IONIA Healthcare Consulting

Episode 21 with Dr. Eric Topol

An afternoon conversation with Dr. Eric Topol on Scripps campus

Dr. Eric Topol is the Executive Vice President of Scripps Research Institute

Dr. Eric Topol is the Executive Vice President of Scripps Research Institute

Full episode here: http://bit.ly/VoiceHealthcare

We need humans. We need doctors. We need care.

Below is a conversation we had recently with Dr. Eric Topol on the campus of Scripps Research Institute in La Jolla, California. Dr. Eric Topol is executive vice president of Scripps Research Institute and founder of Scripps Research Translational Institute.

His work focuses on genomics, big data, and digital health technologies advancing the promise of personalized medicine. He's published 1,100 peer reviewed articles, has more than 200,000 citations to his credit, is among the top ten most cited researchers in medicine, was recently awarded a $270MM grant from the NIH, and has over 170,000 followers on Twitter.

He's authored two best selling books, most recently Deep Medicine focused on A.I. He's widely regarded as the most influential physician leaders in the country.

Full episode here:http://bit.ly/VoiceHealthcare

Dr. Cybulsky [00:00:15] Hi and welcome to the 21st episode of The Voice of Healthcare  podcast. I'm your host, Matt Cybulsky, the founder of IONIA Healthcare Consulting. We focus on performance, quality, digital strategy and design of voicefirst interfaces for healthcare. I'm joined today by my co-host Dr. Reid Maclellan, the founder of Asclepius and instructor of surgery at Harvard Medical School and Boston Children's Hospital.

Dr. Cybulsky [00:00:40] What you're about to hear is a conversation we had recently with Dr. Eric Topol on the campus of Scripps Research Institute in La Jolla, California. Dr. Eric Topol is executive vice president of Scripps Research Institute and founder of Scripps Research Translational Institute. His work focuses on genomics, big data, and digital health technologies advancing the promise of personalized medicine. He's published 1,100 peer reviewed articles, has more than 200,000 citations to his credit, is among the top ten most cited researchers in medicine, was recently awarded a $270MM grant from the NIH, and has over 170,000 followers on Twitter. He's authored two best selling books, most recently Deep Medicine focused on A.I. He's widely regarded as the most influential physician leaders in the country.

Dr. Cybulsky [00:01:33] Before we begin, a quick word from MEDarchon, a proud sponsor of The Voice of Healthcare podcast. MEDarchon assists hospitals, health systems, insurers, and other providers with a secured messaging platform they call QUARC. They approach secured messaging through the lens of effective care team communication essential for ED throughput and length of stay management. MEDarchon has built an exceptionally strong performance case for their platform QUARC.  To find out more about their approach to improving clinical workflows impacting vital outcomes, visit GETQUARC.COM.

Dr. Cybulsky [00:02:09] And now, to Dr. Eric Topol.

Dr. Topol [00:02:11] Thanks,  great to be with you.

Dr. Cybulsky [00:02:19] Wonderful. I'm going to start off with a couple of questions. So, when I meet with folks who have been in healthcare a long time working in technology, especially the edge of the circle expanding tech, which I consider AI and voice to be just that, I'm asked the question several times, which is, do you have to explain what voicefirst is to audiences or to people you meet with? And a lot of times, I do. And that's common, it happened a lot in the 90s for example with internet, and where we went with with web pages and access for patients and data.

Full episode here: http://bit.ly/VoiceHealthcare

Dr. Cybulsky [00:02:51] Why don't you give us your idea of what A.I. is for our audience, and why it's the most exciting thing you've seen in medicine in the last 40 years? 

Dr. Topol [00:03:01] Well, AI has been around even before I got into medicine more than 50 years. But, what's happened in the last decade is the deep learning and deep neural networks and that is the way to take, whether it's images or speech or text. And, take this through machine learning, through algorithms and neural nets to get interpretations that are extraordinary. And so, we've already seen that with images and now we're now seeing it with voice and it's going to have transformative impact in medicine. Perhaps, more than any technology that's antecedent to this point.  

Dr. Cybulsky [00:03:45] The first question that comes to mind when I think about what you just said is the examination of the patient and the relationship that you would have as a healer with the patient and using a tool like A.I. Things like intimacy, privacy, ownership of information. Curious if you could share with us a little your thoughts on just that, healing in the medical arts has largely been about a relationship. So, how does it change that? And, is it necessarily better? 

Dr. Topol [00:04:16] Well, this is really in many levels that we're talking about as far as the actual encounter and it's compromised with the use of keyboards. We need to have the liberation from that, and that's where voice into into text will be extraordinarily important. But, this relationship has been eroding for decades and it's not just keyboards. There are many other factors, but the biggest one, the singular one, is lack of time. And, so, lack of time to cultivate the trust, the presence, the exam, the laying of hands, the sense of care, compassion, and empathy. All these things have taken a hit.

Dr. Topol [00:04:59] So, you know there's many different strategies to use A.I. to restore the care in healthcare and to improve, and hopefully, at some point, get back to the kind of precious patient doctor relationship we had over 40 years ago.  

Dr. Cybulsky [00:05:18] I love that adjective precious, It certainly seems to be a place where that has been eroded. You see that a lot. I mean we've had conversations with folks who talk a lot about the inefficiencies of care and the burden of that on the physician. So, when you talk about these efficiencies being magnified using A.I., and in our case voicefirst, we talk almost exclusively about how can we buy time back for the physician? How can we buy time back for the patients or be more connected? Do you necessarily have a hope for that efficiency to be physician side to reduce burnout? Or, Do you see that more of a patient side to reduce burden on them so they can take better care of themselves long term?

Dr. Topol [00:06:00] Well, it's really a combination. So, on the clinical side that's doctors, and nurses, and physician's assistants, and nurse practitioners. and the whole gamut of health professionals. The idea is that voice is much faster than typing, and the accuracy now is getting at levels that exceed professional medical transcription, and not only that, but there can be machine learning for that particular doctor. So, we're just getting better and autodidactic. So, on one side is a big decompression of a burden with speed and accuracy. The notes that we have today are 80 percent cut and pasted because people don't want to have to type in new information and there's a propagation of error. So, that is an area that can only get better. But, the other side of this is on the consumer patients side because they're people generating their own data through sensors, through their environment, through their genome, all sorts of ways of generating data and that data has no home right now. Moreover, we are going to see, we're starting to see specific diseases conditions like diabetes, for example. But, eventually we'll see a general health coach, a virtual health coach, which you'll choose a voice and an avatar. You personally get to choose your avatar, but they'll be getting coaching and feedback about their data in real time process, and that's going to be for those who are willing, it's not for everyone, and that's going to be another way to decompress the doctor's workload. So, you've got things going on both ways. I see it as a flywheel towards this deep empathy, this better relationship, this gift of time, where we reduce the burden on clinicians, give patients more charge and responsibility, which they actually would like, and try to get healthcare in a far better state than it is today.

Full episode here: http://bit.ly/VoiceHealthcare

Dr. Cybulsky [00:08:13] I love that concept. Deep empathy. In terms of technology and deep empathy, could you go a little farther down that rabbit hole?  

Dr. Topol [00:08:22] Well, I don't like to think of it as a rabbit hole! I actually think of it as something like a treasure. Yeah, but the idea is that Deep Learning, which we've talked about, is the way to take this deep phenotype, which is extensive data about each person and all their inputs.

Dr. Topol [00:08:42] That is, from ideally from the time they're in the womb to the very present moment with all the things that are not just in an electronic record, but in these other fields and nodes of data accumulation, like I mentioned. So, when you take that ability to process that data through an AI, then you have the ability to finally to cultivate deep empathy.

Dr. Topol [00:09:06] What I mean by that is, all these different ways of giving time back to two human beings who are communicating to  establish the trust that used to be there and that was characteristic (of medicine). That's really what we need desperately, right now. Because, what we have is the record worst burn out, clinical depression, and suicide rates among doctors and ever, in history. And there's a  reason for that, which is: why do we go into this profession in the first place?  It was to care for our patients, to make a difference, by, as you use the word, healing. But, you know, that is really the notion, is that you could help people with  health, which, what could be more important?  But, the problem is, people don't feel like their able to execute their mission. And, in fact, they're just burdened, drowning in administrative tasks largely about data entry clerk work, which not only extends to being two thirds of their workday,but also they have to take it home every evening and on weekends. So, something has to give, and that's what I think we need this to restore, take back the profession, the ability to care for people, and re-establish this really important relationship and ability for compassion and empathy to be expressed. Time is so short these days that patients get interrupted within seconds of starting to talk during an encounter. And, that's the last thing we need. We want to give people space to tell their story, which will never get digitized, and never be really captured by AI...a patient's life story. And just listening is step number one, but you've got to have time for that.

Dr. Cybulsky [00:11:12] What about automation for the doctor patient relationship?

Dr. Topol [00:11:16] Right. So, not only the way that I reviewed, but also a lot of automated diagnostics for patients. So, routine things that are not serious, like a urinary tract infection, a skin rash, an ear infection for a child, and a long list of the most common things why people go see a doctor. Those will be done without a doctor, without a nurse. The only time that that may change is when a prescription is needed for an antibiotic, for example. And, in many countries, that isn't even necessary. But, the point being, is that, this doctorless side of routine matters that are not serious are, more and more, going to be used using AI, validated  algorithms. So, between all these different strategies that's how we get this gift of time, which is a core way to get back on track, and I think the problem we have though, Reid, is that we have administrators, managers ,the overlords of doctors, who want to squeeze and have gotten us in the state we're in right now, frankly. Because, the question now still looms whether we'll be able to undo that what has already occurred; that is, get more squeeze, more burnout, more depression, because, see more patients, read more scans, read more slides, and on and on. So, although I have lots of concerns about A.I. in healthcare, things like: worsening inequities, and bias, and the methodological concerns that you might share, blackbox, and explainability, all these other things. But, my biggest concerns are, none of those, but rather being beholden to administrators who only have one thing in mind, which is, more productivity, and not using this gift to turn inwards, and to re-establish care in healthcare.

Dr. Cybulsky [00:13:26] Do you think practitioners and physicians would be able to see patients more thoughtfully with these tools?  

Dr. Topol [00:13:31] Thoughtfully, and at a higher index of need. So, you know, simple things without a doctor or virtual coaching preempting the need to see a doctor. And then when that encounter occurs, it will be with objective data that's been collected, oftentimes. And then, of course, this elimination of keyboards and data clerk functions to a very large, if not complete, degree.  

Dr. Cybulsky [00:14:01] These are all ways that we can do that with the advent of smart speakers and virtual care assistants. What about the patient's privacy and the patient's data?   

Dr. Topol [00:14:11] Well, we have a serious problem because it's not protected as it should be. And the question is: Will medicine and health data get a different level of respect in legislation? Which is, I think desperately needed. It's one thing to sell a person's data regarding their shopping, for example, or their music, or things like that, and another one to to have their very private data regarding their medical conditions and medications and all sorts of things related to it to be sold and brokered and hacked. And these are the problems we're having today. I mean, just this week Quest Diagnostics had some 11 million people's data breached and that's important medical data about lab tests as well as other background about each of those people. Over 62% percent of Americans now have had their medical data either breached or stolen, Cyber Thievery, hacked, held ransom, by health systems.. held hostage. So, this is just an unacceptable environment and now we have the concerns about Alexa listening even when you don't have it activated and other smart speakers. So, something has to be done and we're just not taking this seriously. In Europe, with GDPR legislation that was enacted last May, clearly they have taken some of the right steps to move in that direction and we haven't yet done that here. I'm hoping, because this is so important, that the privacy of medical data is maintained. There's not just legislation that's needed. There's also technological intrusion. As I mentioned earlier, a lot of the data today is homeless and so we don't even have a platform for it. It's not organized, it's not searchable, it's not in a user friendly, or shareable. And, a lot of the data you don't even want in your electronic medical record; you don't want your genome there, you may not want all your sensor data. So, we've got to do better. One of the greatest ways we can get to privacy and healthcare data is for each person to own their data. That is unlike the model of today when doctors and health systems own your data. It should be just the flip of that because, when data is not sitting on these massive servers where they are a target for cyber thievery and hacking, the cyber gurus say that, if you just have it at the unit of one, the chance of that ever suffering a breach is extremely low, if not close to nil. So, though, that's where some of the things we got to keep in mind as we go forward

Full episode here: http://bit.ly/VoiceHealthcare

Dr. Cybulsky [00:17:16] I remember you had a conversation recently with Dax Shepard on what they're doing in Estonia as a use case for the consumer owning their data. Could you talk to us a little bit about that?   

Dr. Topol [00:17:28] Estonia is kind of the world model for medical data. Each person does have all their data and owns their data.

Dr. Topol [00:17:37] It sits on a blockchain platform and it's shared in pieces and components of it as directed by the individual, or in the case of a child, the parent. And, so, it could be used for a medical research study. It could be, some of it could be used for a specific doctor that the patient is seeing. The model is getting a lot of interest in Scandinavia and Finland and other countries. And also, the individual's data ownership is big in Switzerland. So many countries throughout Europe, and even in other places, are starting to cue into the importance. It's doable. Estonia has proven that it's not a big country, but it's actually got the most advanced health I.T. anywhere. And, we can learn from that. And, there was a really great New Yorker article about this last year that delved into some of the specifics.  

Dr. Cybulsky [00:18:34] There's massive economies with data in this country. Legislation ,of course, is one way to get it back into the hands of the consumer. But, I think there also needs to be somewhat of a sort of social movement towards the consumer saying,  "wait a second wait a second!" 

Dr. Cybulsky [00:18:48] I do want that! I don't know, Estonia's small and I don't really know that country well, but I kind of like to own my own information and protect that.  

Dr. Cybulsky [00:18:55] How do you see the social movement of this for consumers and patients? Especially, getting to a place where Senators and Representatives are pressured to look at legislation for the American people?  

Dr. Topol [00:19:06] Well, there isn't any social movement for this yet. I think it's inevitable once data became immanently portable and everything went digital This is something that's a natural step that's going to follow. The question is when? But, you know I have spoken with a bipartisan group that the Commonwealth Fund organized and I actually saw both Republicans and Democrats fully supportive of individual data, medical data ownership. But, we just don't have the follow through and we don't have, you know, when people try to get their data. I have experienced it when you try to get your scans and your labs. They're supposed to be, of course, laws to support that. Now, when you wind up trying to get your data you find out that those laws are not being enforced that there's all this information blocking because health systems and don't want to lose their patients, and so, they don't really cooperate and they still act as they are the owner and purveyor of all medical data and you are basically this beggar trying to show up. And, of course, you have to pay to get your data you know copy per page. I mean, some of the ridiculous things that are still going on today with fax machines and whatnot, it's incredible. So, you would think there would be enough unrest because everybody who wants to get all their data unless they're just using a portal and getting the de minimis data that you get through a portal, and they happen to be the health system that has a reasonable portal but other than that, to get your notes, to get your actual raw data scans, to get things you need for a second opinion and another health system, it's really difficult. But, we don't have people clamoring because they're, I think ,one of the reasons is, they don't have the realization that we're ready for this, but the patient world has been suppressed for so long. Yeah, like over two millennia! They just don't even know they have rights. And, so, you know, it's almost like it's habitual, but eventually I think there will be a social movement to turn this around and to get the rights of individuals to their data established.

Dr. Cybulsky [00:21:27] What about FDA approval of these algorithms and patient safety?

Dr. Topol [00:21:31] Well, there's a lot that's a loaded question and, in many respects, the reason it is is because there's been over 25 companies that have had clearance for FDA approval for algorithms, but they're locked at the moment of approval, they're frozen. So, a lot of deep learning could be far better if we could keep going and be autodidactic. So, if more data was exposed to more input, we just get even more accurate. But, the way we have at the FDA right now that doesn't occur, it's  being treated as if it was a device, or a drug, where, whatever the data is that comes in front of them to review has to stay like that, and any subsequent improvements would have to go through a whole process. So, that's one problem. Another problem is, that you're getting into, is when you have an algorithm that's out there, it's a very different scenario than a doctor with a patient that's a one to one, or, if, you know, an algorithm that's already getting legs and getting used by thousands of doctors that could actually, if it is malfunctioning and has a significant collection error or malware, from hacking, it could hurt a lot of people quickly. So, we don't have a surveillance system for algorithms. We're not very good at surveillance of drugs or devices, no less. We have nothing really for algorithms, so, we're going to ultimately see that happen where there's something, whether it's deliberate from hacking or whether it's just some kind of software glitch, that just wasn't anticipated occurring. We know that can occur, know with our computing systems of today. But, now, when it's a medical algorithm, it's a different story. So, you know, it's unfortunate, it's probably going to wind up being reactive rather than planning this. But, in my view, having AI that's treating patients, diagnosing, treating, involved with patient care. It's going to need continuous, ongoing surveillance. Because, if we don't keep it under wraps and in check, we are setting ourselves up for something to happen bad, quickly, potentially.  

Dr. Cybulsky [00:23:51] What are your thoughts on virtual coaching and behavioral change using these modern tools and voicefirst tools? 

Dr. Topol [00:23:58] The one question is for chronic conditions. Like, right now, we're seeing virtual coaching for diabetes. And the question is: is that going to be effective in having a far better management of the condition than where it is today which is quite sporadic? And, most people with type 2 diabetes, for example, they have high hemoglobin A1C which reflects that they're really not getting good control of their glucose regulation. So, you know. it's encouraging that we've seen people who are now, for a couple of weeks, taking a picture of everything they eat and drink, and they're getting their glucose fed back to them by a short term sensor. So, they're learning what are the things in their life, certain foods, certain activities ,sleep, stress, that are affecting their glucose regulation.  And the question is: will that have a positive behavioral impact? 

Dr. Topol [00:24:57] A nudge!  You know, one of the ways that we can use data is to help people who are willing to promote their own health. We've had a long history of not being to crack the case where people just want to live the way they live, and whether it's eating the wrong foods, or too much drinking, or whatever it is. It's very hard to get people to to move towards a healthier lifestyle. And, the question we don't really know yet. Short term maybe, but long term can we affect the data with virtual coaching? Can we get people to take much better care of themselves to prevent illnesses that otherwise might occur?  

Dr. Cybulsky [00:25:43] What would you say to practitioners and physicians who claim that their role in taking care of the patient will be diminished because of A.I. and voicefirst tools in medicine?  

Dr. Topol [00:25:53] In medicine It's different, because you're not going to have a life threatening condition diagnosed by a machine! You're being told you have cancer by a machine?! I don't think so. And, so what we're talking about, and I already mentioned there are certain specific, routine, non-serious conditions that could get doctorless. But, across the board, the best we can expect is this level 3, which is this balance between machine support, autonomy, and the normal practice of medicine. For anything serious, cancer, heart disease, autoimmune disease, neurodegenerative conditions, things like that, we just can't rely on no human backup system or even minimal backup. We need humans. We need doctors. We need care. We need that care. And so, that's level three, and we are not even close to that yet. But, that's eventually where I think we're headed.

Dr. Cybulsky [00:26:51] Wonderful. Thanks for sharing with us today.

Dr. Topol [00:26:55] Well, thanks for visiting. Good to have you here at Scripps and enjoyed meeting you.  

Dr. Cybulsky [00:26:59] That concludes the 21st episode of The Voice of Healthcare podcast on campus at Scripps Research Institute in La Jolla, California with Dr. Eric Topol. We thank our sponsor MEDarchon and their product QUARC.  Go to get GETQUARC.com to learn more.

Full episode here: http://bit.ly/VoiceHealthcare