Welcome to the interview-series on the socioeconomic consequences of disruptive technologies by Rethinking Economics NL. In this interview, we will be focusing on how disruptive technologies as artificial intelligence, the internet of things, and big data are transforming a very important topic: health care! For this, four experts have been so kind as to join US today: Firstly with us today, is John Halamka. He is the president of Mayo Clinic Platform, and prior to this served as the executive director of the Health Technology Exploration Center for Beth Israel Lahey Health in Massachusetts. Previously, he also was the chief information officer, or CIO, at Beth Israel Deaconess Medical Center for more than 20 years. As a Harvard Medical School professor, he served the George W. Bush administration, the Obama administration
and governments around the world, planning their health care information technology strategies. Dr. Halamka is a practicing emergency medicine physician and has written a dozen books about technology-related issues, hundreds of articles and thousands of posts on the Geekdoctor blog.
Secondly with us today is Dr Jennifer Joe. She is an entrepreneur, having founded companies as MedTech Boston, Medstro, and Vanguard.Health, and is an emergency room physician in the Boston VA Healthcare System. She is a cross-functional executive and innovation leader, focused on areas as digital health ecosystems, innovation challenges, and the telemedicine community. As a daughter of poor immigrants and having grown-up in rural Mississippi,
she is dedicated to community building, providing opportunities to the traditionally marginalised, and providing basic health care to everyone. She has over 70.000 followers on LinkedIn and regularly teaches at Harvard Medical School, the Harvard-affiliated hospitals, and MIT, for example being a mentor at the MIT Media Lab. Thirdly with us today, is Luba Greenwood. She is a Life Sciences, Tech and Health care executive,
investor, and company builder. She had over $5B+ in deals and investments across multiple therapeutic areas and life science, device, diagnostics, and tech sectors. Her company-building experience includes co-founding biotech and digital health companies in the immunotherapy, women's health, artificial intelligence, and microbiome space. She began her career as a litigator at a leading national law firm and recently served as an executive at Google Life Sciences, Verily, and as Vice President of Global Business Development and Mergers & Acquisitions at Roche. She lectures at Harvard University in the School of Engineering and Applied Sciences. Lastly with us today is Umbereen Nehal. She is a paediatrician and MIT Sloan Fellow focused on strategic systems approaches and utilizing human-centred design for improved health and wellbeing outcomes. Previously,
she was a Medicaid Medical Director, where she was clinical lead on a $1.8b Delivery System Reform Incentive Payment design, and a Chief Medical Officer and Vice President of Medical Affairs at the Community Health care Network. She has been four-time “Top Voice” with over 235.000 followers on LinkedIn, and co-authored the United States national curriculum, and contributed to two state public health curricula. She is also an Assistant Professor at University of
Massachusetts, having received her training at Aga Khan University Medical College, Baylor College of Medicine and Harvard. And with that I would like to move towards the very first question for Dr Halamka. Dr Halamka, could you tell us more about how the disruptive technologies as artificial intelligence, the internet of things, big data, are currently transforming health care worldwide? And I was wondering, what are the most important developments for students to know about that arenot specialised in health care, such as students and economics. Applicably, I think, we just had an ambulance that drove by. John Halamka: I was going to say, you must have orchestrated that, perfect. So I joined
Mayo Clinic one year ago. And why did I join Mayo Clinic? Mayo believes that by 2030 the health care system of the United States, and really the health care systems of the world, will be digital-first. And if you're going to be a digital-first healthcare system, there are certain modular components you better put in place. Such as, if you have, as does Mayo, 150 years of historical data, structured, unstructured, images, telemetry, [inaudiable], how are you going to use that in a privacy-protecting, ethical way, to generate new cures and new treatments, new algorithms for patients of the future? And that's as much a policy-question as it is a technology-question.
So my first charge was, create a uniform mechanism of data access for collaboration across government, academia, and industry, with all of Mayo's historical data assets. And we did that. In partnership with Google Cloud, and a series of de-identification algorithms, and security protections that ensure privacy. Then I was asked to create an AI factory, on top of that, so that all the Mayo Clinic staff can say, oh I woke up this morning and I have this notion. Well can you explore that notion using this de-identified data set to generate an algorithm, and validate that algorithm. Well then, how do you evaluate algorithms for bias? So hey, I’ll make this one up but what if, if you've ever been to Minnesota you may know, there are a fair number of Scandinavian-Lutherans in Minnesota. So we create a algorithm based on the population of Minnesota,
and then let's imagine we're going to take that algorithm to Spain, you know. Well, is it going to work so well? Is it, is it fit for purpose, right? So we've got to, had to build that tool. We've had to build the capacity to ingest novel telemetry, from the devices you wear or the devices in your home, and marry that with the clinical data that is from more traditional electronic health records. And finally we've had to figure out, especially in a time of Covid,
how to deliver all kinds of care at a distance. And what's been unique about Mayo in 2020, is we began to deliver high acuity, serious and complex hospital care, in the homes of patients. And we've discharged now over 200 critically-ill patients from their living rooms. And that has taken a whole set of technologies, which combine remote-patient monitoring, and algorithms, and dashboarding, and supply chain, and on-prem people. So I tell you that as context, because as you look to the digital-first healthcare system, manage your data, create your algorithms, get new telemetry, and deliver care of all kinds at a distance. Koen Smeets: I was wondering, you already now brought it up to 2030 a little bit, you already told a bit about that, but what do you expect, do you have any ideas, of what we can that we can expect over the next decades, what we can see? John Halamka: Well, so. Whenever somebody asks me to forecast a decade away,
that's like saying it's 1993 and I hear this web-thing, you know, is it going to go anywhere, right? But a decade from now, we're all going to have Facebook-implants, I mean, who knows, right? But what you could say is this. The experience that all of us have of health care, well maybe not in your country, but let's say in the U.S., is generally bad, right? That is, we don't have the tools we need to navigate the healthcare system for our families.
And so what you would hope, is that, your workflow is going to be quite different. You wake up, your child has a fever. Aha, you pull out the sensor, that is going to take a look at, not only body-temperature, but tympanic membrane, tonsils, you know, other things that may be appropriate.
Initially, that data goes to an AI-chatbot, and does an initial triage. Based on probability of disease, you may or may not encounter some health care practitioner, but the end result is, there will be a very well understood way, to get your child back to wellness. Call it, ways for health care, so to speak. That's got to be the norm, a decade from now. Koen Smeets: I think it is, although it is a decade away, I do think it is a very interesting idea of what we can hope for, or expect. Dr Joe, I was wondering, considering your entrepreneurial background, what is your perspective on the current and future impact of disruptive technologies, as we discussed? Jennifer Joe: Yeah definitely. So, as Dr Halamka highlighted, I think data and artificial
intelligence are going to bring us into a new era of understanding. I think it's just beginning. John Halamka is a leader in the field and Mayo also as a leader in the field they have massive, John you have massive infrastructure and you're brilliant at driving innovation, driving teams, and putting infrastructure in place. So I think you're a leader. So where is artificial intelligence and big data, I think in the US currently? I think we talk about it, I think Dr Halamka and his team are doing some really interesting things, but I think we're just at the beginning. The key points that Dr Halamka brought-up was
data. So getting access to the data, what kind of data is there cleaning the data, and then the cross-collaborations, I think the cross collaboration is profoundly important. Meaning, you can't just throw a data scientist at data. It's very difficult for them to look at the data, know the meaning of the data, but after that, come-up with hypothesis and understand how it would actually fit into the healthcare system. So I think we have a lot of work to do there. I work a fair amount with the federal government, Vanguard.Health holds a contract with them to
drive innovation, and one of their big initiatives is innovation and artificial intelligence, and making a lot of the data that the government, the US government, funds, publicly available in a safe way for data scientists access, and then also innovation initiatives behind driving it. Where are we in terms of health care as a practicing physician and adopting it? I think artificial intelligence hasn't really made it into the practice of medicine yet. Meaning, we're using it, or we're seeing it a lot in claims. Or, were seeing it a lot to potentially optimize workflow. One of the early groups or organizations to come out of MIT was optimizing patient no-show, right.
It's very easy to predict if a patient's not going to show, it's a huge burden to the health care system, how can we maximize just that simple piece? So, is AI in diagnostics? Not yet, not really. Is AI in treatment? Not yet, not really. I think that, is it in clinical decision support specifically with radiology? I think that's where we're playing right now. And then how do we take that and really elevate it to the next level, hold AI accountable to all of the traditional benchmarks that we hold healthcare accountable to. Koen Smeets: It's very fascinating. And I was also very curious about, perhaps how, not, perhaps even broader than AI, general disruptive technologies that we're discussing, how are they changing the general workings of hospitals, do you also perhaps have an example from your own experience as an emergency room physician? Jennifer Joe: I work at the VA healthcare system and I’m just gonna leave it at that. [laughs] And
I love and adore it, veterans are wonderful, we give good care, it's meaningful. We've had some early innovations, but we are, you know, it's a big, bulky system. So some of the innovations that we've had for a long time is in remote patient care, and high risk patients. So high-risk cardiac patients, and high-risk pulmonary patients. For the last 15 or 20 years, we've actually had remote
patient monitoring of these patients. Where we give them a vital signs measurement, blood pressure cuff, pulse [inaudible], scale and they actually remotely load that up into the system, and we have a whole team that monitors that, flags it if something is abnormal, calls the patient and brings them into the emergency room. The VA is unusual for, so in the US, the VA is unusual because the payer is also responsible for the chart, you know, the entire population. That's unusual. So they were highly motivated to roll something out like that, which I think actually underscores that remote patient monitoring has benefits to the patient, but also cost of care.
Koen Smeets: Ms Greenwood, could you also expand on it? What is your perspective on the current and future transformative impact of disruptive technology? And I was expecting curious about this from your experience as both an executive and an investor. Luba Greenwood: Yes, absolutely. So it's interesting Dr Halamka and Dr Joe are really experts and leaders in the in the field, when it comes to, especially when it comes to how care is delivered. And I remember when I was at Google, I was actually, was in the, ended up in emergency room and one of the nurses that was doing the intake, she said, hey what do you, she was basically doing what she was supposed to be doing, and she said hey what do you what do you do? So I thought, you know if I say investor or company builder that's complicated, who the hell knows what that is, and I, so I thought what is it that I do? And I said to her, I do, I do something that hopefully in 10 years you will actually turn around and look at me when you're talking to me. And she, yeah, and when you're taking care of patients. And it goes to Dr Joe's point that, you know we do, and the big tech companies and the amazing healthcare institutions, are working tirelessly together, in order to provide and change the patient-experience, and in order to change, and help, and support the physicians and their decision making, and support other healthcare professionals in decision making.
But we're not there yet, right. So, and we are looking at the tools to do that. I do want to talk a little bit you know, about a different angle, though. Because I come from biotech, and therapeutics, and diagnostics, and tool space. And that's a bit different.
So when we think about what's going to happen in 10 years, is there going to be change? I can tell you with 100 for sure, the way that we make drugs, and the kind of type of drugs that we make, is going to be very different. People are shocked today about, when we think about vaccines, and wow how did they make vaccines so quickly, there are some conspiracy-theories that perhaps they are, they're not done correctly, but one of, some of the things that people are missing, is that, there's, the reason why it takes some time so long to discover and develop drugs and do clinical trials, is because we have enormous inefficiencies. That with the use of big data, and not just artificial intelligence, I’m not a big fan of using ML, that it will solve everything, just sometimes pure, simple analytics, and these, basically, there are so many silos, taking away the silos in development, can actually help you tremendously. It can help you recruit patients
faster, so that you can find the patients, it can help you with actually doing the development faster of the drugs, it helps you with the discovery of the drugs, personalising the drugs. And what do I mean by personalisation, right? We know for patients, for example, as an example with cancer, so you know, the same patient with cancer with the same type of patient, different, two different patients with the same type of cancer, respond very differently to drugs. So we now have amazing tools, and we're using the data that we're collecting, again, in a, in a responsible, de-identified way, so that we're not using patient-data and patient information, in a way so that we can actually predict whether or not a particular patient is going to be responding to a particular therapy. We're also making drugs, so this is on the
treatment side, but we're now actually using that type of data to, from the very beginning, making drugs for the particular population that will truly respond, and not just truly respond, but particular population that will have limited side-effects, right. Sometimes when you hear advertisements for certain drugs, there's that list of many things, many horrible side-effects that could potentially happen. Well, if you make more targeted drugs, they, there, you would limit the side effects that people are have, that people are having. So, there are some really amazing technologies, and novel ways that we're right now, doing drug development. The same thing for another example is autism, right. We say autism is a spectrum. Well, right now we're using AI
[audio-cut] for example to actually find the right population, find the right, not only find the right population, but again make the right drugs for that population. So that's on the treatment side of the diseases that we know, but there are also many diseases that we don't know what in the world causes them, we like to pretend we do, we like to pretend we know the brain, but we don't. We don't really still know, for example what causes Alzheimer’s, how we can treat Alzheimer’s properly. So we are using big data and AI and I’m confident that in 10 years, as an investor and company builder, and this is the kind of technologies that I’m investing in, is to understand the brain better, so that we can make the correct drugs that actually truly work, for everything. From Alzheimer’s, you know Alzheimer’s and dementia, to anti-psychotic drugs as well. And in terms of the discovery of actual medicines, you know as I mentioned the development can be done much faster. You don't need to take a
year you know, year or over a year to recruit your patients, if we can find pa-, if we can find right now you know, cut different you know, if you're a target, and you can find your target-customer. Why is it that, when you're running a trial, you cannot actually find the patient that you need to? So, that is again another example of how we can use, pretty simple basic technology that we already have right now, that we as consumers use, and apply it to a healthcare setting so that we can actually match the right patient to the right trial. And I would say, so that is I would say for the medicines, right.
And if you have been as an investor or company builder in the therapeutic space or biotech space for many years, you view these disruptive technologies as something that expands your pie, right. It's not something that, hey it comes in and now you no longer need biotech, right. That you will, everything will be solved with digital therapeutics. That is not the case, right. That is the case with some things, but that's not the case with everything. So you see these amazing tools, basically expanding the pie of possibilities. There are also many what, what's interesting
is there are also executives, and inventors, and investors, in the diagnostics and tool space. So the disruption, in the diagnostic tool space has been even more exciting, but it's a little more risky than in therapeutics. So, it has been historically a pretty tough business to get into. It has low margins, the reimbursement is pretty low, but actually when you look at the regulatory landscape for diagnostics and tools, it's the same as therapeutics, right. So you're going, it's the same risk, risk profile, it's not, they're the same regulatory hurdles, the same barriers to entry with IP, but then once you're on the market you're reimbursed at a much lower rate, and there's a lot more competition. So, you have, what's amazing has been is that there is this new tech that is disrupting the diagnostics and tool space, but however at the same time you know, a lot of the investors and the start-ups are recognising exactly what Dr Joe said, which is, yes it's great that we have this amazing, these amazing tools, but are they being really incorporated as part of the workflow? So that's a big question, right. You can be reimbursed,
you can be on the market, but are they actually going to be used by patients and by physicians, that's a whole other question. But still, they do hold a lot of promise, we have really neat things, we were working at Google and at Roche, you know incorporating new sensors for example into devices to predict risk of infections, we've used AI to you know, see and help patients manage their chronic diseases you know, you see companies like Livongo on Duo doing that. Many disease apps, and disease management apps, are being reimbursed by insurers, and insurers have embraced this new model. So that will actually continue, continue to go and evolve. And what's been a very interesting I would say, lastly I would say, is that for investors, these are very true, everything I said is the outlook and what investors that have traditionally been in therapeutics and diagnostics. Now, what's been exciting about this disruption, is that it brought in investors and company builders that have never been in healthcare before. And they're now coming, coming in into healthcare, and again building companies. So however,
their space that they're investing in, their space where they're building is a bit different, you know they're looking more clinical decision support tools, they're looking at telemedicine, they're looking in ways to perhaps triage patients, and, or even delivery of medicines. So that's in itself, has been become a very large market. Now it has a lot of landmines, because many of the kind of the items that are, you know in these pockets that has just mentioned, will not necessarily be incorporated as part of the practice of medicine, or the medical workflow, but it holds a lot of promise, and most importantly promise for patients. Koen Smeets: I think it gives a very fascinating overview of the landscape, especially for investors and how the technologies are influencing healthcare. Dr Nehal, could you expand on how you see this, especially from your experience as a Medicaid Medical Director? Umbereen Nehal: Yeah thank you so much. I mean, fascinating discussion, I’m thrilled to be a part
of this. I think I’m just going to pick up on some threads from what's already been, you know woven into this discussion. You know, different things that I’ve heard. So for instance, from Dr Joe we heard about aligned incentives, when the VA is both the payer as well as a provider of care, and has a set population that is not because, you know in Medicaid we often hear about churn.
For those who are not familiar with Medicaid, it's a government agency. If you're in Medicaid you say, because there are 50 states, and that each one's administered at a state level, getting both federal and state funding. So we often say in Medicaid, if you've seen one Medicaid program, you have seen one Medicaid program. So there's an unfortunate, that's another added aspect of fragmentation, as Dr Halamka had referenced early on, that you know a lot of people in the US, we don't, haven't had very positive experiences of healthcare. Majority of Americans get their health care from their employer. But that means that if you change employment, you can have disruption in your care. And then, who owns your data, right. We've talked a lot about data. The data, well
is it owned by the provider, the hospital, is it owned by the health plan, you know, we've heard from Professor Greenwood about, you know the business aspect, the investor-landscape. A lot of times you'll hear about a market's, market segment, you know we heard, you know couldn't we speed-up clinical trials, and get our patients in faster, but then the question is, are we identifying a very specific segment of the market, and we have a very specific customer that we're serving, and we design innovation in that space, for that specific customer. And you know in business side we talk about minimum viable product, or MVP. But then, when you, let's say you are successful in that, then you know, at MIT in the there's the Entrepreneurship Center, there's a, Bill Aulet is really famous as an entrepreneurship professor, and he talks about land and expand. So all of that though is about segmenting,
it's about identifying a specific population, but if you come more from the academic research side, you always think about generalisability. Have you, have you designed the product, have you designed this study, such that it can be gen- reasonably generalised to different populations. We heard Dr Halamka mention, that yeah there's a lot of like Scandinavian folks in Minnesota, there's also a sizable refugee-population from Somalia, right, and we know that AI for instance doesn't work the same for, you know if you're like saying doing dermatologic or facial recognition software, it doesn't work the same across genders, doesn't work the same across races. So I think the the opportu, and actually I’m technically not, I’m technically supposed to be in class right now, so hopefully my professors don't mind that I’m here if they see this, I’m actually taking a risk management class right now. And one of the things we were talking about was you know supply chain, one of the things I haven't heard mentioned yet is what can AI be used for in terms of supply chain, right, that was a huge issue in COVID in terms of whether it was PPE, whether it was reagents needed for testing, whether it was just simple supplies that we needed for our everyday living. That's another application. But one of the things we heard, because we were
talking about which of the companies, which of the industries, had foreseen COVID as a risk, even though it was known, there was you know an entire office of pandemic preparation, nobody had necessarily prepared for it, because again businesses often will think about their budget, the likelihood, the probability something will happen, and they'll often take that risk. Whereas I heard from a classmate who was in the financial sector, because of regulation from the government, they were prepared. They actually had had to test being able to go remote, and have appropriate cyber-security for a pandemic. So they were able to pivot on a dime. And so that's, so as much as people hate government, it's like, oh the regulations, we have like, it slows us down, there's cost, there's a lot of benefit sometimes to having that external authority, certainly when I was a Chief Medical Officer, sometimes within the kind of the gladiator-games that happen in the c-suite, I, you know sometimes if I said it it didn't necessarily have the same impact as being like, well, we might get audited. I used to be in Medicaid I know how this goes, we don't want to
get audited, do we? So, sometimes you kind of need that person on the outside wielding the big stick. And then we've already mentioned around trust, and I think that's a real challenge, and we haven't discussed it, I won't go, get political, but I mean, the elephant in the room is right, that we're living in very, I’m like I would love to live in precedent times again, like I’m tired of the unprecedented times, but I, right now you know there's a lot of distrust of government from all sides. There's the historically marginalised communities, who have, like things like, Tuskegee, but then we also have, on the other end of the political spectrum, people who feel that government needs to be afraid of them, and these you know, there's a lot of distrust. And so as much as I would love for government to be the answer, I also know that, it's complicated, right. But I do think the role government can play, is being that person with a big stick.
Thinking about serving all populations, and finding a way to maybe thread things through different organisations, that maybe the business interests may not necessarily take care of. Koen Smeets: Yes, I think it's very interesting, and it shows really well, as well for students in economics that wish to go on in policy, the importance of the regulation and the other governmental policies. Dr Halamka, could you expand on what you think of the comments of your colleagues, and how you see these things? I was also wondering, especially for students from economics, what they would need to know about, for instance the, related to government and such. John Halamka: Well, glad to talk about that. And so, I always am very careful not to use present future tense. What that means is, we have a product that will. Well either you have a
product that does, or you will have a product that will. So I’m going to be very specific about what Mayo has already deployed. For the last several years, with our large de-identified data-sets, we have developed a series of AI-algorithms that are used in real-time, clinical practice now. And let me give you a couple examples of those. So it turns out you could take a 12-lead ECG, and this is you know, typical heart tracing, and humans interpret them, machines interpret them, but there's much more subtle data, that a machine learning algorithm can find in that pattern than a human or a traditional EKG-machine. We're able to, with a very high AUC, above 0.9, identify your ejection fraction from your 12-lead ECG without having to do an echocardiogram. Now,
just for your audience, basically what that means is we can do a really cheap test, that is not invasive, and have it give us, in effect, a good idea of how well your heart is functioning without an expensive, or invasive test. We are able to predict hypertrophic cardiomyopathy, future atrial fibrillation, look at progression of pulmonary hypertension. So you heard from Luba about this notion of pharmaceutical companies saying, we're going to provide a molecule, but we're also going to provide a sensor, and then we're going to provide a service.
What if that sensor is your Apple Watch. Because we can look at your lead-one ECG, and tell you through an algorithm today, that it is your pulmonary hypertension is getting better or worse. So I tell you all these things because everything they've said is totally right, right, it's the future is already here it's just unevenly distributed, and there are, certainly AI algorithms being used inside workflow today for certain kinds of domains, but it is not normal just to say, I’m going to walk into a doctor's office and AI is going to see you now. They've also said something really important about workflow integration.
So many entrepreneurs say, oh I’m going to create this app, or I’m going to create this website, or I’m going to create this sidecar to Epic or something. Doctors hate that, right. They're not going to go to a different app or website, no. What they need, if you're going to deploy an AI-algorithm, it needs to be passively working behind the scenes inside their EHR workflow, and help them at the point of care. And so that's, very much as we've deployed these
things in production what we've tried to do. My efforts in this Mayo Clinic platform I now oversee is to scale what we've already done at Mayo, to a much larger audience. So when you talk about economics, this whole idea of platforms is that, oh this is not just a Mayo specific adventure, this is just look, effect like Airbnb helps you rent a room if you have a room, the Mayo Clinic platform is to say, oh how do we link telemetry of all kinds, from various manufacturers and a lot of interesting challenges with lack of standards and telemetry today, normalise that data, bring it to algorithms, produce a result in workflow, that goes directly to the physician, or the patient, or the payer, or the pharma, or whoever the customer may be.
And then from the economic standpoint is, how do you ethically, and transparently, monetise that and create value for doing that kind of orchestration. I’ll be honest, Mayo has thought of this three ways. Some is very familiar to my entrepreneurial colleagues here. We co-invest and start companies, and then realize equity-growth as these Mayo generated products and IP are promulgated. Sometimes we will license algorithms, that will then go into the devices as embedded software. Or sometimes we will do revenue share if a transaction is completed through a commercial associate. So, our effort here is really to democratise access to these tools, while also asking the question, as the United States moves from fee-for-service to value-based purchasing and different reimbursement models, how does Mayo position itself in what we hope is this digital-first economy, to be a player not competing with tech, but being a part of that tech-ecosystem.
So that, so I think everything is said is right, and we're just trying to be a part of it. Koen Smeets: I was wondering, if there are students in economics that have become now really interested and wish to go into healthcare, and especially also this tech-side of healthcare, how would you recommend, Dr Halamka, to go into that? Do you have any tips, or ways they can contribute? John Halamka: Well, so here's the sort of interesting future, I think the United States is going to experience, right. Fee-for-service medicine is still the default in many of the areas of the United States. I get paid more for doing more. But on the East Coast and the West Coast, there are a fair amount of risk-based contracts, alternative quality contracts.
So the question for your economics folk, is how do you ensure there's quality, safety, right care, right place, right time, that's also affordable? And these are very tricky issues, and at the moment, it just seems like risk-based contracts are the best way to go, where you're paid for health, as opposed to being paid for sickness. But certainly, we need a lot of analysis, and we need a lot of refinement of those models. Koen Smeets: Yeah I think there's a lot of work at least for students economics as well to contribute here. Dr
Joe, could you expand this on how you see this and especially from the socioeconomic consequences, as, how will this influence the affordability of healthcare for marginalised communities? Jennifer Joe: Yeah definitely. And I think that we're getting at where, what do we think is the big opportunity for healthcare and healthcare innovation in the US, and reflecting I think on COVID and what it's changed. So COVID has rapidly accept accelerated innovation in the US. It's disrupted things, but it's always also been an opportunity for adoption and a change. The number one space that we've seen this happen in the United
States is telemedicine. So we were forced into telemedicine across the United States. The payments were the big regulatory initiative that really drove that through. The fear of COVID and delivering care in a safer way, both for patients and for clinicians, was important. So, telemedicine is now widely adopted. It has many forms, and it's not totally clear what form telemedicine will take. I think there's going to be lots of different shapes. So when we're
thinking about the future, telemedicine's not going away. How much will it decrease, it's a little unclear. The payments need to be addressed, and we have good movement on addressing the payments. There's lobbying groups, there's physician groups, there's patient groups, who are like, this is good, we like it, we want it, we're going to keep it, we're pushing on it. So I don't think it's going away. So I think when we're building, and you're thinking of innovation solutions, you need to build with that being the model going forward, that has a lot of needs. We've already seen that, with what platforms need to be chosen,
how does it integrate into the EHR, but I think we're also going to continue to see it in terms of communicating with the patients. It's really accelerated, there are just text-capability for communicating with the patients, which is crazy! [laughs] We haven't had that before, we do now. But also the add-ons, right. So, I did my telemedicine annual physical with my doctor yesterday. She didn't have my vitals. How do I get my vitals, you know, how does she know that I haven't gained my COVID-19. I have gained my COVID-19, and that should be a factor. But
she probably should have weighed me, taken my blood pressure, make sure that my weight didn't increase my blood pressure, etc. So I think, and I have not ever been formally trained, people call that blue ocean, or blue sky, Dr Nehal will tell me, that's a real opportunity. Remote patient monitoring is a piece of that. Dr Halamka has already alluded to Mayo apparently has very significant remote patient monitoring. I’ve already spoken to the experience of VA, that we've been doing remote patient monitoring for our high-risk patients. It works, it decreases costs, it improves outcomes, and increases patient satisfaction. We've had that for a long time.
The rest of the US, because of the payment models that have been set up, hasn't really had that. We've had movement from a regulatory standpoint, or a payment standpoint, so CPT codes, and also the expectation and the fact that COVID is now here with US for a while. I think that's also going to be an area of opportunity to figure out what kind of technology needs that we have there.
And then a topic that is near and dear to my heart, is that as we accelerate innovation, and technologies, and new care delivery models, we need to make sure that we're providing good care to all Americans. So lower on the socio-economic ladder, and, or who has less access. So what does that mean, what does that look like? That means, we have to as a country get behind funding and supporting innovators and entrepreneurs who are behind that and committed to that, and it's not necessarily a place that big business has always gone. They can go there, I think if the regulatory and payment models are there. In terms of what are examples that I’ve seen and really been excited about in the US, I think the hospital systems have just been so overwhelmed and focused on responding to COVID, that we haven't really had that much opportunity to integrate new innovation in.
The most impressive or interesting example that I’ve seen and I, I’m writing a textbook now on digital health and telemedicine, we're finishing it, the most interesting example is Emory Grady. They're located in the south of the United States, so they have more disparities. They have overall lower socioeconomic status. Georgia is right next to Mississippi, which is where I grew up. Mississippi is the poorest state in the United States.
They have rolled out a tele-EMS system, which is fascinating. They've worked with the public health infrastructure. They've worked with 9-1-1, which is unusual. They've worked with ambulances. And they're facilitating a system where they can give clinical support to the EMS that goes out to the patient. Which is hugely important, because that's how a lot of our lower socioeconomic
citizens access healthcare. So if you can bring a physician out there, you can assess the situation, you can route them to the right place, but they're also providing telemedicine visits. So a lot of those 9-1-1 calls that are EMS, if you don't have constant access to a health care system, you maybe are accessing 9-1-1 at a time when you don't need to go to the ER, you don't need an ICU, but maybe you need your hypertension medicines. And can we handle that with telemedicine?
That's fascinating in terms of closing the gap of disparities, but how many partners had to come together to make it work, providing better care, decreasing cost. So I’m very interested in seeing how the United States can get behind initiatives like that. Koen Smeets: It's very interesting. Ms Greenwood, I was wondering, from your legal background,
could you tell us a bit more about how this relates to the regulation of such technologies? And how does this affect your perspective on these technologies? Luba Greenwood: Absolutely. I mean, one thing that is clear and this probably would be interesting to your students, is we are becoming you know, the United States and other countries are really becoming interlinked. And we're already global economies and it's only going to be growing. United States you know has quite a bit to learn about the regulatory environments
in other countries, especially in, for example in Western Germany, as it kind of relates to regulating of digital therapeutics. So US has actually, did embrace digital technology. Actually I believe Mayo was also quite involved. Verily, Google, was certainly involved in that effort. And and many other companies, so the government did listen to start-ups, and large companies, and providers, and payers about, what is the best approach to regulation of digital technologies. But we're still a bit behind companies like Germany, that are actually making it fairly, fairly easy and accessible for digital therapeutic companies to have their products on the market be reimbursed, and then also get the data, right. Because at the end of the day, if you don't have the right type of data, you know you can't really have a true digital therapeutic that has any validity, clinical validity. I would say on the legal side, so that's on the regulatory side, so yes, still a lot of exciting things are happening, regulatory authorities are very knowledgeable now on digital and are open, but we need more. We need more learning from other countries. On the legal side, so that has been
interesting, Dr Halamka mentioned licensing of algorithms. So that is something that you probably would have never heard about five years ago. So, the legal community was, you know felt very strongly that AI algorithms cannot be patented. However, that has entirely changed. So now, you know you can actually, well there are some ways to do that, but there are you know AI algorithms tied to actual products are patented, and many of them are being patented. So that's another thing that is to, for both economic students and legal students, to understand that there has been a huge shift in, in how we see and how lawyers view intellectual property, what can be licensed, what can be patented, what can be protected with patents, versus what can be protected with trade secret. And absolutely, when you are an organization
like Mayo and you come up with something truly revolutionary, like they're doing, you can protect it with patents, and other type IPs. So that's also important, a pretty big, important change, in how technologies are viewed, patenting of technologies and AI are viewed. Some of the other things that it are important from the legal, regulatory pieces, and this is actually, everybody has touched upon this a little bit, is ethics concerns. Ethics, data privacy, security protection, and ultimately, and very importantly, bias. So, when we talk about data and when we talk about the use of AI, there is still when you look
into who actually puts these the the algorithms together, who puts the products together, there's some inherent bias, not intentional bias, not just in data itself, but even with the software engineers themselves, and the data scientists that are putting this together. So this is actually something that I do each in my class for, you know in the School of Engineering here at Harvard, is to be aware of bias. So to be aware, not just of the biased data but your own bias and think very carefully because, this is no joke right. I mean, when you train a model and
if you train the model wrong, you can have the best intentions, but you will actually be doing, you'll be hurting patients. So this is an area of both bias and ethics. It's another one of these areas that in the next 10 years is going to be an essential component of any product design, right. Just as how today, we've incorporated, for example design, which before, if you talk to anybody in the diagnostics and med-device, and you know, they thought a bit about design but that was not the most important thing for the patient or the physician to actually really want to use the device, you just, hey it's on the market, approve, reimburse, you use it, right. But now we are incorporating of course design into engineering, so the same way bias and ensuring that you'll have a whole group of people in any product design in healthcare that will be looking at bias, that will be looking at data privacy, will be be looking at security, and looking at ethics, and the ethical use of products, is going to be very important. Koen Smeets: Yes, yes. I think we see it also in the other interviews that besides, of course in healthcare it's a very, very primary thing, but ethics is, I think always, and the bias is a very important thing. So considering time, I think we have to move
towards the last question before the closing statements. This is for Dr Nehal. Could you expand for us, as well as on bias and ethics, but I was especially curious on how these technologies are affecting developing countries, in, from a healthcare perspective. Umbereen Nehal: Thank you. Yes, it is, there's you know with everything there's always opportunities,
barriers, and risks. And they they're always in tension, right. So the opportunities are that, there, if there is an understanding of needing to train algorithms, as was just said you know, to account for bias, then there may be you know for instance like, I went to medical school in Aga Khan University, which is part of the Aga Khan Development Network, which has a global footprint, both in different parts of Asia, and Africa, and I know other parts. So that's an opportunity where you could partner with other large systems that care for for diverse populations, and very intentionally develop the algorithms, develop training models, or train the, you know the algorithm that way, such that what you know, to Dr Halamka's point, when you do patent, if you do patent it, you do license it, you want to make sure that you're, then I mean, and this is an assumption of mine, I’m not an expert in this area is that, I’m assuming that there's not going to be a lot of change to it that's the nature of a patent, right. So you want to make sure that what's gone into it a priori, is correct, and you've trained it on the right population. And maybe you want to intentionally build in ways, in which that, even if you're licensing it, or patenting it you can allow some customisation. Because, in theory, machine learning should allow, allow learning, right. So you should be able to apply a new data
set, and it should be able to customise itself, but then the inherent rules built into it need to allow that accuracy across populations. So I think that if there's you know, it from the design, again, underscoring what was just said from the design-perspective, we talk about human-centered design, as long as we're thinking about the full spectrum of humanity. And what, the other thing that's really critical, so when we think about, I think you have to get like really kind of in the weeds about data. You have to be like a data nerd, where you have to not just think about who is the data coming, from but how are you parsing it out, right, how are you defining the fields? You know, we know that for instance gender is a spectrum, it's not binary, right. We know that race is a social construct, and because race is a social construct, so for instance, what the status of those who are black in America or are African-American, may be equivalent to somebody who's of South-Asian origin in the UK, just because of the historical context, who's you know been the historically more deprived populations. So I think, and and we have
these populations, ethnicities, different cultures, in different developed countries. So if there is a, if people are genuinely serious about addressing disparities, closing disparities, closing gaps, being ethical, being equitable, access, not just you have access to the doctor as Dr Joe pointed out, but then, when you're at the doctor are you being treated equitably? Are the data, are the evidence-based guidelines being used, are they are they appropriate for you? And then you know, even like you know Dr, sorry, Professor Greenwood had mentioned about you know personalised medicine. So is that going to become another disparity, where some populations that have access, they have resources, get the personalisation, everybody else gets the kind of what's for everybody. That's another source of disparity. But if we think about our minority populations as being majorities in other countries, is there an opportunity for partnership in developing countries? As well as, you know we know from COVID, supply chains got disrupted, right. What's happening in China matters to what's happening in the US, right. India is a place where there's a lot of outsourcing of many different professional services, there are many you know manufacturing plants in different countries. This workforce in those countries, they matter, I mean
they should matter, just as human beings, but from a business perspective, they matter, right. So I think these are all opportunities to for developing countries, for low- and middle-income countries, but yeah, no, great question, and we could talk forever, I’m sure. Koen Smeets: I think it's a beautiful way to end the the series, the interview as well. So for closing statements, we always have the same question in this series, so it is, if there's one thing you could say to students in economics related to the topics discussed today, what will that be? And I think it makes most sense to do it in the same order as with the starting questions, so first for Dr Halamka? John Halamka: So maybe a quick statement about economics. In this new world of virtual care, some people say, oh, data is the new oil. I would actually tell you, data is the new water.
And that is, it's ubiquitous. What's the oil, is the algorithms. It's turning the data into wisdom and action. But as you've heard from my other panellists, we're going to need labels on those algorithms. This is a a bag of dried fruit. Note it has a nutrition label on it, tells me a little bit about its fitness. We're going to need nutrition labels on our algorithms, if we're going to eradicate racism and reduce disparities, disparities of care across the world. Koen Smeets: I think that's a beautiful statement. Dr Joe, what will be your last statement on this?
Jennifer Joe: I will resonate, I think with what everyone's saying, is that we live in a time of telemedicine, digital, and everything is very interconnected. So I think, what is going to lead to success is cross-disciplinary, cross-collaboration of different perspectives, and working together to move innovation forward. So very excited to have economists working at healthcare, and to give us their perspective, and drive us forward.
Koen Smeets: Yeah I think this is seen really commonly in the series, it's the importance of interdisciplinary work. Professor Greenwood what is for you, one thing you want to give to students in economics? Luba Greenwood: I would say, the most important thing, just as actually both Dr Joe and Halamka said, Dr Halamka said, is access to information. Access to information empowers patients, and we will finally, hopefully in the next decade, get to the point where we can empower patients, and not just patients with means, but really patients that haven't even had access to care. And that empowerment piece is very important, I think, in how we practice medicine, not just in the US, globally, sometimes this is you know, in a way that we undermine the patient a little bit. And so, having, actually empowering the patient to speak for themselves, the knowledge about their diseases, it could help them to be knowledgeable about their health, not even just the disease, that they don't get the disease, is very important. And also, just very briefly, I love the dried cherries that Dr Halamka is eating, I thought I was the only one that gets them shipped from Washington State.
They're very good, they don't have sugar-added, they're good for health. Jennifer: What are these dried cherries? Luba: you have to show them, you have to show, they're amazing, they are incredible. Okay, get them shipped, oh, you're on mute. John: Sorry, it's Chukar Cherries,
in Washington, and I highly recommend the medley, as opposed to just the tarts. Umbereen: Okay now they should just ship it to us all free, given that we've given them some free airtime. Right, they got some earned media time. Koen: I will try to see whether we can have a sponsoring-contract
for the next series. Dr Nehal, what will be for you the next, the last statement on this topic? Umbereen Nehal: So I think that you know, by nature I’m probably not somebody who measures, despite being at MIT, and like the non-data person at MIT. I mean, to me data is a necessary evil, right. I’m more of a humanist, I like to kind of go with my intuition. But, I’m at you know
at MIT getting an MBA in my mid-career, for a reason. And that's because you know, we have to, we have to be able to measure we have to be able to have a common language. And then language often is numbers, and language is values, but I think that when we're quantifying things, when we're putting a value on things, we have to also put a value on trust, right. I’ve heard collaboration, I’ve heard stakeholders at the table, patients knowing themselves, so what is, what is the value of human knowledge, like, lived experience, what is the value of trust, what is the value of collaboration, I don't, I, we, not that we have to have a metric for everything, but that has to be part of our equation. Because when we only look at certain numbers that are, you know usual metrics, usual things on a balance sheet, we discount things that are really important, and we erode that, and then we end up in a big mess. So I would say that, let's not let's not just assume oh we're good people, we'll just do it in empathy you know, like let's let's put numbers, let's put hard numbers behind things, supposedly soft things, that are important, so that we remember to value them.
Koen Smeets: I think it's a beautiful way to end the interviews today. Thank you again, each of you for taking the time, and we hope to see the viewers for the next interview. Umbereen: Thank you so much. All: Thank you!
2021-05-09