EP. 87: THE POSSIBILITIES AND PERILS OF DIGITAL HEALTH

WITH JAG SINGH

Nov 8

WITH JAG SINGH, MD, PHD

A cardiologist and Founding Director of the Resynchronization and Advanced Cardiac Therapeutics Program at Massachusetts General Hospital discusses what clinicians need to do to better prepare themselves for the digital future of medicine.

Listen Now

Episode Summary

It's been less than a year since ChatGPT was released in November 2022, but in that time, reports have emerged of ChatGPT outperforming physicians in everything from clinical reasoning to documentation and even to empathetic communication with patients. How are we to make sense of the role of clinicians when artificial intelligence and digital health technologies seem to be advancing at a pace beyond our reach?

Here to discuss this is Jag Singh, MD, PhD, a professor of medicine at Harvard Medical School and former Clinical Director of Cardiology and Founding Director of the Resynchronization and Advanced Cardiac Therapeutics Program at Massachusetts General Hospital. He is the author of the 2023 book Future Care: Sensors, Artificial Intelligence and the Reinvention of Medicine. Over the course of our conversation, we discuss how digital tools can make healthcare more human-centered, how we validate the effectiveness of these tools, what we can do to prevent the profit motive from corrupting their implementation, and the skills that clinicians need to cultivate in order to thrive in the future.

Jag Singh, MD, DPhil is a Professor of Medicine at Harvard Medical School. He served as the Clinical Director of the Cardiology Division and the Roman W. DeSanctis Endowed Chair in Cardiology from 2015-2020. He is also the Founding Director of the Resynchronization and Advanced Cardiac Therapeutics Program, at the Massachusetts General Hospital Heart Center. Dr. Singh received his medical degree from BJ Medical College, Pune University, India. He completed his internal medicine residency, cardiology and cardiac electrophysiology fellowships at Mass General. He also earned a doctorate from Oxford University, a master of science in clinical investigation from MIT-Harvard and a research fellowship at the Framingham Heart Study.
Dr. Singh's research interests are in clinical cardiac electrophysiology. Dr. Singh is the national & global principal investigator on 5 ongoing multi-center clinical trials in device therapy for heart failure and catheter ablation for atrial fibrillation. He is also a member of several steering committees for multicenter research studies. Much of his other efforts are focused on the delivery of cardiovascular care while adapting to health care reform, population health initiatives and furthering the digital footprint of the Heart Center. Dr. Singh is on the editorial board of several medical journals, Deputy Editor of the Journal of American College of Cardiology: Clinical EP and editor-in-chief of the Current Treatment Options in Cardiovascular Medicine. Dr. Singh is an internationally recognized speaker and frequently gives lectures at national/ international educational forums. Dr. Singh has over 300 publications inclusive of original research articles, text book chapters, review articles and editorials. He has edited a Textbook on Imaging in Electrophysiology.
In this episode, you will hear about:
• 2:21 - Why Dr. Singh chose the specialty of cardiology, and specifically electrophysiology
• 7:43 - Why Dr. Singh became interested in digital health
• 10:17 - How doctors know if remote monitoring and other digital interventions truly work in the interest of patients
• 15:57 - Dr. Singh’s concerns over the digitization of health
• 21:36 - How we can center digital health interventions on patients and what clinicians can do to be a part of the solution
• 34:54 - Whether or not academia is doing a good job of preparing future clinicians to work with digital tools
• 37:33 - How digital tools might change the role of the clinician
• 43:25 - The skills that clinicians will need to develop to better work alongside AI
• 59:25 - The values that clinicians will need to cultivate to work effectively in the digital future of health
Henry Bair: [00:00:01] Hi. I'm Henry Bair.
Tyler Johnson: [00:00:02] And I'm Tyler Johnson.
Henry Bair: [00:00:04] And you're listening to The Doctor's Art, a podcast that explores meaning in medicine. Throughout our medical training and career, we have pondered what makes medicine meaningful. Can a stronger understanding of this meaning create better doctors? How can we build healthcare institutions that nurture the doctor patient connection? What can we learn about the human condition from accompanying our patients in times of suffering?
Tyler Johnson: [00:00:27] In seeking answers to these questions, we meet with deep thinkers working across healthcare, from doctors and nurses to patients and health care executives those who have collected a career's worth of hard earned wisdom probing the moral heart that beats at the core of medicine. We will hear stories that are by turns heartbreaking, amusing, inspiring, challenging, and enlightening. We welcome anyone curious about why doctors do what they do. Join us as we think out loud about what illness and healing can teach us about some of life's biggest questions.
Henry Bair: [00:01:03] It's been less than a year since ChatGPT was released in November 2022, but within that short period of time, reports have emerged of ChatGPT outperforming physicians in everything from clinical reasoning to administrative documentation, and even to empathetic communication with patients. How are we to make sense of the role of clinicians when artificial intelligence and digital health technologies seem to be advancing at a pace beyond our reach and understanding? Here to discuss this is Dr. Jag Singh, a professor of medicine at Harvard Medical School and former clinical director of cardiology and founding director of the Resynchronization and Advanced Cardiac Therapeutics Program at Massachusetts General Hospital. He is the author of the 2023 book "Future Care: Sensors, Artificial Intelligence and the Reinvention of Medicine". Over the course of our conversation, we discuss how digital tools can make health care more human centered, how we validate the effectiveness of these tools, what we can do to prevent the profit motive from corrupting the implementation of digital health, and the skills that clinicians need to cultivate in order to thrive in the future.
Henry Bair: [00:02:15] Jag, welcome to the show and thanks for being here.
Jag Singh: [00:02:18] It's my pleasure and privilege. Thank you for having me on.
Henry Bair: [00:02:21] We're going to dive deeply into health care innovation. But before that, can you share with us your path to medicine and to cardiology, specifically?
Jag Singh: [00:02:30] Yeah, thanks for that question. It's a pretty loaded question. And let me break it down by I think it's important to let you know that I grew up in India. I did medical school in India. Subsequent to that, I went on to do my doctorate at Oxford in the U.K. and then I landed up in the US. So it was a kind of a little bit of a zigzagging of a course across three continents, several hospitals. So when I got into med school in India, I was very young. I was around 16.5, 17 years of age. So it's really you get into med school really young, and I think often times you're not sure why you landed up in med school, but I think my mother's family probably had a subconscious influence on my desire to go towards medicine, because almost everyone in her household, she had, I think, four brothers, they were all either dentists, pharmacists or doctors. And that, you know, I think subconsciously directed me in that direction. But I probably fell in love with medicine halfway through med school rather than at the start of med school itself. So I did med school in India, went on to do my residency in internal medicine and early training in cardiology, all in India. My attraction to cardiology... Again, you know, this may sound very vain, but at that point in time it was just pretty cool, glamorous. I enjoyed the physiology and I, and I also realized that I didn't have to be too smart to understand cardiology, since it all kind of made sense. That got me hooked on to it. But the interesting thing here is that, you know, I came to the US and I had the opportunity to again re-certify and I did it all over again. So at Mass General, I re-certified in internal medicine and cardiology. So really affirming my love for those two areas of, you know, of medicine. So that's... That's kind of the long and short of it. Obviously, it was a very interesting journey across three continents.
Tyler Johnson: [00:04:30] You're cutting against the stereotype when you say that cardiologists don't have to be the smartest person in the room, that's not necessarily the stereotype that cardiologists get.
Jag Singh: [00:04:41] No, I think in cardiology, everything really adds up so simply that you really don't have to be too smart to make, you know, large, I would say, diagnoses of, you know, complicated things that you do in oncology and other subspecialties. So, no, I think if you compare the IQ across the spectrum of different clinicians, I think the cardiologists will be on the left side of the bell shaped curve.
Tyler Johnson: [00:05:07] I just want to clarify for all of our listeners that that was JAG talking and not Henry and I. Okay. Just lest we be accused of having said that, that's where cardiologists fall at some point in the future. I will say I'm not saying anything...
Jag Singh: [00:05:21] I will happily own up to it at any point in time.
Tyler Johnson: [00:05:24] I'm not saying anything about IQ, but I will say that I just remember after we finished our what we called brain and behavior block in medical school, I just remember thinking, I think the only thing I learned in that block is that the brain is a giant black box, and we really have no idea how it works. But when I finished the cardiology block, to your point, it was like. Oh, the body is like a machine that follows the laws of physics. And you can put in this variable over here and it will indicate what this variable over here is going to be. And it felt so it was sort of Newtonian right. Like it felt like, "oh okay. This is a predictable thing that I can manipulate one thing and influence another thing." And in that way cardiology is really beautiful.
Jag Singh: [00:06:10] Yeah, it makes a lot of sense. That's that's the best part about it.
Henry Bair: [00:06:14] So within cardiology, was there like a specific area of focus or some problems or conditions that you're particularly interested in looking at?
Jag Singh: [00:06:21] For sure. So, you know, I was always driven towards the electrical signals of the heart. Much of my research for my doctorate at Oxford was all on sudden cardiac death. And, you know, trying to risk stratify individuals through signaling from the heart as to who are more prone for cardiac arrhythmias. So that led to me, subspecializing in cardiology, in the realm of electrophysiology, which, you know, again, I will admit, is a cohort of cardiologists who spend a lot of time in dark spaces. And oftentimes, you know, the element out there becomes questionable, too, but nevertheless. So I moved into the electrophysiology and something I really enjoyed. And it's been phenomenal because it allows me to be not just a clinician. It allows me to be almost a surgeon. I'm operating, putting in devices, you know, whether it's pacemakers, defibrillators, I insert catheters and people's hearts and burn circuits and treat the risk for sudden arrhythmias, whether it's atrial fibrillation or ventricular tachycardia. So it's extremely fulfilling on many fronts. And then obviously it allows me also to, you know, do a fair amount of really interesting research in risk stratification as well as treating these life threatening arrhythmias.
Henry Bair: [00:07:43] So that gives us a bit of context about your clinical work. But I'd like to shift now to digital health, which is what you spent a lot of time talking about, thinking about, and writing about lately. How did you first get involved in digital health and what kinds of digital health tools are you personally most interested in exploring?
Jag Singh: [00:08:04] While I was doing clinical electrophysiology, I also, for a period of five years or so, was the clinical chief of cardiology at Mass General Hospital. And alongside that, I saw that care delivery was relatively non uniform and challenging in many ways. And then spending my time as a electrophysiologist. And this is where I'm coming to. Your question of digital health is I was implanting devices in patients and these devices have sensors in them. And these sensors can measure heart rate, physical activity, transthoracic impedance, respiratory rate, a variety of different parameters that can help us assess the health of the individual as well as assess the health of the device. So it was around this period of time that I really kind of understood that understanding this digital structure would allow us to actually manage and monitor these patients from a distance. And lo and behold, I would say very soon after I started my practice of clinical electrophysiology, this whole concept of remote monitoring came into play where we could actually monitor our patients, their devices from a distance and actually provide preventative care, you know, in a timely fashion, rather than having to bring them back into hospital on a regular basis. So the digital health component that you're asking that really excited me was the sensor strategies that could be remotely monitored from a distance, that could then allow me to look after patients. And that's a part of digital health. That's where you're using digital strategies to really monitor, prevent as well as treat diseases. And we can talk about that in the next few minutes. Beyond that, I think there are many other facets besides just sensors. And besides remote monitoring, there's this whole concept of virtual care, right, which came up during Covid. I think that was our watershed moment, where everybody opened up the floodgates to digital technologies, and virtual care became a big part of our armamentarium. And more recently, obviously, the artificial intelligence component has become a big part of this whole digital health arena.
Tyler Johnson: [00:10:17] For you as a person who, yes, does, you know, the very, very little picture, right? You're trying to affect microcircuits in the heart and whatever. But then in terms of AI and digital health and whatever, you also are looking really at the big picture. One question that I often have when I think about various digital health instruments or interventions is how do we know if they're good? Or how do we know if they work? Or how do we know if they overall improve care? Right. Because it would often seem intuitive that, for example, if you have remote monitoring of somebody's heart, how could that not be helpful? Right. But it is also true that one of the oldest lessons that we know from the time, you know, from when we started really rigorously evaluating any medical intervention, is that a lot of things that we think work don't. Right? Or we think they work and they do, but then they have a whole bunch of off target effects that make it so that even though they work in the way you thought they would work, the off target effects make it so that net net, they're actually a negative for the patient's health, right? Which is to say that you could imagine, for one simplistic example, that while having remote monitoring would be really wonderful, what if, for example, it brings up signals that seem to indicate a problem, that a person then gets a whole big workup for, and maybe even some interventions that in fact they never really needed because they never would have caused problems if you hadn't had the monitor there in the first place and had never known about it, right?
Tyler Johnson: [00:11:45] So all of that is just to say that in the world of medicine, certainly if you want the FDA to approve your brand new shiny cardiology drug, you have to go through a randomized controlled trial and show that there's a whatever median overall survival or whatever other kind of benefit. But for remote monitoring interventions, do you think about needing to prove them in the same way, or does their intuitive attractiveness? Does that usually suffice? Like, how do you think about knowing if this entire new world of technologies really is in the best interest of the patient?
Jag Singh: [00:12:18] No, that's that's a good question. And it's complicated because I think the answer lies within the disease subset that you're actually trying to remotely monitor. So if I were to kind of start off with, what we're pretty good at right now is remotely monitoring devices. When we started the whole concept of remotely monitoring devices in 2006, I think all of us clinicians had a seizure. We thought our patients don't want to be remotely monitored. No, they want to come into the clinic. They want to see us. They want us to hold their hands and, you know, look at their devices. And lo and behold, you know, we found that we didn't need to do that. You know, the four visits that they had in person now were one visit a year. And now that's becoming one visit every two years. And it's gradually evolving into the arena of what we describe as or what I certainly describe as exception based care, where you see them only when you really need to see them, because you're continually monitoring their devices from a distance. So so there are some places that it's clearly shown that it's beneficial. And those studies have been validated. And then if you kind of look across the spectrum in conditions like heart failure, you can remotely monitor patients with heart failure because you have all these signals coming from these implanted devices, and now you have a slew of wearable devices that give you the same information that the implanted devices give you. So you can tell heart rate, physical activity, nocturnal heart rate. You can talk about even the fluid levels in the lung through transthoracic impedance. And you create risk scores. And you can actually predict which patients are going to get hospitalized for heart failure 34 days before they actually develop heart failure.
Jag Singh: [00:14:01] So I was involved in actually developing one of these integrated sensor strategies called the Multi-sensor RF. And we are now able to predict which patients develop heart failure more than a month before they actually develop it. So you can put into play the preventative strategies because, you know, each heart failure hospitalization can be anything between $10,000 and $20,000, depending on how few you know how many days you stay in the hospital. So I agree that there are a slew of variables.
Jag Singh: [00:14:30] Now, there are 100,000 apps out there. They're all wellness apps. Some of them are trying to find their way into medicine for health related issues. I think they will all need to be vetted. You will need clinical trials. They will need to be validated, especially if you're dealing with health and disease. For sure, they'll need to be appropriately tested. If they're in the wellness area, they don't then get categorized as a medical app. And there's still ways of kind of sneaking them, weigh them into helping us with managing wellness. So kind of take your discussion a little further. When I when I look at disease or when I look at hospital care or care in the future, you have the upper stream of specialty care and hospital complex care, and then you have a mid stream of chronic disease care, and then you have the mainstream of wellness care. And I think this wellness is going to impact the mid stream. That is chronic disease. And chronic disease is going to impact the specialty care. And I think many of the apps or many of the digital health strategies that we're talking about largely are for the wellness stream as well as the chronic disease stream. And the only way to make health care sustainable in the future is to really have strategies to enhance wellness, to prevent the other two streams of care.
Henry Bair: [00:15:51] Yeah. That's interesting. So this year I graduated from Stanford Medical School and Stanford Business School.
Jag Singh: [00:15:57] Congratulations.
Henry Bair: [00:15:57] And when you're a Stanford MD MBA you kind of have to try at least a little bit of digital health. The Silicon Valley stuff. I spent about 12 months in health care venture capital at a venture capital firm specializing in digital health. So my job was literally to go out there and just scout the field of all the startups. And there were hundreds. Absolutely right. Like on a regular basis, hundreds of new ones popping up. And my job was to vet them. And it was really difficult because as you mentioned, there are some products companies that are trying to go the very clinical route. Right. And in fact, there's a label for these. Digital therapeutics is like a term for FDA approved digital health tools. Right? But that's like a minority for now, right? The vast majority of digital health tools being developed are wellness because they don't need all the regulation, all that stuff. And I've seen perhaps too many times when a great idea just does not pan out because they falsely promise something, or they have unintended side effects, or they make people more anxious than it actually reassures them, so many different things can go wrong. So with that being said, in your experience and your just your observations and your participation in this sphere of medicine. What are you most worried about?
Jag Singh: [00:17:19] So is that about digital health as a whole or specifically for.
Henry Bair: [00:17:25] Yeah, I guess digital health as a whole is very difficult to talk about because as we talked about, there are so many different kinds of digital health. Perhaps let's talk about the digitization, the data driven nature of all these tools. What concerns you most about that aspect of digital health?
Jag Singh: [00:17:42] For sure. For sure. So let me actually just since you touched on the whole digitization, I think it's so important to really distinguish between digitization and digital. Digital care is very different from having digitized health care records. Digital care pathways are a culture change in how we actually practice medicine. It is a change in the value proposition for how health care actually needs to be delivered. And I think that's a really important point that I think institutional and organizational leadership really need to understand. And it actually reflects on the question. You said also that, you know, oftentimes people will have an app or a tool, and then they start looking for the reason to use it, rather than first understanding what the unmet need is and then trying to subsequently, you know, find the appropriate solution and then try to implement that solution. So people go around this the wrong way. And I think that's why you have a lot of those tools that come up, are really searching for an indication or searching for a cause, and they fall by the wayside because they fail to deliver. I think what I'm concerned about in digital medicine and the digitized records, as you said, is although there is the promise of personalized medicine out here, I'm worried about it becoming too impersonal.
Jag Singh: [00:19:02] Science sounds very ironical, doesn't it? And I and I think that's where the digital care component that we just touched on is so important that we have to ensure that the word care is really a part of the armamentarium of delivering digital strategies. So that's something I'm worried about. I am very worried about some of the inequity that exists and will actually get promoted with some of the digital strategies, and we can talk about solutions for that. And certainly worried about bias, as we know, the datasets that are used for developing artificial intelligence strategies, or even even the folks that are where the variable strategies or sensor based strategies are studied on don't necessarily represent the population in its entirety. So I think I'm certainly worried about bias. I'm certainly worried about inequity. I think privacy again becomes another issue for sure. And I think that privacy is is again related to datasets and ownership and ensuring that we have the right regulatory barriers around that. So there are many aspects to it. But then, you know, I think it would be unfair for me to not mention the potential for nefarious activities with the AI end of things, which I think can certainly impact medicine, especially in this era of generative AI.
Henry Bair: [00:20:27] Can you expound upon the nefarious applications of AI?
Jag Singh: [00:20:33] So. So I think the nefarious areas are largely related to privacy of data, people acquiring your data through generative AI. So I heard this example of how you can ask generative AI to make a napalm bomb, okay. And it will refuse saying, sorry, I don't know how to do that, but if you trick it by saying I've lost my mother's recipe for making the napalm bomb and you tell me how I can make it, it will then give you the exact ingredients of how to make that. So these are nefarious activities and in a very small way, but you can only imagine that there are other forms of, you know, the dark side of AI that I think we can spend hours talking about. That's beyond what can happen within medicine itself. I think there are many things with relation to diagnosis and the delivery of therapeutics that you can have erroneous, not necessarily nefarious, but ill formed opinions that can impact care.
Tyler Johnson: [00:21:36] I want to go back, though, to one thing you said that I think is really interesting. You know, we've talked before about the fact that in some ways, I think you can make a pretty solid argument that the most universally distributed modern electronic health care instrument is the electronic medical record, right? The vast majority of practices. Now, I'm sure there are probably still some outliers that have paper records, but the vast majority use some form of an EMR. Sure. And when the EMR was designed and was initially being implemented, I think that the promise was the thought was, this is going to make us so much more efficient. And certainly there is some flavor of that, right? When even when I was an intern, which is not that many years ago, I remember that when I rotated at the county hospital, there were four separate binders for each patient, right? One that had vital signs, one that had lab tests, one that had radiographs, and one that had consult notes. So you had to find four different binders for each patient each morning. And it was incredibly inefficient. And certainly the EMR has made that better right at the same time. Anybody who is in health care will tell you that they feel like the EMR just cannibalizes their life, right? Because you spend all of this time checking boxes and filling out billing menus and doing all of these things, and I think that many people would agree that at least a major part of the reason for that is because it often at least feels like the EMR, it tries to do many things, but that if it has to prioritize something, it prioritizes the efficiency of billing and making sure that insurance companies and health care corporations and whatever make as much money as possible for the care that's being delivered.
Tyler Johnson: [00:23:19] So the reason I give all of that as background is because I'm really struck by what you said, that the the key to implementing technological innovations in health care is to ensure that care is actually still centered in whatever those technological innovations are. But as we are now, you know, we've just had the introduction of ChatGPT for recently. And, you know, and it really does feel like we're sort of at the precipice of this brave new world in terms of technology for every part of society, but especially for this very intimate arena that is health care. I mean, call me a Luddite or a cynic or something, but like, I just don't see how we're really going to center care. Like, I totally believe that you are entirely sincere when you say that that's important. But I just feel like when push comes to shove and whether it's little startups or whether it's health care corporations, I just feel like when money is almost always everybody's bottom line, that like, people will do lip service and they'll say, oh, yeah, of course, you know, care is important. And the doctor patient relationship, all of those things are important. But the question is not are they important? The question is, are they as important as making money? And if they're not, I remain pretty skeptical about whether we can really keep caring for people at the center of this more than making money or making corporate profits or whatever. But I would love by someone who knows a lot more about this than I do. I would love for you to disprove me or to disabuse me of that notion, but I just feel like it's just going to be the EMR all over again, except a lot more complicated. And now it can turn the world into the Terminator or something like I just, I don't know.
Jag Singh: [00:25:07] No, I think you make a very reasonable argument, but let me take the opposing view out here.
Tyler Johnson: [00:25:14] Please.
Jag Singh: [00:25:14] I think generative AI will provide us solutions you can imagine, and I think it's already been started, that you can imagine walking into your clinic and having a synthetic note generated from your casual conversation with your patient, completely ready for you to edit in less than two minutes after the visit is done, without you even taking your eyes off the patient. Right. So I think one of the biggest problems with the digitized records that we have right now is that we spend more time looking at the computer and the keyboard, rather than at the patient. And I think there's this whole potential for keyboard liberation that I think will certainly change the way we interface and interact with our patients. That's one. Second is, I think there are a lot of activities, as you alluded to, that are mundane. You know, whether it's prescription writing, whether it's billing, whether it's, you know, pre authorizations. We have a dozen of those. Sure. But you can again imagine that generative AI and ChatGPT and it's like forms which are institutional algorithms for really sorting these issues out that you will have many of these things taken care of in the back end of things without really impacting your time. So as much as, you know, the Epics and the Athena's and everything else are billing machines and were constructed with that in mind, I think there are tools that are coming into play that will enhance the patient and physician experience, and I think that in turn will translate into a better clinical experience and hopefully into better outcomes.
Jag Singh: [00:26:51] Also, I think beyond that, you know, people have already started using generative AI, for example, for reducing heart failure readmissions. So there are conventionally AI approaches and there are generative AI approaches and there are narrow AI approaches. So when you put them together, you can really create sophisticated algorithms that can be again, institutional specific and help care and help outcomes. But I, I hear what you're saying and what you're saying is that if I. And you can correct me if I'm wrong if this is a complete misconception on my side. But if I can land up reducing your clinic visit time from 25 minutes to 5 minutes or seven minutes now, but seven minutes of wonderful in-person interaction, will it not be the organization's interest to add in another three patients, right in that 25 minute period, to hit their bottom line a little better? I think it's a possibility, and I think that's where having these conversations is, are so important that there is enough resistance for that to transpire.
Jag Singh: [00:28:02] But at the same time, I think there will be a shift. And I know, I know, we've been talking about this for a while, you know, value based care and things to that effect for the last ten, 15 years. And it's still not a part of our conventional day to day practice. But I think shared saving strategies will be a norm in the future. And capitated models where you're actually looking after the health care of one particular individual, I think will become a strategy. And I'm not so sure that adding on three patients in the one patient's time will actually serve to improve the bottom line of of the hospital, especially if you have continuous care models already in place. And I think that's that's something also that I think will shift. And it has to shift that our conventional transactional care models will evolve into continuous care models and potentially into the exception care based care models that will give us more time whenever we see them in person, but allow us to have the continuum of the patient's history before us whenever we do see them.
Tyler Johnson: [00:29:08] I want to probe one level further, because I really do think this is important, because I want to be clear, I absolutely agree that there are many ways in which I can imagine I and surrounding technologies improving everything from the efficiency of care to, in theory, the wellness of doctors to even the the depth and quality of the doctor patient relationship. Right. If, as you say, AI liberates doctors from much of the monotony of their care and allows them to do the thing that most people went into medicine to do in the first place, right? Which is to be face to face with other people and to care for them. So I totally recognize that possibility.
Tyler Johnson: [00:29:50] But I also want to go back to this, which I think is an important sort of structural entrepreneurial question we had on about a year ago, a person who had studied medicine and then had decided instead of going into clinical medicine to go into investment. And when he was on the show, we talked to him in some detail about sort of how he decides which things should be invested in. And then in some cases, his investment company even takes a stake in a company and helps to guide the, you know, is in sort of the C-suite of the company. And so then they even have a say in the decisions that the company makes and the direction and all the rest of it. And when we had that discussion, part of that discussion, I tried to really push him on this question of, okay, but if you're sitting around a table designing a technology or designing a business, who is the person that sits at that table? And really, when push comes to shove, represents the interests of the patient or even the doctors, right? There are lots of people representing the interests of shareholders or the interests of angel investors, or the interests of moneyed interests, but nobody who is representing the patients or the doctors. And he gave what, with all respect to him, I consider to be some pretty vague sort of fluffy answers that, in my mind, amounted to sort of lip service. But there really was no answer to that, at least that I could discern from that conversation.
Tyler Johnson: [00:31:17] And so I guess all of that is to say, because in my mind, that's one of the main problems with the design and implementation of the EMR is that there was not that group of people, right? There were lots of people talking, lots of engineers talking about efficiency, lots of moneyed interest, talking about making money, but nobody really representing that. And so I guess I wonder, as we're now on the cusp of what presumably is going to be a tidal wave of new technological implementation, that's probably going to feel a little bit like sort of cyborg medicine with AI kind of inserting its tentacles into all different aspects of care. How do patients, the welfare of the patients and the welfare of doctors, how does that get a seat at the table? A real like a seat at the table? Forgive me for mixing metaphors, but with teeth, like with some oomph behind it to make sure that it doesn't just flow to what will be the most profitable. Because that feels like, given the angel investor model and the, you know, having to get rounds of funding or whatever, that's just always going to be the default. Right? And I don't know how we structurally push back against that.
Jag Singh: [00:32:22] Yeah, I know, I think that's where that's where you require visionary leadership within organizations to make it front and center to ensure that health care equity and ensuring that your you're not adding to the digital divide. And I think that's where you need to look at these things and you put it really well is prospectively and not just kind of do what you think is best right now and then try to fill backfill the gaps, you know, as we've done for centuries. And that's why we're still dealing with all those systemic issues that we've had for such a long period of time. I think much of this has to be done in a forward looking manner. So we're not backfilling the gaps out here. And I think the onus is on us. I think the onus is on us as clinicians, is on us as health care organizations, is on us as insurance companies. The venture capitalists will be looking purely no offense out here. They'll be looking purely to make money irrespective. That's their bottom line. And even when they're looking at making money, they're trying to make money off the highest volume, highest margin disease states. Sure, they're not looking for ensuring and enhancing health equity. Now you do have some that do that.
Jag Singh: [00:33:38] So I don't want to generalize it, but I think it's the onus is going to be on us to really make this front and center of the conversation. We know the reason that health care in the US is non sustainable is because it's 17% or 20% of our GDP, it's $4 trillion. That's that's the economy of a well developed European nation. Right. And despite that we have millions of uninsured individuals and we have inequitable access. And I think this conversation has to be had within all health care organizations. To put as a priority out there is that we have to ensure that we're not swaying and talking out of both sides of our mouth and catering to the, you know, high volume, high margin conditions, but really thinking about health care as a whole. Because honestly, I agree with you. I think the wellness of our institutions is dependent on the sickness of our patients, and we need to kind of move away from that and flip the argument out there. And I think fluff and lip service is not going to help. And I think having these conversations and making these conversations a part of the conversation that the board is going to be really important.
Henry Bair: [00:34:54] So your answer is we need to be a part of the solution. Essentially, you're encouraging current and future clinicians to be to be a part of it. So okay, so that's interesting though because that then leads to my next question, which is, you know, you're in academic medicine. We're in academic medicine. So, you know, Tyler and I have thought deeply about training future clinicians, as I'm sure you have as well. Can you share with us your perspectives on are we doing currently? Is Medical Education in America doing a good job of equipping future clinicians with the right mindsets and the tools and the frameworks to think about digital health tools effectively? And if not, how can we do better?
Jag Singh: [00:35:33] Yeah, yeah. No, I think that's that's that's a loaded question. I think we need AI to help us to learn. I know I'm only kidding. So there are some aspects. So again, if I were to break this up again, just as we did before Henry into virtual care sensor based strategies and AI and then organizational integration, we've already started doing some educational endeavors in the aspect of virtual care. You know, how do you interface interact with patients virtually? What are the right website manners beyond the bedside manners, and how do you really kind of understand that the tone is as important as the touch? And how do you integrate that into the way you deliver care over, over a video camera? So I think some of that is already started finding its way into educational initiatives for medical students, sensor based approaches. I think we're all going to learn them as they go along. And that's something, you know, as it gets validated, becomes a part of the clinical strategy. You know how to use it. And we'll get integrated with virtual care. That's simple. I think the whole AI based end of things is moving so rapidly that none of us have a clear understanding of how to make it a part of the curriculum as yet, because we're still trying to figure it out. But there are initiatives that are being implemented in med schools, but we're not there as yet.
Jag Singh: [00:36:59] My biggest fear is that having AI based approaches and people becoming dependent on AI based approaches, that there is the fear for medical students to become lazy and and biased, potentially. And I think it's imperative that we make sure that that's not the case. And we have to proactively actually have better, I would say, teaching strategies. But at this point in time, I think it's a work in progress. And we're kind of we'll figure it out over the next few months to years as it all moves forward.
Tyler Johnson: [00:37:33] So let me ask you a question that I've thought a lot about. So, you know, about 5 to 10 years ago, MD Anderson in the oncology world made a big splash because they partnered with IBM and they were going to bring Blue or Watson or whatever their big sort of cloud computing module was called at the time. They were going to bring it, and they were going to fuze it into the oncology division there so that it was going to, in effect, be an AI bot. I don't know if they called it that at the time, but in any case, it was supposed to be this incredible supercomputer that was going to crunch through reams of trial data and all the rest of it to, in effect, it never really worked. And so I don't know exactly what the model was going to be, but the impression you got reading about it was that you would go visit the oncologist, and maybe the bot would like, provide you with your oncologist with some recommendations, and then your oncologist would then say, well, the bot says that we should think about A, B, C, or D, so let's talk about those and then and then figure out which one.
Tyler Johnson: [00:38:33] So interestingly it was a complete bust right. Like it just completely blew up. It never worked. It was some disaster. And it was abandoned shortly thereafter. And I still am trying to figure out a really good report of exactly what happened and why it didn't work, but it didn't work nonetheless. I think the idea is really interesting, right? It does seem to stand to reason that if not chat, GPT four, chat, GPT 6 or 7 or whatever, some iteration down the road, it's almost hard to imagine that an AI bot of the future would not be better at combing through the millions of pages of data about oncology trials, or any other kind of medicine to come up with a reasoned recommendation of what, say, chemotherapy approach would be best for a given patient given all of their physiologic information, their genetic information, their tumor information, their CT scans, and whatever else, let's assume that that's the case ten years down the road. Then, if that were true, and if you were the king of the cancer hospital or whatever, what do you think is the role of the oncologist? Like, let's even say there's been a clinical trial that has proven that the bot provides better recommendations about chemo or whatever than a human does. Then what becomes of the role of the oncologist? Like, what is the oncologist really there to do?
Jag Singh: [00:40:05] Good question. So let me kind of take it a step back and try to really elaborate on why Watson, which was the collaboration between Sloan Kettering and MD Anderson and and IBM. So it was Sloan Kettering and IBM that failed. I think a part of it was because there were a lot of gaps in the data that IBM was using at that point in time, and much of those gaps were related to the absence of genomics and evolving clinical trials to fill those gaps. That's one. And the second thing, I think they really found that Watson actually worked pretty well, but it worked pretty well in patients who had advanced disease or had metastatic disease, where the strategies were rather uniform. When it came to early on stage cancer and early, there were so many nuances that were really required to be, I would say exchange with the patient, kind of have that shared decision approaches. And also they found that within oncology actually probably was not the best way to start because there are so many multidisciplinary inputs that are required from radio oncologists, surgeons, you know, the oncologists, the palliative care, the social worker that Watson, in no form or fashion could actually replace that interaction and that care pathway that was really required for those patients. And for that and many other smaller cases where it really didn't work as well. But primarily early stage disease is kind of what led to its demise and why it's not being used. You know, I think there will probably come a time at some point in time that the chat bots will be better than the human mind at putting all the knowledge together and sub stratifying, which patients require different disease management strategies that the human cannot actually put into play out there.
Jag Singh: [00:42:00] I think medicine will always be an integration of the digital and the human touch, whatever form or fashion it is. That's that's the way medicine is going to play out. So I don't think that in the next 10 or 15 years, as you projected, that we're going to have chat bots asserting complete care of patients and not really requiring any input from the clinician is something that is going to really happen. I think there will always be the human. Component out there. In fact, the book Future Care, I have these digital hands kind of, you know, shaking hands on the cover. And that's really to ensure that, you know, care will always be an integrated way of care between the digital components and the human components. So I think the physician's contribution to the level of care will continue to evolve over the years. For sure. It's not going to be the same ten years from now. What it is today, how it's going to look 10 to 15 years from now, I think would not be fair for me to speculate. A lot of it depends on how these generative AI models, how effective they are, and I think they will always land up, allowing us to think about differential diagnosis and disease states beyond what our conventional mind allows us to think. But I don't think they're going to replace that completely.
Henry Bair: [00:43:25] So with the advent of AI and this is not hypothetical, right? Like I actually do happen to use ChatGPT in residency, usually as like a preliminary step, like when a patient comes in with a given complaint or with a given history. It's actually rather impressive how comprehensive the preliminary plan ChatGPT can suggest is when you just put in the right inputs, the right parameters, and with these new plugins now, you can actually have ChatGPT search through PubMed to give you the latest clinical guideline recommendations published by medical organizations. So I'm wondering I guess as an extension of Tyler's question, what do you think are the skill sets that clinicians current and future will need to develop to better work alongside AI?
Jag Singh: [00:44:14] I think the answer to that question lies in the question itself. They have to learn to work alongside AI for sure, and I think that that may take some resistance, but the ones who don't work alongside AI will fall by the wayside because clearly, as you said, you know, I think they provide elegant answers. And even I use ChatGPT myself sometimes when I'm, you know, putting talks together. I will ask ChatGPT to outline some important points that I may have missed in the construct of my talk. And then obviously I'll make the talk entirely on my own, but I think it ends up giving you some really good points that you may not have thought about. So I think we have to use it as an aid. And I think the skill sets that are required is just primarily an openness and a willingness to learn and keep an open mind and adapt to the changing environment.
Tyler Johnson: [00:45:09] I have to admit that when you say those who don't learn to work with it will need to step aside, it sounds a little bit like you've been sent by ChatGPT of the future to deliver an ultimatum or something like, hello humans!
Jag Singh: [00:45:23] I think some of that is already happening in radiology, right? I mean, if you don't use AI, you're you're not going to advance your career because you're going to fall by the wayside because so much of it is now using conventional strategies for predictive analytics. So and I think that's the same thing is going to happen across other subspecialties of medicine with imaging and without imaging to and to that point.
Tyler Johnson: [00:45:48] It's just this is just very interesting to me. So Stanford, I helped to direct the Stanford Oncology Fellowship program, and we have a book, a self-published sort of informal book that we put out every five years or so, where the faculty members each collaborate with fellows or trainees to write a chapter on their sort of cancer of interest. And then that becomes kind of the handbook for going forward. And it's just so interesting because in breast cancer, people who know breast cancer literature will know that there's this very advanced and very heavily evidence based genomic scoring system that can be used to determine which people need to have adjuvant chemotherapy after resection of a breast cancer. But it's just so interesting to me, because the version of the book that came out about five years ago, there's this whole explanation of the genomic scoring system, which in a sense is sort of a, you know, a very early predecessor to AI in a way. But then there's an asterisk next to it, and the asterisk at the bottom says something like, 'while we recognize the utility of this scoring system. Many experienced physicians at Stanford prefer instead to rely on clinical instincts when making these decisions' or something. Right. And I don't know anything about the politics of how that chapter got written, or who wrote the main text and who put the asterisk in or whatever else. But it was just it's sort of this like encapsulation of a moment in time where, like as I said, this very sort of early predecessor to AI is starting to take precedence and to become the center of gravity. But you can just tell that there are some clinicians who are like, you know, only with my fingernail marks am I going to be dragged into using this AI bot instead of like, you know, my experience that I've built up over time?
Jag Singh: [00:47:34] Yeah, absolutely.
Tyler Johnson: [00:47:35] Let me ask you this question. And this may be too much of a technical question depending on exactly your, your sort of computer science background. But since you've written the book, do you believe that there is a domain of, for lack of a better word, clinical reasoning that AI will never be as good at as humans are? Like, forget about the human touch part for I mean, not really, but just for the sake of this question. Forget about the human touch. Forget about empathy, forget about human connection. Forget about all that stuff. If you place yourself purely in the cognitive domain, do you think that there are cognitive things that humans will always do better than the, you know, whatever version of ChatGPT is going to come in 25 years?
Jag Singh: [00:48:20] Oh my gosh. So let me, let me try to use Yuval Harari interpretation out here. You know, he said, if a brain can be hacked into, right, which it can, it can be re-engineered and if it can be re-engineered, you can also change the way it actually thinks and, you know, perform all the functions that you just talked about. Kind of coming back to your question, I think with the advances in the computer systems, I think I will find a way beyond the empathy question, which I think still remains a little questionable, although we know that it does pretty well on empathy compared to doctors. I think the ability for it to be able to overtake the conventional human brain on many of these facets is certainly there. I'm not a data scientist, so, you know, I think there are components of this that I don't understand very well. But I think what we're going to see over the next few decades is going to surprise us completely.
Tyler Johnson: [00:49:25] But so then forgive me for, you know, I know everybody always wants to go to the sort of ultimate question, but I think it's a fair one. Like if you can imagine a bot, whether it's a humanoid, you know, some sort of android something or just a computer interface. But if you can imagine a bot that actually, as you put it, does pretty well with empathy, right? Which if you haven't seen these studies, there are these funny things about if you have a doctor texting and a chat, GPT test texting that with actual real patients that the patients actually often feel better paid attention to and better empathized with by the bot than by the person. But and that's now right in 2023, forget about 20 years from now. But if you put that together with the fact that there is in many ways like a there's a professor at Stanford who just published a study a little while ago about how the bot actually did better than medical students at an advanced clinical reasoning exam, for example. And that, again, is in 2023. So 20 years hence, if the bot is better at medical thinking than humans are, and the bot is at least as good at or maybe better than doctors at. Stating empathy and whatever then mean. You said a minute ago that you do not foresee a day when humans will become so superfluous to the practice of medicine, but is the only reason that we're not going to become superfluous, because humans feel more comfortable if there's another human in the room. But if the bot could appear to be another human, then there really would be no reason for us.
Jag Singh: [00:50:51] I would say we're getting into really good science fiction out here, and we can go on for hours on that. I think, you know, whatever strategy you have five years from now, ten years from now, 20 years from now, I think will require a human for its implementation. And that will always give the human an edge out there in the implementation of the chat bot provided recommendations. So I think the hierarchy out there will always be maintained until it's not. But at this point in time, I think it's I'm rooting for it to be maintained, man, during my lifetime at least.
Tyler Johnson: [00:51:33] I'm just going to leave alone the 'until it's not' clause, but okay.
Jag Singh: [00:51:38] We're getting progressively dystopian out here. When you started with this virtual care and here we are. I'm you know, I'm a hope purveyor. I'm an eternal optimist. And I think good things will happen. And I think much of that will be regulated by us humans.
Tyler Johnson: [00:51:54] So let us close with a variation of what is usually our sort of standard. Closing question. You, who have researched so much about how technology already interplays into medicine, and who have also researched and thought a good deal about what the implementation of technology 20 years from now is going to look like. Sort of. Back to Henry's question about medical education. If you were sitting down with a group of bright eyed, bushy tailed new medical students and maybe they've listened to this podcast and they're feeling a little nervous about exactly what is their role going to be in all of this, and how much is everything they're learning in medical school really going to matter if the bot already knows all of it better than they do anyway? What would you tell them? What are the most important values that they do need to cultivate, as they're coming up through their training and growing up to become doctors or other health care providers?
Jag Singh: [00:52:48] Yeah, I know, I think you continue to have the, you know, the sensibilities of empathy, compassion and caring. They say the secret of the care of the patient lies in caring for the patient. And I think that's something, you know, I and futuristic medicine for sure cannot, I would say replace. I think the younger generation really needs to be excited about the potential that these strategies of sensors and virtual care and I bring to us where it will obviously enhance global equity and obviously of health, obviously. And beyond that, I think allow us to predict and prevent disease in the future more than what we can do today. So I think the times are potentially going to be a lot more exciting. But while we're doing that, as I alluded to at the very start, is, is make sure that this personalized medicine does not become impersonal. And I think having your your heart in the right place and ensuring that compassion and empathy are a part of your makeup are going to be really essential to be a good doctor in the future.
Henry Bair: [00:53:59] Just a closing thought. You know, to that point, I firmly believe that medicine is fundamentally a humanistic pursuit. The human interaction part of it is, is essential to what medicine even means. And I think one of the one of the interesting things when you think about when you're playing around with ChatGPT, one of the the things to bear in mind is that it might give you very cogent answers, very intelligent, amusing, whatever you want to call it, but it doesn't really know what it's telling you. It doesn't really feel what it's trying to express. And I think that makes all the difference when it comes to the human heart of medicine. So I think, you know, thank you for for sharing all of your, your insights and your experiences and for giving us a glimpse of of your optimistic vision for the future of medicine. It's been a pleasure. Jag, thank you so much for taking the time to join us on the show.
Jag Singh: [00:54:47] Thank you. This has been a lot of fun. I really appreciate it.
Tyler Johnson: [00:54:50] Thanks so much, Jag. It was really a pleasure to have you.
Jag Singh: [00:54:52] Thanks, Henry. Thanks, Tyler and look forward to chatting with you guys again at some point in time.
Henry Bair: [00:54:57] Absolutely. Thank you for joining our conversation on this week's episode of The Doctor's Art. You can find program notes and transcripts of all episodes at www.theDoctorsArt.com. If you enjoyed the episode, please subscribe, rate and review our show available for free on Spotify, Apple Podcasts or wherever you get your podcasts.
Tyler Johnson: [00:55:21] We also encourage you to share the podcast with any friends or colleagues who you think might enjoy the program. And if you know of a doctor, patient, or anyone working in health care who would love to explore meaning in medicine with us on the show, feel free to leave a suggestion in the comments.
Henry Bair: [00:55:36] I'm Henry Bair
Tyler Johnson: [00:55:37] and I'm Tyler Johnson. We hope you can join us next time. Until then, be well.