Interview with ClosedLoop CEO Andrew Eye

0:00:11 - David Williams
They say an ounce of prevention is worth a pound of cure. Yet it's not usually obvious in advance exactly which patients will benefit from targeted proactive interventions. Payers and at-risk providers have long used predictive modeling to try to figure out where to act, but big success stories are rare. New AI approaches should be a game changer, however, especially if applied to the right datasets and coupled with effective follow-through. Today's guest, andrew Ai, is CEO and co-founder of Closed Loop, which is using AI to predict health events like hospital readmissions and appointment no-shows. 

Hi everyone, I'm David Williams, president of Strategy Consulting firm Health Business Group and host of the Health Biz podcast, a weekly show where I interview top healthcare leaders about their lives and careers. Please leave a comment, subscribe or leave a rating or review. Andrew, welcome to the Health Biz podcast. Hey, david, thanks for having me. Great to see you. Andrew. You've been doing a lot of really interesting stuff at Closed Loop, but also a lot of interesting stuff before that, so I just want to explore that a little bit to understand your background, maybe starting with your childhood, any notable events in childhood or any childhood influences that have stuck with you over time. 

0:01:21 - Andrew Eye
Oh boy, there's quite a few stories I won't share with you today, but if you do some digging you might find some horror stories. It was probably a little difficult for my mom to raise, but in terms of influences I'd say my mom was my biggest influence. You know, single mother, public school teacher, went back to school after raising my brother and I, so I got the distinct honor of watching her graduate and walk across the stage. She really instilled in my brother and I kind of a work ethic as well as kind of a real focus on education. So she used to tell us you can be anything you want to be in life as long as you get a college degree. I said well, that's great mom, but what if I want to be a trash man? She said you'll be the smartest man on the trash truck. Yeah, so we were really lucky to have our mom kind of bring us up right. 

0:02:09 - David Williams
That's excellent. You know, if I look at your background and a lot of the entrepreneurs that I speak with, they probably weren't the easiest in school, and so I know, if your mom had probably a classroom of people like you, she probably wouldn't have been able to manage too long. And yet at the end of the day, if we didn't have a few like that sprinkled about, we wouldn't have the kind of dynamic economy that we do have. So when you got on to, you do have a college degree. So did you do that just because your mother said you had to? Or what were you thinking at the time? 

0:02:39 - Andrew Eye
Well, there wasn't a choice in our household. I was going to college, whether I liked it or not, but yes, I'm a proud Hokie Went to Virginia Tech, did my undergrad and kind of a business slash technology background. So came out of school as a software engineer but also had kind of a basic business education as well. So you know, that was kind of my my route. I was lucky to be at a school, it was a lot of fun and had a lot of school pride, but that was my journey as far as college was concerned. 

0:03:10 - David Williams
Got you. So how early on. You worked at, I think, a place called I2, and then you blasted off to NASA. 

0:03:16 - Andrew Eye
Yeah, that's right. So you know I2 technologies. It's funny. Now we talk about artificial intelligence and machine learning. You know I2 was an algorithms company in the supply chain optimization business. You know they were trying to figure out how to load packages onto trucks to get the best gas mileage out of them. We didn't call it AI or machine learning back then, but similar technology, and actually met my co-founder there, dave DiCaprio, almost 20 years ago at I2 technologies. Then, yes, you're right, spent a little time at NASA working on Earth science satellites and then, you know, ended up starting my own businesses thereafter. 

0:03:50 - David Williams
Yeah, it's interesting you mentioned meeting your co-founder at your first job. That first job could be so important and I see it. I have my own kids are recent college graduates a couple of them and I remember back. You know people that I met my first job I still work with today, 30 years later. It's a really formative time and if you can find the right person, it is really worth a lot. 

0:04:14 - Andrew Eye
You know, I was very fortunate. Our campus at I2 was about a block and a half away from MIT, so when I first got there I thought, boy, they made a big mistake. I must be the dumbest guy here. You know, I worked with all MIT grads, you know, at the beginning of the Internet, and so there are a lot of smart young people working there that have all gone on to be CTOs and CEOs. So, yes, you're right, quite formative, and I was very fortunate to have that first experience. 

0:04:41 - David Williams
Now, you mentioned going and founding a couple of companies, and I saw two companies that you, I think, were involved in founding, scythand and Boxer, which then both were acquired by larger organizations that you spent some time with as well. What was that process like? 

0:04:55 - Andrew Eye
Yeah, so you know, the first company I kind of fell into had a friend who had kind of encouraged me to leave my day job and start a business with him. That was in the information security space Scythons, as you mentioned. So that was a bootstrapped company. We started it from nothing and grew it to number 16 on the Inc 500 list, you know, had a bunch of success there, ultimately sold that to Acuvant, which is now Optiv, the largest information security reseller in the country. So really fortunate to grow that business. 

You know the second one, boxer, was a consumer mobile business. That was Outlook for your smartphone before there was Outlook for your smartphone. That was the first venture back to business that I had started. So went out to Silicon Valley, you know, went down Sand Hill Road doing the rounds and pitching the idea and ultimately built one of the first third party mobile applications, third party mobile email applications for iPhones and Android, ended up selling that to VMware. So yeah, that was kind of the first journey Information security, then consumer mobile email and, of course, after you do those two things, what else would you do but start a health IT company? 

0:06:02 - David Williams
Right, of course I was going to say well. So I mean, it's obvious, right? So if you start the first third party app, it's not like you're going to start the third first party app. So that's, where else do you go other than healthcare? Usually people go into healthcare because they don't know what they're doing. They think it's going to be easy. A lot of times you have people from the technology field and they say you know, we solved this, that and the other, you know moving boxes around, et cetera. How hard could it be to go and fix everything in healthcare which is so broken? Were you one of those, or were you a little more eyes wide open going into it? 

0:06:29 - Andrew Eye
You know, I'd like to think I was eyes wide open, but maybe kicking and screaming. You know people always ask me how did I end up in healthcare? And the short version is my co-founder made me do it. 

Yeah, dave DiCaprio had worked on the Human Genome Project at the MIT Broad Institute under Eric Lander, spent some time in the payer market doing predictive analytics and machine learning, and so Dave and I had wanted to start a project together. You know, after I was leaving VMware and originally we were working on something completely different we bought the first two Microsoft Holo lenses in Austin and we were going to build, you know, augmented reality applications because we thought that might be the next big platform. Yeah, but luckily, about three months into that project, dave said hey, actually, andrew, I think this stuff we're working with and kind of deep learning and machine learning could be really applicable in the field I came from healthcare and I said, boy, that's a terrible idea. I really don't want to do that. I heard that's a really hard market, but Dave roped me in. 

I spent about a month getting kind of acquainted with the industry and got really excited about value based care in particular. Yeah. So to me it felt like a real opportunity, with a major market shift and a lot of new entrants who might be receptive to new technologies and maybe not stuck in their ways, and so that was kind of where we started applying machine learning and artificial intelligence first in the value based care segment. So, yeah, what was not my first choice was not excited to enter the industry. But you know, after having some personal experiences with my own daughter getting sick, I caught the bug right and got really excited about. You know working on mobile email that. You know at the end of the day we were saving white collar workers five seconds a day by swiping on their emails. Working in healthcare, you know you're applying the most influential technology of our lifetimes. You know artificial intelligence to improving health and saving lives. It's pretty easy to get excited about that Sounds good. 

0:08:22 - David Williams
Well, I don't need machine learning to see that pattern and got enough examples of my podcast to people who say how did they get into healthcare? And you had, sounds like two reasons of you know, a co founder and then a personal experience. The personal experience is the usual reason for it. So what was the why? Was there a need? You mentioned value based care in particular, but once you came to see that healthcare made sense, I mean why? Why is that a place that has a real need for what you're doing? 

0:08:47 - Andrew Eye
Yeah, you know, for us, when we were starting, the reason we went after value based care and accountable care organizations in particular at the time. You know we have a lot of investors, a lot of people saying like this is the dumbest place you could start. You know, if you've seen one ACO you've seen one ACO six years ago you know only one in three ACOs was actually making money or succeeding in those value based care contracts. And so you know we had a lot of pushback that maybe this wasn't the right place to start. But our thought was this is the part of the pie that's growing. You know we're going to see more and more dollars shifted to value based care and these are early adopters. These are folks who, by definition, are trying new things. And so we saw a green field opportunity in value based care, the real problem to work on. You know, if you want to succeed in value based care, if you're a risk taker, you know whether that's a self insured employer or payer or, you know, value based care provider. Really, you know there's lots of games you can play. You can kind of look at kind of trying to get coding correct. You can look at how you design your network and all of those things can really have a benefit in building a good business, but at the end of the day, what the federal government needs us to do, we need to spend less money and get better results. If you really want to succeed in value based care, you've got to get better results with fewer dollars. 

That's what we're all after is kind of bringing down the total cost of care and improving health, and so to do that, I always tell people there's really two fundamental things you've got to be able to do. You've got to be able to predict the future. You're going to see the bad thing that's going to happen before it happens and you've got to be able to change it. You've got to have programs and interventions that actually are successful in reducing that total cost of care. And so you know, if we talk about just pop health and if we talk about just kind of the core tenants of value-based care as it pertains to driving down total cost of care, those are a lot of pop health use cases that you hear about avoidable hospitalizations, readmissions, chronic disease onset progression, et cetera. So targeting those programs predicting the future using machine learning and AI was where we saw the first market opportunity and we've had a lot of success there. 

0:10:50 - David Williams
Why do you call it closed loop? 

0:10:53 - Andrew Eye
Yeah, you know it was interesting. I have a history of being pretty terrible at naming companies, actually but closed loop, we had looked at a couple of different things. You know we looked at deep medicine, which ended up being or deep health I think you know playing this deep learning idea way back when. But you know, for us, closed loop was this idea of not only predicting the future but actually demonstrating whether or not you were able to change it right. The closed loop. 

And closed loop is all about hey, I said something bad was going to happen, or I predicted some future event, or I said somebody was going to know a show for appointments. Then you're going to do something different. Whenever you make a prediction, the whole idea is you spot something in the future that you either want to happen or you don't want to happen, and then you make a new decision based on that insight or prediction. And so the closed loop. And closed loop is all about hey, I made a different, I took a different course of action based on that prediction. Did I actually impact the outcome? And so that's a whole different branch of what we do, which is not just about prediction but about outcomes evaluation, which is traditionally thought of as like an actuarial science. But that's another kind of aspect of what we do here at closed loop not just prediction but outcomes evaluation. 

0:12:06 - David Williams
The concept of closing loop in healthcare, I think, is a very good one. If you speak to a physician and they make a lot of referrals, you can then try to say well, you're referring me here. What's their success rate? What have you seen? And a lot of times they don't know because they didn't actually get the feedback. Part of it also wasn't analyzed and presented in the right way. But they send things out and they just don't know. And I even remember back in the day you go and get lab tests and they say well, we'll let you know if there's a problem, but they don't let you know. But then if I didn't get the postcard, is it because they didn't send it or why? So this sort of lack of closing loop is one of the things that holds healthcare back, even at a broader level than what you're doing. So I like the name. So you get your history. You're getting better on that one Now going back to-. 

0:12:51 - Andrew Eye
The key is you've got to get somebody else to weigh in. So Dave and I bounced a few ideas off of our wives and they zeroed in on closed loop and helped us make the right call. 

0:12:59 - David Williams
No, that sounds like the way to go. You mentioned that when you were looking more recently in talking about machine learning, you realized that back even at I2, that's what you were doing, even if you didn't call it that. Now we're talking now about predictive modeling, and predictive modeling is something that's been talked about in healthcare for a long time and it was absent AI and machine learning and I think it was actually because it wasn't using those techniques. It was using more basic techniques in the past. But when I hear about predictive modeling, I mean that is something I've heard about being used in healthcare around in the 90s. So what's different about what you're doing now and what you're still calling predictive modeling compared to maybe, what has been done in the past? 

0:13:41 - Andrew Eye
Yeah, you know. I love the way you framed the question, david, because at the end of the day, you know, especially if you back up four years ago or so, there were a lot of debates. Every time I would go to a conference. Every time people would get up to speak about this. The first thing that they would talk about what's the difference between artificial intelligence and machine learning? And they would draw these Venn diagrams. And AI is this and ML is this. And you know I would get this question as a speaker often and I'd let everybody else in the panel speak and then they'd ask me Andrew, what's the difference between AI and ML? And I would say no one cares. 

At the end of the day, these are marketing terms, whether as technologists, we like that or not. Ai, machine learning these have all been adapted as marketing terms. So when I think about what's different now versus, you know, 10, 20 years ago, it's just better math. At the end of the day, what we're able to do today that we couldn't do 20 years ago is use more variables, use more data points, synthesize those variables together in different combinations. The math has gotten better. The techniques have gotten better at taking all of the available data for a given organization and using it to make the best possible Prediction for a given outcome. You know, in the old days what you would say is well, we've got this rules of thumb based approach. It uses these 30 variables and we weight these variables in these combinations and we come up with a risk score. 

Today, what we're able to do is say, hey, let's look at the unique footprint of data that a given organization has. If I'm a payer, I'm going to be anchored in medical claims, prescription claims, maybe I'll have an ADT feed, you know mission discharge and transfer. Maybe I'll have some social determinants of health data. But if I'm on the provider side, as an ACO, I might be anchored in EMR data plus labs data. I've got the same question who's going to be readmitted, who's going to have an avoidable hospitalization, etc. But I'm working with different data assets, and so what machine learning allows us to do is build the spoke predictive models Based on all the available information and signal available to a given organization, to make the best possible prediction, the most accurate, most explainable, most actionable prediction for a given organization. So that's really what's changed is this ability to kind of leverage all the data that an organization has, rather than kind of Looking for just a handful of variables. Now how about? 

0:15:55 - David Williams
the data sets themselves a lot of what you're describing claims, adt. That structure of those hasn't changed that much, although there actually has been some change on the least on the claims side. Are the data? Is the quality of the data better, now that it people realize what it's being used for and it's being Collected in different ways, or the data sets different? Or is it most of the difference and improvement? You know, based on the math and then also the perhaps the change, so that you've got value-based care and people actually have a reason to act on it? 

0:16:22 - Andrew Eye
Yeah, so there's a couple of big trends going on here, but you know this goes back to your question of what's different about machine learning. So you know there's this idea in healthcare, particularly a few years ago. I hear about it less that you know. Oh, our data is too messy. The first thing we have to do. Our data and healthcare is so messy and you don't understand, and we've got to clean up the data before we can do anything with AI and machine learning. It turns out machine learning is really good at handling messy data. 

The the idea that I've somehow got to get everybody perfectly coded, I've got to get everybody perfectly classified before I can then use AI or machine learning. It's just false. And so, if you take a simple example, I might have a patient who isn't coded as diabetic. So I look at their claims data, look at their EMR data and maybe somehow I'm missing that diabetic Diagnosis code, and yet I see a prescription for insulin. Machine learning is really good at figuring out that most people who have prescriptions for insulin are probably diabetic, and so it's going to treat those two patients very similarly, regardless of that missing data. And so when you have a rules-based approach that says, if I have a diagnosis code for diabetes. Your answer to that is gonna be no right, but when you're able to take in more data feeds, you can make up for that missing or incomplete data. 

So, yeah, the other thing that we're seeing is the push towards interoperability and the policy changes that CMS has enacted about. Shoutshare data has really made a big difference. We're seeing a lot more utility in the data feeds that are available, and I'm very excited as well to see what CMS is doing in terms of making their data more timely. So if you look at kind of the traditional 90-day claims lag, oh, claims data's got a lot of information, but it's so delayed. With new movements from CMS, with things like the BCDA data feed, you can get that down to as little as 14 days, and with their pre-adjudicated claims feeds they're not perfect, but you can get that down to as little as four days. So you're seeing this not only are we getting more data, but we're getting it faster, and that really is very powerful when it comes to kind of making better decisions and surfacing insights. 

0:18:34 - David Williams
So you have more data, you can have it faster, you can have it cheaper from the standpoint of just interoperability. And then also, if some of the things that we've been done before, which used to call data cleaning, can be just skipped because you can do it, machine learning, that's improved. And I think the machine learning can go beyond what you're describing as your simple example with the insulin and the diabetic, to find, for example, patients who had certain tests or certain symptoms, and even if they weren't officially diagnosed, there's probably a likelihood that they may have something. And then I think the business model side of it is all going together All right. So now it's a no-brainer that you went into healthcare. What about this question about funding the organization? You said you'd had success on a bootstrapped one, and then you did a venture, backed one. I think you raised money for closed loop as well, but what did you learn and how did you apply that to your financing decisions for closed loop? 

0:19:22 - Andrew Eye
Yeah, sure it was funny. The first company, as you mentioned, Bootstrapped Second company. I thought, well, geez, I've got this company that was Inc 500, number 16, and I've sold a company and so I'm just gonna go out to Sand Hill Road and they'll throw money at me. Yeah, boy, did I have a wide, a rude awakening. I got 60 nos, that's six zero nos before I got one, yes, when I raised money for Boxer, fortunately for closed loop, having two wins under your belt is a little bit different than having one. So the early funding gets a little easier when you've had a couple successes. But everybody's on a level playing field once you get past that early funding. So it really comes down to metrics and kind of adoption. So for us, going from that seed round to an A round and then to a B round was all about market traction and for us kind of the two big kind of points we're winning the CMS AI health outcomes challenge that really put us on a map, beating out the likes of IBM, Accenture, deloitte, merck, mayo Clinic, and a head-to-head contest that put us on the map for a lot of folks. And then analyst reports, so folks like Klass who cover these categories and talk to your customers, and so I think that's where we've been fortunate in being able to raise some of these other rounds. Obviously, ai is a hot sector, so we've been fortunate in having folks to kind of believe in the vision. 

Who is the customer? 

Yeah, so for us we have two types of customers those who have data science teams and those who don't. 

For those that have data science teams, we provide a platform for them to build, deploy, monitor predictive models better, faster, cheaper. So that might be the chief data officer at a payer think of like a mid-tier blue on the payer side. And then we have customers who don't have data science teams. So in the ACO market you've got folks that are building new teams rapidly growing, and they might use our platform plus our bench of data scientists as kind of a fractional FTE, a fractional part of their team, to build some of these same predictive modeling use cases as well as some of that outcomes, research or actuarial work that we do as well. So sometimes a technical buyer, sometimes a VP of Pop Health or a VP of Quality Chief Medical Information Officer. But our big two segments are kind of payers, value-based providers, and then now some of the folks in the traditional kind of fee-for-service hospital world are waking up from that COVID coma and are starting to make investments in this area as well. 

0:21:50 - David Williams
You know, what you mentioned upfront was that you've got some organizations that are new and ready to try something different. You know you're sort of getting ahead of one question that I had, which is that these health care organizations tend to be fairly overwhelmed and not necessarily cynical but kind of battle-hardened, and you know there's just so much coming in that they have even trouble assessing it, just doing anything with it. How is it that you get these? If I'm right that you know a lot of overwhelmed health care organizations out there, how do you get them to try something that's new? 

0:22:23 - Andrew Eye
Yeah, you know, again, when we started it was all about finding the people who had a little bit of a different mentality. So if you look at kind of value-based care, by definition, these were people who were kind of starting with a different business model and they had the very acute pain point which was I'm going to make investments in proactive care. I need to make sure that proactive care is going to the people who actually need it. So I've got to be able to predict the future if I want to be able to change it. So you know, a lot of this is about picking the right people first, the right customers first. Then, once you've demonstrated success, it's not that big of a jump when people see, gosh, you're having success with one medical or IARA. You know, you're having success with Palm Beach ACO, the number one MSSP ACO in the country. You're having success with Southwestern Health Resources, number one next-gen ACO in the country. When they see that success, then others who might not be the first adopters start to say, gosh, you know, maybe this is something that makes sense for me as well, and so you know, for us it was about going from value-based care to payers and now, you know, back to the traditional hospitals. 

You know, I think part of this is also you can't come at this with like a shiny object and like you don't want to be a hammer looking for a nail. And so, for us, our first question, you know, is always let's assume I can predict the future what are you gonna do different based on that prediction? Right, and how is that gonna drive your bottom line? Right, whether that's improved outcomes or whether that's financial outcomes? But we've gotta have a tangible ROI associated with whatever use case you're interested in, because otherwise this program's gonna get shut down for all of us. And so, whether that's quality and safety or, you know, population health, you've gotta really find those tangible ROI opportunities in order to kind of get past the hurdles that you mentioned. 

0:24:15 - David Williams
So, Andrew, where does the company go from here? 

0:24:19 - Andrew Eye
Yeah, you know, for us, the thing we're really excited about right now, you know, it's not just about kind of again, as we talked about, predicting the future and machine learning and AI, what we're really excited about is closing the loop and being able to not only target these programs, the proactive care, but help organizations demonstrate that they are actually bending the cost curve, that they're actually improving outcomes. 

You know, when you look at, particularly in this pop health space, which you know we've kind of focused on that today, but there are use cases in risk adjustment and prior authorizations, lots of other opportunities. But at the end of the day, if you're gonna do something different based on that insight, that AI prediction, you've gotta know whether or not the action you took actually had the intended outcome. And so that's all about this outcomes research. That's all about what we call evaluate, which is program evaluations that really, in a statistically sound way, tell you is your chronic care management program working, is your transitions of care program working, et cetera, because otherwise you're just looking at it and saying, well, we did better than we did last year, so I hope our program is working. But doing that in a statistically rigorous way, through program evaluations and actuarial science is what we're really excited about. Closing that loop Great. 

0:25:37 - David Williams
Well, my last question for you is about whether you've had a chance to read any books and recently, if there's anything that you would recommend, or if you recommend anything that people avoid. 

0:25:47 - Andrew Eye
Yeah, sure, gosh, a few that come to mind. Obviously, in the space for folks who are interested in AI and chat, gpt and large language models, great book, the AI revolution, gpt at four and beyond. Peter Lee had a research from Microsoft, maybe only a couple of months old. I'm reading right now the Song of the Cell by Siddhartha I'm going to butcher the last name, mukherjee and then another recent one in the AI and ML space is A Thousand Brains by Jeff Hawkins. So those are a few favorites and maybe one more. My daughter's favorite was the Stranger in the Lifeboat by Mitch Album. So if you want to break from some of these deeper reads, that's a great one as well. 

0:26:31 - David Williams
I get a good mix of those Usually adult ones and people with younger kids have. Yeah, they say I'm reading this one, but this is actually the one that's probably the most profound and have the biggest impact. 

0:26:39 - Andrew Eye
Stranger in the Lifeboat was a good one to have a few dinner chats about afterwards. So a good quick read Sounds good. 

0:26:47 - David Williams
Well, andrew I, ceo and co-founder of Close Loop, your mother may have found you to be difficult as a kid, but I'm sure she would have predicted as well that you were definitely going to college and that you're going to then do something big and good, which you have. Thanks for joining me today as a guest on the Health Biz Podcast. 

0:27:01 - Andrew Eye
Thanks so much, David. Great talking to you. 

0:27:03 - David Williams
You've been listening to the Health Biz Podcast with me, David Williams, president of Health Business Group. I conduct in-depth interviews with leaders in healthcare, business and policy. If you like what you hear, go ahead and subscribe on your favorite service. While you're at it, go ahead and subscribe on your second and third favorite services as well. There's more good stuff to come and you won't want to miss an episode. If your organization is seeking strategy consulting services in healthcare, check out our website, healthbusinessgroupcom.