June 20, 2023: In today's discussion, we are joined by a special guest, Dr. Anthony Chang, who is an esteemed expert in the field of AI in healthcare. Dr. Chang shares valuable insights into the integration of Artificial Intelligence (AI) in healthcare and provides his expertise on the immense potential of AI technologies to transform the medical field, focusing on key topics such as its role as a 'clinical GPS' and the importance of AI education in medical training. We also explore vendor selection for AI implementations, scalability considerations, and the practicality of AI solutions in real clinical settings. Join us as we delve into the fascinating world of AI in healthcare.
• How AI serves as a 'clinical GPS' in healthcare, assisting without disrupting clinical routines • The importance of including AI education in medical training, and how this dual education is contributing to the field • How healthcare organizations should approach vendor selection and collaboration for AI implementations • The possibility of AI scalability meeting the expansive needs of the healthcare industry considering current computational power and algorithm maturity • Why it's crucial for data scientists to be exposed to real clinical settings, and how this influences the design of AI solutions • The resources available for AI education in medicine, and how these can be accessed
• Understanding AI's role as a 'clinical GPS' in healthcare • Recognizing the importance of AI education in medical training • The approach to vendor selection and collaboration in healthcare • Addressing AI scalability in healthcare • The importance of exposing data scientists to clinical settings • Exploring resources for AI education in medicine
• "AI can act as a GPS in the clinical setting, offering directions without interrupting the journey." • "The blend of clinical medicine and data science education is a promising development for the healthcare sector." • "While choosing vendors for AI implementations, beware of those who might overpromise and underdeliver." • "The scalability of AI in healthcare is largely backed by current computational capabilities and the maturity of algorithms." • "Exposing data scientists to real clinical settings can significantly influence the practicality and impact of AI solutions." • "A variety of resources, including AI Med, offer a wealth of information for those interested in AI education in medicine."
Hey, podcast listener. Thanks for listening. If you're enjoying this week in Health it, we just wanna give you some information on how you can support these conversations to keep them going. This week In Health, it's goal is to keep you, your organization and your employees updated with the emerging thought and TR trends in the healthcare industry.
Through our conversations with healthcare and technology leaders. The best and easiest way you can show your support is to go over to this week in Health IT on iTunes and leave us a review. Also, you can subscribe on iTunes, Google Play, or Stitcher, or go over to our YouTube page and subscribe and hit the notification bell.
Again, we really appreciate you spending your valuable time listening to this podcast.
Welcome to this Weekend Health It where we discuss the news, information and emerging thought with leaders from across the healthcare industry. This is episode number 20. It's Friday, June 1st. Today we do a deep dive into the world of healthcare ai. This podcast is brought to you by health lyrics. Visit healths.com to schedule a free consult.
My name is Bill Russell, recovering healthcare cio, writer, and consultant with the previously mentioned health lyrics. Today our guest is Dr. Anthony Chang, one of the leaders in healthcare ai. Welcome to the show. Thank you, bill. Um, I'm going to give people a little bit of your background. Uh, we are, first of all, I've gotta give context to our background here.
So we are at the Innovation Institute in Newport Beach on the seventh floor overlooking the, uh, Pacific Ocean. And, uh, And today I think you could see Catalina. I see cata. It's kind of nice. And we have some guests in the audience. Our first time we've had guests in the audience. Yeah. No clapping, . Uh, so let me give people a little bit of your, uh, bio real quick here.
So, uh, Dr. Chang attended, uh, uh, John Johns Hopkins for his BA in molecular biology prior to entering Georgetown University School of Medicine, uh, for his md. He then completed a pediatric residency at Children's Hospital. Uh, national Medicine Center and his pediatric cardiology fellowship at Children's Hospital of Philadelphia.
He then accepted a position of attending cardiologists in the cardiovascular intensive care unit at Boston Children's Hospital and as assistant professor of at Harvard Medical School. Uh, he has been the medical director of several pediatric cardiac intensive care programs, including Children's Hospital of la, Miami Children's Hospital, Texas Children's Hospital as well.
He served as Medical Director of the Heart Institute at Children's Hospital served. Are you still serving as the, as the Director of the Heart Institute at Children's Hospital of Orange County? No, you're not. So what's your role at Orange County right now? Um, well, I'm the Chief Intelligence and Innovation officer.
And I'm the medical director of the Heart Failure Program here, and I'm also the founding director of the Sharon Disney Lund, um, supported Medical Intelligence and Innovation Institute, or MI three for short. So you have the three or four jobs going at once, right? Right. It has a lot of fun. And the reason we're here at the Innovation Institute is you had a team meeting today and you're, you're talking about the AI Med conference and, and this year you've been at, um, On three continents and you're looking to expand that, right?
Uh, even further. In fact, I'll, I'll give a, a shout out at the end of the, at the end of the episode for the website cuz I went to the website and you, at one of the few conferences, you put up every piece of material you can on that. And it's, it was great. As I was preparing for this, all the, uh, all the slide decks are there and all the background.
It's, it was really helpful. And the reason we do that as opposed to, I think most meetings that only. Put up a teaser trailer or a talk, whatever, cuz we really want this to be, uh, widely available around the world. Yeah, you're almost like open sourcing. AI knowledge. Correct. Which is, which is phenomenal.
Correct. Um, so you completed your, uh, you completed your MBA as well from Miami School of Business and graduated with the McCall Award for Academic Excellence. Uh, completed your Master's in Public Health in Healthcare Policy at the, uh, Jonathan Fielding School of Public Health at ucla. I graduated with the Dean's Award in Academic Excellence and you graduated with a Master's in Science in biomedical data science with a sub-specialization in artificial intelligence from Stanford School of Medicine.
And you're also a computer scientist in the residence and a member of the Dean's Scientific Council at Chapman University. Um, So do you have, uh, student loan debt then? Actually, I'm proud to say I don't have a penny of it. That's amazing. That's where the MBA came in handy. And, and your, uh, your, you went back to Stanford.
That was, there was a little bit of a time period between there. Did, did you like all of a sudden decide, Hey, AI is where it's at and I need to know more, and Stanford was the place to go, or, yeah, basically as a, um, senior pediatric cardiologist, I realized, From, uh, going to meetings and giving talks and writing books on the topic that a lot of decisions, a lot of the management issues are not always data, data science, um, driven.
And that, um, senior physicians tend to espouse their own principles without sometimes data science or artificial intelligence. And a lot of the controversies in my. 30, 40 years have not been, um, uh, resolved. So I wanted to resolve them and, and think about another dimension as Einstein says, you know, insanity is doing the same things over and over and expected different result.
And I feel like we have these, um, Debates at various meetings year in, year out, but no one ever resolved them. And I felt like, well, maybe I should think about this differently. And from a data scientist per perspective, it's, it's really interesting as we start to talk about artificial intelligence, when, um, when we really start to bend the paper for people, they, they almost get nervous, right?
Cause things are going to significantly change. And we'll talk about the different levels of artificial intelligence and where we're really at and right, the hype curve and those kind of things, but, You know, once it gets through the hype curve and the, and the trough and what we're looking at out there is, is pretty fantastic.
Yes. In terms of precision medicine, in terms of robotics, in terms of just, it's overwhelming in terms of the, what the dividends can be. And I think, um, I know you have a lot of CIOs in the audience and I think 10 years from now, We'll be talking about chief intelligence officers in different healthcare organizations and not only chief information officers or that job will be just transformed into someone that understands the intelligence as well as the data.
Yeah. And, um, as if you didn't have enough going on, you, you have a TEDx talk here, singularity University faculty. Yeah. So did you speak at Singularity University at their, uh, There're conference down in, I have, um, a couple of times there with, uh, good friends with Daniel Craft. Yeah. That's a phenomenal conference.
I, yes, it's, uh, in a beautiful venue. Yes. Where, where's, where's your, you have one in Laguna, right? Right. Our, um, us um, AI med meeting is usually in Southern California and last two years it's been at the, uh, Ritz Carlton Resort in Dana Point. And that's, Awesome. Yeah. For anybody, anybody who wants. Well, I think when we talk about something as intense, uh, and esoteric as you know, artificial intelligence, it helps to have that wide expansive ocean in front of you to decompress when your brain brain is hurting from all that it, it did, uh, important information from smart people.
And did I read this right? Did, did I see this right, that you actually did like a shark tank out on the beach? Uh, or no, or not quite? No. , we had a slide about that and people were vying to go on the beach, but we did have a Shark Tank event, um, where we had five startups in AI and medicine, healthcare space pitched their companies.
Yes, man, that's, that's, that's a lot of fun. And then you're also ceo, co-founder of Two Startups. Uh, do you wanna tell us a little bit about those two? Sure. The two startups, one is, um, cardio genomic intelligence, and it focuses on precision medicine for cardiovascular disease. Since I'm a cardiologist and my co-founder, Dr.
Spero Mus, is a genomics PhD. We thought that was a natural, um, coupling of our strengths. And then, um, medical intelligence is a, uh, resource company for artificial intelligence and healthcare organizations. So if I'm a healthcare leader and I don't know where to go with this, I give you a call. You'll come in and Sure.
And work with us. I mean, do you have a team there or is that primarily We have, we're putting together the team and as you can imagine, There are not a lot of people well versed in this space. So we're trying to put together thought leaders in AI as well as in it, and putting together this, uh, team of, I think, uh, incredible people.
Um, so great. So you offer a service for healthcare leaders, but you're gonna bring, you're gonna bring me and our listeners up to speed. So here we go. Um, the crash course in 20 minutes. Crash course. 20 minutes on ai. I'm gonna know everything I need to know. Um, Actually, I looked at the materials on your website.
There's no way to know everything there is to know. Yeah, I, I see that there's subspecialties and people are really focusing in on cognitive general intelligence. It just don't, it's no different than a doctor being a subspecialist, basically. So, yeah. So, we'll, we'll see what we can do. So here's what we're gonna do.
I, I'm gonna, I'm gonna play the healthcare executive. And we're going to, the framework's gonna be three things. One is primer on why ai, what it is, what its potential is. Uh, the second area we're gonna look at is, uh, where should I be thinking about it and where is it in use today? And then the third is, how do I get started and how do I, uh, how do I get started and how do I stay relevant in this space?
So let, let's start at the beginning. Um, what's the promise of ai? Why, why are you so excited about it? Why are you, uh, looking at it? Well, I think, um, after four years of education, I realize that it's kind of like wearing a different lens and looking at the world. Um, you see so many little places where data science or, uh, computer programming can really make impact.
And if you think about our world in general, so many facets of our life is are already being replaced by automation. Um, how we order books and how we get referred books is all algorithm driven. It's not perfect, but it's pretty good. Um, how we, uh, have now autonomous driving vehicles now, at least, uh, in the picture.
And I think 10 years ago, if you were to ask people, do you think there would be autonomous driving vehicles? And everyone would be thinking this is like 20, 30, 40 years down the road. Right. But it's not, uh, uncommon to see one in, in Silicon Valley now, and I think there's so many aspects. And then it was, um, February 14th, 2011.
Valentine's Day. So it's a special day for cardiologists that the human contestants were beaten soundly by the supercomputer Watson from ibm. And that was the night that I downloaded the application for the data science and AI program at Stanford. I realized that, um, this time AI is here and it's probably gonna be here for a long time.
Yeah. And those, uh, Those stories just continue. I mean, you know, one of the things you shared was, I think I have a picture of this. Yeah. So yeah, and I'll, I'll put this up on the screen. So, how a robot passed China's medical licensing exam. Yeah, I think uni knows about that. Um, so it's a, uh, obviously a robot with, um, what we call natural language processing, uh, capability and understanding.
So NLP that you hear about a lot is natural language processing or how a. Computer can understand the human spoken language or written language. And then N l U, which is uh, even more important is understanding. So obviously, just like Watson on the show Jeopardy, um, this Chinese robot is able to understand the context and the content of the questions in a medical exam and pass.
But still not ready for, we're not gonna put that robot into a clinic and have people go see it just yet. No, I think, again, I think one of the misconceptions is that, AI is all about the robot, and it's not helping that the public media is often having pictures of robots and then having ai, uh, headline.
Right. And what I say, and I, my intellectual partner, uh, spare ramus and I kind of made it up together, is that AI is about making the visible invisible. So by that I mean if you walk into a doctor's office and you sat down. Inevitably that physician's gonna be tapping away on a computer, right? And not paying attention to you a hundred percent of the time.
I'm lucky my doctor doesn't do that, but, but, um, in my exam rooms, um, computers are not allowed in the exam room because I think that's distracting. So if AI is really good at some point in the near future, then the entire conversation, the exam will all be extracted onto the medical record automatically with AI and N lp, and you wouldn't need.
The doctor tapping on the computer. So it's making those things go away. Yeah. You, you, in the last conversation we had, you liked it to, um, the, the, the doctor isn't, um, Sherlock Holmes, it's more like Watson, right? It's your assistant. Right. It's there to make the invisible visible. Right? Right. So it's, it's going through reams of data, uh, genomic data, whatever the data and it's coming back and things that you couldn't read if you took a week, it's saying, Right, just process this.
So that's making, that's the, um, co uh, sort of the corollary, which is making the invisible visible. So now, uh, medical information is doubling at incredible rate every three to four months or so. So there's no way a doctor can understand all of the medical literature. Pour through all of the patient's record, even if it's just the record in front of him or her.
Um, and then come up with the best decision. So we have to make the invisible visible by data mining and data science. That's amazing. So you give, um, again, this was a great slide for me. You gave us Yeah. You know, three, three different categories. Can you give, you know, just break it down for us Yeah. So that we can think about it, right?
Cause it's not just the robot and the autonomous cars. Right. Right. So basically artificial intelligence can be. Sort of subdivided in a number of ways. One way we can subdivide artificial intelligence, which is the generic general term, is think about the different types of artificial intelligence. So there is the assisted artificial intelligence, which means the machine's doing the work, it doesn't really interact with humans.
And it's something that's symbolized by the, um, the robotic vacuum cleaner that we have now. Oh, so like the room by. Bumping into each thing until it figures out the room. Right? It's on sort of automatic pilot. It's doing its own thing. The humans don't need to teach it. Um, the intermediate kind of artificial intelligence is something called augmented intelligence.
So that means there's an interaction interplay between the human and the machine. So the example, um, would be something like the, um, Um, book choices that Amazon may have for you. So that's in sort of an, uh, augmented intelligence. It machine learns what you are doing and it teaches itself to relate to you in that way.
And then you give a feedback. So it's a constant interaction. Where, where does like RPA fall in? Okay, so that's like, we're, we're. We're actually telling the machine, right? Here's the, here's the redundant repetitive task that we normally do, right? I think rpa, um, and it's a very, um, popular term right now, robotic process automation, rpa.
So for the clinicians out there, it's not right pulmonary artery anymore. So, um, RPA is, I think somewhere between assistant and augmented. And as you know, we have a, uh, computer science PhD here sitting behind me. So, um, I think he can, uh, give you his, uh, opinion too. But I think it's automated in sense that we're just telling the machine what to do, but there is a little bit of element of learning there.
So, Um, I think it's mostly automated, but perhaps an element of assisted. And then, um, the third category sort of is the autonomous intelligence category, and that's symbolized by the autonomous driving vehicle. We don't have anything like that in medicine yet, per se. So in medicine there's only assisted, uh, ai.
Um, and that's something like a robotic, um, Pharmacy service where the robot is done delivering medications. Right. Yeah, that makes sense. Um, augmented would be some of the decision support, um, medical image, computer vision type of project. So my, I think that by far the most exciting area is gonna be in the next decade or two, is gonna be the autonomous AI in medicine.
So, so you told two stories. One was that the robot passing the exam, the other you told was the, uh, machine actually beating the champion. And I assume that falls if Yeah. The game go so that, that falls into that category. Yeah. That would be more, um, autonomous because the computer is playing a game on its own and, and figuring things out.
Looking at, looking at, you know, years of playing and then saying, okay, this is the best way to do it. Right. That's what we can look forward to. Is that? Yeah. Um, well, except the, um, so that's a great story about the AlphaGo software program from, uh, Google being the human contestant. And go, uh, so handily actually.
Um, but, and everyone publicized the second game, 31st move because it seemed like the computer made a move that it had not learned from any human based on hundreds of thousands or millions of games before. It was like a move out of nowhere. And then, but yet that move was instrumental in winning the second game.
What's not publicized is in the fourth game. Uh, one of the moves the human champion made was sort of in the category of that really creative move. So some eerie moments there, right? Because maybe the computer thought or was creative for the first time and now the man, the human champion is learning from the computer.
But I think it's a wonderful, uh, example how man and machine can learn from each other. Right. Agree. And still the human brain is, is unbelievable. I saw some slides on, on the Amed and it talked about, you know, processing power and the exabyte, right? I mean, that's just the Yeah. Well the, the number one takeaway for me after learning for four years in school is just how amazing and how cheap the human brain is.
Yeah. It is. It actually, the, the more you dive into this, you realize, um, The things we're trying to get computers to figure out and to learn, which is an amazing thing in and of itself. You know, an infant is doing right. They're crawling around, and so we really are still at the beginning stage of this.
Right. Well, it's, um, things, there is, there's some really interesting analogies. One is, you know, when man wanted to fly, it looked at birds. And in the Greek times that, you know, people even got dressed like birds, thinking that that's all you need to fly. And in fact, we needed to understand aerodynamics, right?
And then build machines that are now flying much higher, much faster than birds. And I think I see that for ai. So AI right now, um, and I disagree that people say all the time computers are fast and stupid and humans are slow, but smart. I think we are getting to, you know, middle ground where. We both can contribute to overall intelligence.
And one of the things I said at a recent interview is that artificial intelligence plus clinician intelligence equals uh, something new and special medical intelligence because wouldn't you want a man and a computer combined to take care of you in the future? Oh, absolutely. Because the, the computer can process every New England Journal of Medicine.
Right, right. And have that Yeah. Readily available 6 million pages in second. Yeah. It's, it's unbelievable. So I, you know, I'm, some of the pushback you're gonna get is here's Gartner Hype Cycle and just about everything on this left curve here in ai. Yeah. Uh, so it really hasn't gone through the, uh, the, the virtual realities coming out, augmented realities coming out of the, the hype cycle.
Um, it looks like the autonomous vehicles is actually also coming over the curve. Yes. But all these other things are just on the other side, so we have huge promise. I mean, usually what that says is we can see the promise, but we just can't get it yet. Right. And I think, um, this is a wonderful graphic that you had just before showing you saw the evolution of ai.
With, with the last few decades and in the future. So rpa, which is making a comeback ironically, has been around for a long time. We just haven't really, we've sort of been reawaken to the potential of rpa and something about AI that's good to remember is, you know, don't go for the moonshot project with millions of dollars.
In fact, the, the, the biggest and the fastest ROI could be the mundane project that a little RPA can take care of. In your revenue cycle or in your decision making process? Absolutely. At the JP Morgan Conference, Tony Tarini, uh, CEO for Ascension got up there and started talking about RPA and how they applied it to their call center and saved millions of dollars.
Right. In fact, yeah. What was interesting to me is he said, you know, not only are we taking this service that we now done to other health systems, we're actually doing it for. Organizations outside of healthcare because they Yeah. Had, uh, done such a good job with, so, and actually when people say to me where I should start with ai, we're getting ahead of ourselves, but where I should start with ai, I, I rarely point them in the clinical direction.
I almost always point them in, um, you know, claims processing, fraud, detection, security. There's so many areas you can point at that it's not life and death yet. Right. Um, Not that you shouldn't be looking at those other cases and we'll, we'll get into those, but there are a lot of places that improve the cost of healthcare.
Improve access. Yes. It's sort of like, um, multidimensional. So if you think of, um, artificial intelligence as a orchestra right now, machine learning, deep learning is getting a lot of attention. So I sort of equate that to the string section. And obviously there are many sections to the orchestra, so.
Someday soon, um, hospitals, healthcare organizations gonna be able to have amazing music with all the sections playing in the wonderful symphony and, and also together. Cuz right now there's a lot of cacophony. Everyone's doing their own thing. Um, you can have it in the same hospital, you know, perhaps the management is doing a little RPA and not know.
That their clinical data science team is working on the septic, uh, sepsis prediction model, and yet it's all under the umbrella of artificial intelligence. I, one of the things that surprised me as I was going through the material from your website is we were using AI and I, what? I didn't really categorize it as ai.
Yeah. Which, which kind of surprised me. Um, I'm gonna show you three different things and I'll, I'll put these up on the screen. So the first is, this is the Accenture. Uh, A diagram and it talks about the opportunity within AI in the different categories. Um, the second one, I'm, I'm not sure what the source is for this one, but that's me,
Oh, that's it. You're the source. Exactly. So AI and medicine and, and, and, uh, the, the third is hospital operations and workflows. Okay. Actually, I was wondering if you could just sort of walk us through those in terms of, um, you know, where. Where is it in practice today and where, where should we be thinking about it?
Yeah, I think, um, I'm looking at the Accenture slide for the first time, so I have respectfully have some disagreements. I think AI is really up to our imagination in terms of what the return's gonna be. Um, I don't think anyone can just tell you right now yeah, that the numbers, what the return is. I think these are perhaps projected projected.
Um, but I think for instance, I think, um, the area of. Uh, cybersecurity could be much bigger because I think as we learn about data protection and cybersecurity, there's uh, a lot of, uh, instances where we need to have that, especially if we're gonna share data. So that, I think that's underrated. And if I could drive that one home, we, we had implemented a lot of tools and each one of those generates log files, the log files got so massive.
Right. I couldn't build a big enough team to go through those log files. Yeah. We had to implement. Um, uh, the, the ability to essentially a big data store and then the ability for computers to identify anomalies and pull those out, which is a form of right. And I think, uh, if you look at automated image diagnosis, and this ties into the radiologist's concern that they may be out of a job.
Um, and I think medical imaging is exponentially increasing in volume and complexity. Um, most of medical imaging data. It's really been informed. If you look at just sheer volume has really been generated in the last three to five years. So if you can see that exponential curve, there's plenty of work for everybody, man and machine combined.
So I think, um, that's gonna be a growing area as well. And then the future areas, RPA will be really big in terms of helping with the administrative burden as well as, uh, I think the eventual sort of AI promised land. Is gonna be using cognitive architecture on top of deep learning so that you have something I, I call deep thinking, which is sort of in a way, barring Gary Kasparov's book title.
But I thought, you know, deep thinking or deep cognition is gonna be what medicine's gonna need because in the Game Go was impressive that the, uh, deep learning, uh, software was able to be the human contestants. But biomedicine and healthcare is kind of like, Playing hundreds of, uh, games of go with thousands of players, and the board configuring is changing.
Um, it's, it's much more complicated than a simple game go, even though it's not simple. So what should I be telling? I'm a healthcare leader. What should I be telling my radio? So computer vision now can read these images. In fact, you know, again, another slide here. Algorithm better at diagnosing pneumonia than radiologists.
Yeah, right. So, um, The other thing I heard about, you know, looking at these images is that the computer can look for a a hundred, a thousand things, whereas usually the, the Cardi or the radiologist anyway is looking for one specific thing or whatever, and isn't looking for some of those things. And so even having a, a computer go back and look at all these images could identify, uh, some things that weren't found.
I, I think, um, what we could reassure the radiologist is that you actually need. Both man and machine to have the best result. And I think, uh, the machine can help the radiologist by relieving the burden of particularly the normal studies. Right. And I think the radiologists could focus on what's abnormal and what the repercussions of that funding would be, which the, so it really is moving them up to practice at the top of their license.
So Really? I think so. Yeah. Um, and this is what we're seeing today, even with autonomous cars, right. So we're saying, okay, we're gonna, I was in an Uber car and a guy's like, oh, these autonomous cars are gonna put me outta business. Yeah. I'm like, well, not for the foreseeable future. I mean, even the, even the autonomous cars in, in Phoenix, they had a pilot going, uh, they had drivers right behind the wheel.
Yeah. Because there are gonna be situations that the computer can't figure out. We're not sure, uh, you know how long that's gonna be before the computer Right. Takes that step. Um, I said that the humans need to be alert though. Yes. As we found out. Yeah. In Phoenix. Uh, in Tempe actually. The uh, so, and the same thing's true here.
We're, we're not at a point where I'm gonna go, Hey, yeah, just have the computer give me the read. Right. I'm good with that. I'm still gonna, if, if there's something serious I'm still gonna want. Right. I'm gonna want someone to look at. Well, we did an interesting audience survey. Um, about the question, and the question was, would you, who would you trust with interpreting your CAT scan?
And the choices were radiologists alone, machine alone, radiologists, and machine. It was 85, 90% wanted all, both, uh, parties to be reading the, the ct. So give us some, give us some clinical examples. You're, you're, Pediatrics. So where's, where's AI being utilized in pediatrics today? Well, there's some exciting areas.
One is real time analytics, um, in the pediatric cardiac intensive care unit, which is, as you know, my domain, uh, previously. So we could look at, uh, take all the vital signs from the pediatric patient after, before or after heart surgery, particularly during the unstable periods, and we can actually forewarn the clinician.
What the, uh, the composite is, um, telling us in terms of, um, impending deterioration or cardiac arrest. So obviously the key to successful patient care is the avoidance of cardiac arrest, and that's a very useful too, and if you combine that with other, um, Information like in the electronic records and lab values, that's gonna be immensely useful in the future as well.
So that's just one snapshot example of how that can be used. So if I'm a hospital administrator, I should be looking at, uh, imaging today and it, or, so let, let's, let's, let's get to, let's get to the pragmatic part of it, which is, I made you CEO of a health system. Where are you looking at? Where were the first couple of places you'd be looking at it?
Mm-hmm. . What would you be doing with your team to maybe bring them up to speed? Yeah. On what's going on, those kind of things. Well, actually in the process of coming up with what I call an AI score for a healthcare organization, because that's a great question. Like where do you spend the money to get, not necessarily a best return, but the best value.
Right. And I think ironically, um, I try not to look at AI projects because I think. The, one of the major areas of deficiency as we think about AI projects is actually the quality and the management of the data, right? In healthcare. And I think you remember I said that before, so yeah. I think my first investment, it just like, if you look at the pyramid, the graphic, uh, showing the bottom being the data.
Yep. The next layer being, uh, information. On top of that is knowledge. And then above knowledge is intelligence or wisdom. So I think in order to do good intelligence, you wouldn't want to do a lot of, uh, projects in intelligence and realize that the foundation data layer is actually problematic cuz that's gonna topple.
So I would build a very strong foundation. Make sure your data. Is, um, very sound from the IT perspective. And then the it, the, uh, AI part is actually quite straightforward in just someone get, someone well versed in AI projects. I will focus on medical imaging and decision support because those are now, um, maturing.
That's areas of AI in the healthcare domain. What are, what are some of all, right, so I do. I'm the cio, you're the ceo. You're looking at me saying, how good is our data? Right? And I'm telling you, ah, you know what? Some of these physicians are not great data entry clerks. Right. And our data's really all over the place.
Not only that, um, a lot of the information we're getting is from our clinically integrated network and they don't even work for us. Mm-hmm. . And so we have all this data coming in. So, so I have a data cleanup project. That's one of the things I have to figure out in order to get that data ready, uh, for these projects.
Um, But there, there are some data sets that are really good already. Right? Right. So the financial data set is typically relatively clean. The, um, uh, the monitors, the bedside monitors, I mean, do you use that data? I mean, it's, it's, uh, you know, you have time series data. It's, you know, a lot, a lot more data points.
Right? No, I think the, the, that's exactly right. Thing ICU monitor data's fairly straightforward, is relatively complete in terms of not having many. Uh, missing points of data. There's even publicly available ICU data and adult ICUs called the Mimic three that's available. So, uh, and you can easily do projects without actually involving your own patients.
So there are publicly available databases in healthcare. There're just not many. Ideally, you know, if you were to ask me 20 years from now what I would like to see, Uh, I'd like to see every patient's data imported into the cloud, and it'll be universally available for any healthcare, um, stakeholder to look at that's anonymized, um, perhaps with blockchain or other types of security, um, mechanism.
And it'll be all available because that's gonna help. Hopefully we'll figure that out by 20 years. I'm very optimistic it'll be tackled within 20 years. So you would like to see, um, you, you'd essentially, what your, your belief is that within 20 years we're gonna have AI, machine learning, deep intelligence in, in the ability to point these things at that dataset and come up with all sorts of new thought processes in terms of how to, how to attack certain disease states.
Well, I call it a, um, perhaps like a clinical gps. For the clinicians so they can actually think perhaps even, um, more creatively than the gps, like as you would in a driving situation. You like the gps, but you may or may not, you know, want to adhere to it. So, and also occasionally, like it happened to me, uh, just a week ago, the GPS wasn't working, so you have to now rely on your human intelligence to get used.
My, my kids laugh at me cuz I, I get in the car and I put GPS to our home and they're like, you don't know where, how to get home. Right, right. I'm like, no, I'm just more comfortable with the gps there. Yeah. Well if, and, and that's how I like physicians to eventually think about AI and medicine is it's a GPS that you're just gonna get used to.
It's your routine. It's not something that is so esoteric or advanced that you can understand it's gonna be there quietly as your partner. Um, if you looked at a Microsoft commercial 25 years ago, Okay. Um, there was actually a segment saying, you know, imagine the future. And I laughed because it says, imagine the future traveling across the country without fold out maps.
And people were laughing, they just did not think that was gonna be possible. And now with gps, you don't think twice about driving across the country. No. And I've, I pulled out a fold out map. Uh, not too long ago with my kids in the room, and they're just like, what is this? Yeah. Is that a new puzzle or
What are you doing? Um, so in, within 10 or 20 years, as more clinicians are, um, getting educated and aware of ai, I think it's just gonna be part of their routine. And I, that's why I like to see it's actually embedded in your clinical routine without disruption to your workflow. Without any distraction from their usual routines and just be their part silent partner.
Do you, do you think we need to change how, um, how doctors are educated? Do you think we'll start to see AI start to get built into those programs? Yes. I think, um, I'm starting to see clinicians that. Are becoming more seriously interested in data science and, um, the younger generation considering, uh, dual education in both clinical medicine and data scientists.
Do you think like Hopkins and, and, uh, Stanford, gw, my alma mater, Stanford, um, ironically at Stanford, I didn't meet too many doctors in the data science program, but I think, uh, we'll see more, more, right, just as clinicians that may want to go back to school and get a data science, um, education to be. More of a specialist in this area.
Like other specialties. I, interesting enough, had a data, a young person was a data scientist who got so interested in healthcare. She's thinking about going to medical school. So it can be the other way around too, which is great for healthcare, that we have a cohort of young people that are gonna be dedicated to data science.
I couldn't be happier to hear that. Yeah, that would be phenomenal. So the, again, I'm playing the healthcare administrator, I'm gonna say, you know, This, this last year, I've had five different vendors come into mind. Right. I, I mean, how should I be thinking about these vendors? I, you know, cuz even some vendors from outside of healthcare are coming in and going, Hey, we've done this in other industries.
Right? The complexity of healthcare is such that I'm not, I mean, I'm not saying they can't do it, but I'm just saying the learning curve for them on the, on the healthcare side's pretty high. Well, I, how should I think about vendors? Yeah. I think, um, be careful and, um, be very vigilant for. Um, then there's that overpromise and underdeliver.
Obviously, I think there've been some very big, uh, companies that, um, you know, that yeah. Over-hyped and underdelivered and actually became unsuccessful, those big institutions. So I think, um, those are cautionary tales I think we need to pay attention to, and I think invest, you know, a relatively small amount of, uh, revenue into sort of the space and get to and use it.
Cautiously as, as, and as a way to learn about the limitations of AI in this space. But I think the promise is tremendous, and I think the future is very bright for this area, but I think, um, you know, don't overspend your resources now and just get to know it. And education is key. I think it's, it is the same thing we did around big data.
It's, it's small projects, um, uh, with defined outcomes. And you give those vendors like tasks to do and once they do it, you can then expand it and grow it. Um, is there gonna be a problem scaling AI at this point? Not at all because I think, uh, the, the, um, potential is really limitless because of the computational power being so cheap now and so many algorithms are maturing and I think it's gonna be an amazing portfolio of things are available.
And I think if you look at how. Inefficient and how badly run healthcare is. It's just gonna be amazing transformation in the next 10, 20 years. The other thing that's amazing to me is how much, uh, open source there is out there. I mean, Google, Google is open source, right? Uh, you can tap into Amazon, you can tap into Microsoft.
There's a lot of ways to tap into this without building it out on site, right? So there are, there are inexpensive ways to. Play around with it. You have to get your data right and get it to the right place. Right? Once you do, you can do some things. I think, um, as the common saying is, you don't have to be an engineer to drive a car, and I think that's valid for this too, and.
But I do think you need to understand the, the general mechanics of the car and also the, the rules of the road. Just like, uh, with any, just like with ai, this is one of those things where the technologists and the, uh, clinicians are gonna have to come together and what, right. Um, I think I saw one of your slides that c h l, the, the, uh, data scientists actually goes on routes, right?
Yeah. That was, I'm a very strong advocate of that. Um, our data scientists, um, at chalk will. See patients with me in a clinic and truly understand the clinical culture, just like I did with, uh, my school years where I spent a lot of time with computer scientists. So what will the data scientists get a pull out from that?
What will they, will they identify something and go, Hey, we could, we can run the numbers on that. We can Exactly. Or they, uh, understand the nuances that when. We have healthcare data. It's not as exact as they might think it is or why data is missing. And so I want them to focus not on just the AI potential, but also the acquisition and management of data.
So our tagline for the show is for the next generation of, uh, health IT leaders, right? What am I gonna do for my, my team? How am I, I clearly not everybody's going to get up to speed on this, but how, how do I. Where's, what are some resources, obviously the ai, ai med website, uh, what are some other ways I can get up to speed?
Well, I think ai, uh, med website's gonna have the, the ebook I made publicly available for free. That's a good start. Has a glossary of about 400 words, if you wanna look up words. Um, we're in the process of producing, um, A short video series on hot topics, like what's the difference between AI and deep learning, things like that.
We'll have that available by the end of the year. Um, there's a, um, an academic magazine, I call it academic magazine because it's not boring to read like an academic journal at the same time is very entertaining with education. So an academic magazine is gonna be free also to everybody on a bi-monthly basis.
We focus on a specific theme. Every other month. So, and this is all through AI Med, this is all through us. We wanna be, um, your sort of educational source for AI in medicine. And there are breakfast briefings all over the world. There will be seminars starting here in Orange County, but that will also spread all over the world.
So I think just a sense of awareness and, um, some education will go a long way. So I would also, um, recommend that. Every person in the hospital just have some awareness that this is sort of the new paradigm in medicine. And of course you'll start your new podcast here shortly, and we'll all right, we'll all tune in.
Yes, that'll be great. Um, well, you know, thanks, thanks for your time. I, I'm, I'm actually gonna interview your friend here in a, in a minute, but I'm gonna do the close and then we'll come back to this. So, um, how can people follow you? Are you on social media or? I am. Um, I'm just, um, that's maturing right now, but I think the best source is still the website.
It's AI Med md io. It pretty much has everything to get started. Has the ebook, the, the magazine that you can go through, it has all the upcoming activities loaded, um, most likely in your, um, near your region. We have three big meetings this year in three continents, and we'll go to five continents next year.
So you'll, um, Be seeing a lot of us. Yeah, I well, I'm, I'm definitely gonna look you up, down in Dana Point. I know when it came up last year, somebody, uh, asked me to go and I was out of town. So, yeah. And also I'm, uh, finishing a book project with Elsevier. So there'll be a textbook that's written for everybody, cuz I don't want this to be read by only clinicians or data scientists, so I'm trying to write it for.
Everybody and, uh, with enough, um, you're writing it for me. So is it a picture, picture book? Is it like, I'll have you a picture next to me when I write from now on is what I, no, I think, um, what I'm really happy to see is there's no longer a emotional pushback and there's more now a natural curiosity and the sense of, um, I wonder about what this can be someday.
So I'm happy to see that. Yeah, it is exciting. So, um, uh, you know, just to close out, you can follow me on Twitter at the patient cio, my writing on the Health Lys website and health systems cio.com. Uh, don't forget to follow the show at this week in this week in ht, and check out our website this firstname.lastname@example.org.
Uh, if you like this show, take a few minutes and, uh, give us a review on Google Play or iTunes and you can catch all the videos. We're now up to 120 videos on the YouTube channel this week in health it.com/video. Or you can just go to YouTube and search for this week in health. It. Um, that's all. Uh, please come back every, uh, Friday for more news information and commentary.