July 11, 2023
Rick LeMoine: Hello and welcome to Day Zero. Today my guest is Elliott Green, who is the CEO and co-founder of Dandy Line Health. In full transparency remember that my employer is Sharp Healthcare and Sharp Healthcare. Has had a relationship with Dandelion Health for almost two years now. Also any re-broadcast, retransmission or account in this interview without the express written consent of Major League Baseball is prohibited. Elliott, welcome great to see you. Again, even if it’s by electronic means. Elliott and I have broken bread together at least once. Elliott the way we usually start this is with a a brief not too brief review of your background education and usually stop before your kind of first work experience.
Elliott Green: First of all, thanks for the invite, Greg. It’s always nice to see you. So thank you. Thank you for having me on. In terms of my history as it were. Born in, in London, England. Hopefully that’s obvious by this point. Was educated there. Went into university, did economics and international studies. Found myself by way of a couple of gap years and lots of time in various countries around the world. Working at JP Morgan in finance doing fixed income derivatives, tradings, trading of all things. Yeah, it feels like long way from US healthcare and as you came to America the first time in 2007. For kind of what I occasionally refer to as what I thought was the end of the world, being on a trading floor when the great financial crisis was happening. One of the most remarkable experiences I’ve ever had on a, on professional level. And then left after about three years, went back to the UK and then decided, I think I enjoyed my time in the us. Maybe I should go back. My now wife agreed. And so I found myself here and was very fortunate to walk into the, an office of as Saint Mario Schlosser, who was the CEO of Oscar Health, the insurance company just dying in New York around 2012 and 13. And so it was one of the really early members there. Just a, an incredible experience. People jokingly say that, and this is relevant for this podcast for sure startups are very hard. You gotta be crazy to do. Something in US healthcare to start an insurance company, in US healthcare. That’s a different level, day one, you gotta have everything ready.
Rick LeMoine: Yeah. Yeah I remember the startup of Oscar and how it sent shock waves. Through a lot of the insurance providers in the country and was a wake up call actually for our organization who sharp has its own health plan and with a great team. But we looked at these guys and said, wow, if they make this work, we’re in big trouble.
Elliott Green: Yeah. No, I’ve got nothing but respect for Mario and Josh to kick that off in the first place. Just, just an amazing vision and appetite almost. And I think what was so exciting for us at Oscar when we began it was this idea of what if you could, what if you could really alter the way that people interact with the healthcare system? What, if you could really provide. Value and a human touch and create, the outcome benefits that we all hope are there. I think it’s such a labyrinthine system that to some degree Oscar has found that yes, it can do it in pockets, but my word that’s a tough industry to, to revolutionize.
Rick LeMoine: It cer It certainly is. And there’s a lot of inertia in terms of treasure that doesn’t want to change.
Elliott Green: Yeah. I think what I’ve learned in healthcare over, over a decade now is the funny thing is you have to follow the money more than you would like, and you start on this and you hope that you’re gonna be following patient outcomes and experiences and. Then what you realize over time is that actually, if you work out the alignment of incentives, because there are just so many players, you can often get to the answer that’s gonna get you to the right place. And I always used to, the fun stat, the fun game I play, and fun is probably a relative word in this context. You go out for dinner and someone would say, you’d say I worked for a US insurance company. Be like, oh, they’re the devil, terrible. Et cetera. Then you’d say what do you imagine the profit margin is? And I would get anywhere from 20% to 60%. The average pro, the average person in the public thought these guys were just raking. And then you’d be like, it’s three,
Rick LeMoine: Yeah. Yeah. Yeah.
Elliott Green: It’s the same as, the broker that knocks on your door or comes to your kitchen table and sells you healthcare. It’s the same profit margin. And so it just an incredible education really. And then after that, I was lucky enough to work at. A couple of unicorn kind of startups on the northeast coast trial spark and clarify health. Really like life sciences and clinical trials focused, and then much more enterprise data. And then I pause Rick cuz we get to the beginning of dandelion and all in there.
Rick LeMoine: And was dandelion your first? Place where you were holding the reins
Elliott Green: Yeah.
Rick LeMoine: and now you’re the co-founder. Watch your team there now.
Elliott Green: So there’s 17 of us. In total there are four co-founders actually. I can give you this, I can give you a story of how we all came to
Rick LeMoine: Oh, please, yeah.
Elliott Green: Yeah. So one of the great relationships I had from Oscar was actually not within Oscar. It was a gentleman called Neum Gandhi who was running Oscar. Mount Sinai did a partnership together and. Nim was running the Sinai side of things and myself and my boss, Oscar Joel Kleer, running the kind of Oscar side. And we would interact, as you imagine, pretty frequently. And Nim said I’ve met this fascinating chap called Send Malay Nathan. He’s tenured professor at University of Chicago. He’s done this and that and he’s also interested in AI and healthcare. And he’s got a research partner, Ziad Meyer. We should probably talk to them. This is interesting. And so began this kind of slow burn and I resigned from clarify just before covid. I’ll never forget that kind of late February, 2020 thinking. I’ll have a couple of months and, finish some of the deals on the table. And then of course Covid hits about two weeks later. And I think we’re moving un slide faster now. York shut down. And and so the conversations between myself, Sendal, Nim, and Z had accelerated. And what we really were trying to work out was a following. Why was AI in healthcare just so non-existent? It seemed this incredible opportunity. The world’s largest industry basically. And then one of the most exciting, if not the most exciting. So in the last six months, technology that had been around, why wasn’t this being utilized more?
Rick LeMoine: What does DA Daddy Line do and how do you do it?
Elliott Green: So we are in a nutshell a data platform with a consortium of as of now three L systems, including Sharp, your employer, and we create a healthcare data platform that is focused on clinical AI and precision analytics. What that means is we are creating what we think will be the largest, maybe even the world’s largest and highest fidelity. Training data set of all of the clinical information that is generated within those systems. So not just electronic medical records, not just notes, not just pathology slides or electrocardiograms or MRIs or cts, but everything. And some people say why? Everything? And the answer is healthcare data is plentiful. You could, was petabytes upon petabytes of this stuff. But it’s often in silos. So you can go and get electronic medical records or you could get images, or you could get ECGs, but it’s very rare that you can see a longitudinal patient journey. And so why is that important? If you really think about ai, one of the great things it can do is take an incredible amount of information and give you a relatively simple output, almost in the same way a physician does, right? Doctors do you ingest tons and tons of information. You put it through your own kind of internal algorithms, and then you spit out a diagnosis, whatever may be relevant at that time. So why don’t we use machines to do some of the calculations that are just beyond humans, and to do that, we really needed to have as much and complete and high fidelity and complete information as we possibly could.
Rick LeMoine: So this notion of data as a treasure for health providers, particularly hospital organizations now that kind of everybody is up on some kind of e m r has been around for a while, and we hear stories about how successful. People like Mayo Clinic, Cleveland Clinic have been in terms of selling their data. I’m not certain they’re all true. What’s your experience and perspective on, on, on that side of things, and then if you would go into how you think you can commercialize this? I.
Elliott Green: I think the key difference of dandelion compared to what’s come before is. A lot of the consulting companies went around and told hospital systems such as yourselves, your data is a gold mine. It’s worth a fortune. You guys are gonna work out to use it. And everyone went, that’s great. How? And I think that was where, no, it was like everyone went, I’m not sure yet, but they will work it itself out. And I think what Dandelion said was, we don’t know what it’s worth either. And actually we don’t want to buy your data. What we wanna do is we wanna get into partnership with you. Because we think this is best done from within the system, not from without, so we very much felt actually this is a case of the value of the data is combined with the understanding and experience of the people generating it. And that context was so important and so what we are doing, I think is different to what other people have done before. Instead of saying, Hey, shark, can we buy your data? We’re saying let’s partner. We’re gonna take your data, we’re gonna de-identify and tokenize it before it leaves your environment because privacy isn’t a real thing. We have to be very respectful and it is an important aspect. And then we’re gonna make it available within a cloud environment. But that data’s not gonna leave, so I’m not gonna sell it. What we’re gonna do is lease access to it, okay? Because it’s so valuable, but it’s, and Z AD often talks about this. It’s a public good to some degree. We should be using this for innovation. We should not be, I think as a patient, if someone said, I’m gonna take your data, I’m gonna allow some of the smartest minds in the world to work on it and see if they can predict lung cancer earlier than they currently can, I would be like, great. I, that’s a worthy use case. And so that’s how we are really thinking of commercializing it.
Rick LeMoine: What. What experience, what eye openers have there been in terms of trying to get information from three different sources and then putting that together?
Elliott Green: Health systems were not built to be data generating companies that would then share it. They were built around the patient and the physician and the practicalities of providing care as they should have been. And so a lot of the architecture decisions that were made with that in mind. And it’s only the last 10 or 15 years I think people have realized the value of the data, right? No, not until
Rick LeMoine: Yes.
Elliott Green: So we call it, we jokingly call it a game of finding Bob, and Bob is the person who owns the server. Where the pulmonary function tests sit, you and I have had this conversation already or where the EEGs sit or where this and why, because no one thought, wow, that could be really insightful. And so I it’s a big challenge of locating it, finding the right person, trying to work out what kind of servers they’re on trying to extract it. And then sure. To your point, once it’s actually in our environment, how do you take three very different care systems? We were incredibly deliberate about the systems. We chose Sharp in California, Sanford, in the Dakotas, Texas Health in Dallas. Those represent three very different types of places types of people, and also types of care given.
Rick LeMoine: Do they ever? Yep.
Elliott Green: To the kind of capitation and value-based care model that you guys practice compared to a lot of what Sanford and T H R probably do, right? And so that, that wide variety was very much done on purpose because what we’re really trying to do is not just harmonize the data for, that’s a challenge, but we’re trying to create a data set that’s scalable. That’s the big thing. You need to build that data set so that the algorithms and the products that are built are for everybody. I think that’s been the biggest challenge is how do we do that In terms of the data, I’m very lucky. I have a great colleague called Dr. Zd Obermeyer, who you know well, who’s pretty well class at doing this, and and he is the one who is leaving the charge to make sure the data from those three systems ends up as one harmonious AI ready dataset.
Rick LeMoine: I want to ask you I’m gonna get a little technical here and throw a couple of acronyms around that I don’t think you’ve heard of before. P H I and hipaa. Tell me about your experiences with that involving three different healthcare systems.
Elliott Green: Yeah. I think this comes back to what we just discussed. There are. Why is healthcare innovation, especially with data not being more forthcoming, right? Why do you and I both get way too many emails about synthetic data than we should do? Because there is a technical and compliance and security fear about using this kind of data. Where we started was, is the benefit that you would get from this? Sufficient that this is worth the struggle of trying to make sure that we do this. And I think the answer was a resounding yes. Like the innovation that’s possible is without doubt worth it. But the key thing, and this is really important, is privacy is an important issue. It’s not one of those things that we look at and we think, oh, this is just something we have to tick off it. It is a genuine aspect that we have to respect and be ethical around, and we are. And What we do is we identify within the three systems and you know this a great deal is their HIPAA came up with 18 different identifiers and you could either get rid of them all in that safe harbor. If you do that, you destroy a lot of the data value to the point where you can’t get the innovation. So we have this balance of how do we get enough information out to make sure that we create the innovation, but not so much that we put people under risk. And so a few key things. We get at something called an expert determination which is given by a person or a body that is seen as an expert in the field of de-identification. So the information that we take, we understand, and there’s even things like, what’s the cardiologist’s workflow look like? Is this a field that someone’s likely to write something in or not? And, you’ve gotta get into that level of detail to feel safe. And we’ve been very lucky, the three systems we work with, and you know yourself very much included. Everyone wants to work on that and make sure, because that’s what gives us the right, once we’ve done that well, to then go and do the innovation, but not the other way around.
Rick LeMoine: I have to bring up chat. G P T, ai that kind of stuff. Where do you see things like the large language mo modules being augmented with more current data so that. You could take advantage of, an individual patient’s care if you had access to the whole smoke.
Elliott Green: Such a great question. I think just to say, LLMs have been wonderful in so many respects, right? I think what’s interesting is people haven’t yet quite worked out how an LLM is gonna affect healthcare. And I think the LLM world, if I were to be very basic about this, and this isn’t all the uses is obviously involving language. So notes. Whether it’s physician, nurse, things like medically unlikely edits, things like, pre-auth. So many great use cases for it, but not for things like imaging or waveforms, which you and I have discussed because got so much potential knowledge in there that we haven’t even vaguely explored, and we’re not pretending it’s 100% accurate, but I don’t think anything is 100% accurate. So the point is, and I think this is the big change in zeitgeist, people have realized that AI doesn’t need to be perfect. It’s okay if it’s just better, and that’s fantastic.
Rick LeMoine: Can you give us a an example of without selling the farm or giving the farm away, but can you give an example for folks who are listening as to the kind of thing that they could do with dandelion.
Elliott Green: Yeah. You bring up a great point, which is just. It’s almost to, to the vastness of medical discovery. The amount of data that, that we are stewarding, I would like to say is so big that the, who are we to say where you’re gonna find interesting or correlations interesting insights. And so we’re much more of the case of, come one ka all and tell us what you’re looking for now. Maybe we’ll be able to help you and refine some of your searches or help you understand, how it may interplay or what the right variables are. But I think the key fact is, yeah, you may have a perspective on healthcare that I don’t have and therefore have something that you wanna, investigate. To give you an example, someone may come in and have that experience and be like, I wanna do that. I vs. Someone may wanna do that. Other examples that we’ve had an e c G, very common in the emergency department. Someone comes in, gets it. A ten second E cg. And then the physician has to make, and Ziad would say, this is an emergency department physician a call on, what do I do next? Is this indigestion or is this, a serious cardiac condition and he’s got a paper on this. Sometimes you can’t tell, and so use the phenotyping, but wouldn’t it be amazing if you had a data set that analyzed that E C G and gave you a really good understanding of, actually this is 96% likely that the potassium levels are very high and this person. It’s bad to have a heart attack. It’s 93% on troponin levels. They just had a heart attack. That tremor wasn’t indigestion it, it was actually a mild heart attack. Quickly get them admitted and those are the kinds of products. And so the big thing that we just did, and thank you for your helping in getting this off the ground, Rick is an EEC G validation project. Where we are then saying, Hey, not just what can you build, but why don’t we help you understand if what you’ve built is good. Because that’s the next frontier in ai.
Rick LeMoine: Elliott, before we close. Tell me a little bit about what as the kind of intended arc for dandelion over say the next five years.
Elliott Green: Number one thing is just can we get this data in the hands of as many people as possible to further healthcare innovation? I just think we’re in a golden time for that, in a way that we just, we haven’t been before. And so it’s really about how do we facilitate that so that over the next five years is everything from how do we get more people building algorithms, some of those already in the healthcare industry, some of those who don’t even exist yet. You and I have discussed could we do it Sharp and Sanford and Texas Health in terms of getting these algorithms into responsibly into care. The CMO at Sanford very kind wrote to me about the ECG validation project and said, This is one of the most wonderful things to happen with an EKG in 50 years. And those are the kind of emails you get as a ceo, and you’re like, wow, I wish I could blow this up and put it on my walk.
Rick LeMoine: Elliott, it’s been a pleasure to do this just as it’s been a pleasure to get to know you and to work with you. Good luck you’ve got. Incredible team with you and I look forward to your future success.
Elliott Green: Rick, thank you so much and thank you so much for everything up to now in the future.