If you’ve ever tried to find a doctor in the United States, you likely know how hard it is to find one who’s the right fit—it takes quite a bit of research to find good information to make an informed choice. Wouldn’t it be nice to easily find a doctor who is the right fit for you? Using data, Covera Health aims to do just that in the radiology specialty.

Ron Vianu: I think the tools are really improving year over year to a significant degree, but like anything else, the tools themselves are only as useful as how you apply them. You can have the most amazing tools that could understand very large datasets, but you know how you approach looking for solutions, I think can dramatically impact. Do you yield anything useful?

Ginette Methot: I’m Ginette,

Curtis Seare: and I’m Curtis,

Ginette: and you are listening to Data Crunch,

Curtis: a podcast about how applied data science, machine learning, and artificial intelligence are changing the world.

Ginette: Data Crunch is produced by the Data Crunch Corporation, an analytics training and consulting company.

If you’re a business leader listening to our podcast and would like to move 10 times faster and be 10 times smarter than your competitors, we’re running a webinar on February 13th where you can learn how to do this and more. Just go to datacrunchcorp.com/go to sign up today for free.

If you’re a subject matter expert in your field, like our guest today, and you’re looking to understand data science and machine learning, brilliant.org is a great place to dig deeper. Their classes, help you understand algorithms, machine learning concepts, computer science basics, and many other important concepts in data science and machine learning. The nice thing about brilliant.org is that you can learn in bite-sized pieces at your own pace. Their courses have storytelling, code writing and interactive challenges, which makes them entertaining, challenging, and educational. Sign up for free and start learning by going to brilliant.org/data crunch. And also the first 200 people that go to that link will get 20% off the annual premium subscription.

Today we chat with Ron Vianu, the CEO of Covera Health. Let’s get right to it.

Curtis: What inspired you to get into what you’re doing, uh, to start Covera health? Where did the idea come from and what drives you? So if we could start there and learn a little bit about you and the beginnings of Covera health, that would be great.

Ron: Sure. Uh, and I, I guess it’s important to state that, you know, I’m a problem solver by nature, and my entire professional career, I’ve been a serial entrepreneur building companies to solve very specific problems. And as it relates to Covera, the, the Genesis of it was understanding that there were two problems in the market with respect to, uh, the healthcare space, which is where we’re focused that were historically unsolved and there were no efforts really to solve them in, from my perspective, a data-driven way. And that was around understanding quality of physicians that is predictive to whether or not they’ll be successful with individual patients as they walk through their practice. And so if you, and we’re focused on the world of radiology, which today is highly commoditized and what that means is that there was a presumption that wherever you get an MRI or a CT study for some injury or illness, it doesn’t matter where you go.

It’s more about convenience and price perhaps. Whereas what we understand given our research and the, the various things that we’ve published since our beginning is that one, it’s like every other medical specialty. It’s highly variable. Two, since radiology supports all other medical specialties in a, as a tool for diagnosis, diagnostic purposes, any sort of variability within that specialty has a cascading effect on patients downstream. And so for us, the beginning was, is this something that is solvable through data? Could we understand for an individual patient as they’re looking for medical care, what is the right physician for them that would yield the most accurate diagnosis related to their condition.

Curtis: Got it. And I’m assuming you have some experience in the medical field. Do you usually have the companies, you’ve started been in the medical field and so you had insight into this issue or where did that come from?

Ron: Yeah, I mean, my background, I was a premed student actually, uh, in New York and I, at the time, I felt like going to medical school really wouldn’t be solving problems as the way I saw, uh, the life of a physician. And so I decided that business was probably a better perspective to solve problems. And ironically I ended up solving problems within healthcare my entire professional career. And so I have a fairly deep knowledge base, if you will, around clinical medicine for a lay person and obviously a lot of experience around starting businesses and using data to solve problems. And so it really, for me it’s an interesting combination of skills that allows me to tackle these things in a way that perhaps a physician or a business person, uh, independently wouldn’t be able to do.

Curtis: And where did your expertise in data come from? You seemed to approach things from a very data-driven perspective. Where did you get that from?

Ron: I think that’s honestly something that one is innately born with and then one finds the tools to help them explore that. And so in college I studied chemistry and philosophy, and I think part of it is because I was trying to approach different parts of the way my brain functioned. And so when I solve problems today, I try to solve them in a very data-driven manner, generally speaking. And so when I find tools like statistical modeling or AI and so on and so forth, that can further enhance the approach that I would take in solving a problem. Those tools are extraordinarily useful for me. But I don’t think it’s something that I, you know, and one could argue, maybe others have this where you take a course and you’re like, ah, this is an interesting science and I could use this science. For me it was how do I kind of expand the very way I generally function.

Curtis: One of the things that we see as the tooling and the understanding around machine learning and analytical practices is becoming better and better. As someone that didn’t study this, you know, computer science, this kind of stuff. Have you found it accessible? Sort of easy to pick up and apply to problems?

Ron: Right. So I guess two points I would make there. One, I’m, I’m not a data scientist per se in any traditional way. My background is comp sci meaning back in a, in kind of an untraditional way, meaning both in college and pre college I was programming. And so I have a little bit of that background even though I didn’t study it in a formal setting, but I think the tools are really, uh, improving year over year to, uh, to a significant degree. But like anything else, the tools themselves are only as useful as how you apply them. And so I think, you know, you can have the most amazing tools that could understand very large datasets, but you know, how you approach looking for solutions I think can dramatically impact do you yield anything useful.

Curtis: And do you have a specific approach that you do? Is this, does it come naturally to you or do you have some sort of framework or approach that you use to look at things and figure out how you, how you could solve it?

Ron: Right. So I’m agnostic from a data science perspective with respect to the actual approach we’re taking, uh, meaning what tools are we going to be using? But moving aside technically, you know, there are two different approaches one can take when one broadly thinks about data science and analytics. And you know, the, the big approach that I think has been very popular over the last, call it five, seven years, is around big data as people call it, which is now that we have access to lots of data and we have access to all these interesting tools and algorithms that can analyze that data, what can we ultimately understand from that data, what patterns can emerge that perhaps we haven’t seen in the past? And I think that’s very productive and useful in many contexts in healthcare where it’s very difficult to understand what data you’re looking at to begin with.

And so you have very dirty datasets and cleaning those up becomes half the challenge. And so for me, my approach with respect to healthcare data analytics has been more hypothesis driven rather than that big data approach. And what I mean by that is if you speak to physicians around this thing called quality, which is what we’re trying to solve, how do you understand what physician is ideally suited for a particular patient in order to yield the best outcome? And so as we approach that problem, we work with many experts across the field and we ask to understand their intuition around quality, what makes a good physician. And once we have a unified sense of what the experts think, then we start attacking the data in a way that explores those theories and understands if we can ultimately find some signal with respect to those theories or rather correlations with respect to those theories. And so it’s, it’s a little bit of a different approach, much more hypothesis driven than big data-driven.

Curtis: So instead of sifting through the data to find random signals and then seeing if those are useful for some application, you then make some hypotheses, uh, bring domain knowledge and then see if you can find some signals in data that, that, that you have available. Is that accurate?

Ron: That’s accurate. And, and I can give you a concrete example. If you think about the world of radiology and if you ask radiologist what makes a good radiologist, what gives them the skills and capabilities that ultimately will drive higher quality care? And in radiology, quality is very simple. If you go to an imaging center and you get an MRI, are you walking out of that imaging center with an accurate diagnosis? And so it’s fairly binary, unlike other medical specialties where it may be a little bit more gray in terms of defining quality.

And so if you ask physicians around that, you’ll often hear things such as, uh, the physician needs to be subspecialized, which means that they have either training and or experience that is focusing on a particular area. Or you’ll hear things such as the physician should be in an environment where they’re training residents and, and you, you have countless examples of what different folks will say around what they think would yield a higher quality physician. But it’s very difficult to difficult ultimately to prove this through either clinical practice because there’s no way to ultimately objectively identify errors prior to our efforts in the market. And then to find those correlations with these sorts of things that they’re describing in a scalable way, which then requires a lot of AI is challenging. And so the, that’s a perfect example of us taking the instinct and intuition of physician experts.

Then applying that to ah the most up-to-date data science tools and uh, and models and so on and so forth to understand whether or not the things that they’re describing makes sense. And of course we will pick up other things at the same time. Meaning the big data piece sometimes happens incidentally as a byproduct of what you’re doing because we’re looking at datasets that are millions and millions of, of, of records long. So other things may emerge. But ultimately what we’re trying to understand is are the intuitions as described by these physicians, correct, which we believe them to be, but how do we really narrow them down such that we can have a better understanding of what are the specific features that drives better care for patients, which is our goal.

Curtis: And that’s a really interesting data science question and problem; I want to dig into that. But first I maybe want to just set the stage here for people that aren’t familiar with Covera Health, you guys have had some pretty big things happening recently in terms of working with Walmart and having this large network of radiology centers that are quality centers. Can you talk about what you’ve been able to achieve and then we can then get into how you were able to get there?

Ron: Absolutely. And, and the first thing to state is that there are a group of employers in the market and they’re making a headlines on a weekly basis where they’re really trying to innovate in healthcare for their employee base because they don’t see those activities happening as rapidly or as effectively through their insurance company partners. And I think Walmart is probably the best example of that, where even over the last two weeks there’s been a lot of news around various things that they’re trying to do to improve the quality of care that their employees are receiving. And so for us, Walmart was an amazing partner with respect to improving the care of radiology for their members in particular because as they roll out dozens and dozens of programs to improve musculoskeletal care, oncology care and all these other domains, fundamental to those things are radiology. Again, if you think about the patient, most patients resolve with time and maybe some Tylenol, Advil for most ailments, they go to their primary care doctor, they’re not feeling well and typically that, you know, not much comes of that.

But for those patients where additional testing is necessary and typically that is in the form of radiology, those patients are at higher risk for some complicated illness or injury. And that’s really where we help Walmart and others to improve the likelihood that those patients will be diagnosed correctly so that all those other programs and physicians who are then encountering that patient downstream have the most accurate information. So we spent quite a bit of time working with the Walmart team to both design this program, stand it up on a national basis such that, you know, we now today have access from a physician perspective in most communities across the country to service their population as well as the population of other employers. Um, and so that’s been for us in really incredible journey, uh, given their level of support and innovation in the market.

Curtis: What would you say was the biggest challenge to achieving that? Was it, uh, figuring out what data to collect or was it a data issue? Was it more of a business side? What, what was some of the biggest hurdles you had to overcome?

Ron: So there are two pieces to that, right? There’s bucket number one. How do we know the scientist working me? How do we know that ultimately improving the care of a patient at that stage is ultimately going to have some benefit to them longterm. And then bucket number two is even if you have all the evidence to show that that improvement will yield meaningful benefit downstream, how do you actually get the patient engaged with that program? And I think they’re both very meaningful challenges which we’ve been able to overcome as it relates to the first approach. For us as a data science company, the challenge wasn’t just how do we define quality?

And again, just to emphasize that really means how do we understand which physician is most ideal for each individual patient based on their injury or illness in order to produce the most accurate diagnosis. Um, and that in itself was a challenge and we had to partner with physicians across the country in order to accomplish that. But then relatedly, how do we ultimately understand that doing so is going to impact patients? And then how do we quantify that impact? And so we, we have teams of folks specifically dealing with that issue, which is understanding the errors that physicians make and understanding how those errors ultimately correlate to the treatment they receive, the, the ultimate outcomes of those patients by outcomes. You really thinking about their overall wellbeing, how long does it take them to get back to their regular activities of daily living, total cost of care, uh, so on and so forth.

Are you able to reduce the level of invasive care that they receive that may have been unnecessary? And so we spent a considerable amount of time, both with Walmart and prior to Walmart working on that problem and have been able to show, uh, in a fairly extensive way that improving care of patients at this pivotal junction in their continuum, which is radiology. It has a profound and outsized impact, their longterm wellbeing, uh, total cost of care and outcomes. And that it’s that data that we have to work with Walmart for a year or more to really understand how that applied to their population. And they were only willing, if you will, to move forward with a cohort program when they felt sufficiently that has been sufficiently demonstrated that implementing this program for such a large population would ultimately yield a meaningful benefit to, to their members.

Curtis: What would you say was the biggest, uh, maybe data challenge? Is it collecting it? Uh, you know, I imagine all this data has to come from lots of different partnerships and lots of different institutions, right? Cause not one institution is in charge of a patient’s care. There’s several that they have to go through or maybe, I dunno, electronic medical records helped in that. I don’t know too much about the space. I’m curious, what were the biggest data challenges you run into as well?

Ron: Right. So I would say as a startup individual, all challenges that we face I think are solvable. And maybe that’s obviously not true and I’m sure there are many challenges I faced that have not solved. And so when I think of those sorts of things, I kind of stack rank them. What were the most to least difficult challenges that we had? And access to data wasn’t actually the higher level challenges for two reasons. One, physicians were very supportive of our efforts. We were going to the physician community and saying, look, we’re trying to solve this really important problem that’s going to help you, one, not be viewed as a commodity to hopefully provide you data such that you’ll understand the level of care you’re providing to the patients in a way where you can then improve that level of care. And three, use that data with the payer community such that better physicians would ultimately have access to a larger patient population.

And so the physician community has been overwhelmingly supportive of our efforts and while certainly there’ve been technical difficulties with integrating with thousands of sites across the countries, those are, I would argue more of in-the-trenches type problems that we have to solve. That doesn’t keep me up at night. The the, the challenges from a data perspective that I think about is, uh, because this is such a complex problem and because even despite the success that we’ve had thus far in answering some of these very fundamental questions I see, I still see us very early on in this journey. And I think as we look at the data and have access to more and more data, the things that I worry about is that we’re not asking the right questions. And it’s very easy when one is engaging in data science and looking at, again, millions of millions of records to one, draw the wrong conclusions to perhaps not ask why those conclusions are ultimately being derived from the analysis that we’re running.

And really what I describe as how do you sanity test your results against real world input and real world experience. And so I very much work very closely with our data science team and others and our clinical experts to constantly reinforce that message. What are we missing? What should we be thinking about? Do these results make sense? And I think that’s probably one of the biggest challenges in data science is always to keep yourself centered around that, to make sure that you’re not just producing what seems to be statistically valid results, but that you producing results that seem to actually make sense above and beyond the, the methods and methodologies that you’re applying.

Curtis: Final question here. Where do you see all of this going, next five years? Where do you want to take Covera Health? What’s the ultimate uh, uh, you know, goal that you have in mind?

Ron: I think the near term goal, and by near term it’s within the timeframe you’re describing. So let’s call this three to five years is to provide to the market a, and this is my cofounder coined this term a currency of quality, some data-driven objective manner by which physicians are being evaluated that is both useful and beneficial to the physician community, such that they’re willing to accept it. Similarly useful and beneficial to the payer community such that they too are willing to accept it. And of course that ultimately drives better patient care. And then I think about my family members when they’re choosing a physician and the chaos of the selection process today or the randomness I should say and the goal for us is that what we’re doing will ultimately create a foundation that it is the standard of how one thinks about physician quality such that patients don’t have to think about very complicated things when they’re selecting their physician. They could see a Covera uh, designation, if you will, and understand these physicians are appropriate for me based on the sort of designation that they have.

Curtis: That’s awesome. I hope you guys are successful in that because it is, it’s, it’s a major problem and I’m glad people are working on it. Thank you again for taking the time to talk to us and a great perspectives. Really, really good stuff.

Ginette: A huge thank you to Ron for being on the show and as always go to datacrunchcorp.com/podcast for our show notes and attribution’s. If you’re a business leader listening to our podcast and would like to move 10 times faster and be 10 times smarter than your competitors. We’re running a webinar on February 13th where you can learn how to do this and more. Just go to data crunch Corp com slash go to sign up today for free.

Attributions

Music

“Loopster” Kevin MacLeod (incompetech.com)

Licensed under Creative Commons: By Attribution 3.0 License

http://creativecommons.org/licenses/by/3.0/