This transcription is provided by artificial intelligence. We believe in technology but understand that even the smartest robots can sometimes get speech recognition wrong.

 Thanks for joining us on this week in Health IT Influence. My name is Bill Russell, former Healthcare CIO for 16 hospital system and creator of this week in Health it a channel dedicated to keeping health IT staff current and engaged. We are talking with Dan Rifkin, the CEO of Vers, adjunct professor of surgery and biomedical informatics research at Stanford.

Special thanks to our influence show sponsors Sirius Healthcare and Health lyrics for choosing to invest in our mission to develop the next generation of health IT leaders. If you wanna be a part of our mission, you can become a show sponsor as well. The first step. Is to send an email to partner at this week in health it.com.

Just a quick note before we get to our show. We launched a new podcast today in Health it. We look at one story every weekday morning and we break it down from an health IT perspective. You can subscribe wherever you listen to podcasts at Apple, Google, Spotify, Stitcher, overcast, you name it, we're out there.

You can also go to today in health it.com. And now onto to today's show. Today we are talking with Dan Rischi, the CEO of Vatos, adjunct professor of surgery and biomedical informatics research at Stanford, and also a nerd MBA in bioinformatics from MIT. You don't mind me calling you a nerd or, or a geek because you I take pride.

Great. That's what I, for man. Great background on. Well, thank you. It's such a pleasure to be here and, uh, love the show and happy to contribute where I can. I appreciate that you're a serial entrepreneur. It looks like you've had a couple of startups and you're the CEO of a new one. I would imagine with your background, you're coming up with a new idea every.

What, every week, every day. There are so many great ideas out there. There are so many people that come to me each year just to ask for career guidance, and we all ask for guidance from people who have been there, done that. I built up my career off of some very smart mentors. When they ask me, they say, oh, I've got this great idea, and my feedback is, well, out of a hundred ideas, 10 may be great ideas and one may have a business model.

And it's hard. We work within a capitalist system and need to figure out what's possible and do those things that are possible that have the greatest benefit for patients and the business model.

It's complicated. Try figure out what the financial model is, support these businesses and go in the, in the healthcare direction, insurance direction. That's I. Flow into. Plus it's a multi-trillion dollar part of our economy, which is kind of amazing. We started this conversation prior to going live, and I do that from time to time where I'm like, alright, stop.

Let's, let's hit the record button. So you start talking about epics in healthcare. Let's start there and let's kick off the conversation there and bring the the listeners up to speed on what you're talking about. In terms of the 50 year epics within healthcare, there is such great research in business and disruption.

The randomized trials can never look at subgroups, are not doing much comparative effectiveness. That's needed now with people having multiple comorbidities and multiple treatments available, and so there's a desire to tailor therapy. What's happening is doctors are making guesses. We've gone from medicine as a science 40 and 50 years ago to medicine as an art again.

Is that the most, uh, I mean, is that a good handle for it? I think it's a fine handle. Quite frankly, we never really know what the world will name something until years later. , whether it becomes tailored health or precision medicine or data-driven or personalized healthcare, it will get named something.

And so you start talking about the enablers here and there's technology enablers, there's cultural enablers, there's regulatory enablers for what's going on today. Can you touch on those various things? Yeah. It's so powerful and there are also fears, and I'll touch on those as as well, the enablers, the data in the system back when we started health data reform in the us.

Almost all technology was on premise and servers. We were just beginning to start to use Amazon Cloud. Look now, and we have multiple cloud vendors that offer unlimited compute power. It's a really powerful situation. So the foundation for data driven has to be data. Right. What about the aggregation of data?

So let's regulatory been involved

That's powerful. Other incentives that have been created, there are efforts in 21st century cures to do everything from making NI h's focus, more future looking to efforts in real world evidence and a request of FDH create pathways for regulatory use of real world evidence. These are incentives and pathways to try to encourage innovation.

I wouldn't say they're perfect by any means, but it's nice to see the stakeholders, whether it's provider, payer, regulator, or Congress working to try to make healthcare better. You, you've used this term now real world evidence, is that when we're actually tapping into this big data.

And common.

To the personalized medicine? Is that what that is describing or is it describing something else? It is what it's describing, and quite frankly, these terms are fairly overlapping, whether it's precision medicine, personalized medicine data-driven healthcare, or real world evidence, they have slightly different connotations.

For example, precision medicine is more often used in genomics. Real world evidence is more often used in routinely collected data. But what we are saying here is our cts randomized trials will never get us the subgroups and comparative effectiveness. We need to tailor therapy. They're too expensive and you can't run studies on every person in the us On the other hand.

Routinely collected data has a huge amount of information within the different approved therapies. Which one is working better or worse for which subgroups? And if we get to the point where we can analyze that and say the 73 year old woman who has osteoporosis and heart failure will do better on this heart failure drug and it will hurt her osteoporosis less, or help it.

Even if we can get to the point where we understand how drugs work between different conditions or devices or procedures can benefit individuals. That's a powerful place to be. Interesting. I'm gonna come back to that. You mentioned the word fears when I talked about culture as another factor in enabling this, you, you talked about some fears as well.

Expound on that a little bit if you can. So there are rational fears and irrational fears. And irrational fear would be a doctor who says, I don't wanna have to use a computer. I'm gonna retire because those computers are so scary. That happened. There was a huge wave of retirement with electronic health records.

The rational fears coming from patients, I look at, one of them is, you are going to harm me. That's a rational fear. We look at tailored therapy and low quality efforts, and we saw hormone replacement therapy created off of real world evidence, and it's for a lot of time, doctors prescribed women in menopause hormones.

For many of whom there was no benefit and there was an increased risk of cancer, meaning people died based on that. If we look at Covid, early Real world Evidence in Covid led to publications in New England Journal and Lancet that suggested it wasn't that dangerous. People went out based on that information that they, from scientific sources and some of those people who went out caught Covid and some of those died from that.

There is a real risk of bad evidence making it into the system, changing the standard of care and harming patients. Another real risk is that people's information will be used inappropriately identifiers attached to their information sold on the open market, so that now an insurer can know about a patient.

Change their behavior based on that. Or a or a researcher will that shouldn't have access to an individual's information and name will know this patient has cancer. This is a real fear that we won't handle privacy and security well as a industry, and these fears have to be openly discussed and addressed.

This ends up on the front page of the New York Times, relatively often now. And we want it to end up on the front page as healthcare is doing good and using data appropriately to make your life better. Not people are selling your data without your consent and potentially causing you harm. Yeah. There's also this conversation around bias, which I'm biased, which I.

Create some of that incorrect real world evidence that's out there. You have bias in the data and then you have the, the quality of the data. No, no offense, but doctors aren't exactly the best data entry clerks, nor do they want to be, uh, great data entry clerks, but for the quality of the data in the system that we're gonna be using to generate this, this evidence.

And then the bias that exists, doesn't that also create some challenges? Yes, there are two types of bias we should consider. One is when laypeople hear bias, they are thinking healthcare inequities. For example, are you gonna make the system worse Where someone who is a little poorer or who is a um minority will get worse care or will have care that is applied to them but not tailored to them?

That's one kind of bias. Another kind of bias is. Bad data, leading to bad information in the system, leading to a bad change in standard of care. Let's try to tackle each of them separately. I, I would say for the health inequity, one real world evidence is good for that. Randomized trials historically, or choosing from wealthier people because they're done in referral centers, they're done in high-end centers and cities that are academic, where wealthier people visit.

The randomized trials have had a historic inequity where they give an answer for wealthy people and then that answer is applied to poorer people and it may not be right. Real world evidence offers an equalizer where we can take information from the general population and understand the subgroups and treat more clearly.

I'll pause there before I go on to the bias question 'cause I think it's so, so separate. Sense to, doesn't sense to.

In New York City, they're, they're gonna serve both populations, uh, pretty equally. Maybe not. Rural America is probably not taken into account, but even Chicago Medical Centers, LA I mean, they would have a representation, wouldn't they? Let me give an example of the kind of thing that may happen. You are trying to understand for a type of cancer what the best treatment is.

There's one treatment that requires the person to go to the hospital or the clinic. Every day for a month, and there's another treatment that is a pill taken once a week, a randomized trial is performed and finds that the intensive one that requires you to go in every day has a 10% better effect size.

10% better. So standard of care says, now use that intensive one. That's the correct one to use. But did that study ever look at the 25 year old mother of two single mother working two jobs who really has no ability to get to the clinic every day for a month? Will that person really do better? Or was it worth looking at that subgroup and saying actually for them.

We should offer them a pill once a week. That is not exactly perfect, but we'll do better than if they miss half of their appointments because they need to go to their job to feed their children. This is the kind of real discussion we need to have with doctors, with patients, and we'll only get there if we actually study those people where they live and are seen, which is not in randomized trials.

Interesting. Is there a movement to, I assume there is, there's always progress in healthcare, so is there a movement to take that population into account? To change the way we do clinical trials, to be more inclusive? There have been efforts in clinical trials for the last 30 years to try to create subgroups and to try to equalize the health inequities, and they have been mildly successful.

There's just a limit to what you can do when you have to spend a fortune on a trial. You can't enroll in unlimited number of patients. You can't get all the subgroups. There are limits and, uh, I've gotta compliment pharmaceutical industry for, um, making efforts to do so many trials. But the fact is, the overall approach is not going to be successful in subgroup analytics and comparative effectiveness.

It cannot be. On the money it takes to do this. So you also gave us another example of bias. Let's delve into that a little bit. Could you expand on that? So the second kind of bias is harder to understand. Health inequity. Everyone understands and many people feel passionately about. The second kind of bias is bad data.

We. Often run studies, real world evidence is not new. We've been doing it for decades and real world evidence often tries to get quick and dirty answers. We will take a single source of data and we will analyze it and it'll be cheap and it'll be fast, and we will get a sense of trend of what the population is like, who needs this kind of care, or generally how they do.

The challenge is that kind of low accuracy, low validity, real world evidence, which is perfect for trial recruitment or trial design or marketing insight, does not apply to the new requirements in real world evidence, which are making clinical assertions. They are enabling clinical payer or regulator decision making called, called market access or called regulatory decisions.

This new real world evidence, which is changing the standard of care needs to be high validity, but we're only doing it over recent years. We don't have a lot of experience and no one really has a complete answer of what validity is. Uh, as you talk about this, the other thing that sort of strikes me is we, we used to talk about the whole patient profile, and when we talked about the whole patient profile, we would have this percentages and we'd say healthcare has maybe 15% of the information on the whole person profile.

And then there's this other 85% that exists that really defines who I am, how I act, how I make decisions. Why I chose to buy the Big Mac instead of going next door and getting the salad. Just all those factors that make me up. And we don't have access to all that data yet, or are we starting to get access to that data?

This gets back to the issue of a hundred good ideas or a hundred ideas where 10 are good and one has a business model. Let's, in healthcare, do what is realistic and possible in the short term. In the long term, we can try to create the things that seem crazy and unrealistic. So in the short term, over the next five or 10 years, we are not going to be able to get all information on all people.

It would be nice to ask everyone how they feel about their treatment and when they feel healthy. That would be wonderful for running these studies. But realistically, what we can do. Is use the information in the medical record and in the claims and in the other areas, put it together in a high validity way.

Start making assertions as to this treatment works better than that treatment in this subgroup. I'm not going to say that over time I don't want to have people, all, people in society contributing all the information they feel comfortable contributing to run the best studies possible, but in the short term, we have to work within the medical system and the information that's there.

The challenge is we've only begun to touch the surface of the information that's actually there in the medical system. Alright, so let's talk about the next five years then, and what is going to be possible, what do we need to get right before we talk about what aspects do we need to get right before we start talking about what's possible?

I mean, we talked about the, the quality of the data. Are we gonna start putting systems in place? Increase the quality of the data at the data entry level. What do we have to put in place in order to enable the next five year, five years of progress? So recognize where we are now. We have a huge amount of information in the system.

We just can't access it in any reasonable way. We have information written in narrative notes. We have information in an EHR that can't be connected to a claim. We have information in claims that represent outcomes that may have happened in the person's home health system, or may have happened in a whole other state.

There's so much information in the system and we have a lot of trouble using it. Combine that with confusion. Over what high validity looks like. For example, if you look at the last 30 years of real world evidence, always single source data, never accuracy, determination of the information. If you look at the next 10 years, I would argue that we can now get to a point where we can understand the information from the EHR and check accuracy.

We can use artificial intelligence to pull it out. We can create gold standards to ensure highly accurate information. We can link with claims and registries to understand outcomes. We can manage privacy and security through, through Safe Harbor determination or expert determination, where we take out any identifier and we can aggregate the information to have meaningful insight that's possible.

Today, you just need really good technology and a commitment to do really high quality work, and we're, we're not seeing a lot of groups that are really putting in that effort. Because quite frankly, the field of real world evidence is growing so fast. Just traditional efforts are already being required at a level that can't be served by the vendor community.

So when we talk about those enablers, the technology is pretty sophisticated these days. We're seeing what Google is doing with Ascension, which is really interesting, but what they're doing is really fascinating. Taking a hundred disparate EHRs, pulling that information together and creating a health record that has meaning across all those a hundred different facilities.

And so the technology is, is there to aggregate the data, to use NLP and other technologies to make sense of the data and then pull it back out there. So help us.

Landscape. Of course it's gonna be better in 10 years, but how is it today in terms of its ability to do what you're talking about? Yeah, so it's like an iceberg. The base of the iceberg is all traditional. It's claims data all queried in traditional ways, a little bit higher up, A smaller group that are, uh, making these efforts use unstructured data and natural language processing to pull out pieces.

Never check accuracy, but generally get a richer, uh, view of the patient and the tip of the iceberg. The smallest group. Is doing really high validity where they're bringing different data sets together, figuring out the intersection between the data sets, and then using source data to determine accuracy and defining in their protocol what required accuracy looks like.

That's considered advanced real world evidence or high validity, real world evidence. It includes deep phenotyping and linkage. Yeah, I'm sorry. I'm actually taking notes, which I don't do in a lot of these podcasts. I appreciate it. So the high velocity or high validity, that's the part that you're saying is not done all that much today?

There's no reason it would've been meaning. It's only over the last few years that regulators, payers, and doctors are accepting or expecting high validity evidence. They're starting to use it to change the standard of care. They're, the payers are deciding what they're reimbursing based on it. The regulators are deciding what they're approving.

The doctors are deciding what they're prescribing as these groups use real world evidence to make these decisions that defines the standard of care. What is approved, what is paid for, and what is prescribed is the entirety of the standard of care for medications. And so, as these groups use real world evidence, they're starting to ask the question.

Is this believable or are we gonna get into another hormone replacement therapy situation where you're giving me bad evidence, doing things that harm my patients, just because you didn't do the work to get really good data quality and really good generalizability in the study. And so. Firms are starting to think through what that would look like.

So Dan, is it better to have more data or less data to focus in on a smaller LA amount of data? Or is it better to have significant data sets? Yeah, to to work with. I'm asked that question all the time. I run a company for Antos that does high validity, real world evidence, and customers say to me, can't I just get more data and it'll be more accurate.

That's not the case. If you're using low quality data, doubling the amount of data or taking it by a hundred x does not give you a different answer. It does not give you a better answer. The key is to use high quality data, um, whether it's a small amount or a large amount, and power it sufficiently for the assertion you're trying to make.

So how do you get higher quality data? Well, that's the trick. We have a belief of how to do it. We're currently under a study with FDA sponsoring that's looking at this specific approach and the approach is. Take from the EHR, the phenotype, meaning the information about the patient that says these diseases, these subtypes you get from the EHR.

Since the source of truth in the EHR is the unstructured data, the narrative the doctor writes. You have to work with that. Use NLP and Inference Technologies to pull out the information. Then like we do everything everywhere else in healthcare. Check the accuracy, create a sample that has a gold standard and check that you've achieved high accuracy.

This is the step, almost no one does, but we do it everywhere else in healthcare is we check to make sure what we did worked. That's part of it. And then the other part is outcomes, where you link to claims and, uh, registries to understand how the patient did. Knowing the subgroup from the EHR phenotype and the outcomes from the claims, you can actually make an assertion as to what's happening.

But the key is to get a really high validity phenotype to know the accurate description of the patient, and that's the hardest part of this. So who is using this kind of high validity data today in the administering of care? Right. So. Everyone is using a little bit of real world evidence right now. In terms of high validity, real world evidence, we're seeing some come out in oncology.

We hope it'll be out this year. This is the beginning of an effort, not the end of an effort. So it's just the beginnings of using this information. My audience is gonna be a fair number of healthcare providers and people in IT departments across the. What can they do to enable this progress or be ready for this as this comes to their health system?

Well, I think the first thing to do is no good from bad. If we are indeed going to feed information back to the health systems and say, this subgroup should be treated in this way. You can find this subgroup for pop health or disease management. Here are the ways to look at subgroups. If we're feeding the information back, they need to know good from bad.

They need to actually think about high validity and accuracy. The second thing they can do if they want to be part of the future is work with some of the companies that are running high validity studies. We have engagements with some of the largest health systems, and as part of our engagement, we are teaching how to safely move sensitive data, how to extract it in a high validity way, how to feed the information back into the health system.

We're happy to do that work if a health system wants to work with us. They can feel free to go onto the VATOS website and reach out to us. We don't mind doing the engagement. We don't mind teaching what we know. We think that's important for the industry, and there are always plenty of studies that are being done.

We're doing studies right now under National Science Foundation and FDA who are doing studies. We just announced a partnership with Amgen. We have other firms we work with. There are always lots of studies to use as test beds. So that health systems can learn. So Dan, I've started to close out my interviews with a very odd question.

It's sort of a catchall, I've gotten such good results that I keep doing it, which is, is there any aspect of this that we haven't talked about or anything that you think the community would benefit from a discussion about? Yeah, let's talk about the future. The future as we look at this 50 year epic.

We're just at the beginning. The future should look better than it looks now, or we haven't done our job well. I was in Congressional retreat after my 21st Century Cure's testimony, and I was asked, what's the biggest thing that we can do in healthcare to make it better? The most disruptive, most beneficial thing.

And I said, make the patient the customer. A large part of my career, not only understands the science, but has to understand where the dollars go and the patient has to be the center of this In our industry, I think all of us know that vendors work with provider organizations or payers, and they defer to regulators.

And the money is flowing between these different stakeholders, but the patient doesn't have the control of the dollars and decision making or the attention of the different stakeholders that they deserve. And so what the future should look like is real world evidence and other parts of healthcare, um, deferring to the patient where the patient has the expertise.

The Dr. May be the expert on biology. The patient is the expert on that person. The woman who has two kids and works multiple jobs and is a single mother. Healthcare should not treat her like every other person. Healthcare should Ask her, what is your life like and how do we give you healthcare that works for you?

Maybe coming into the clinic every day for a month doesn't work for her. Or maybe she wants to make that choice. She needs to be informed about the subgroups and the outcomes, and she needs to be the center of this discussion. So as I look at the item that we need to make sure to keep in this, it's patient centricity and a focus over the next 50 years on giving them the tailored therapy they deserve.

In 50 years, we're gonna see engaged patients. Who are informed, have their information, have choice, and they are seeing physicians care providers and that maybe who knows what it's gonna look like, but they're gonna be seeing care providers that have real world high validity information available to them that can create a essentially.

Come up with a diagnosis and come up with a treatment plan that is really custom to that individual that close would add. Bill, I'm gonna do everything I can to give you that in 10 years. I can't even tell you where we're gonna be in 50 years. I know that this effort, we're gonna be getting better and better at it.

Using genomics and using sensors and using all the information and tailoring care, we're gonna be getting better and better at it. And 50 years from now we'll have another disruption, and hopefully that'll be even better than this one. Yeah, but I gotta tell you, we're having a lot of fun on this one and we are going to make healthcare better.

Well, Dan, thank you for your time. We'll have to check back in in a year and see how much progress we're able to make. Thank you so much. I really enjoyed the discussion. If you know of someone that might benefit from our channel from these kinds of discussions, please forward them a note. Perhaps your team, your staff.

I know if I were ACIO today, I would have every one of my team members listening to this show. It's it's conference level value every week. They can subscribe on our website this week, health.com, or they can go wherever you listen to podcasts. Apple, Google. I. Overcast, which is what I use, uh, Spotify, Stitcher, you name it.

We're out there. They can find us. Go ahead, subscribe today. Send a note to someone and have them subscribe as well. We wanna thank our channel sponsors who are investing in our mission to develop the next generation of health IT leaders. Those are VMware, Hillrom, Starbridge advisors, Aruba and McAfee.

Thanks for listening. That's all for now.