Dr. Michael Abramoff:   Yes, and so, a few things to start with, it’s a prescription device, meaning a physician or provider needs to give the order. The patient experience is that in this case, it’s for diabetes. It doesn’t replace a full eye exam. It’s specifically for someone with diabetes to know whether they have a problem with the retina from the diabetes, called diabetic retinopathy. Most important cause of blindness in the US.

Bill Frist:                       Just statistically, how many people have diabetes?

Dr. Michael Abramoff:   30 million people have diabetes and we know that all of them are at risk of going blind from diabetic retinopathy, and that’s why the diabetic eye exam is so important. But studies show, the most recent study show for the Medicare population, only 15%, 15% of who need it every year, they actually get it. So 85% are not getting the diabetic eye exam, and that’s not good because we know the diabetic eye exam allows you to catch the disease early and treat it in the early stage when visual loss is still reversible.

Bill Frist:                       And prevent the blindness?

Dr. Michael Abramoff:   And prevent the blindness. And so, if you let it go on, they come in too late. There is detachment and other problems, and then we can try to do what we can but usually, it’s very late and there’s irreversible damage.

Bill Frist:                       So this procedure is a screening procedure, or is it-

Dr. Michael Abramoff:   We call it a screening procedure, but it’s early diagnosis.

Bill Frist:                       … [crosstalk] early diagnosis, and there are about 25,000 people that go blind every year.

Dr. Michael Abramoff:   It’s actually more, the most recent studies show it’s more like 50%, sorry, 50,000 people a year go blind directly as a result of diabetes.

Bill Frist:                       Wow!

Dr. Michael Abramoff:   Almost entirely preventable if caught early, because we have really good treatments now to treat this disease, which is what my day job as a retina specialist… but back to the patient experience, let’s say you have diabetes and you go for your diabetes management to your primary care provider and endocrinologist, you would show up. They do your vitals, your blood pressure, they measure the standard things, the A1c, and then you go to sit in a chair. The IDX-DR camera is there, so a robotic camera, essentially anyone with a high school graduation can operate it, it’s really easy to use. It takes a few minutes, it takes a few flat images and then to diagnosis itself takes about 30 seconds. So the AI does a really fast job. In five minutes, maybe eight minutes, you’re done. And what is exciting is it’s instantaneous.

Dr. Michael Abramoff:   So, it’s autonomous, but instantaneous. The diagnosis is already there, when by then, you as a patient will show up and discuss with your provider that you need to stop smoking if you do, but you don’t. And eat more vegetables, and in general, yeah, you’re doing great with your diabetes and we just took your diabetic retinopathy exam and it’s good. Which is great, and will happen in most cases. In some cases, the diagnostic report from the autonomous AI will show you actually have the disease, refer what we can’t manage, and then it’ll say, “Well, you actually have the disease. You are high risk of going blind, you really need to see an eye care provider now.”

Bill Frist:                       And it’s the machine that is doing this-

Dr. Michael Abramoff:   Making the diagnosis.

Bill Frist:                       … making the diagnosis. So, basically once you’re looking, the machine flashes or done, within four or five minutes, you get a print out that says, “You’re free and clear,” or this is the diagnosis and this is what you need to do. There’s not a physician at that point involved in terms of the autonomous performance of the read out?

Dr. Michael Abramoff:   There’s no physician involved. So the operator’s doing all of this, but typically, the operator would not explain these results. The physician or provider needs to discuss because the patient, if they hear, “Well, you have this disease that you may go blind from,” they will have questions, and you will need to explain it but again, the expertise of the doctors here explaining what it means for the patient in terms of their diabetes. [crosstalk] Go ahead.

Bill Frist:                       And is this better? So this could be in the primary care doctor’s office? So in my whole description of getting prescription, going across town, waiting in an office, [crosstalk] schedule, dilatation and all. [crosstalk] So that’s pretty powerful, but how does the machine do that? It’s a computer, is there a continuous learning? Is that scan being compared with millions of other scans? Are there algorithms? How does the machine do that?

Dr. Michael Abramoff:   So now we’re getting to what you asked earlier. Essentially, artificial intelligence and machine learning, how do you actually do that? I already mentioned, we tried to mimic what the brain of a clinician like me does, but then better. And so, if I look at a patient’s retina, I look for markers of disease, hemorrhages, exudates, abnormal vessels, other things that tell me this patient has diabetic retinopathy rather than not having it. What we did was build detectors, many detectors for hemorrhages. So these are computer, with computer codes, computer algorithms that look at an image of the retina, so it’s like a color image. And they look for hemorrhages, and they either find them or don’t. And then, you have other detectors for exudates, just like my brain works because it looks for these things.

Dr. Michael Abramoff:   And there’s a number of advantages with that that we’ll go into, and these detectors, we typically make, but there’s different ways of making them. You can do it mathematically, but typically we do it with machine learning, where we take examples of hemorrhages, millions of them and we slowly train an algorithm to detect these. For example, the hemorrhages really well. So, it is shown an image of a hemorrhage, and then an image without a hemorrhage, part of the image. And then we teach it, “This you should detect as a hemorrhage, this you should not detect as a hemorrhage.” And if you do that a million times, it essentially, the algorithm contains all the knowledge essentially statistically learned from the data. That’s machine learning, so I’m explaining a lot now.

Bill Frist:                       No, that’s really, and it makes it hugely clear because typically, and I think in terms of the individual physician right now that’s going to come to how good the machine is and how good the physician is, is my question. But if you use specificity or sensitivity or the combinations of accuracy, is the machine which is quick and convenient and easy hugely less burdensome, but isn’t it better to have a human being making the diagnosis? Surely, a physician is more accurate than a machine?

Dr. Michael Abramoff:   And no, and so-

Bill Frist:                       No? Okay. Now-

Dr. Michael Abramoff:   [crosstalk] I consider myself a pretty experienced retina specialist, and it’s better than me. And so I’ll explain you why. And so, so you compare it to physicians, so typically what you see in studies of AI like this is that it’s compared to physicians. If you ask two different physicians to diagnose a patient, typically they will differ in about 30% or more of cases.

Bill Frist:                       Wow, wow.

Dr. Michael Abramoff:   I know that from me and my colleagues. We differ in some cases. Already, if the AI and the [inaudible] is agree, you don’t know whether the AI’s wrong or the clinician’s wrong.

Bill Frist:                       Right.

Dr. Michael Abramoff:   You just don’t know at that moment, so rather than comparing to clinicians, we look at patient outcome, because that’s what the patient cares about, and what we should care about as a healthcare system. And so, that’s easy to do in the case of an ear infection, because you treat it, the AI does it and a day later, you know whether the patient is doing better. You can ask them, “Are you doing better or not?” That’s the outcome, right?

Bill Frist:                       Right.

Dr. Michael Abramoff:   In a chronic disease like diabetes or diabetic retinopathy, you need to wait 20 years for that, so that’s why you say, “Well, how are we going to do outcome?” Well, for 50 years in retina, we have been doing studies looking at the natural outcome of retinas with very specific biomarkers in terms of hemorrhages, so we know exactly if we see this in the retina, the risk of going blind five weeks later is that. So, we literally can say, “This image means that outcome.” We call it the surrogate outcome, that’s used a lot in drug studies. Was never done in AI, and so we compared the autonomous AI to surrogate outcome, and that meant because we had early studies that compare surrogate outcome to physicians, to experienced board certified ophthalmologists, that showed that the sensitivity was about 30-40%. AI sensitivity is 87%, and it shows you that if you compare these two studies, that in this case it’s more accurate than the ophthalmologist.

Dr. Michael Abramoff:   The specificity of the ophthalmologist is about the same as the AI. More interestingly maybe that we did not only measure sensitivity, which is safety, specificity, which is efficiency, how many… but also a third, which came from almost ethical principles of equity. Because like people recognized early on when they started building AIs, you can make easily an AI that is 100% sensitive. You just make everything abnormal. It’s 100% safe, because everything is normal, it’s useless because the specificity will be zero. And so they said, “You need two,” like you already mentioned, “You need two measurements, two end points.” Sensitivity and specificity both need to be high for the system to be useful. But then we realized, you can make an AI that works for 1% of the population really well, it’s 100% sensitivity, 100% specificity. But it’s useless because it doesn’t work on 30 million people with diabetes of different races, ethnicities, ages, we need to show that this AI, autonomous AI works in the vast majority of cases. That became the third endpoint.

Bill Frist:                       And so where are we now? So, put all that together, I’ve got it. It’s a lot more convenient coming in. It’s instantaneous, you’re not waiting for results. We’ll come back to the affordability of it as well. But now you’re telling me that the machine can be the ophthalmologist in terms of the overall measures that are used. And I’m asking this because as a patient, do I trust the machine or do I trust the doctor. You trust them both, but the machine is equal to or better than your typical specialist.

Dr. Michael Abramoff:   On this very narrow task, it doesn’t replace an eye exam. It’s not for everything that an ophthalmologist or an eye doctor does, but on this very narrow task, in terms of sensitivity, yes. In terms of specificity, the doctor is still better. But overall, it’s interesting to see that if you look at the numbers, we need… The AI needed to hit these three end points, I as a retina specialist would not be able to be FDA approved, can you imagine?

Bill Frist:                       Yeah.

Dr. Michael Abramoff:   So I’m just not good enough.

Bill Frist:                       You don’t the standards of FDA approval?

Dr. Michael Abramoff:   I don’t meet the end points.

Bill Frist:                       All right, so let’s talk about that. For this to come out, luckily in this country we have an FDA, a Food and Drug Administration. It looks at safety and some efficacy. Where does this AI world today, in this podcast, as you and I have talked about, we looked at policy, we look at innovation, which is we’re talking about both of those, and then we looked at the medical or clinical care which we’ve talked about. Let’s go back to the policy. Does this have to be FDA approved? And if so, I would assume so, but if so, where is IDX in getting that approval? And what have the findings been? How hard is that process? You always hear about the FDA is so difficult. Tell us a little bit about that.

Dr. Michael Abramoff:   FDA, this IDX-DR was authorized, as it’s formally called, by the [inaudible 00:22:29], which is a complex term on April 18th, 2018. So that means FDA approved since then.

Bill Frist:                       And then how many autonomous sort of learning diagnostic machines in other fields have been approved?

Dr. Michael Abramoff:   There’s only one autonomous AI that has been approved so far, but we are aware of many others that the FDA is currently working on clinical trials, we have several where we discuss with the FDA when to start a clinical trial.

Bill Frist:                       So you were the first-

Dr. Michael Abramoff:   We were the first.

Bill Frist:                       How do you say it? The first autonomous what?

Dr. Michael Abramoff:   AI.

Bill Frist:                       A artificial intelligence in the medical field?

Dr. Michael Abramoff:   Yep.

Bill Frist:                       That’s pretty impressive. Congratulations.

Dr. Michael Abramoff:   Thank you.

Bill Frist:                       And that comes from a background of what? We’ve talked a lot about the machine and sort of where we are today, and we’ll come back to that. But where in the world did you come up with this idea? And tell me a little bit more about your background.

Dr. Michael Abramoff:   Sure, from my accent, you can hear that I was not born in Iowa, which place I love to [crosstalk 00:23:28]-

Bill Frist:                       Where you live today-

Dr. Michael Abramoff:   Where I live today, where IDX is.

Bill Frist:                       Yeah, yeah.

Dr. Michael Abramoff:   Originally from the Netherlands. Trained there as an ophthalmologist, but I also had a background in computer engineering. So in fact, I worked for many years in the software industry, and so I was always interested in combining how computers work with the human brain, and with medicine. 30 years ago, I was in Japan during neural metric research in Tokyo, in the post doc. And I sort of tried to combine these two interests, and see how they could benefit patients, and that was interesting because everyone is always telling me, “That’s a great combination. Computer engineering and medicine, that’s a match made in heaven.” But I could never seem to make it work, something that’s compelling. And then, I saw as a resident in ophthalmology, I saw so many patients that either came for the diabetic retinopathy exam and I did everything wrong, like you described. Everything is good, come back next year, so eventually you don’t do that anymore or they were too late, and I realized well, maybe a computer can do better than what I’m doing here.

Dr. Michael Abramoff:   And so, that started work on my PhD, so I have a PhD in computer engineering of AI. And so, I’ve been doing this for a while, came to the US, got NIH funded, and it’s interesting. I thought in my naivete that if you do scientific publications, people say, “Wow, this works, it’s great. Let’s do it.” That’s not of course, how healthcare innovation works. But as a researcher, NIH funded scientist is [inaudible] what you think. And so I applied the neuroscience I’d learned even 30 years ago, of neural networks, continued applying that, but then realized well, there’s this magic company out there that if I have some patents, they will come in, license the patents and they will make it work. That never happened.

Bill Frist:                       That didn’t happen either. Yeah, yeah, yeah.

Dr. Michael Abramoff:   And I thought well, I will do philanthropy. I will find some philanthropies, and I will demand money and we’ll go through this process, because I realized by then, the highest standard to me is FDA. In the world, it’s the highest standard. And I realized that I want to get it through FDA, rather than come through some other country where there’s no regulation and trying it out there. So that’s what we do. We needed and I wanted to go through the FDA.

Bill Frist:                       How did you end up in Iowa?

Dr. Michael Abramoff:   Iowa has the best ophthalmology program in the world, and they invited me to come for a sabbatical, and then invited me to become faculty, and I’m honored to be the Waskey Professor of Ophthalmology and Visual Sciences.

Bill Frist:                       Fantastic.

Dr. Michael Abramoff:   It’s an awesome place.

Bill Frist:                       And you’ve been there for how long?

Dr. Michael Abramoff:   17 years now.

Bill Frist:                       17 years, and has the academic center and academic community been encouraging? Because this is real cutting edge innovation, as you said. It’s the first FDA approved autonomous device, and to get to that point takes a huge amount of work but has the university been supportive of that endeavor?

Dr. Michael Abramoff:   Extremely, extremely supportive. In general, the people of Iowa have been so welcoming to me and my family. It’s just been amazing and I really want to give back to Iowa by creating this company and creating jobs, and this different aspect of it. But yeah, it’s just a tremendous environment to do cutting edge research. Lots of support, but I will mention one thing. I mean, there’s so many advantages of AI, but some of your listeners may think, “Oh, aren’t all doctors going to lose their jobs because this AI is now replacing what I do?” My answer is no, and let me illustrate that. 10 years ago now, the chair of ophthalmology at Hopkins, Johns Hopkins, Pete McDonald was a great friend. He made an editorial about my research saying, “The Retinator: Revenge of the Machines.”

Bill Frist:                       Oh, gosh.

Dr. Michael Abramoff:   Which is supposed to sound like the Terminator for the retina, because he expressed the fear of ophthalmologists losing their jobs because of all this AI research. And that turns around, by being very transparent and open about what we were doing, doing lots of scientific publications, and showing there’s actually this win/win/win/win situation. Some people like to call it only win, which I don’t know. Yeah, so interesting term, but if you’re a patient, we already talked about advantages.

Bill Frist:                       Right.

Dr. Michael Abramoff:   High quality care, better access, lower costs. If you’re now a primary care provider, you can expand what you can do with your patients. Better care for your patients and you can actually bill for it, and then if you are a ophthalmologist, rather than being fearful of job loss, it’s interesting. You’re actually getting more patients referred that actually have disease and you can treat and actually be meaningful to rather than saying, “Everything is normal, come back next year.” So it’s much more exciting for ophthalmologists and eye care providers. And so, they are literally so supportive now. But can you imagine?

Dr. Michael Abramoff:   We went from the Retinator to well, it’s currently on the American Medical Association website, these ophthalmologists are doing AI the right way. So, you can address the concerns, and there’s other concerns people have about AI, like the bias, by being open, transparent, and account for it.

Bill Frist:                       Right, yeah.

Dr. Michael Abramoff:   And so, there’s a win there. And then for the payers of course, there’s that win of better quality measures, lower costs, and so it’s really a win for everyone.

Bill Frist:                       And I mentioned the convenience, we’ve got that. We’ve got the accuracy, we’ve got better results, and I mentioned the cost. How affordable is this? And let’s start that discussion in terms of a great invention that can apply to hundreds of thousands and millions of people, but it does have to be paid for and how do you make it affordable? So first step would be, you’ve got to get reimbursement somewhere, in these insurance markets, the way insurance is structured here in the United States. Walk me along that journey in terms of how you get reimbursement for this.

Dr. Michael Abramoff:   Let me pull back a little bit from there.

Bill Frist:                       Yep, yep.

Dr. Michael Abramoff:   And talk about how nothing in the healthcare system was prepared for autonomous AI. Everything is built on doctors and nurses and other providers providing care. Humans providing care. The payments, the approval, the medical boards, everything. So, when we got approval, first challenge was, how are we going to tell doctors to bill for this? Thanks to the American Medical Association and American Academy of Ophthalmology, within seven weeks, we had a breaching code. So there was a code, it was not for AI, but we were able to at least use the technical components so doctors can bill for it. So they’re billing and collecting now, which is very exciting. That helps adoption, of course. And so, but we’ll talk about how you pay for this.

Dr. Michael Abramoff:   The whole goal of IDX is to save healthcare costs using autonomous AI like I mentioned in the beginning, so we need to make it lower cost than my cost. And so, how do you go about this? Well, clearly it’s very scalable, you can do it anywhere, and whether you do a million or a million and one in terms of cost is very different, so you can highly scale it. So that drives down cost enormously. You can offer it anywhere from a technical perspective. Because we were the first, everyone is looking at us saying, “Can I get through FDA? Can I get reimbursement and code?” And so a lot of work was done by the American Medical Association and many other groups, the FDA, NCQA, to help us create a CPT code which is essential if you are in healthcare. You need a CPT code, it’s essential.

Bill Frist:                       So a billing code that can be used-

Dr. Michael Abramoff:   [crosstalk] Yes, it describes what it does and that only is created when it’s valuable, when… One thing that was very important to us was we as a company assume liability, medical liability for the autonomous AI, just like a doctor does. I mean, if I say that I do these narrow tasks that a doctor normally does, a doctor is liable. You and I have medical liability. This AI has medical liability.

Bill Frist:                       Interesting.

Dr. Michael Abramoff:   And so all these aspects work, and the ethical rules we use, they all work to allow the AMA editorial, CPT editorial panel to create this new code in May of 2019, a category one code for autonomous AI. It’s coming out January 1, 2021.

Bill Frist:                       Wow.

Dr. Michael Abramoff:   And so, that’s very exciting because like I already mentioned, there’s no doctor work involved in autonomous AI. Normally, there’s an RVU component, Relative Value Unit, and it’s very important. And so, everyone is always talking about, “Yeah, this new surgery, there’s so many RVUs.”

Bill Frist:                       Right.

Dr. Michael Abramoff:   There’s no RVUs for this. So, the CPT needed to accept that there were no RVUs for this new code, and it was a long discussion which I cannot share more about because I needed to sign lots of NDAs to be at the meeting. And so, but let me tell you, that was a long discussion with the realization that this was a precedent for other AIs to get CPT codes.

Bill Frist:                       So this is the real, first code for AI-

Dr. Michael Abramoff:   Yes.

Bill Frist:                       … in medical field [crosstalk 00:33:01]-

Dr. Michael Abramoff:   Autonomous AI, autonomous AI.

Bill Frist:                       Exactly.

Dr. Michael Abramoff:   Where again, there’s no RVUs for a doctor. So, we created something AI work, and it may [inaudible] you given your background as Senate Majority Leader, that we briefed in the Senate Finance Committee, the hearing room, which you probably know.

Bill Frist:                       Yes, I’m aware.

Dr. Michael Abramoff:   We did this briefing on how we’re going to pay for that, because if we don’t get a payment model, R&D will dry up. Because there’s two risks as an investor. There’s the FDA risk, you’ve spent all this money, will you really get your trial successful? And then there’s no payment for this AI, so there’s too much risk. So, investors say, “Oh, maybe not so interesting.” We have very supportive angels that helped in the process.

Dr. Michael Abramoff:   But so we went through the different ways you can pay for AI, and I can go through if you want the different models you can do, but to decide the best way is just to take a percentage of currently a doctor gets for a similar task. Now, you have immediate cost saving and it’s still workable for AI companies like us.

Bill Frist:                       Yeah, well [crosstalk 00:34:10]-

Dr. Michael Abramoff:   And that was a lot of debate and excitement, and that has made a difference, because now you see, investors are seeing, “Well, they created an FDA regulatory path, they created the payment path.” By the way, we also created a quality measure path, which was another discussion that no one anticipated, and finally it’s now part of the standard of care for diabetes. So the American Diabetes Association has what is called Standards of Diabetes Care, that they publish every year, and we are up next for 2020. So AI is mentioned in there, autonomous AI for diabetic retinopathy, as part of the Standards of Care, which is very exciting.

Bill Frist:                       Yeah, no. Exactly. This, it’s really important. Our listeners are very interested in this intersection of innovation, new products, new ideas which obviously, IDX is right in the middle of with a type of scan that does the autonomous diagnosis. And we’re looking at the intersection with regards to policy, so I’d very much appreciate your really walking us through, because the policy includes the FDA. The sort of safety, the efficacy, the clinical standards, the accuracy, the sensitivity, to make sure it is a benefit, quality benefit to the patient. And then this last component, which is also a policy component, which is the reimbursement.

Bill Frist:                       In our healthcare system, to really have things sustainable over time, you have to have a code or a billing, what we call a CPT code, and what’s really interesting to me because it intersects with innovation is that you’ve had to do this, you as a scientist, you as a running… a great company, but a small company, a startup company, you’ve had to walk and all that rests on your shoulders as an individual to set up the policies, both in terms of safety and efficacy to set up the policy and the ethical guidelines, the quality guidelines and ultimately the reimbursement. And people often times wonder, where does all this come from? How does it happen? And I think your story is a magical one in many ways.

Bill Frist:                       Tell me, where are we today in terms of you’ve gone through these processes, all of this is real time in the last few months and the last few years in terms of the policy adaptation to it. But where are we now, and at what point do you see our viewers and our listeners being able to go to their primary care doctor, have this scan and 15 minutes later, have all the advantages we’ve been talking about?

Dr. Michael Abramoff:   They already can, and I mentioned grocery stores in Delaware having this, and they have sort of a primary care clinic within the Safeway grocery store. So there’s one place you can get it, but there’s many other systems that are adopting it. We were just discussing in the elevator another system. And so, I think a great example which was actually on NPR last year when we went live was in New Orleans. So, downtown New Orleans after Katrina, eye care was a big problem in downtown New Orleans.

Bill Frist:                       Yep, sure.

Dr. Michael Abramoff:   So all these people with diabetes had their diabetes care and the waiting period for getting an eye exam through human, an eye care provider, was four, six, eight months and people just didn’t go because you explain just the process of you get referred, and you need to make an appointment, but then there’s snow. I’m from Iowa, or you get sick or the child or something, and you may forget about it. It’s not because you don’t want to go, but it’s just complex to do that. And so, you can imagine very few patients have an eye exam. We put the AI in, and now they can see the eye care provider if they need it the same day. So we went from months of wait time to same day for patients today for the past year. Very exciting. And so, you see many examples of that already.

Dr. Michael Abramoff:   I think, the start of 2020, the real story of this year will be very rapid adoption, so it gets to as many patients as we can this year. So, very rapid expansion in place if we already are. Signing up new customers and implementing it there, which is interesting to see the company adapting to this much more rapid rate of implementing it and making sure it works for the customers which are healthcare systems and retail clinics in many other places. So this is the year of rapid scaling as it’s called, rapid adoption. It’s very exciting to finally get it to patients, which is where it all started with. We want this to make it better for patients, specifically with people with diabetes in this case.

Bill Frist:                       Yeah, and how many… I know you just said it, but how many people have diabetes, and how often should they have, or is the recommended standard in your clinical practice to have an eye exam?

Dr. Michael Abramoff:   30 million people with diabetes in the US currently, rapidly rising and the recommendation, for example, the American Academy of Ophthalmology is an annual eye exam.

Bill Frist:                       Annual exam. And today, how many people do not get that annual eye exam roughly? [crosstalk 00:39:34]-

Dr. Michael Abramoff:   The estimates differ depending. If you ask patients, it’s about half of them because they like to answer the right answer. But I think the most authoritative study was of the Medicare population. They took a big sample of about 300,000 if I remember correctly people with Medicare insurance and in that, only 15, about 15, 15% go to diabetic eye exams.

Bill Frist:                       Yeah, and the reason they don’t for the most part, is going to be affordability or convenience or lack of access, and it seems to me this is explosive in how revolutionary it is, the fact that you really address the affordability, as you said, lower costs. And the autonomous nature of it means that the convenience and the quickness is going to be making it available to so many more people, which is truly transformative when, if you make the diagnosis early as we said, it is preventable and treatable and avoids the ultimate problems, blindness being the worst.

Dr. Michael Abramoff:   And interesting enough, for a person with diabetes, studies show that their biggest fear is going blind, more than dying and more than losing a limb, which is a common complication of diabetes unfortunately. So, it’s really important for people with diabetes, I mentioned instantaneous. It may interest you that access is really the biggest problem. We see that in A1c and primary care. So typically, everyone with diabetes needs an A1c, HbA1c, for the blood glucose level and the metabolic control. And so, if they are sent to a lab, typically compliance is about 50% if they have the new desktop machines, meaning they can do it with the primary care physician still in the office, it goes to 95%, studies show.

Bill Frist:                       Wow.

Dr. Michael Abramoff:   And so this access problem is important. What I always try to challenge the people I’m presenting to, cost it’s very important, cost is very high, we need to bring it down. But interestingly enough, if you pay people with diabetes, they did it at Hopkins, if you pay people with diabetes to get their eye exam they actually go less because they get suspicious about, “Why am I getting paid to go do this?” And so, cost is important for the larger system because it needs to be affordable, and so we can use that money to do the better treatments we now have and to treat more people rather than doing diagnosis. So, cost is important, but access I think is the key.

Dr. Michael Abramoff:   So that’s why I want this to be everywhere this year. Everywhere where the patients are is where it should be, and the same for the other AIs, autonomous AIs we’re developing. A pipeline, as we call it [crosstalk 00:42:20]-

Bill Frist:                       You just mentioned a couple, there are other types of AI. You were the first in the eye-

Dr. Michael Abramoff:   Yes.

Bill Frist:                       … but I know you’re looking at a number of other areas, but just so our listeners will know, where do you see the next, most common applications of AI being in a diagnosis and treatment recommendation?

Dr. Michael Abramoff:   I will give you a few examples. Again, I think there’s a few things. It needs to improve outcomes, right? In some evidence-based way. It needs to be doing that in a way that’s close to what clinicians do, because that’s safer and we can talk about it. So I see those AIs, in the eye because we already talked about why it’s such a wonderful organ in terms of being connected to the brain and part of the brain. In skin, in the ear, so all places where people go and they want to receive that type of care in the primary care environment or retail environment.

Bill Frist:                       And the skin would be if you have a nevus or a mole or a skin abnormality, that some type of a scan, driven by algorithms would make an autonomous diagnosis in and of itself.

Dr. Michael Abramoff:   Yeah, exactly.

Bill Frist:                       Same with the ear.

Dr. Michael Abramoff:   Yes.

Bill Frist:                       If there’s fluid behind the ear or there is infection in the ear, then autonomously, a diagnosis can be made through artificial intelligence and machine learning over a period of time.

Dr. Michael Abramoff:   Yeah, and then if it’s abnormal, if there’s some exception, then it still needs the care of specialists. But in many cases, it won’t. In most cases, it won’t.

Bill Frist:                       And I think you’ve made the point in terms of the great fear, is that well, you’re going to run doctors out of business or you don’t need doctors because the machines are smarter, and I think your points, of very narrow areas, the importance of having ultimately that human perspective, knowledge-based decision really outweighs any of the great fears that are out there. One final question being from the medical, and so you have gone through this whole system of invention, then implementation and execution and policy. What does that say about America in terms of you have done it, you are doing it, you’re making huge contributions that are out there to be realized by millions of people, and making their lives better? But America, what has it done? Has it given you this opportunity? Has it been-

Dr. Michael Abramoff:   It is an amazing country. I could never have done it anywhere else, and the fact that I talked to Congress as an immigrant a few years ago, doing all of this, having access to all these people, and the support, you cannot do it anywhere else. It is just an amazing country, I love it. So, my kids love it. And so, nowhere else.

Bill Frist:                       Yeah, yeah. It’s a great statement. I think your story, which I know is just beginning, and just being felt by people here and indeed around the world is we haven’t talked a lot about but the global impact that you’re having right now, but it is a great story and gives me a lot of hope and a lot of optimism about the future of innovation and policy and medicine and clinical care in our country today. Michael, thanks a million for being with me and with our viewers and listeners today. And I look forward to many, many more conversations and engagement, which we know is really going to change the direction of health in America. Thank you so much. Appreciate it.

Dr. Michael Abramoff:   Thanks so much for having me.

Bill Frist:                       Thank you.

Dr. Michael Abramoff:   Thank you so much.

Bill Frist:                       This episode of A Second Opinion was produced by Todd [Schlosser 00:45:57], the Motus Creative Group, and Snapshot Interactive. You can subscribe to A Second Opinion on Apple Podcasts or wherever you’re listening right now. And be sure to rate and review A Second Opinion so we can continue to bring you great content. You can get more information about the show and our guests and sponsors at ASecondOpinionPodcast.com. A Second Opinion broadcast from Nashville, Tennessee, the nation’s Silicon Valley of health services, where we engage at the intersection of policy, medicine, and innovation.