Eon Webinar

Eon Expert Webinar Series: Eon + Mark Cuban, Gerard Silvestri, Peter Mazzone & George Eapen

It’s not everyday you get to sit down with a tech tycoon and three of the best minds in pulmonary medicine. Listen to some of the challenges doctors encounter and how each of the stakeholders involved in healthcare (industry, providers, insurers, pharmacies, the government, patients) can be part of the solution.

Synopsis

Synopsis

Mark Cuban & Top Pulmonary Doctors Discuss Healthcare Technology Opportunity at the American Thoracic Society’s Annual Conference

Healthcare Technology Company Eon hosted a rare opportunity to challenge current healthcare technology, analyze its shortcomings and review pragmatic solutions.

Dallas, TX Release Date: 6/05/2019, For Immediate Release

Eon co-CEO’s, Dr. Aki Alzubaidi and Christine Spraker, sat down with Mark Cuban, Dr. Gerard Silvestri, Dr. Peter Mazzone, and Dr. George A. Eapen for a live webinar hosted May 20, 2019, at the American Thoracic Society conference in Dallas, TX. The panel met to discuss healthcare technology issues that have led to patients, health care providers and hospitals facing an ever-widening gap between what technology can do and what is currently implemented.

Eon, a complex patient management technology company out of Denver, Colorado, is dedicated to creating sea-change in the healthcare industry and is determined to use advanced technology to improve complex patient management of every kind. Technology has improved processes and is now less expensive in every other industry – why not healthcare?

The webinar addressed issues in healthcare technology, such as the cost, the failures of EMRs, the siloing of data and misaligned incentives, as well as solutions to mitigate these issues. Eon launched the discussion with a question on how providers, as well as hospitals, can educate themselves on what technology should cost and why it has been okay for the industry to overcharge for technology that is now fractional to develop.

While some healthcare technology companies repackage and repurpose technology from the 1990s and sell it to hospitals for hundreds of thousands of dollars, it’s not surprising technology now contributes 40-50% of the annual cost increase in healthcare. As Dr. Eapen points out, it’s not necessarily the cost of the software that is the issue, but what the value is. He says, “It’s really not about driving down health care costs. Something costs what it costs, but if it provides value, it is either worth it, or it’s not.”

The physicians on the panel encountering these challenges were able to pose questions about patient data and the utility of EMRs from a user standpoint. Cuban, as an entrepreneur, Eon investor, and Artificial Intelligence (AI) aficionado, brought a unique perspective to the conversation and allowed it to naturally move in the direction of siloed data by both hospitals and insurance companies.

“Stop siloing data,” said Cuban. “My perception of it [referring to doctors using electronic medical records] is that you try to fit whatever you can into the square and round holes that are already there, and that’s not necessarily optimized for process.”

As the technology gap continues to grow, so does the understanding of what AI in healthcare actually means for clinical utility. Google recently published a study in the journal Nature Medicine titled “End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography.” The researchers trained a Convolutional Neural Network (CNN) on a publicly available dataset from the National Lung Screening Trial (NLST). The researchers compared their model to six thoracic radiologists and concluded that their model outperformed the radiologist. The study was subsequently picked up by the New York Times in an article that proclaimed “AI Took a Test to Predict Lung Cancer. It Got an A.”

The lack of understanding in AI becomes apparent when you see proclamations like this, and it’s important to understand who is looking at the study and analyzing the results. With data from 578 patients, and only 70% of it being used to train the model, the potential for overfit and incorrect diagnosis grows as you apply the model across larger populations. A meaningful AI model for true computer vision in lung cancer will require millions of labeled data sets. And while the hospitals procure labeled data sets, the value of de-identified data and the data silos themselves have made it nearly impossible for researchers to create models that result in clinically significant results.

To build on this point, Cuban added, “One of the biggest jobs in healthcare is going to be labeling data. You want to have a data scientist on staff that you are going to be able to go to and say, I’ve got this data, what can we do with it?”  Hospital systems have limited teams dedicated to understanding and developing deep learning models from their data sets and should partner with industry to open the repositories and create meaningful AI.

“What’s happening now is that [healthcare companies’] marketing departments are selling AI, right?” poses Dr. Alzubaidi. “What they’re really selling to these guys is a software that can do what is called segmentation, CAD [computer aided detection], where you take a CT scan and then reconstruct a pulmonary nodule. They’re calling it AI, and it was built in 1990.”

Doctors essentially have to fight for new technology in budget meetings each year. Unfortunately, technology companies make big promises to solve clinical problems, but more often than not, they do not deliver pragmatic clinical solutions to providers. “Peter and I write a lot of guidelines for the “how to’s” that say we’re going to solve it. But my god, we know that 50% of the time doctors don’t follow the guidelines because it’s just not easy enough,” stated Dr. Silvestri. He continued by posing the question, “Most physicians aren’t writing their own software, they’re trying to take care of patients. How do we bridge that gap?”

Dr. Mazzone adds “I couldn’t build a software program, I couldn’t do data science. But I know what’s necessary for clinical practice and so I want to be able to partner with those folks doing that good work to ultimately help the patients.”

The importance of translating verbal and interactional patient data from experiential to sequential is the answer. What doctors will be able to take from properly developed natural language systems is the ability to translate interactions with their patients in their entirety. This will allow for data that may not have had previous labeling to be identified and tracked for potential use later. More information will be collected, aggregated into EMRs, and used for enhanced diagnoses and tracking.

Data is the main driver of current technological change. Current healthcare systems are logic based, “if this, then that,” and do not necessarily allow for the predictive needs of healthcare. Now, companies like Eon use machine learning to approach and anticipate needs. Cuban said, “Platforms like Eon allow you to aggregate data, learn, have the data teach you, have the data become more predictive. So that you can be smarter.”

Irresponsible tactics used to develop and sell outdated healthcare technology has warped the value of new software development and created frustration for all healthcare stakeholders. As the industry leader for complex patient management software, Eon is changing the way the healthcare industry views technological advancements and is improving patient care, all while lowering costs for users across the country. Patient management and outcomes can be improved, saving time, money, and most importantly lives. There is no time for expensive, out-dated technology. Together we can defy disease.

Paige Ruppert

paige.ruppert@eonhealth.com

Eon

400 South Colorado Blvd.,

Suite 380

Denver, CO 80246

Transcript

Christine Spraker:

Welcome to the EON Expert Webinar Series, we’re broadcasting live from the Mark Cuban Company’s home office in Dallas, Texas.  We’re going to introduce everybody in just a minute, I’m Christine Spraker- co-CEO of EON.

Sitting next to me is the founder of EON, co-CEO as well, Aki Alzubaidi. Aki, you’re an interventional pulmonologist, you founded Eon just to solve your own clinical problem that was a result of inadequacies in healthcare technology.

Dr. Aki Alzubaidi:

Yes that’s right.

These two guys here (points to Gerard Silvestri and Peter Mazzone) came out in 2013 with the ACCP (CHEST) guidelines for managing incidental pulmonary nodules, which is just an abnormality that can be found (pulmonary nodule) and that could be cancer. I wanted to just follow their directions honestly. In 2014 I went down to my radiology department and I said send me every single incidental nodule so I could manage them per the evidence based guidelines and that was impossible.

I tried to fix this problem with my EHR, with Excel spreadsheets, with hanging file folders. And then I went out into the market trying to see if there was a solution but there weren’t any.  And so I met up with Christine and found out that this was a broad problem and really just said why not me. And so I decided to try to solve this problem on our own.

Christine Spraker:

I’m so glad you did. Do you want to introduce everybody?

Dr. Aki Alzubaidi:

So, Mark Cuban, in terms of leaders of digital technology, I always knew that you were going to be the guy that would help health care in bridging the gap in what technology could actually do and what was actually implemented in healthcare.

And so I emailed you at all of your portfolio companies by looking at the email structure, and you finally wrote me back five years ago on June 25th, 2014.

Mark Cuban:

2014? Wow 5 years ago.

Dr. Aki Alzubaidi:

In that email, you wrote “wish I could help but cant you find a local software developer you trust” and then you went ghost.

But over the years, I kept giving you updates and honestly, thank you so much for investing in our company.

Mark Cuban:

You didn’t give them the pay out. You just left it at ghost. [laughter]

Dr. Aki Alzubaidi:

You told us what to do to be in. And I just wanted to say, I’ve been waiting a long time to tell you, thank you so much.

Gerard Silvestri is the ex-president of CHEST – American College of Chest Physicians, and is an excellent scientist. He practices at the Medical University of South Carolina (MUSC). I started emailing Gerard (I like math) and so started emailing him about predictive modeling back in 2014 or 2015. I had sent him so many emails, and he wrote me back and said I have this goal,  and that goal is to know if a pulmonary nodule is cancer or it isn’t and thankfully he decided to work with us.

Gerard, you’ve been a huge influence in my life and I wanted to thank you too – thank you for being here, Gerard.

Peter Mazzone is at the Cleveland Clinic and is really responsible for creating evidence-based guidelines. Peter you are a wonderful human being and you’ve been excellent to me. You also have a goal to understand what abnormalities actually are. You have been the best doctor to us over the past four or five years and have really helped us get it [EonDirect] to this point. Thank you so much for being here as well Peter.

George Eapen is from MD Anderson and is the incoming president of the American Association for Bronchology and Interventional Pulmonology (AABIP). I learned yesterday that George is in the Army Reserve and has been deployed, and I just wanted to thank you formally for your service. I also learned a lot about how there’s a need for doctors to go into the Army Reserves and just wanted to give this a quick time to say that if you’re a physician and just coming out of fellowship, its the best time to do this, sign up for the Army Reserves.  I can’t believe that you do that, being a physician, and everything that you’re doing. Thank you.

Christine Spraker:

Thank you so much, now let’s talk about today.

We divided the webinar into two sections. The first half we’re going to talk about the issues in healthcare technology: the cost, the failures of the EMRs, how IT departments within hospitals are really paralyzed because of technology, and then finally misaligned incentives. On the back half we’ll dig into some of the steps we can take to mitigate these issues.

So we always talk about how healthcare is unaffordable, and Mark you’ve been really outspoken about the payer system and how it contributes to the cost of health care.

Another big issue that’s contributing to the cost of health care is healthcare technology, which now contributes 40 to 50% of the annual cost increases.

Aki and I see it all the time where technology from the 1990s is repackaged, repurposed and then sold to hospitals for hundreds of thousands of dollars, where if we created it today, it’d be fractional.

So our question to you is: how can providers on the panel as well as hospitals know what technology should cost, and then why do you think it’s been okay for industry to overcharge for technology that is now fractional to develop.

Mark Cuban:

That’s a loaded question!

Just to give you a little about my background. My first company was a systems integrator that wrote custom software back in the ‘80s believe it or not. I did work for hospitals and probably overcharged a few back then. And the nature of it hasn’t changed – it’s hard because when you’re writing software, developing technology, you typically have a goal in mind that you are trying to solve and trying to get all the pieces in place. So that not only you’re writing the software so it works but that it integrates to what’s already there. And works and operates with the people that are going to be using it.

It’s hard and I think we’re going through a transitory period right now. In 2014 you didn’t really have a lot of conversations about machine learning, you didn’t have a lot of conversations about AI because the processing speeds weren’t quite there yet. Video was doing its thing with GPUs, but it was just starting to introduce it into technology, making itself available or being used by technologists, that could solve problems using AI.

We saw this in the early days of machine learning, now we’re extending AI to so many different areas, including what Eon’s doing, and so what’s changing now is IT, the role of IT is changing dramatically.

So it’s no longer IT. The role of data is changing IT dramatically. It’s no longer just about writing software, solving a problem, but really creating platforms like Eon that allow you to aggregate data, learn, have the data teach you, have the data become more predictive. So that you can be smarter. In the past it was really about logic driven, you know: if this, then that. And so that’s really what’s changed and so it’s difficult for hospitals to evolve because you don’t turn over your IT people or think you’re going to change your IT every six months or every two years and even though you know the nature of technology is to change, seemingly, daily. It’s really changed dramatically, and dramatically for the better. So providers of any sort are going to have to start recognizing that capturing data and being able to understand how to utilize that data is going to drive as much of what you do as the processes that are used to automate changes in software.

Dr. Gerard Silvestri [8:40]:

Can I just follow up and ask a question?

I hear all of this, and I hear the AI piece, machine learning, and I do some of that. My question to you is how do you take it from the system’s level to the individual physician. When I go on my smartphone and order something, it’s all there and then I don’t have to think about it. I think that’s where we are struggling. How do you do that to have me help manage a patient?

Mark Cuban:

So first of all you stop siloing data. Right now, because of the insurers networks and the way they’re designed, the complication comes at administration first. So when you’re, and again I’m not a physician and I don’t deal with this on a daily basis, but my perception of it is that you try to fit whatever you can into the squares and the round holes that are already there, and that’s not necessarily optimized for process. That’s part one, that’s the administration part.

Part two is the decision-making part of it, right? People talk about evidence-based medicine, people talk about using your experiences, doctors just leveraging their experiences, and that’s really an aggregation of data.

The third portion of this is letting machines calculate something using AI, using machine learning, so that provides a third input. There are two discrete parts, so when you talk about ordering it’s an administration function. You know that’s going to happen, that is an IT driven issue. In terms of implementing care, that’s a data driven aspect and I think that they are separate if that answers your question. And if not, tell me.

Dr. Gerard Silvestri:

I think it does. The problem is getting these two people together, and then putting it in a usable format. Again, most physicians aren’t writing their own software, they’re trying to take care of patients. How do we bridge that gap? Peter and I write a lot of guidelines for the “how to’s” that say we’re going to solve it. My god, we know that 50% of the time doctors don’t follow the guidelines because it’s just not easy enough.

Mark Cuban:

It’s hard just taking the time to learn the guidelines, let alone implement them. It’s difficult. You know I was just having this conversation yesterday with an epidemiologist, if I can say that right. I think we’re going to get to a point where, hopefully sooner than later, where you’re going to be using a device to capture your natural language, right? Then there will be natural language processing that will parse all those things, distribute it and disperse it into the feature that you’re looking for. So, if you’re treating a patient and you’re just talking naturally, it’s recording what you say to the patient. Obviously, you can take out the patient’s name and there are a lot of privacy things that you have to take care of. Then that becomes kind of your play-by-play, to use a basketball analogy, of everything that you’ve done on a step by step basis. Then if you take AI natural language processing and your guidelines and apply that to just what was said, then all of that information not only becomes your personal experience that the patient gains benefits from, but also becomes inputs to all the data sources that need to happen. In the past, as a programmer, what we would do is we’d say, okay, we’re creating the database. Each database has records and each record has fields in it. Then, we’ll determine based on what you tell me, and five different physicians will tell you five different things that they need for their practice, but based off of what you tell me, we will create an application that accommodates those things. What’s changing is rather than being database driven like it traditionally has been, ie here’s what you told me you need and here’s what you’re telling me now. I’m now going to give you an output, based off of what you told me, that will be a lot more AI driven, natural language processing driven. You’ll be able to take anything that’s been said by ten different doctors, even accounting for regional differences and specialty differences, and translate those words into data that can be filled in. While you’re talking and saying I need to order X, Y and Z, it will process that and recognize the word order, and record every word you train it to, and then respond accordingly. Almost like Alexa right?

Dr. Aki Alzubaidi: 

But we’re so far off. What’s happening now is marketing departments are selling AI right? So what they’re really selling to these guys [gestures to panel doctors], we all deal with pulmonary nodules, so there’s a software that can do what is called segmentation [computer aided detection or CAD] where you take a CT scan and they can reconstruct a pulmonary nodule. They’re calling it AI. It was built in 1990, straight up, right? And now it’s open source and I could probably build the whole thing for less than $10 grand and then you know, sell it for a hundred dollars. They’re selling it for $150K. What you’re talking about isn’t there yet. How do we educate doctors, physicians, providers and hospitals to not buy shitty technology from 1995?

Mark Cuban:

It’s outcomes. It’s purely outcomes right? Because if the shit’s shit from the 1990s it’s not going to give you the outcomes that you’re hoping for, and then the beauty of data is that you’re tracking outcomes. That’s what you guys are doing – right? Because you know you’re able to be predictive, right? You’re able to be proactive. That’s the big difference from the old school programming that was done in the ’90s by guys like me that is all logic based. If this, then that. And that’s the difference. That’s what differentiates. Logic based is completely different from what AI and its derivatives are doing.

Dr. George Eapen:

I have a question for you. You talk about outcomes, and I agree with you I think it all comes down to outcomes. How as individuals with all of these platforms coming in and promising us certain things, then there is a hope that they are going to do this or that, and a lot of times when we try it out, it turns out to be hype? What we’ve seen so far, and I’d like to get your thoughts on the solution to this, is if it’s difficult for me to use it, it’s difficult for him to use it [points to Peter Mazzone] but it’s easy for him to use it [points to Gerard Silvestri]. How does that get aggregated up so that the ones that are value added for the greatest number of people gets to the top.

Mark Cuban:

Yea, that’s the question of all medicine, right? This drug versus the outcomes verse not testing that at all. A lot of that comes down to the resources you have to test right? And in some cases you’re winging it, right? Which vendor do you trust? You trust Aki and Christine to get the right outcomes. That then builds trust and that trust extends into other locations and that’s the reality of software today. Then, in doing your research, you know that when you read research on specific medical problems you look at the peer review in a lot of respects because if a vendor writes something you question it just inherently. In software it’s the same thing right? If a vendor is writing something, I always question it because no vendor is going to call themselves an idiot. But if you ask for references, and then go out and research those references to see what the outcomes are, then you’re going to really know. Now the challenge is your time constraint, right?

The second challenge is that the nature of doctors is to find all of the information themselves, right? So you’re always looking for shortcuts to try to figure it out yourself. I used to hate to write software for doctors, hated it, because they knew more than I did and they would tell you [laughter]. And so that’s part of the challenge, but again I think in decision making now you want to look for platforms that you know will evolve. And when they’re using old school software written in old school ways where it is logic driven then you know you’re going to have challenges because you will always have to update the software with the program.

The nature of AI isn’t so much that you’re changing, it changes because libraries [data libraries] and research change things, it’s a lottery ticket for neural net, whatever, but at the same time it’s data-driven right? You want systems where the data makes the system smarter as opposed to programmers being required to alter the system to make it stronger. That’s how you tell the difference between a system that’s going to be a foundation for your future, verses not.

Christine Spraker [17:01]:

So, that’s a great segue so we can move into EHRs [electronic health records]. Part of the problem with traditional EHRs is that all of the data is siloed. When you’re talking about AI and data we’re assuming that it’s all easy to access, right?

The problem is that it’s not currently [easy to access]. It’s held hostage in these silos across every hospital system and you have the same EHR as another hospital but they still can’t talk to each other, right? So that’s a lot of what you guys are experiencing. The technology is there, the NLP is there. We can do all of that today, it’s integrating it into the workflow that is a challenge.

Mark Cuban:

Part of the challenge is insurance companies right? Everything is driven by a different contract and every year the contract changes, whether it is for an employer base or ACA or Medicare. Maybe with Medicare/Medicaid the contracts don’t change nearly as often, but that’s part of what you need to talk to your representatives about. When I work with HHS, with their blue button program, that’s a big part of what they’re trying to do, to minimize the number of contracts to work with.

If you’re in an environment where you have to renegotiate contracts every year to determine your networks and determine how everything works, you’re already against the eight ball because you’re siloing to the insurance companies. Then you have to fight to get it back. I’m just not a big fan of how the insurance companies are dealing with those issues right now.

Christine Spraker:

What about hospitals who have invested a billion dollars into a medical record system that is failing them?

Mark Cuban [18:3-]:

You know, shit happens. I’ve got friends that are CEOs of big hospital systems, and if you go back to the history of technology and healthcare, there’s always inflection points. And during those inflection points, or at those inflection points, there’s disruption and there’s difficulty.  One of my first jobs out of college was an internship where I was converting banks from physical systems to digital systems and it was disruption. Hospitals want to do the same thing. How do you go from physical, paper driven systems to digital, and then how do you integrate networks? That was a big deal.

That’s what my company did, hooking PCs together so now PCs can talk to each other, that was a big deal. Then you got to the cloud and that was a big deal. So this is an inflection point right now where data is your best ally, and part of the transition isn’t just about getting rid of old software systems but designing new systems and enabling systems that don’t silo data right? But instead can create data that can be used elsewhere. But that’s part of where you’ve got to be.

Dr. George Eapen:

Let me ask a question regarding that. I think you just very well articulated it. Part of what I see with the medical record, or the electronic medical record, is it means different things to different people. It could be a record of things that have been done.

You’re describing a much broader picture of how it can be used to learn the future and improve things. When we set up designing these platforms, a lot of times these platforms are designed before, like we talked about the banks. Well, all of their software is designed around making them money, so they have a mission and they understand that mission and everything drives towards that.

For us with their medical record, some of it is set up for billing purposes, which is what you talked about with insurance. Some of it is set up so that we can communicate with each other and with our patients. Then some of it is set up to learn, but this is very minimal. How do you see this overarching mission as to what the medical record ought to be? How do we agree on that?

Mark Cuban:

Well that’s what’s changing, right? That goes back to what I was saying. Traditionally with EMRs they are database-driven, and for the most part, relational database driven. I’m a little bit out of date on some of my technology there but it’s like what I was saying earlier. Someone designs it, you write software that reflects that design and then you try to keep it updated as new things come up, right?

That’s what becomes outdated and so for right now, this day and age, it’s hard because you’re running into that conflict that the only data you really have access to is the way the program was designed with the logic that is behind it. And so they’ve tried to integrate it and extend it into as many places as they can so you’ve got the foundation of patient information which then drives codes, which then drives billing, and then the billing doesn’t really need to be shared outside the payer, but with the patient information there is a way to send it to a cloud so that other people can retrieve it, and then those other people can use it. Our bodies are basically one big math equation, right? And there’s X number of variables that we’ve been able to identify but once it gets to another doctor there may be a whole other variable or there may be another element that’s important in that circumstance that wasn’t covered or wasn’t even considered, right? So anticipating nodules is different than a database driven environment where it is directly recording what’s already been identified, and that’s what’s changed, right?

So that’s why companies like Eon are starting to really take off because they’re going from logic driven to data driven. And logic driven, meaning predefined with huge IT departments dealing with change requests from doctors and administrators, to now being in an environment where the data drives everything. So if I introduce new data, immunotherapies. If you have a system that was designed seven years ago, you didn’t incorporate immunotherapies in any type of your considerations at all, right? With the new system, the more we get into CRISPR and there’s an altered gene, or here’s your genomes that this patient brought in from 23 and Me, and it says they have this gene. And with the gene, you know that we’re gonna take a look because you’re more likely to have X, Y or Z. Well, now you start getting into data-driven systems where the data is the repository and it can grow. It kind of takes on a form of its own and then you later, on top of it, take extractions from that data.

The data drives everything and that’s why when you talk about AI, they call them neural networks because it reflects how we approach problem solving and thinking. Now, we’re still only at the stage of maybe a two-year-old, if we’re lucky, in terms of logic. But that two-year-old knows enough not to put their hand on the fire. He’s not doing that a second time, that’s what’s changing.

[23:45] So the challenge is that you’re going from an area where there are incumbent businesses who are trying to, to Aki’s point, to tell you we’ve got this covered, right? We’re just going to layer. We’re just going to extract data from the database, and they can to a certain respect. If you extract all the data that you have and position it in a way that you have open access to it, okay. Export it, put it up in the cloud, let some logic deal with extracting the data and then you’re putting that into a new format and then letting your AI deal with extracting what’s important and tell you what you’re missing. But you need a company that is built that way and thinks that way and that’s the challenge.

So it’s not the best news and if we were having this conversation five years ago, I wouldn’t be able to say the exact same things and that’s what’s changed. So what comes next, the real foundation of why we can have this conversation, is processor speed which is going like this, we’re on that hockey-stick of processor speeds and the ability to not only take one processor but a variety of GPUs and connect them together to do things that five, seven, ten years ago would have taken five hundred hours, or five thousand hours, we can do in 50 minutes. That’s what has changed. Now if you look at the number, there’s a website Archive,  it’s like roman numerals [arXiv.org]. There used to be a hundred papers a month about AI, now there’s thousands.

It used to be about, okay now you just aggregating big data. Now it’s about how do you create predictions with less data, right? How do you build a neural network with less processing speed so that your IT department can deal with it and that’s why data scientists are becoming so important. That’s why labeling data is far more important. One of the biggest jobs in healthcare is going to be labeling data. This is a water receptacle, right? This holds water, so does this, and being able to recognize different data types and be able to recognize what something is. Now that they’ve got something called generative adversarial networks, that allow you to take the output of one neural network, and have it processed by another and have it tell you what it would it see. Is it a cow or cat?

There is so much that’s advancing so quickly right now. It’s almost like the old days, when I was programming, where there’d be new programming languages but then there’d be process speeds that were faster and there would be faster and larger hard drives. So when I first started programming, a 20 megabyte hard drive from IBM for six megahertz processor, that IBM PC costs $5,000.

You guys remember those things, right?

That’s what you dealt with and that was a huge step forward. I remember having to go get training for Lotus 1-2-3 so I could sell it as a spreadsheet for $500. That was a great step forward. You could say what if I did it this way? Now, the difference is when you have it, you can source on these unlimited amounts of data. What AI will do for you is examine patterns in an unlimited amount of data. Where in the past a physician, a doctor, surgeon or whatever it may be just had the benefit of the experience and the people around them to look to identify things.

That opens a door for thousands of options and that’s what’s changing dramatically.

Dr. Aki Alzubaidi [27:12]:

I don’t disagree with you but, let’s talk about the data. First of all in these EHRs, let’s just take Epic for example, they have 500 different implementations where there are variations on all the data. So I agree with you about the need for translation, right? A cup has to be a cup. We’ve all gone through this, especially in the Watch the Spot Trial, where there’s a data codebook. And then across 26 different sites and the variations in what those data fields are labeled as, and how they’re labeled, and what they actually mean needs to be translated to get good insights [because they are all labeled something different.

Now, Peter, you and I tried to do a computer vision product, a computer vision experiment, and what we found was (and I don’t if you remember but Google had a Kaggle $1 million dollar lung nodule model contest, right?) and so we used the same data that was used by the Kaggle data scientists and also worked with Peter and what we found was 1) theres so much variability in ground truth, right?

Mark Cuban:

In terms of the labels themselves?

Dr. Aki Alzubaidi:

If they’re labeled. And they live in these different places. To get true clinical value from AI, the number of ground truths and curated data has been grossly underestimated, right?

So it’s kind of like we need to create an alliance among foes to try to actually get to the point where AI actually brings clinical value.

Mark Cuban:

That’s a different issue. So what you’re talking about is domain expertise, right? So if you don’t even agree on what something is, what you called your ground truth, right? Then it doesn’t matter.

Dr. Aki Alzubaidi:

No, what I’m saying is that we could agree on ground truths, but right now the data sets that we have are like you said.

Christine Spraker:

So for example, a ground truth is is it cancer or is it not. But right now the radiologists will say it looks like it could be cancerous and then they were using that as the label as the ground truth in the Kaggle study.

Mark Cuban [29:14]:

But that is domain expertise, right? Versus CYA, right?

Dr. Peter Mazzone:

I think what’s exciting though is what you just said there. How do we get clinical benefits from AI or these other technologies? We felt 10 years ago like technology would show up to solve a problem that really didn’t impact us, whether it was billing or whatever it might be, it was necessary. But now I feel like we’re more at the table. We’re able to say what do we really need and we hear how people with expertise using these solutions can help us. When the solution is made without real clinical knowledge, it’s going to fail us. Do you see that happening more?

Mark Cuban:

That’s what they were talking about, right? So if the radiologist looks at the MRI or whatever it may be, and says I’m not sure, right? Then whoever is responsible has to go through and look at all the other data and then you take your best guess right? So when you label data, that becomes part of what feeds into your, in this case machine learning, right? Where you want somebody. So what I think is going to answer your question is that you’re going to have radiologists who are able to identify nodules, go in and your resident radiologists will spend a lot of time labeling, and going through verifying the veracity of the data sets that you have, so it feeds the data in a way that solves the problem that you’re referring to.

Dr. Aki Alzubaidi:

Labeling is one problem, but then the second problem is commercial interests.

Mark Cuban:

That’s a battle that you and I talk about all the time. How do you overcome the incumbency right? Because when somebody puts their butt on the line and says I think this is the right solution they’re not all going to all of a sudden turn around and say yeah I was wrong, right? But that’s where you get forward thinkers, right? Who are willing to bring you guys in and say you know what, we’ve recognized that lung nodules is a specific domain expertise that when it comes to AI, machine learning, however you’re approaching it. You need somebody who recognizes it. And fortunately, even with the incumbent systems, they are not mutually exclusive, right? As long as you can source the data for what you need you could do it side by side. That’s the good news. That’s why I invested, right? Because it can be done side by side and you can demonstrate results. It doesn’t mean that big guys that are having this conversation aren’t going to try to squash you. That’s their job right? But what did Clayton Christensen do. That’s the innovator’s dilemma. Where it’s just going to be that people want to keep selling what they’ve been selling and that’s always going to be a challenge. But that’s the challenge of every start-up.

When you run with the elephants there’s the quick and the dead. And that’s why you guys are quick. Once you can apply that domain expertise, and like we’ve talked, extending it beyond just lung nodules because now you’re able to take that same approach. So data is the new oil and when you have domain expertise and work with these smart people who have domain expertise, and you can start to define data sets. Then you can start expanding those data sets. There is a point of diminishing returns but your systems get smarter. That is the direct opposite of the traditional legacy systems where they give you databases of what you’ve done but they don’t really get smarter. They can make you a little smarter by telling you what’s already happened, so your knowledge improves, but the systems themselves don’t make you smarter and I think that’s what you’re saying.

Dr. Gerard Silvestri [33:21]:

Can I just jump in here. So you may not know this but they sent me your entire background. So I knew then you were selling paper. So what I got out of it is, and you just sort of articulated, that I need to ask these questions. We are in these purchasing rooms. We go in, we have to beg if you will, they have a certain amount for this year’s technology. You go in, you can bring in Aki or  not, you can tell them this is the need. Then they’re looking for the return on their investment. But the thing that is impressing me so much that I need to ask you how you got through? Because I think of all physicians, which is this. There seems to be a generational difference when it comes to technology, right? So people were scared to take a chance. Look I’m sitting here talking to you and you could sell me anything right now and I would buy it. But the truth of the matter is that I can’t do that when I am in a technology space. And there seems to be this technological, this generational, difference. And you’ve had it in your businesses coming through, the newer businesses, where it actually extends to the research arm of this. For example when we write grants, Peter and I’ve written grants around AI. We had reviewers of those grants at national levels say we just don’t really know about this. You have to show us the stepwise approach. You’re not able to tell us why this works even, right? Because the machine is teaching itself. So how you get through that generational gap?

Mark Cuban [34.49]:

I am 60 years old. I’m old! Well I guess 60 is the new 40. The new 30 is what I tell my wife and kids. I have to learn too. If I’m going to be a real geek I’ve got to be able to geek out right? So that’s true I’ve got to be able to talk to the kids. I’ve got another healthcare company that takes the output of every organ in your body, which emits an electrical pulse right? Their sensors capture that and they do cardiac analysis. They’re able to tell you within 30 minutes if it’s one of five different cardiac diseases with 97% accuracy. Well how do I know to invest in them and these guys? Tutorials. I literally go on Amazon and I do the machine learning tutorials and it’s really straightforward. I go into how to build a neural network with JavaScript or Python. I took a Python tutorial. The thing about technology that I’ve told myself, and tell my kids, and everybody I have worked with, is that there’s the person or people who invented it and created it and then there’s everybody else. You’re tied for second place the day it’s released, and that I could keep up with everybody else if I just put in the time. Now I’m not a doctor, I don’t have to deal with all the medical stuff and all that, but I just put in the time. So when it comes to down to now being able to explain it, because I’ve had a history of all that technology, I’m able to give it context. But you’re right, even kids right now, if you talk to Gen Z or Millennials, they don’t know this stuff either because it’s still within the last five to seven years. We’re literally, here and in one of my other offices, have taught classes to underprivileged high school kids on machine learning. How you can use a spreadsheet, and what machine learning is, because AI is going to change everything. They’re telling those kids, not to get out too much on a tangent. They’re telling these kids that if you’re a coder you’re going to be okay. In fifteen years, math is math right? The system is going to do the math. Programming is just math. It’s just logic driven math, and that’s going to be done by systems. Here’s what I want and I’ll just say it. The NLP will parse it and convert it to logic and boom you’re out of a job. So understanding that it is a process but actually it’s like every other tech I’ve had to learn. It’s not that complicated once you spend some time with it. Machine learning is like a big spreadsheet that just fills in some blanks. I’ve got ten thousand records that tell me here are the five features of a plant. Well I’ve got 900 records where I’m missing one feature, what kind of plant is it? Tell me what kind of plant it is, that’s machine learning. That’s all it is.

Dr. George Eapen:

I have a question to ask, if you don’t mind? It goes back to what you talked about with data. I’d like to hear your thoughts as to who owns the data? Patient level information clearly, to me and this is my opinion, belongs to the patient, right? It’s my name, it’s my data, so I’ll be able to take it wherever I go right? The predictive analytics…, so the billing component, probably belongs to whoever is doing the billing. But the predictive analytics, whatever platform somebody has created and this neural network and has learned all these things that has used this data to learn that. How do you prevent the natural sort of siloing that happens when you’re looking at a proprietary formula of some kind?

Mark Cuban:

Well an individual’s data isn’t of much use to a bigger picture, so you can anonymize it and that’s fine. So if you feel like your privacy is being violated, you can extract it and you can take it. So a) the patient should own all of the data. B) then it becomes very, very valuable. I’ve gotten into big arguments, I’m a big believer that you set your own personal benchmarks through testing. So I want to own all of my personal data but when it comes to training data it can be anonymized. It doesn’t matter who’s data it is, if its Aki’s, George, Mark, whomever. So that part really is safe. Now where it starts to get finicky is if, when you walk into a hospital and they are using facial recognition that’s so advanced it knows to compare the skin tone of your face to the last time you were there and says oh this person might have a fever. Somebody go walk up and talk to them, and oh by the way, they weren’t a patient, they’re the aunt or uncle of the person who’s bringing in a patient. Then we start crossing lines that are going to be more difficult. I think that will happen in hospitals from a security perspective, a proactive analytics perspective, because you want to know if there is a 78% chance that the person who you don’t even know who they are because they are bringing in a patient, that there’s a 78% chance that yeah they have a fever and it might spread. That’s where it starts to get finicky.

Dr. Peter Mazzone [39.59]:

The other question that goes along with that is what is the value of that data? Health systems are working on small margins now and looking to say how much value is there to this well-annotated data set, what should I be asking for in return?

Mark Cuban:

Let me give you an example. With this company Genetisis, it’s kind of like what Eon is doing in some respects. I gave you the example where they use a sensor to determine one of five cardiac conditions, right? Guys our age, you feel something in your chest and you freak out because it’s the big one, right? You don’t know what it is until you start to get comfortable with whatever discomfort you had. So what steps do we take? Get an EKG, get an ECG, get the sonogram thingamajiggy right, do whatever nuclear test you do. All these different things, not to find out what you have but to exclude what you have. Maybe you have a-fib and it’s easy to tell, right? Well with this, 30 minutes and you don’t have to do all those different tests. And it’s because they were able to train with 10,000 examples to know that when you put somebody, and the sensor that extracts an audio file, essentially, of the electromagnetic field. That’s being reported here. You’re going to know with 94 to 97 percent accuracy which one of the cardiac conditions you have so you don’t have to take those other tests, right? You don’t have to sit in the hospital. That’s where you start to get the value of it. So it’s like Eon, right? You get to be predictive, right? You need to go back and check with this person.

Dr. Peter Mazzone [41:33]:

So that’s what I’m asking. Of those 10,000 examples, what was the value of those examples? Should the hospital system be charging a lot to partner with the developer of the tools, at least that you’re describing, to have a separate version?

Mark Cuban:

Data gets its value from using AI whether it’s machine learning or neural networks. Machine learning is more linear, this or that. Neural networks are nonlinear, so they can tell you to be more predictive or to generate things that you didn’t think of. That’s the way I define it anyway. So it’s not that hospitals need to go out there and recreate the wheel, they don’t. Simply because, and Aki mentioned it earlier, there’s a lot of off-the-shelf libraries that do all this basic blocking and tackling. It’s really if I’m trying to use machine learning to determine if something’s a flower or not a flower. A cat or dog. If you follow Silicon Valley, hot dog or not. Those are all the same libraries. So nodules, are they cancerous or non-cancerous? As long as you have enough training data that gives you the probability levels that you feel confident in. That’s the beauty of what is happening with AI. You don’t have to do everything from scratch. You can use the existing libraries to go out there and tell you. It’s the same with neural networks, and it is just amazing how open it is. The biggest companies, from Facebook to Google, to, not so much Netflix, but maybe Amazon. They’re publishing and open sourcing all of this because they want it to advance more quickly.

So for the Mavericks here we’re trying to be able to look at an NBA video game and extract as much data as possible. We’re hiring kids right out of college with data scientist, data backgrounds and Python basic programming backgrounds and trying to find new ways to extract data, and create data that we can apply our domain expertise to. We are using off-the-shelf libraries. So data, whatever data you can aggregate, it’s not going to get its value from just one rocket scientist, person, that can only figure it out and you’ve got to go hire someone. 90% of the value is going to come from off-the-shelf libraries and the other 10% is going to be the competitive domain knowledge. That is why Google or Uber or Lyft, for their autonomous cars, go out their and hire super-expensive million-dollar-plus data AI experts to be able to give you something that they can’t get elsewhere.

Dr. Gerard Silvestri [44:18]:

Peter, were you asking should the hospitals, because some things are not off the shelf, if Peter has curated a database of a thousand nodules at Cleveland Clinic, if they were to partner with someone like Aki, should the health care systems who have aggregated like this data, be in the business of partnering? And should they be charging?

Mark Cuban:

To answer your question, if you’ve got a thousand nodules, you’re going to have to be in the business of not necessarily programming, per se. You want to have a data scientist on staff that you are going to be able to go to and say I’ve got this data, what can we do with it? Then they’re going to need somebody who knows and understands the technology to be able to say our best response or our best results are going to come from this vendor or these alternatives. I, as an investor, I get pitched AI every minute of every day, right? And I can ask five questions and know if they’re full of crap or not right? Because everybody talks that they have AI when in reality 99% are just using off-the-shelf libraries and are just integrating. So to answer your question, having somebody who is a data scientist that takes responsibility for capturing all the data and making sure you retain that data, looking for up side by partnering. Each of you can partner and work with partnering vendors on a cost benefit analysis, build versus buy, and then making the choice. More often than not in this early-stage you’re better off using vendors, and that’s self-interest in a lot of respect, but we’re still so early that you just have to have somebody who understands AI and understands the value of data so you can capture it. You’ll find that between a few hospitals in a system, over a period of years, you’re going to have all the data you need. It’s just a matter how do you use it and what can you learn from it?

Dr. Peter Mazzone:

I think my fear is that the hospital’s value the data at too high a level to foster partnership with tech vendors.

Mark Cuban:

Yes, I can’t say for certain but unfortunately the reality is there’s going to be people with lung cancer all around the world right? If the data comes from Kazakhstan it’s just as good as the data coming from Cleveland, right? That’s just reality. So that equilibrium is already in place. So do you pay? There are companies that are popping up that are saying, I’m going to go out and find as many MRIs as I can and label what’s in them and then I’m going to sell that data to people who may not have captured it. Hopefully your hospitals are saying we’re capturing it ourselves. Maybe we need a little bit of that data to get us to the next level, or Aki will buy that data so he doesn’t have to go through the labeling process. That is just build versus buy, like any other decisions.

Dr. Gerard Silvestri:

So the universities have TTO, technology transfer offices, so that an investigator like me, I’m not a basic scientist, I’m in health services research. But let’s say I work on a new drug and it got done through labs through NIH funding and they’re selling that. Do you ever envision for example that someone like Dr. Peter Mazzone, who has curated a huge database, or Georgie, or myself, who have curated this database that says cancer or no cancer, and all these other primers about the patient working through a TTO and selling data? Because that’s not what’s happening right now.

Mark Cuban:

There’s not a lot of value in that. The value diminishes over time. The longer you wait, the less valuable it is unless it’s just such a huge database you don’t need anybody else’s database for the reasons I mentioned. Now if you have a specialty and you uniquely acquire and retain data for people with a heart on the wrong side of their body, that’s going to be a small volume of people. That’s different. Then there’s more data if that is applicable to how people are using it. Data is the new oil and how you process that data becomes…  The next big challenge, right now the fight in terms of AI, is bias and the ability to audit it. Where I think the bigger issue is going to be is hospitals having expertise to say we need to be able to ascertain whether or not the model you use for your machine learning is accurate or not. Is there bias that diminishes the impact of African Americans or people of color or you know whatever, versus… If you read some of the research where facial recognition is miss applied to certain body types or certain skin tones, somebody has to be able to audit them. You don’t want bias in your Al and that is going to be the biggest challenge. The rest is basic math.

Christine Spraker:

I have a feeling we could talk about data forever.

There are  unique challenges in healthcare for sure, one of them being contracting, and who owns the data, and privacy, and all of that. Is it okay if we talk about some of the misaligned incentives in healthcare? So we’re all stakeholders sitting here right? Everything from industry, providers, insurance companies, the government, pharmacies…

Dr. Gerard Silvestri:

Let’s not forget the patients.

Christine Spraker:

And the patients. We all have a financial interest in the business of healthcare but our incentives diverge  and none of us are coming together and saying “No we are the ones willing to change.”

I worked for large healthcare corporations and they are well intentioned but at the end of the day their incentives don’t necessarily match up with driving down healthcare costs. I think what we’re trying to figure out is how does industry partner with physicians in a way that we get to some clinical utility that makes sense and it creates clinical value but how do physicians remain unbiased after taking money from industry to help provide these services. And is it possible to remain unbiased.

Dr. Peter Mazzone:

I think we all go into medicine because we want to do good, we want to help people, “First do no harm” and I think in our field, at least 99 plus percent of everybody, I trust.  Nothing big gets accomplished without a team of people who have expertise in different areas. I couldn’t build a software program, I couldn’t do data science. But I know what’s necessary in clinical practice and so I want to be able to partner with those folks doing that good work to ultimately help the patients. There have been examples where people who help then give bad advice to the greater community. And then it sours that kind of relationship building. So now people are very, very careful. I think ideally these teams are built and safeguards are in place so there is not just one person saying something is the right thing to do. But I don’t think you’re gonna achieve all of the grand successes we’re talking about without clinicians being involved in guiding scientists and industry.

Dr. Aki Alzubaidi:

I’ve seen the other side. You know, we started this company five years ago and just even yesterday there was a presentation that had our company compared with some other bigger companies. We’ve never paid a physician for any consulting, we just can’t figure out how not to make them biased. And then there was this competitor, you know, where the competitor had all of the check marks and then Eon had zero check marks in terms of capabilities, which is obviously a lie, right. But to me, you know, there’s a saying I’ve heard physicians use called speak the Kool-Aid but don’t drink the Kool-Aid. And it’s a true thing. We’re at this conference, right, it’s the American Thoracic Society, and people are here to learn. And there are some key influencers, Peter, that speak the Kool-Aid but they’re not truly drinking the Kool-Aid.  What’s your advice to me George, I know you have some things to say about this. How do we compete with that and stay above board?

Dr. George Eapen:

So I think, first of all, I believe you because clinicians have to be at the table because we’re the users, we’re under the user base, right, and therefore whatever solution that comes up has to fit. It has to be driven by our experience of it. You know, the issue, which I actually don’t like that phrase speak the Kool-Aid but don’t drink it because it’s very disingenuous for somebody to speak the Kool-Aid but if you’re not willing to drink the Kool-Aid as well…you’re letting other people drink the Kool-Aid, and expressing support, so that’s why I don’t like that phrase. I’ve never heard that phrase before but its not a good phrase.

And I think when you look at where we are in terms of who’s experience becomes important, you ask yourself, we talk about driving down health care costs – it’s really not about driving down health care costs. Something costs what it costs, if it provides value, it is either worth it or not worth it. The reason we’re talking about driving down healthcare costs is because we all have a perception that the value we’re getting from whatever it is that we’re paying for is too much. It’s not sustainable. So it doesn’t matter if it’s increasing that value and hopefully it is. You know value is an equation, it’s what you pay divided by cost.  So if the denominator goes up and the value keeps going down, it’s not an equation anymore. And so somehow or the other, when we’re talking about all these things, what we drive to the same, you can reduce cost by reducing access, you can reduce costs by rationing, you can reduce cost by a whole bunch of different things. And there is always going to be a component of rationing, which I know we do not like talking about, but the fact of the matter is when you have limited resources and their resources are finite, you want to squeeze as much as you can out of it, but it’s not going to be everything to everybody it just isn’t. You can ration either geographically if you live in a remote area, you’re not going to get the MRI scan because there’s none close to you. Your 80 years old, you can’t get into a car to drive, there is rationing all of the time, we have to recognize that. We have to figure out how do we build a system, or how do we participate in building a system, that it provides you know, I don’t want to use the overused term justice…

Mark Cuban:

But what is that first step, more doctors?

Dr. George Eapan

Yes, I mean that is one thing, and the question is “is that always going to be the case.” I don’t know.

Mark Cuban

Well I’ll tell you this. We have 2.3 doctors per thousand, 885,000 thousand doctors in this country, and to compare that to the Nordic countries where people like to use them as a primary example in healthcare, whatever that might be, there’s 15% greater. And I won’t get all political on everybody but Senator Schumer’s sponsoring a bill I think it’s 332 or something, that provides more Medicare dollars for more residency’s to hospitals you know, and I’m not a big fan of his, but I’m a big fan of legislation like that because it requires, and part of it is paying off student loans, and testing student loan payoffs for doctors if you go into healthcare deserts. So you can start to use some basic math and some common sense. I think because we don’t have enough doctors you have to ration, particularly if the Medicare for all, for America, passes because then it gets worse. You know and then to extend it into overall costs, I agree with you a 100% on value, but part of the challenge is the cost of healthcare is the cost of healthcare  an important anymore because of the extraneous costs the insurance companies introduce. Under the ACA there’s medical loss ratio of 15 percent that they punch out the way anyway by taking enhancements that they plug through there. Just the administration concept they put on you guys and then work for 21 to 30 plus percent. So the cost of health care isn’t the constant of health care, and so there’s a variety of ways to deal with it. That’s not a topic for today for today. I think we’re complicating things for your organization’s by forcing things to insurance companies as the payor. The biggest lie ever told to the American people when it comes to health care is that insurance is a metric for determining wellness. The number of people insured reflects the wellness of the population. Insurance is a financial instrument. They really give your money to the insurance companies or the government gives our tax money to the insurance companies and we fight to get it back. Again a soapbox on that [laughter].

Dr. Peter Mazzone [58.24]:

I think expanding more physicians, and it might be even more just clinicians, there is growth in the advanced practice provider community. And matching those clinicians to what their needs are, we are using technology to bring specialty care to areas where it’s not.

Dr. Gerard Silvestri:

I live in Charleston, South Carolina, and 70% of the counties in the state are rural and underserved. You talk about having enough physicians, the problem is that physicians who graduate from medical school, and my son is a second year medical student, want to practice in Charleston, Columbia, Greenville which are cities. So what Peter says is absolutely true and we’ve been using telemedicine to bring some of those places for cancer care. To get back to this team-based thing, an alignment of incentives, it is different than a lot of other businesses. We have people on completely different ends of the spectrum of alignment in terms of big insurance having a certain amount, the clinicians do, patients certainly have additional perspectives, about the alignment. I do agree with you Peter, having those people at the same table is important and as it relates to the question about conflicts of interest, at American College of Chest Physicians [ACCP or CHEST] we have a Professional Standards Committee that has to vet people who want to be involved in the guidelines and say “no you can’t be involved because you have too close of a relationship with industry” or anyone else. The thing that comes out of that effort, the head of the PSC (the Professional Standards Community), says “We do the New York Times test.” If the physician might turn up in the New York Times for having taken a ton of money from industry… And that pendulum has swung both ways, right? For a little bit, actually for a long time, at universities you couldn’t do anything with industry. Anything. I think people recognize now that we need to be working with people with different skill sets and leverage people at the table. Even research organizations like the PCORI, which is the Patient-Centered Outcomes Research Institute, a federally funded Institute, wants to have patient stakeholders on these committees, interviewing brands. So I think it’s moving back a little bit to making sure that all the stakeholders are in the same place. If you stay close…Interestingly enough, what we haven’t had are the insurers at the table for many of these meetings. Maybe that’s good, but maybe its not. Maybe we should have them there and make a better case for what we do, and at least to understand what they do and to combat them, if it needs to be combatted. But not having them there at all, we lose out on who went to install the decisions downstream.

Dr. Aki:

One last question, we’re just about at our time. I want to get this one in. Amazon figured out, and you have told me many, many times [to Mark], make customer happy right? So in healthcare the patient isn’t the customer in terms of technology purchases. So the patient experiences it’s value right? So why can’t healthcare to be like Amazon?

Mark Cuban:

Well, it is. I think your outcomes are a lot more definitive than even Amazon’s.

Dr. Aki:

So he [motions to Gerard] doesn’t have a thousand reviews. I can go look at the paper cup, there’s a thousand reviews on the paper cut right right?

Mark Cuban:

Yeah for better or worse, right? Whether or not I like a cup, there’s a lot less at stake and that’s the difference. That’s the challenge. It doesn’t mean though that, to your point, patients can’t be part of the evaluation process. That they probably should right? If you’ve got the patients in the room when it came to IT decision making they probably bring some common sense in ways…. that you talk about self-interest… I’ve for decades said the head of IT always has the greatest amount of self-interest because they rarely can admit they’re wrong. There’s just so much at stake. So maybe introducing the patient… I don’t know how to solve the problem, but it goes back to outcomes right? If you guys are demanding you need Eon because you want the ability to know when you need to contact this patient next and what’s going to put the patient in the best position to get healthy again. That’s as strong of demand as I think anybody can make and I think that’s kind of the Amazon.

We went back to Amazon 20 years ago when they first started because it was cheaper and easier. What [Jeff] Bezos always says is “your margin is my opportunity” and effectively that’s the same with hospitals too because when a big IT company or big software company is charging millions of dollars or whatever it may be and has huge margins, that’s your red flag. You know that everybody else works on minuscule margins, and I really think that probably is the same from Amazon that fits the best. Your margins, talking to the software companies, is my opportunity, and we’re out there to get it. And you’ll know if you’re getting it by the implementations and outcomes.

Dr. Gerard Silvestri:

I get the analogy of Amazon and I appreciate all of the things they do that are incredible right. Such as, if you don’t like that cup you can send it back. So if you have an illness or are in a life-or-death situation, it’s very personal. It’s personal for them, for family members going through it. That’s where I think the breakdown in the analogy occurs and why so many people have a difficult time solving the health care issue.

Aki:

So thank you guys so much.