July 9, 2021: It’s astonishing how many things have changed in healthcare over the last 18 months. John Halamka, President of the Mayo Clinic Platform shares their concept and vision for the future. How is this new rolled back regulatory environment affecting investment and consumer behaviors? How do we take data and use it in novel ways? How do we incorporate algorithms and workflow? Episodic medical care often falls short. How is digital going to help? He also touches on remote patient monitoring, the Vaccine Credential Initiative and his brand new book The Digital Reconstruction of Healthcare: Transitioning from Brick and Mortar to Virtual Care.
OPEN NOW! HRSA Announces New Loan Repayment Program for Behavioral Health Providers – Apply to the Substance Use Disorder Treatment and Recovery Loan Repayment Program
Every day you’re using skills to help end substance use disorders (SUD) within your community. The Health Resources and Services Administration is here to help you with the new STAR LRP (Substance Use Disorder Treatment and Recovery Loan Repayment Program).
Pay off your school loans with up to $250,000 from the STAR LRP in exchange for six years of full-time service at an approved facility. Behavioral health clinicians, paraprofessionals, clinical support staff and many others trained in substance use disorder treatment are encouraged to apply. Applications are open until Thursday, July 22, 2021 at 7:30 p.m. ET.
How 2021 Became 2030: AI and Innovation Post-Covid with Mayo Clinic’s John Halamka
Episode 422: Transcript – July 9, 2021
This transcription is provided by artificial intelligence. We believe in technology but understand that even the smartest robots can sometimes get speech recognition wrong.
[00:00:00] Bill Russell: [00:00:00] Thanks for joining us on This Week in Health IT influence. My name is Bill Russell, former healthcare CIO for 16 hospital system and creator of This Week in Health IT, a channel dedicated to keeping health IT staff current and engaged. Today’s guest is Dr. John Halamka President of the Mayo clinic platform. Special thanks to our influence show sponsors Sirius Healthcare and Health Lyrics for choosing to invest in our mission to develop the next generation of health IT leaders. [00:00:30] If you want to be a part of our mission, you can become a show sponsor as well. The first step is to send an email to [email protected]
[00:00:36]I want to take a quick minute to remind everyone of our social media presence. We have a lot of stuff going on. You can follow me personally Bill J Russell on LinkedIn. I engage almost every day in a conversation with the community around some health IT topic. You can also follow the show at This Week in Health IT on LinkedIn. You can follow us on Twitter Bill Russell HIT. You can follow the [00:01:00] show This Week in HIT on Twitter as well.
[00:01:03] Each one of those channels has different content that’s coming out through it. We don’t do the same thing across all of our channels. We don’t blanket posts. We’re actually pretty active and trying to really take a conversation in a direction that’s appropriate for those specific channels. We really want to engage with you guys through this. We are trying to build a more broad community. So invite your friends to follow us as well. We want to make this a dynamic conversation between us so that we [00:01:30] can move and advance healthcare forward.
[00:01:32] Today’s guest is a former guest of the show, someone who I follow pretty closely quote often and respect greatly. That is Dr. John Halamka, President of the Mayo clinic platform. Good afternoon, John. Welcome back.
[00:01:44] John Halamka: [00:01:44] Well, great to see you.
[00:01:46] Bill Russell: [00:01:46] Yeah, I just for the record, I pronounce your name Halamka and someone corrected me once. Am I saying that right?
[00:01:52] John Halamka: [00:01:52] The answer of course is depends on the country you’re in. And so How Lamb Kah is how a US person would say, it’s all [00:02:00] fine. They might say something slightly different. But if you’re in Hawaii, I’m sure it would be Halamakmakah.
[00:02:09] Bill Russell: [00:02:09] I’ve heard a couple pronunciations. I just want to make sure that that I’m not saying saying wrong.
[00:02:15] John Halamka: [00:02:15] Just as it’s spelled. It should be Smith. It’s a bad last name to have.
[00:02:19]Bill Russell: [00:02:19] Everybody in healthcare is pretty busy. So I don’t want to minimize that, but you’re getting hard to keep up with. You have the medically home co-investment with Kaiser, vaccine credential initia two new data [00:02:30] companies launched, COVID-19 healthcare coalition and oh, by the way, I did just get your latest book, The Digital reconstruction of healthcare, transitioning from brick and mortar to virtual care that just arrived yesterday from Amazon. So you’re staying pretty busy.
[00:02:46] John Halamka: [00:02:46] Well, so look at this way, I’ve written a couple of articles over the last decade on what I call the perfect storm for innovation. And that is when government academia and industry align. And there’s a sense of urgency to change. I could argue [00:03:00] that this era, which is maybe we’ll call it a COVID new normal, or at least approaching the COVID new normal has regulatory rollbacks, and waivers has a culture change that people are expecting more care at a distance, the rise of digital health and investment like we’ve never seen before in the life sciences. This isn’t just a perfect storm for innovation. This is a once in two lifetimes opportunity to change.
[00:03:28] Bill Russell: [00:03:28] Yeah. It’s interesting. I was [00:03:30] doing a little research for this in Becker’s. You talked about that a little bit. I guess they’re quoting the Mayo report and you were talking about healthcare 2030, and you said but what we’re seeing now is that 2030 is going to arrive in 2021 because of those things that you just said. We’ve had a lot going on. Give us some examples of how far we’ve been able to come over the last 18 months because of this rolled back regulatory environment and investment and changing consumer behavior. It’s amazing how many things have changed in [00:04:00] 18 months.
[00:04:01] John Halamka: [00:04:01] And of course, I can give you plenty of details, which I will in a second, but let me make it very personal. My mother is just about to turn 80. Do you know that my mother was not a user of digital health, virtual health? In fact really felt that this iPad or laptop was really getting between the doctrine. Well now of course she had the experience of the last 14 months. You couldn’t get care unless you accepted [00:04:30] virtual health. I mean, for the usual kinds of stuff, obviously, but emergent surgery or something. Sure.
[00:04:35] She says, wow I’ve been able to get the care I need with the same quality, safety and outcomes, lower costs. Don’t have to drive. Don’t have to wait. This is great. So how many times in our history have you seen an 80 year old person who’s typical technology growing up was a oh, I don’t know, Smith Corona, typewriter [00:05:00] say this virtual care thing, i’s actually the future. Right? So the cultural change is profound. At Mayo, just to give you a sense of 2020, we had 3% of our visits virtual in January. 95% virtual by April. And now about 25% is the new normal. So sure we’re not still at 95. But we’ve gone from three to four to 25 in one year again. How many times in our history have you seen [00:05:30] adoption of a technology that goes up by 5, 6, 7 times in a year?
[00:05:36] And part of that, as we found with medically home, as you say, and the advanced care at home is the regulatory rollbacks enable us to have care to distance, enable us to have different licensure and different levels of practice, but reimbursed. That’s near parody, right? So the problem with telehealth before COVID is you could never cover your costs. So you couldn’t scale. [00:06:00] And I can tell you from every conversation I have with HHS, CMS, ONC, there’s no desire to roll back these changes. Cultural imperative, absolute reimbursement and policies that suggest we can do this forever.
[00:06:17]Bill Russell: [00:06:17] We’re going to go into all those things. We’re going to go into medically home in depth, but I want, wanna, I want to lay the foundation for this. So the Mayo clinic platform, the concept of a platform is probably why we’re [00:06:30] talking about in addition to all these things but a platform enables us to move quickly much more quickly than individual applications, individual data stores, and those kinds of things. So talk about the concept and the vision for the Mayo clinic.
[00:06:45]John Halamka: [00:06:45] So Bill, this is probably one of the hardest things for people to understand. They say, oh, well, I see you put a whole bunch of businesses together under one roof and it’s a platform. Well, no, that’s a collection of pipeline businesses. So let me give you an [00:07:00] example, which you’re going to find very odd but I am at unity farm sanctuary surrounded by 300 animals.
[00:07:07] So. This is a John Deere 2032 tractor happens to be the tractor. We use John Deere with last ethic pipeline company. Which is it bought a bunch of parts from all over the world, assembled this thing, sold it to a customer. Done. That is a very linear, if you think about it, supply chain flow [00:07:30] from the tire manufacturer to the tractor builder, to the customer and there’s no additional revenue several years ago. And a Harvard business school case has been written about this. John Deere said, let’s become a platform company. Say, well, wait a minute. How can attract your builder be a platform company? So here’s what they did. They instrumented every one of their larger tractors with various sensors that could measure GPS.
[00:07:55] Where are you on your field? How fast are you going? How’s the engine running. [00:08:00] Oh, and then they got water telemetry. So they understood. Let’s see how much rain you had and then how much work you did. Oh, man, I started getting commodity prices of wheat and soybeans and corn. Today they have the largest database for precision agriculture that exists in the industry.
[00:08:20] And you say, wait a minute. Yeah. Precision agriculture. Yes. In fact, we can tell you what the commodity price of corn two months from now will be. Because we actually know who’s [00:08:30] harvesting, how much rain they had, what the RPM of their tractor is. So it isn’t just John Deere selling data to its customers or building better tractors.
[00:08:40] Now there are third parties coming to that platform and building all kinds of new apps, new analytics, even stock market forecasting. John Deere is now a platform data company. And for every new tractor, they had more [00:09:00] telemetry comes to the platform, making it even more valuable. So that’s a way to think about a platform.
[00:09:05]Bill Russell: [00:09:05] We’ll get back to our show in just one moment. Every day you’re using your skills to help end substance use disorders within your community. The Health Resources and Services Administration is here to help you with the new STAR LRP program, which is substance use disorder treatment and recovery loan repayment program.
[00:09:26] Pay off your school loan with up to $250,000 [00:09:30] from the STAR LRP in exchange for six years of full-time service at an approved facility. Behavioral health clinicians, paraprofessionals, clinical support staff, and many others trained in substance use disorder treatment are encouraged to apply.
[00:09:44] Applications are open until Thursday, July 22nd, 2021 at 7:30 PM Eastern time, which is right around the corner. To learn more and apply to join the STAR LRP. You can use the link in our show notes or visit [00:10:00] bhw.hrsa.gov to learn more. That’s bhw as in behavioral health workforce dot HRSA.gov. Now back to our show.
[00:10:12]So talk about the elements that go into that for healthcare. I think this is This Week in Health IT. I think people understand the elements in that it is data, it’s advanced analytics, advanced capabilities, artificial intelligence and machine learning. I’m trying to envision what [00:10:30] other aspects it’s probably also processes I would imagine as well. I mean, what, what are the elements that come together around this?
[00:10:36] John Halamka: [00:10:36] So let me give you four broad categories and these are just categories, right? These aren’t specific products, but they’re categories. Would you agree that we have a huge amount of new data sources because of the devices we wear the devices, we carry the devices in our beds, the devices in our homes, right?
[00:10:55] All this data, lots of it. Continuous high velocity data. [00:11:00] How are you going to gather it? How are you going to curate it? How are you going to normalize it? Put it in standards. So the first set of platform components, I called gather. Right. It’s gathering all these new data sources from wherever they may be.
[00:11:14] And unfortunately, lots of them aren’t in standard form. So there’s just a lot of work to do to clean them up and to put them into a universal schema so they can be used for the next step, which is discover. Imagine you [00:11:30] have 60 petabytes of data. That’s a lot of data. How are you going to find the signal on that noise?
[00:11:36] How are you going to answer a question a clinician has today, how are you going to do precision medicine? Well, you need large de-identified datasets again in a normalized schema and the tooling that is necessary so that you can create an AI factory. If you woke up this morning and said, oh I have this theory. I bet at the people who have a positive COVID test probably had certain [00:12:00] characteristics oh and long COVID. Maybe that has certain characteristics. Well, you need the tooling to be able to explore that on very large scale de-identified datasets. And then you develop an algorithm. How do you know it’s fit for purpose?
[00:12:15] And Bill the joke I tell which it’s just, of course not true, but it illustrates the point. We are going to use the data of Minnesotans. So it’s a million Scandinavian Lutherans and we’re going to create an algorithm to look at some [00:12:30] disease and then we’re going to apply it in Georgia. is it going to work? I don’t know.
[00:12:36] Right? So you need a set of componentry that looks at bias, efficacy, the area under the curve. How well does this stuff actually meet the need of the patient frankly and finally, we’re gonna, once we have the algorithms and the analytics and the visualizations, how do we deliver them in a workflow?
[00:12:56] So that a nurse, a doctor, some kind of [00:13:00] extender can get the benefit of whatever decision support you’re offering. So when you look at Mayo Clinic Platform, it broadly falls into those four categories. When you look at our co-investments and code developments, the things that we’ve launched as joint ventures.
[00:13:14] Bill Russell: [00:13:14] Yeah. Side I’d like to go into the Kaiser co-investment with medically home and I mean, there’s so many things. This is really exciting. Why don’t you first talk about what medically homes doing and how you got to this. How you [00:13:30] got to this point, because you started working with them before the significant investment with Kaiser. Talk about how this came to fruition.
[00:13:38] John Halamka: [00:13:38] So ask yourself the following question. Suppose you want to put a patient who’s serious and complex and acutely ill into a care process in their home. What are some of the problems you’re going to need to address? Well, first is that even a good idea? And so medically home has worked for eight years and actually [00:14:00] answering the question of who is going to do well in a non-traditional setting.
[00:14:04] Now I’ll give you an example, congestive heart failure, chronic obstructive pulmonary disease, pneumonia, all serious disease, but you know, you’re not going to go bad in two-minutes. Right. You’re going to have signals that say, oh, shortness of breath, getting worse over 24 hours or fever, spiking or something.
[00:14:23] How about somebody with Vtech? Right? You get a heart arrhythmia that one minute you’re fine. The next minute, your heart’s not beating. [00:14:30] Not a good candidate for hospital at home. So the answer is 30% of patients in hospitals today are good candidates for care and a non-traditional set up.
[00:14:44] Bill Russell: [00:14:44] So that just sets them up. That’s just saying that that population is candidates but it still doesn’t establish that we’re going to care for them in the home yet.
[00:14:55] John Halamka: [00:14:55] Absolutely. Right. 30% are candidates. And then you have to do things like social determinants of [00:15:00] health or what is the nature of the home? Is the home safe? Does the home have reasonable power? Does the home have access, right? If it’s a fifth floor, walk-up that’s not so great. Or do you have any kind of internet capability, whether that’s cellular broadband, satellite, whatever. Right. So what we have to do is say disease, state, comorbidities, demographics. Okay. You’re in the 30% now let’s go do a site assessment.
[00:15:28] And that site assessment [00:15:30] is either yes or no. And then of course, think of reading generally do you have a skilled staff you need, and Mayo clinic has established a training and certification program. So we’re building a whole new workforce of say, EMT is they get upskilled so they can do community paramedic, work in the home, deliver, bring supply chain, setting up electronics, putting the bed that you need in place, all that stuff.
[00:15:55] And so what you end up with is [00:16:00] as long as you hit these criteria, you can deliver very effective care at a distance as we have for over a thousand patients in Florida and Wisconsin, and we’re expanding next month into Arizona.
[00:16:15] Bill Russell: [00:16:15] Wow. So the eight years of history and we have we have other systems that have done similar work to this, obviously up there in Massachusetts, as well as others. What’s the data to support [00:16:30] that it is better or it is comparable and less costly to do it this way.
[00:16:35] John Halamka: [00:16:35] So as you can guess, we’ve done substantial analytics on every patient we’ve seen. And the mantra at Mayo is start small, think big. Right. So we did one patient at a time and then we did a deep dive. Oh, what was the outcome? What was the quality? Oh, do we have a supplychain issue? And then we [00:17:00] started doing measurement, like not only the usual quality indicators and such, but patients and families going through a full net promoter score evaluation would you recommend this to your friends or your family members?
[00:17:15] And so mathematics. After the first several patients and series of process improvements and then ramp up and all the rest, what we end up with is to your point, the analytics that say, oh, every quality measure, same. Outcome same. Safety same. [00:17:30] Cost. We, Cost computations and healthcare are a science, but you don’t have the bricks and mortar sunk costs.
[00:17:42] When you’re talking about a home, you don’t have heat, power light, you don’t have all the janitorial staff, all that other stuff. So you do a cost computation and it looks like a about half. Net promoter scores, you measure them 97% of patients and families who went through this [00:18:00] experience would recommend it to family and friends. There is no way any hospital in this country is going to get a net promoter score of 97%.
[00:18:09] Bill Russell: [00:18:09] No, that’s, that’s amazing. But now I’m thinking, okay, let’s scale this up and would I bring this to my health system and what, what would I have to overcome? I think the biggest thing I would have to overcome is in the administration, I’d have to go into the administration and say, look, we’re going to be giving up a 20,000, $15,000 visit, whatever it ends up being. [00:18:30]
[00:18:30] And we’re going to get a $7,000 visit. Isn’t that one of the challenges?
[00:18:34] John Halamka: [00:18:34] Well, not necessarily. And let’s talk through this in a couple of different ways. So first, if you had a risk of rain, Any kind of Medicare advantage, right? If you use say, I can now deliver the same cost, same carer with lower costs in any kind of risk rate, but they say, oh, that’s great.
[00:18:50] We need more of those. That’s wonderful. But here’s what we found, especially in a time of COVID you found that hospitals were saturated, right? There [00:19:00] was no capacity. And so if you say, oh, I actually can take my non COVID patients, put them in non-traditional settings, deliver care, fill the hospital with just COVID patients or high acuity patients, whatever people say, oh, wait a minute. What that means is I actually am able to use my bricks and mortar facility to its greatest extent. And that actually from a business sense makes it easy to convince people. But there’s one element of this. And that other element is, [00:19:30] imagine that you have an immunocompromised patient 87 years old. This is actually a true story.
[00:19:35] And you’d say, oh, no problem. Ma’am, I’m going to bring you to a place with lots of people coughing on you. Right. It turned out to be medically, even better to care for a whole lot of these patients in a non bricks and mortar facility. So it made sense to a hospital administration.
[00:19:54]Bill Russell: [00:19:54] We’re looking at hospitals are still putting up towers. And part of that is we just came through a pandemic [00:20:00] where we were worried about not having enough beds but in your book, you talk about, hey, you know what you’re to want to take this new account because again, 30% are eligible. So let’s just say 20% of those bed visits might be done out of the home.
[00:20:15] You might increase the number of overall beds that are Mayo clinic or whatever beds in that community without adding a single room in the hospital itself. How do you sort of balance those two things? In the pandemic we were worried about not [00:20:30] enough rooms or the right type of rooms, I guess. And the nature of healthcare is changing. We can do higher acuity out of the home with a certain level of effectiveness. And it’s going to look different moving forward. How do you balance those two things?
[00:20:45] John Halamka: [00:20:45] Well, so to your point in several of the sites, we’ve rolled this out, they say, oh, well, we have a bricks and mortar facility. We can’t add beds. It’s physically constrained, right? Don’t have enough land. Or the capital that is [00:21:00] required to add beds is so high. We just can’t do it. You say, how about this? What if you could see 20% more patients. Without investing capital or worrying about your physical footprint and use your existed staff is, oh my God, that’s amazing.
[00:21:16] It’s like one of the only ways for a healthcare system to expand its economic impact for a. And so we certainly also found that that was a very valid use case for hospitals that [00:21:30] really just didn’t have a lot of capital as most community hospitals don’t at
[00:21:33] Bill Russell: [00:21:33] At a 97% net promoter score as well. Which is kind of amazing. So talk about the logistics of it. You say using your existing staff and earlier you talked about using EMT or training EMTs. Are we literally seeing doctors and nurses go on rounds to people’s homes? Are we utilizing a different model in this case?
[00:21:56] John Halamka: [00:21:56] So typically the model is this. What you have is they EMT [00:22:00] doing the physicality of the home configuration, a visiting nurse who is doing medication administration, bandage changes, and those sorts of things. But the clinicians, the specialists and the hospitalists, the intensive, they all are working remotely. And so they work out of a command center or run with this cloud hosted software from wherever they might be. The dashboard is available in a [00:22:30] cloud hosted, accessible fashion. And so what that also means is that you can take an individual clinician, you could scale them larger. And so I would argue that as we train our next generation of docs, we’re going to see a specialty called a virtualist right. Like a hospitalist, but you’re a virtualist. You got all the telemetry, it’s like an F15 you’re flying and you can be anywhere and you can deliver the same care and you get up people who are in the home, who can be [00:23:00] your hands and feet if need be.
[00:23:03] Bill Russell: [00:23:03] Yeah, I do want to talk about the book a little bit. When did the book come out?
[00:23:07]John Halamka: [00:23:07] Came out on June 8th. You’re very timely.
[00:23:11] Bill Russell: [00:23:11] Wow. Okay. So talk about the title a little bit digital reconstruction of healthcare, transitioning from brick and mortar to virtual care. That’s what we’re talking about right here.
[00:23:19]John Halamka: [00:23:19] Yeah. Cause what we’ve seen is that just as John Deere became a data card, I am going to argue that in the next decade. And it may be even faster than that, that [00:23:30] hospitals in general, healthcare, as an industry has to recognize it’s a data business, right? It’s figuring out using algorithms who needs, what care in what location at what level of intensity and cost at what time.
[00:23:47] And so you’ll see an explosion of algorithms, of data, aggregation of new apps and workflows that give the patient and their family access to what they need when they need it at the [00:24:00] cost that is understood and predictable. And it’ll be based on evidence. I, and this interesting thing as a doc there, I’ll tell you, right?
[00:24:09] So I have been a doctor for 35 years. We were all trained as apprentice. We weren’t necessarily trained based on data. So when we make decisions, we make decisions based on the cases we’ve seen or what our chief resident taught us 35 years ago. And so what that means is that all [00:24:30] the doctors are amazing and of course are trained in physical diagnosis.
[00:24:35] They don’t necessarily always get that right care plan based on purely the experience of being a doctor, you need more, you need data and that’s where hospitals are transforming. And that’s the title of the book too. Will. How do we take our data and use it in novel ways? How do we incorporate algorithms and workflow? How do we think of new care models? Because that’s really what healthcare has to be.
[00:24:59]Bill Russell: [00:24:59] It’s [00:25:00] interesting. Back in 2011, when I came into healthcare, we talked about increasing the number of touch points. And then I learned this term longitudinal patient record. Once you get into healthcare, they load you up with all the terminology and the jargon that you need.
[00:25:15] But it’s an important concept of that data. And just the first chapter is central reconstruction necessary. You answered some of the questions. But you have a section where you talk about episodic medical care often falls short. And I want you to talk about that a little bit. In [00:25:30] this context, I was, I listened to Jonathan Bush, talk about Firefly health, which you’re probably familiar with.
[00:25:38] And he was talking about it sort of in its pilot space. And they were saying on average their customer, talk to a clinician of some type 65 times. It’s just an interesting concept in that they made it so easy that people started to use it. They were doing things like if they were in the grocery store aisle, they would use that [00:26:00] service to say, Hey, I’m thinking of buying whatever is should I, is this good for me given my condition or whatever like that.
[00:26:08] So anyway 65 times they, so they’re getting a ton more information. And he now just launched a new company called Zeus health. It’s longitude, which is addressing that longitudinal patient record as well. You talked about data, really changing it and the challenges with episodic care let’s in that context of those two [00:26:30] things.
[00:26:30] And I’m not sure how to phrase this question real well, but I’m sure you have an answer anyway. Episodic medical care often falls short. How is digital going to help? How does it fall short? And how is digital really helping that?
[00:26:43] John Halamka: [00:26:43] Well, and to your point, the data we are increasingly putting into our AI models is multimodal, right? It isn’t just, you saw the doctor one time it’s oh, what of your lab has been over the last decade? What’s your weight been over the last decade? What are the comorbidities you had not [00:27:00] this visit, but three visits ago, right? In the model. For example, we have 60 algorithms that we’ve developed and Mayo clinic platform, but one is a breast cancer prediction.
[00:27:10] It has 84 inputs. Right? And so we going to have predict which women will develop breast cancer and offer them treatment today. So they never develop breast cancer, but it requires 84 input period. [00:27:30] And those are gathered over a life of experience. I mean, it’s right. It’s structured. It’s unstructured. It’s telemetry.
[00:27:36] I’m going to use a funny term, the exposed zone. Yeah. Right. What did you, where do you live? Oh, you’re live in a dusty place where you’re breathing grime. All of these things have to go into the algorithm. Let me give you a silly example. This is actually true example, but don’t worry. It’s me. So in January I developed a cough and a fever [00:28:00] in January of 2020.
[00:28:02] And do you know it was nothing, right. It just like a cold kind of thing. So who would worry about that? Oh, did I mention December of 2019 I was in Wu Han China. Really? I really was. Now would you say that if I went to the doctor and had an episodic visit in January and Hey, whatever, it’s the flow, it’s a cold, it’s not enough. Right. You need to know my travel history for the last 90 days. I didn’t have COVID by the way. I’m still COVID [00:28:30] negative, vaccinated. All good, but right. You could see how it’s your longitudinal experience? Your travel history, your, oh, what do you do for a living? Oh, you live on a farm. Oh, well, I didn’t think about this sheep disease that you could possibly be carrying.
[00:28:46] You have to integrate all that stuff to get the right diagnosis.
[00:28:50] Bill Russell: [00:28:50] Yeah. And do yuo know honestly, I want to get back to the farm because there’s this new thing that’s happening, where people are foraging for stuff. And I thought these people are going to [00:29:00] kill themselves if they don’t have Dr. Halamka’s knowledge there just to go out and like pick poisons and we’ll get back to that later.
[00:29:06] It’s just a side thing for my knowledge, I was reading this article, thinking, I think these people are, I understand what they’re trying to do, but I’m afraid they don’t have enough knowledge to go out and pick the right mushrooms. And they’re going to end up doing a televisit with you at some point in the future
[00:29:22] John Halamka: [00:29:22] I do 900 televisits a year. I’ve done five today. And so to your point, I mean, again, we [00:29:30] can talk about this later, but here, if I were to just take my email for the last half hour or so and I’m just, I’ll find you a good mushroom photo but just coming to like my three year old ate this, what am I going to do?
[00:29:47] Bill Russell: [00:29:47] Yeah. Let’s get back to an area where I’m more comfortable. We go into chapter two and you talk about the merits and limitations of telemedicine. And I want to skip over the the virtual [00:30:00] visits the urgent care visits and whatnot. That’s been talked about really ad nauseum. Not that it’s not important, it’s really important, but I want to talk about remote patient monitoring, because this is one area where I think you guys are at the forefront of this.
[00:30:16] And there’s a couple of things about this one is let’s first of all, talk about where has remote patient monitoring Bennett and effective, or where have you implemented it where it’s been really effective for [00:30:30] for the Mayo clinic or the Mayo clinic platform?
[00:30:33] John Halamka: [00:30:33] Yeah. So there’s two ways to answer that question, right? One way is to say, I’ve got a patient. And I’m going to actually look at longitudinal high velocity data while they’re in at home hospitalization, pulse, pulse-ox, blood pressure would cometary and that kind of thing. And you can use that to say, when is a patient starting to have an issue? Where do they need an intervention?
[00:30:56] Right. So that capacity we’ve done [00:31:00] effectively over this time of COVID. And we’ve been able to say, oh, well the patient needs their Lasix increased or their oxygen turned up or that kind of thing. And that works, but here’s, what’s really interesting. Two weeks ago, in nature medicine, we published a randomized controlled trial on the use of using telemetry, like your apple watch, telemetry to be able to predict failing heart pumps, ejection fractions, diminishing over time.
[00:31:28] And so what [00:31:30] can we do? Well again, for the, I know that not all of your listeners are statisticians. But with an AUC of 0.93, that’s really, really good. Right? Perfect is 1. We can actually predict your ejection fraction from your apple watch. Wow. And why is that interesting? It’s like, okay. You can take an asymptomatic person and look at body parameters to [00:32:00] predict future disease. And do that with very high accuracy. And that’s really interesting.
[00:32:07] Bill Russell: [00:32:07] Thank you for that example because that’s going to help me. That’s a great example. Point 93. I mean. That’s a phenomenal example but who’s looking at the data. Is it a machine that’s looking at the data? Is it a call? Is it a command center? Cause it can’t be a physician who has or even an office staff that has normal kind of [00:32:30] functions. How are we going to take that data and act on it?
[00:32:33] John Halamka: [00:32:33] Sure. And of course, this is a brilliant question. Bill I have no conflicts of interest and I am always exceedingly careful about disclosure Mayo clinic developed all these algorithms and spun out a company called located in Boston to provide the algorithms to the industry to do this in an automated way. And so, as you point out, it has to be a combination of automation, [00:33:00] looking at millions of incoming signals. And for most of them are just completely normal. And if there is something abberant, then you’d get a human review and Mayo does have a command center by the way, for the review of not only the remote patient monitoring activity, but also for incoming talent.
[00:33:21] So, so in fact, to your point about what hospitals is the future after B, you’re going to be these massive diagnostic centers using [00:33:30] humans plus AI, to look at all these signals and be able to identify disease more rapidly. You’ll find by the way, in the Eagle trial that we published in nature medicine, the algorithm plus human was 30% better than human alone.
[00:33:46]Bill Russell: [00:33:46] That doesn’t surprise me. You spent a lot of time in this book giving data studies feedback and you’re fair on both sides. You will point out studies that show one direction and studies that [00:34:00] show another, and you talked about the false of the different studies, but I would imagine that’s just the nature of the world that we live in. We can take AI and throw it at whatever. We can throw it. Information security and we could throw it at AR and AP and those kinds of things and not worry too much. But when we bring it into the clinical setting, we’re going to have to have these conversations of, of what’s actually what’s actually behind the curtain.
[00:34:26]What is the algorithm? How did you come up with the algorithm? How’s it making the [00:34:30] decisions? Because now they’re a part of the clinical decision-making process. And in order to get trust, we need that transparency.
[00:34:38] John Halamka: [00:34:38] You are so right. So I could to say this, that as we use more and more AI in medicine, that we are going to have a credibility. And that’s because well, at the moment, the FDA through software as a medical device will say that an algorithm is safe, safe. Well, that’s good, safe, [00:35:00] but does it do anything, does it actually work for you? Right. The FDA isn’t actually measuring. So what I have proposed, and I think you’ll probably see in the upcoming weeks potential announcements about this, you’ve talked about all these various coalitions collaborations, joint ventures, and consortia that I have co-led. We’re putting a national coalition together for the standards around evaluating AI efficacy. [00:35:30] AI appropriateness AI ethics and AI bias. And what is going to come out of this is a nutrition label. And you say, why, what do you mean a nutritional aid? Well, you buy a soup can, and I’m going to just guess you can decide if you picked up a can of soup and it’s a thousand milligrams of sodium. 59 grams of fat. 500 calories, a serving. You might say mmm, nah, but what [00:36:00] if no soup can in the aisle had any nutrition labels and you would just pick a can cause it looked good at a picture. But that’d be awful. Well, welcome to the world of AI algorithm. No one tells you today, what his ingredients were, tell me how effective it was.
[00:36:16] And so this nutrition label will include race, ethnicity. We’ll include age and gender we’ll show the algorithm, that approach that was used, convolutional, neural network, whatever AUC stratified by the various [00:36:30] populations. You can say. Well, Hey, this one is 0.9, three, and oh, it’s like 60% male, 40% female. It’s got mixtures of white, black, Hispanic. Yeah. I think I’ll try this one.
[00:36:43]Bill Russell: [00:36:43] Yeah. And it almost has to be done. It has to be done at that level. I wouldn’t when I talked to some of the smaller health systems and they’re like, yeah, we just bought some AI tool. I’m like, all right. What does it do? How does it fit? How does it flow and all these things. And [00:37:00] to be honest with you, the smaller organizations, even if they have a governance group, that’s looking at it, I’m not even sure they know what questions to ask. It’s not like they’re an academic medical center with resources and, and or Stanford, where they can tap into the Hey, come on over and tell us what you know, and, and help us to understand this.
[00:37:17] I mean, a lot of these smaller organizations rely on consortium like that. To provide some guidance. Otherwise it’s, it’s really shooting that shooting in the dark.
[00:37:28] John Halamka: [00:37:28] Exactly right. And so this [00:37:30] will be a nonprofit benefit to society had Gemini of non, just like all these other coalitions of academia and industry where we’ll of course bring in collaborators from government when the time is right.
[00:37:44] Bill Russell: [00:37:44] I want to give you kudos on chapter five. I don’t know if that was your chapter. I don’t know how you write your books together with. But I don’t know if that was your chapter or his chapter, but it explains the exploring the artificial intelligence machine learning [00:38:00] toolbox. It’s probably the most clearly understood layout of all the different tools in the toolbox I’ve I’ve read. So kudos. And if, if people want to pick up the book, it’s worth it just for that chapter to understand where we are going and what some of these tools are. And I’m not going to, I’m not going to go into some of these tools because to be honest with you, I was really fascinated by this, but I’m not sure our listeners would be assessed students with all the details of it. I wish.
[00:38:28] John Halamka: [00:38:28] But what you hope the listeners [00:38:30] come away with is AI isn’t magic. AI is not a computer that’s thinking, right? It’s not sentient. It’s probability. It’s statistics. And you have to be careful because it doesn’t imply causality. Right. And so here’s a quick example for you, my mom, when she was growing up.
[00:38:51] So did you know, I heard that ice cream causes polio. Well, what do you mean ice cream causes polio? Well all of [00:39:00] us would see our friends get polio during the same months of the year, they were eating right. Well, it turned out polio was transmitted by respiratory droplets and people went swimming together in the summer and there was water and breathing, et cetera.
[00:39:14] So if you ran an AI algorithm and you didn’t choose your variables well, and it was just like months of the year polio cases and ice cream sale. You would get a really good model. It’s meaningless. However,
[00:39:29] Bill Russell: [00:39:29] Yeah, you make that point. It’s [00:39:30] math, it’s data science. It’s all. It’s it’s it’s not magic. It’s we talk about the Oz behind the curtain, but at the end of the day, that’s how we think about it, but it’s not actually what’s happening. It’s just really smart mathematicians, statisticians and data scientists who are putting together algorithms based on the data they have the data that’s built.
[00:39:54] John Halamka: [00:39:54] Yeah. And so it’s going to be a wonderful future, but only if it’s regulated and analyzed and transparent, [00:40:00] and that’s the next adventure you’ll see in the next few months, we’ll move forward with getting that in place.
[00:40:08] Bill Russell: [00:40:08] All right. So I want to go back to a couple things. So you were a part of the COVID-19 healthcare coalition. You were part of the vaccine credential initiative and BCI initiative. As well talk, talk about some of that work during the during the pandemic. What are you, what are you most proud of that those groups were able to do?
[00:40:30] [00:40:30] John Halamka: [00:40:30] So what happened in 2020 that will live with me for the rest of my life was the camaraderie. Right? So what you found was it was March 13th, 2020. Look in my inbox. Oh, all of these companies, they want to do something. We had a few phone calls by March 20th, 2020, we had 1200 companies working together for free. We did things like, oh, [00:41:00] we do contact tracing and you had Google, apple, and Microsoft working side-by-side for free.
[00:41:08] Right. You just don’t see that in most exemplars in our history. And now as you’ll see in the next few weeks, the vaccine credential, which I’m going to call validated clinical information or verified clinical information, right. We don’t like to use the word passport, right? There’s a lot of controversy [00:41:30] about what it means to be able to display such a credential.
[00:41:33] We’ll just say this, you want to go to a sports game. You want to go to a country. And you choose totally up to you. Okay. Then to get into a certain section and the concert, you’re going to just show something about your immunity. It’s up to you, right? Do it, or don’t you’d want a standard to do it. As of today, 598 companies have come together and adopt that standard.
[00:41:58] You might have [00:42:00] seen that yesterday. Walmart and Sam’s club announced that every vaccine they get comes with a digital credential that follows that smart health card standard that vci.org has promulgated.
[00:42:16] Bill Russell: [00:42:16] John. I have, I have the really cool looking card. I don’t have it on my desk. I think it’s, I think it’s in my wallet, but I have that card because I went to the department of health. I didn’t even go to a health system. It was one of those drive-thru events, department of health. [00:42:30] And I, when I was getting it, I said, is this going to go into my medical record in some way? And the guy just looked at me and sort of smiled and he’s just like, no, probably not. But he goes, here’s your card? How does that turn into something that’s digital?
[00:42:43] John Halamka: [00:42:43] Let me tell you the two realities of that cause it’s all good news. So do you know that every state has an immunization information system and whether you got your vaccine at a Walmart, a CVS or Walgreens, a mass vaccination site or a doctor’s [00:43:00] office.
[00:43:00] They all end up in your states iIS. The immunization information registry. So where you’re going to see in the next few days, literally a few days states will begin to announce they are making available to every citizen in their state, a vaccine credential that draws on the state’s IIS. And so, yes, it exists in digital form. Now it just give you a sort of quick view of Mayo. Mayo is in five states. Do [00:43:30] you know that we have 98% of the vaccination records of our patients purely because either we’re getting it from a doctor’s office or another hospital side or something, or a state IIS. So actually behind the scenes, the interoperability is better than you think it is.
[00:43:49] Bill Russell: [00:43:49] Yeah, no, that’s fantastic. My follow on question to that is. We have actually planned a family vacation next year overseas. So am I going to be [00:44:00] able to flash that at the airport and then slash that in where whatever country I’m going to?
[00:44:06] John Halamka: [00:44:06] Complicated answer. So when you actually look, you’ll see that clear, for example, as one of the providers has chosen to use the vaccine credential initiative, smart health card standard, many of the airlines have. But in Europe, for example, the world health organization has proposed a different kind of credential. [00:44:30] So some of the work to be done still working on this is to bring the world health organization and everything that’s been done by VCI together, because we don’t want two standards. We want interoperability and crosswalking so that you can do exactly what you’ve said.
[00:44:48] Bill Russell: [00:44:48] So John, before we get to, and I want to talk about the farm a little bit and that kind of stuff, but before we get there, is there anything else that I I’ve started to close my interviews with this question? Which is, is there [00:45:00] anything I didn’t ask you that I should have, or that you think that the community would benefit from a conversation around?
[00:45:06] John Halamka: [00:45:06] I know that every time I chat with you, I always end on a privacy note, which is that if we’re dealing with more data for more devices used for more purposes, We have to get consent privacy and security. Right? So a couple of the things that we’ve done. So Mayo hosted a conference on April 22nd, brought together 80 experts from around the world to [00:45:30] look at the next generation of consent.
[00:45:32] And so what have I said, Hey, bill, I’d like to use your medications, your problems in a de-identified aggregated way for an AI model. That’s going to help people get curious, you’d say, oh, well, that’s, that’s that sounds. Well, I’d also like your genome. Oh God, no, I I’d rather not have my genome part of that. It’s fine. Right? You should be able to decide how your data is used for what purpose with some level of granularity, not insane [00:46:00] granularity. I mean, it has to be finite. And so coming out of that conference, we have a consent we’ve proposed five levels of granularity. So you can see Mayo start to pilot, a new kind of consent about data use that gives patients choice.
[00:46:13]Bill Russell: [00:46:13] That’s really fascinating. I mean, cause you didn’t hear my rant, but I had a rant on one of the groups that has come together and huge anonymized de-identified data source and there they’re gonna use it for the good of me in time. Well My data’s in there and my rant was, [00:46:30] is anyone going to ask me, can I, you know what? I will probably give it. I will probably allow it, but I just want to be asked.
[00:46:36] John Halamka: [00:46:36] So our view is you need consent. You need DID algorithms that are really good at stripping out the things like rare diseases, things that would be easy to re identify, and then you still need control. Right. So I’m asked every day. Oh can you ship the de-identified data outside Mayo? Yeah, there is no I’ll invite you to bring your [00:47:00] algorithm into a secure computing enclave where you can run the algorithm against the did data and get a result. But no, I’m not going to give the data to some third party. Right? So there’s all these privacy and security things you’ll see in the next several years.
[00:47:17] Huge numbers of new hardware and software approaches to keep data safe and to bring algorithms and data together in a way that’s privacy, preserving and ethical.
[00:47:28] Bill Russell: [00:47:28] Yeah, I guess I’m going past my last [00:47:30] question. Are you concerned with the amount of ransomware events that are going on right now?
[00:47:36] John Halamka: [00:47:36] Oh, of course. Right. And so whenever I lecture about security, I describe it as a code. Right. Everyone innovates to say, oh, look, we’ve put up the world’s best encryption. We got a moat, we got hot oil. We got people shooting arrows. And then what do you know, the hackers invent the missile. Right? And so if [00:48:00] this cold war constant escalation, and it’s not at all a project, it is an ongoing process. And we need to innovate every single day.
[00:48:13] Bill Russell: [00:48:13] Now they’re not even inventing missiles anymore. They just, they keep using the here’s an email click on this. And that seems to keep working. I’d love to see us get really good at defense around fishing and those kinds of things. Cause it seems like they’re all still starting at [00:48:30] that very easy level. Not that the other attacks don’t exist on medical devices and other things. We need to be concerned about, but it just seems that that’s the entry point at this point. You got it. All right. So tell us about the farm what’s new on the farm.
[00:48:47] John Halamka: [00:48:47] Well, of course, it’s now that lovely time of year 70 degrees in Boston. We’ve had surprisingly a good amount of rain. So as the west coast just parched the east [00:49:00] coast at the moment is not as good. And so animals are really quite happy when it’s 70 and moist. And so the usual part of running a farm is that I’m constantly dealing with illness, disability, right?
[00:49:16] As we are getting animals that have been injured or abandoned or abused. So every day, either our new adventures in dealing with oh, this is an infected wound, or this is an orthopedic [00:49:30] injury you didn’t expect today. I have a duck that clearly has a broken neck, but yet no spinal cord damage. And so know, we take all of our emergency medicine skills and all of the community’s expertise, bring these animals back to help the best.
[00:49:48] Bill Russell: [00:49:48] So you have the largest animal sanctuary in New England. So I mean are you trained in veterinary medicine or do you rely on others from the outside as well?
[00:50:00] [00:49:59] John Halamka: [00:49:59] Well, and so it’s a triage, as you can imagine, mechanism, which is a, it turns out emergency medicine, trains you in a lot of things. Right.
[00:50:09] So the physicality of closing and bills, wound or Ripley the horses wound turns out to be fairly similar. Yeah. The medications you’d use for you getting conjunctivitis and a goose with conjunctivitis are the same. Right. So, so we can do that, but then there are [00:50:30] certainly specialty surgeries. I am not licensed nor qualified to do. And so we have Tufts veterinary school just 20 minutes away. And so we can rely on a series of visiting vets for each species that we care for.
[00:50:45] Bill Russell: [00:50:45] And so it’s a sanctuary, it’s not a place where people are coming and doing tours and that kind of stuff, right?
[00:50:50] John Halamka: [00:50:50] Yes. We actually have tours every day. Yeah, well, and we ran an event last Saturday with 400 people. And the way we did it is [00:51:00] we socially distance. We brought them in for two hours of intense touring and lectures and animal experiences, and then the next group, and then the next group and the next group. So on average, we’re running about three tours.
[00:51:12]Bill Russell: [00:51:12] All right. Are you going to do anything around the health conference when it’s up there in Boston?
[00:51:15] John Halamka: [00:51:15] Th that’s a thought, yeah, no one has asked yet, but because I have been a speaker at health for awhile they always do these characatures. So this is the John Halamka mask.
[00:51:35] [00:51:30] Bill Russell: [00:51:35] So do you think at HIMSS is it going to be masks and social distance?
[00:51:40] John Halamka: [00:51:40] Well, remember both are required, vaccination credentials,
[00:51:44] Bill Russell: [00:51:44] Right. Well, that’s what I’m wondering. So, I mean, I’ve been vaccinated and they’re going to require credentials to get in. So everybody there will be vaccinated in theory, you wouldn’t need masks and social distancing. Right?
[00:51:56] John Halamka: [00:51:56] I think the answer is they’re going to say indoors. Do what [00:52:00] you feel is proven. And so you’ll find some people mass, some people socially distanced, but everyone on prem will be vaccinated, including staff.
[00:52:10] Bill Russell: [00:52:10] Fantastic. Hey John, thank you very much for your contributions to the industry, as well as taking your time out for spending some time with us. We really appreciate it.
[00:52:20] John Halamka: [00:52:20] Well hey Bill, always happy to talk anytime and I’ll close with one sort of silly thing. So while we were talking, a three year old, just take that. What are you going to [00:52:30] do?
[00:52:32] Bill Russell: [00:52:32] Three mushrooms.
[00:52:34] John Halamka: [00:52:34] It’s morass misoriendees. I recommend garlic a little onions and maybe some olive oil.
[00:52:43] Bill Russell: [00:52:43] So you get pictures of mushrooms all day. That’s what people are sending. Wow.
[00:52:48] John Halamka: [00:52:48] Exactly. This 3 year old will be fine.
[00:52:55] Bill Russell: [00:52:55] I anyway, I’m an example of, I had a phone [00:53:00] call with an emergency medical doctor and it was in the middle of the night. And I was thinking of going to the ED. I called them up or we went back and forth and texts. We had a phone conversation and he said, you don’t need to go to the ed. And he was right. I didn’t need to go to the ED based on things. That’s that’s the value of, of telemedicine. It’s it’s really putting me at ease, putting my mind at ease. Because I don’t know at that point I just feel pain and he’s like, no, that, that will pass. You’ll be okay. And I’m like, [00:53:30] okay, that’s all I needed to hear.
[00:53:32] John Halamka: [00:53:32] This is the new normal.
[00:53:33] Bill Russell: [00:53:33] It is the new normal. John thanks for your time again.
[00:53:36]What a great discussion. If you know someone that might benefit from our channel, from these kinds of discussions, please forward them a note, perhaps your team, your staff. I know if I were a CIO today, I would have every one of my team members listening to this show. It’s conference level value every week. They can subscribe on our website thisweekhealth.com or they can go wherever you listen to podcasts, Apple, Google, [00:54:00] Overcast, which is what I use, Spotify, Stitcher. You name it. We’re out there. They can find us. Go ahead. Subscribe today. Send a note to someone and have them subscribe as well. We want to thank our channel sponsors who are investing in our mission to develop the next generation of health IT leaders. Those are VMware, Hill-Rom, StarBridge Advisers, Aruba and McAfee. Thanks for listening. That’s all for now.