Using Real-World Data the Right Way
Pramod John, PhD, CEO of VIVIO Health, discusses a recent presentation where he spoke about using real-world data to drive better outcomes and lower costs, as well as weighs the benefits of value-based care.
Please introduce yourself and tell me about your background.
I am Pramod John. I am the CEO of VIVIO Health. My background includes a few ventures in the high tech space. I switched gears because I felt that health care was our biggest national social problem affecting our economy. I ended up going to McKesson, one of the largest health care companie in the world.
It was a fascinating experience because most of our experiences about health care come from being in consumer health care. We all hear the public debates and everything else—whether it’s health care insurance, single payer, a free market system—but we’ve all got these assumptions that we make about how the system works.
It was a fascinating experience, being in the throes of understanding where the dollars flow. You come to the understanding that how dollars flow has very little to do with what we think about how the system works.
I left McKesson in 2012 to work on a start-up venture. It was the first venture that was focused on the script RX raw space of receiving scripts and transferring them. That was sold to PokitDok which was bought by Change Health care. When I left in 2016 I began work on the specialty drug problem, which is what VIVIO is focused on.
Can you give a brief overview of what your recent presentation focused on?
In early September, I gave a presentation focused on the issue: how do we use real-world evidence? That makes an assumption that we’re not using real-world evidence already.
In reality, what we’ve found was there was a big disconnect between, for example, a clinical trial, the data that comes out of a trial, and how that data is actually used in real life. We’ve come to the conclusion that part of that disconnect is caused by the fact that we all have misconceptions about what a trial actually is, even more misconceptions about what the FDA actually does.
We start with the misconception that an FDA approval equals that a drug works. It’s even more complicated because the term that we always throw around is efficacy. It turns out that even our use of the word efficacy, in some ways, highlights our misunderstanding of the drug trial process and the FDA approval process.
Even though we typically use it to mean the question of, does the drug actually work, efficacy is technically the definition of how many people out of 100 or the population response rate of meeting an endpoint, or in this case the description of a drug, is actually working.
The problem with that is that when we don’t understand the difference between those things, we’re often talking about the wrong thing. As a result, it not only affects our understanding of what an FDA approval is, it also affects how these drugs are being used by physicians in the market.
We find that it’s not only a consumer problem that we don’t understand the data behind a drug trial. Physicians often don’t understand the drug trial data either. More concerning is the fact that we find that physicians often have not read the clinical trials themselves on the drugs that they’re prescribing.
When we talk about real-world data, the real question is how do we translate the data from a trial into being able to objectively assess and understand for every patient, who’s on one of these therapies, are they responding, and how do we collect the data along the way to know that they are responding in the way that we would expect them to.
What examples of real-world data are the most valuable in making significant outcome improvements and lowering costs?
In health care, we see a lot of discussions about the need for more electronic health record utilization, we need to collect more data, and what we find is exactly the opposite. It’s not that we need more data. It’s that we’re not doing anything with the data we already have.
For example, when a drug trial comes out and indicates that there was no improvement in overall survival, for example, for an oncology therapy, why is it that we’re not using that data?
Why is it that we’re not able to say, hold on, that drug shows no improvement in overall survival, why are we using this as a first-line therapy or as a
second-line therapy or as anything other than saying this is an experiment when we have no data.
Let’s start by using the data we have.
In the trials, if we were to go back, we find that there were objective measurements that were used to assess, did this actually improve the quality of life in this case for someone who has rheumatoid arthritis? How did we measure that?
It turns out that when we talk about real-world evidence, most of the time, we’re not taking that same standard objectively for every patient and asking, I’m going to put you on a $50,000 or $70,000 a year drug. How do I assess and know for sure, objectively, that you’re actually responding and benefiting from being on this therapy?
We’re not even doing the simple things of saying, couldn’t we at least go back and compare it to what the drug is supposed to do? It turns out, in real life, even the basic simple things like that we’re not doing.
A lot of our focus is on just saying, let’s take the real-world evidence we have today and apply that to every patient we see.
What role does value-based care play in using data to drive better outcomes and lower costs?
One could ask the question, what do you mean by value-based care? Isn’t all care supposed to be value based? If there was no value to it, then why exactly would you be providing any care?
Do we want people to get care that doesn’t have value? More importantly, do we want to pay for care that doesn’t have value?
We would argue that there is no other care or there should never be any care other than ones that we can demonstrate value for, especially in a third-party payer scenario because our whole system is a third-party payer system whether it’s the public sector in which taxpayers pay or it’s the private sector in which employees lose their bonuses to pay for health care.
In any one of these circumstances, it’s the employed people in America, who pay for health care, one way or the other. From that perspective, in our view of the world, it very much is that objective connection meaning, do we have data that demonstrates that there is value for this individual patient? If not, then the answer to that is, it is valueless care.
Imagine a drug comes out. A 50-50 coin flip or a 50% response rate is actually really good, as low as that sounds, because most drugs don’t achieve that. We see drugs that are being approved that have a 2 of 100 response rate. A 50-50 response means that one person is going to meet the endpoint, one person’s going to fail.
We can unequivocally state that there is a zero value for that person for whom the drug is going to fail. While the drug may intrinsically have value, it has no value for the person it’s failing for. We should never have to pay for something that doesn’t work for somebody.
That would be like for us to say, in the rest of our lives that we, as consumers, would have to buy three cars every time or two cars because only one of them would ever work, but we have to buy at least a couple of them and we still have to pay for the car that didn’t work. That’s how we pay today for drug therapies, which makes absolutely no sense.
What are some examples of technology and processes that have been utilized to improve outcomes?
What does technology do really well? One, it takes the process and it improves it. For example, we used to have four manual steps and use five people, and now, a computer can do that with lower statistical variability or error at 100th the cost of what that used to take and time.
The second area where technology improves upon our capabilities is that it can look at data in ways that human beings cannot. As humans, we have limits on how we understand things, but computers and algorithms have no such limits.
Why that’s really interesting in the case of health care is that things like drug therapies, by their fundamental nature, are mathematical/statistical problems. Every drug has to go through a trial or a scientific experiment.
We have a large data set that we always start with. What we’re really doing is we’re taking that data set and we’re saying, “Let’s take that data and apply that to an n of one patient who walks in the door.”
Imagine that you’re a physician. What’s the first problem that you have to solve when a patient walks in the door? Accurately diagnosing the patient.
Diagnosis, in this case, is between all possible differential diagnoses. What are the most likely things and what’s the overlap between competing diagnosis? Because the higher the overlap, the higher the rate of misdiagnosis.
Even that comparison, if you were to think about it, is a statistical comparison. It’s all about data. It’s not about what you believe. It’s not about how many times in the last hour have you seen the same thing. It has nothing to do with that. It has absolutely to do with the probability of the relative probability of those events and the overlap, and how that determines a diagnosis.
So when you think about diagnosis, that is a statistical problem. It has nothing to do with biology or chemistry. If you look at a drug trial, that is a statistical problem. If you look at the question of once a diagnosis has been determined, what is the best trajectory or outcome that a patient can achieve?
What is the cost benefit and risk associated with each one of those trajectories? That too is a mathematical statistical problem. If there’s one thing technology is really good at doing, it’s mathematics and statistics. It’s far more accurate, if you will, than a human being.
We should be using technology where it can do something better than we can. And this is something that technology can do something better than we can.
And so, it’s a fantastic area for us to figure out how do we use technology to get people to better outcomes faster.
What are some of the other challenges you face with using real-world data?
Most of the challenges that we face are not related to data or technology, ie, in some realms you get into problems that are so complex that we don’t have the technology necessary to be able to solve those problems. Health care as it turns out as much as we talk about things like big data, it’s a small data problem.
The biggest problems that we see are not about technology, it’s about our adoption of technology or the economic incentives to adopt technology, which are really human problems.
We see a lot of artificial restrictions to trade and in the same way, we see large amounts of artificial restrictions to things like the use of technology that obviates the need for physicians.
If we were to come up with a system that could diagnose and could assess data for treating a patient, in areas that we understand well, like oncology, for example, if we have the data then, of course, we could build an algorithm that would do better than a human being. It’s a given because it’s a data problem.
If we could do that, what are we doing? We’re obviating the need for the oncologist. We’re commoditizing the value of the oncologist and we’re saying that, “Hey, anybody in primary care using technology can get to a better outcome than your average oncologist can do.”
The bigger problem in these cases is not if we can do it, it is the opposition to us doing it, changing the current economic models and how people make money. We see those things being the biggest impediments.
We see the physician who runs, for example, an infusion center in their office and they want to charge five times more for exactly the same drug that you can get outside saying, “Look, if you don’t use my infusion services, sorry, you’re going to have to find another doctor.”
We see those kinds of issues as the biggest impediments to change, not technology.
