Crunching the complex data in health care

By ConvergenceRI
February 3, 2020

PROVIDENCE – Neil Sarkar, Ph.D., the interim president and CEO at the Rhode Island Quality Institute, occupies a unique position atop a mountain of health data analytics under way in Rhode Island.

The Rhode Island Quality Institute is the home of CurrentCare, the state’s health information exchange, which manages the flow of information around clinical health data for more than 50 percent of Rhode Island’s residents, some 537,000 individuals.

Sarkar is also the founding director of the Center for Biomedical Informatics at Brown University, where he has played an integral role in helping to analyze the data from the All Payer Claims Database to help the statewide compact manage the rate of growth of health care costs in Rhode Island. [See links below to ConvergenceRI stories, “Annual cap of 3.2 percent put on health care costs,” and “How data analyses may bend the curve of health costs in RI.”

In his work as an associate professor of Medical Science and an associate professor of Health Services, Policy and Practice at Brown University, Sarkar’s recent research has been focused on the transition of care of newborns in the Neonatal Intensive Care Unit at Women and Infants Hospital to federally qualified community health centers, to better understand the gaps in knowledge and practice around such care.

And, as a member of the R.I. Senate Special Legislative Commission To Study the Impact of Insurer Payments on Access to Health Care, Sarkar provided a data analysis at the first meeting of the Commission, “Identifying Top Billed Diagnoses in Rhode Island.” [See link to slide deck below, as well as the ConvergenceRI story, “Will anecdotal and data evidence converge into legislative health remedies?”]

As deeply immersed as Sarkar is in analyzing health care data, he also voiced a healthy skepticism, if that is the right phrase, acknowledging the current limitations of health data.

“Health data,” Sarkar said in a recent interview with ConvergenceRI, “by and large, is one of the most misunderstood entities in health care.”

When you say “misunderstood,” what do you mean? ConvergenceRI asked.

“When I started my graduate education, more than 20 years ago,” Sarkar began, “there were all these promises folks made about [health data]: if we had access to electronic health records, if we had access to health data, we’d be able to enable real change.”

You can pick whichever science fiction genre you want, Sarkar continued. “We honestly thought that as a community, these computers would not just transform how we dealt with data but would replace humans for many rote tasks, including medicine.”

As role models for such expectations, as it were, Sarkar explained, we had the holographic doctor in “Star Trek,” and Darth Vader in “Star Wars,” put together by droids, using knowledge from the Jedi library.

When promises didn’t match the hype
Sarkar spoke about the struggles that computer science and Artificial Intelligence went through when the promises did not match the hype. “Artificial Intelligence [AI] went through what we call the AI winter, where the promises didn’t match the hype,” he explained. “Some of that was purely computational; we didn’t have enough computational horse power.”

With health data, Sarkar continued, “We’re sort of running into that same moment again, where there are a lot of promises being made, and people are not fully understanding the complexity of health data and how we turn that data into information, which can be used as knowledge that practitioners – whether you are a clinical practitioner, or a payer of health care, or whether you are a consumer of health care – can benefit from.”

Brokering missing data
What the Rhode Island Quality Institute is most known for is CurrentCare, Sarkar explained, which he described as moving data from one clinical site to another clinical site, and essentially supporting the notion that the right data will be available at the right time to the right person, in the right format.

“Every HIE [Health Information Exchange] in the country, every organization that runs an HIE in the country right now, has the same problem,” Sarkar said. “We’re able to provide some data, most of the time, in a format that a clinician is willing to accept.”

Sarkar talked about what made medicine so difficult, from a health data perspective: “If medicine were easy, meaning from a data perspective, if we could say that everything that we need to know about this individual is actually recorded in a chart – I don’t care if it a stone tablet or on paper or electronic. And, if I just take that chart, and I give it to someone, and all that information about an individual is going to be transmitted, there is not enough time for a clinician to really do that.”

Instead, what most clinicians do, Sarkar continued, is that they just keep track of key facts, and there is lots of missing data. “So, what we are doing is brokering missing data.”

Here is the ConvergenceRI in-depth interview with Neil Sarkar, Ph.D., the interim president and CEO of the Rhode Island Quality Institute, talking about the challenges and opportunities in health data analytics in Rhode Island, a state Sarkar believes that can play a major role in changing the way that health care operates.

“What we are trying to take on in our state is really trying to revolutionize how health care operates. And, frankly, it’s because we don’t have a choice. We all know that. Not just in our state but in our nation,” he said.

ConvergenceRI: Thanks for making the time to talk with me. What is it that you like so much about ConvergenceRI?
SARKAR: In our organization, we see all of health care. As part of that, we need to understand everything that is going on. It’s hard to find sources where there is a good distillate – here’s something interesting, here’s something that is important. And, it’s fun to read. We talk about it at our regular weekly senior leadership meetings. What’s the latest thing on Richard’s mind? It’s important.

It’s one of many sources that we have. It’s one that we value, because it’s local, it’s real, and it’s genuine. And, that’s not meant to be disparaging to any of the other press folks around the community.

There’s a true likeability to ConvergenceRI, just calling it out as it is. That’s hard to find sometimes [in the news media].

ConvergenceRI: The tendency is that some folks want you to serve as a mouthpiece. And, if you are unwilling to be a mouthpiece, it can prove to be difficult.
SARKAR: This is a hard business. It is a really hard business. Honestly, I believe that every single one of us who works in health care, we’re in it because we want to make a difference.

I’m a pure academic. So, for me, failure is OK. Say, I did something dumb. That’s OK. I can deal with it.

ConvergenceRI: Failure is an important part of the innovation process.
SARKAR: Ninety-five percent of the things I do fail. And, I’m proud of that. And, that’s OK.

What I always tell my students, and it’s really important for myself, is: You don’t just stop, and say, “Oh, that didn’t work, now I’m going to give up and move on.” Then you become a failure. That’s not the approach you want to take in life.

Rather, it’s: “Well, that didn’t work; what can I do better.” How can I improve on that?

One of the things that, as a scientist, is: How do we look at what to some might be seen as a failure? How do we look at the same result and say: “Well, what’s something valuable that came out of it?”

ConvergenceRI: Two years ago, the president of Epic, Carl Dvorak, spoke at Brown. In that presentation, “Technology for Accelerating Changes in Health Care,” focused on the role of health IT in shaping the future health care system, there was a physician who spoke up and challenged him, saying that the basis of all relationships in medicine was the interaction between the clinician and the patient. Because that was where trust was established. [See link below to ConvergenceRI story, “Is health care all about the data, the data, the data?]

If medicine is all about the enhancing of the relationship between patient and clinician, how that does that play out in terms of health data?
SARKAR: Data is just another tool, how we provide the data for a clinician to make a decision.

We had the same debates when the stethoscope was first introduced into clinical practice. Clinicians were like, wait a minute; you want me to stop putting my ear to the chest of a patient?

As a physician, they would say: I need to touch the patient. I need to feel the patient. You want me to put this device between me and the patient?

The answer was yes: it will make it so you can hear the heart of a patient better. We’ve come a long way, since then.

ConvergenceRI: I’m not sure.
SARKAR: The stethoscope is still an important tool. Actually, what we do, which is done with purpose and intent, is when students go to medical school, we train them how to use a stethoscope, so it doesn’t get in the way of the interaction between the patient and the clinician. In fact, depending on what specialty you are in, you learn different techniques,

So, when a pediatrician interacts with a patient, and they are taking the vitals, they are doing all these things very differently than an adult physician. And we teach folks how to do that.

I want to get back to Epic for just a moment. What we oftentimes forget are these organizations’ origin stories.

Epic, now a big behemoth in Wisconsin, was founded by a woman [Judith Faulkner] watching her significant other, a physician, struggled with managing the data for patient care.

She said: I’m a computer person; I can solve some of your challenges.

The very first electronic health record -- everyone always debates who had the first system – but amongst the first, up in Boston, at Massachusetts General Hospital, developed in 1968, was a system called COSTAR [Computer Stored Ambulatory Record], was designed with the same notion in mind.

Two things we do as a civilization, whenever we have a new technology, and computers are no different: We figure out how to blow people up, and we figure out how to improve health care.

You can find that those two things happened with almost every major technological advance. When we could write, the two things that we wrote about and recorded, whether it be on papyrus or paper, or whatever, were: how do we beat up my enemy, and these patients are coming in, what do I need to record how to provide better care.

The EHR [electronic health record] is no different. When I was starting in this discipline, maybe 10 percent of the United States had electronic health records. And those places, if you go to them now, they still love their electronic health records.

If you visit Kaiser Permanente, they talk about how much they love Epic, and the reason why they love Epic is because the system was installed and was used to support patient care.

What we did in the late 2000s was to say: that is a really great idea, and let’s deploy it, really quickly, and let’s hope for the best because it worked really well in some of these smaller places. Well, that’s a recipe for disaster in any industry.

ConvergenceRI: Electronic health records are not going to fade away. One of the growing trends is that physicians are using transcribers so that the collection of data does not distract from their interactions with patients. Research conducted by Healthcentric Advisors has shown that the practitioners who are most stressed by the demands of EHRs are often primary care providers. As a result, it seems, sometimes the voice of the patient can get lost.
SARKAR: It’s easy for us to point to the electronic health records as a point where the system has issues, The real problem, in my opinion, is the documentation burden that we put on our physicians, not for the sake of documenting for clinical reasons, but for payments. We have some glimmers of hope, now, that people are listening.

What the EHR has absolutely done, because we deployed it very quickly, it has put a huge magnifying glass on a major problem we have right now. The joy of medicine is completely lost and buried in paperwork. Plain and simple,. And now, that paperwork cannot be avoided.

As a health care organization, [such as] a hospital or a clinic, I can mandate that you, as a physician, [need] to fill out this form that’s in front of you, and you won’t be able to discharge this patient if you don’t fill this out.

When you look at specialties, documentation to level of detail does not matter as much. Surgery is a great example of this. If you look at a surgeon’s notes, they are just the facts. We did this procedure, these are the complications, because their payment does not always and does not completely depend upon a complete review of systems and so on and so forth, that a primary care physician has to do.

The primary care physician – and any of the primary care providers, physician assistants, nurse practitioners or otherwise, are really burdened by: you’ve got to check the boxes, even if it is not relevant for the patient in front of you.

For these 12 minutes or 16 minutes, depending on how you are counting, you have to check these boxes, if you just want to get paid for this visit. It’s absurd. I probably can’t run for political office [after saying that]. [laughter]

ConvergenceRI: The new Senate commission on studying rates of reimbursements by insurers is an effort that you seemed to be very involved with. Your initial presentation documenting the highest-cost procedures was fascinating. Will you have the opportunity to drill down further?
SARKAR: Once again, I truly think that anyone who is in this industry, we’re all trying to do the best that we can. And, if you’re in it for the money, there is a saying: you’re going to get caught, eventually. I think that happens.

One of the things that we’re starting to have, as an opportunity here in Rhode Island, is we have the data.

And, I’m digging more and more into that data. I’m hoping to finish the analysis in time for the Feb. 25 meeting so I can present where we can at least see some hints toward understanding what are some of the driving factors are for many of these costs.

It’s not simple; there are these complex interactions. But, now that we have data, we have numbers, and we can get into the details, because in health care, we [have to] deal with research on the networks of care with complex co-morbidities.

I think one of the reasons I was invited to serve on this committee was because of my work on complex co-morbidities, and the relationship to complexity and cost.

A patient with diabetes, if they have depression, there is a high probably that it is going to cost more. A patient with diabetes, if they have hypertension, it is going to cost more.

Talk to any physician, especially in primary care, and they will say, yes, the dichotomous patients [with co-morbidities] are important to monitor.

But a patient isn’t just [a series of] dichotomous pairs [of co-morbidities], actually, they are a set of those, a network.

And, what a clinician is challenged with, especially for many of these chronic conditions, is that the complexity increases over time, because there are more things happening. We’re all aging; we all come with an expiration date. And, depending on which path our expiration date is pointing us, the complexity of co-morbidities will vary.

What are the co-morbidities that the clinicians need to put their energies on in the clinical trajectory? Where is that window? If you miss that window, it explodes.

ConvergenceRI: Anecdotally, advocates have told me that alcoholism is a co-morbidity of drug overdoses about 75-80 percent of the time.
SARKAR: That sounds about right.

ConvergenceRI: But there is little or no data on alcoholism – or acute alcohol intoxication – that documents its relationship to drug overdoses, creating the data link to deaths of despair including suicide, alcohol and drugs, linked to economic conditions. What would it take to try to mine that data from the All Payer Claims Database?
SARKAR: That’s a really great example. And, it gets at some of the complexities.

APCD is a wonderful place to start. The challenge of working with the APCD is the root definition of what it takes to be in APCD. Which means, what is being billed for?

Many of these complex factors are not billed for. So, if it’s not billable, chances that it will be in APCD are pretty slim.

Alcoholism is great example. Alcohol use, until it becomes to the level of something that someone needs to be enrolled in some kind of treatment program or it leads to some exacerbation of an event, isn’t going to show up in APCD.

ConvergenceRI: It may show up in the number of transports by EMS services to hospital emergency rooms under incidences of acute alcohol intoxication. But how it gets billed is another question.
SARKAR: Yes, that’s my point. The APCD is one lens. But it’s missing some important pieces of data. So it may show up in other parts of the system. What we have at RIQI is the other part of the system.

The way I like to think about it is: the APCD tells you want happened; it gives you some hints, some ideas about why certain things might happen. But it doesn’t tell you why it happened, or how it happened.

That’s where you need to get to the clinical data, the actual raw clinical data. And that’s what we, at the health information exchange, have at least a lens into. It’s not all, but it’s most.

ConvergenceRI: The number of people enrolled in CurrentCare, is it 40-50 percent?
SARKAR: It’s over 50 percent; we have [roughly] 537,000 individuals enrolled.

But even for those individuals who are enrolled, the part of the chart that we have is, in many ways, dictated by the type of EHR system we are interacting with, and what each of the providers are technically capable of pushing out.

ConvergenceRI: Another question, related to the APCD, is whether or not it captures denials of claims. You have all the procedures that have been paid for. What happens when a claim is denied? Is there any way to capture that data?
SARKAR: It is an interesting question. I don’t know the answer.

ConvergenceRI: And, is the rate of denial higher with Medicaid or Medicare, or commercial insurance?
SARKAR: I don’t know the answer. That is a really interesting question.

And it gets to another really important facet of clinical data. Clinicians write down what they observe and what they see, not what they don’t see and don’t observe. That’s actually a very important bias that we have in clinical data.

It’s just what the clinicians, at that moment in time, notes as what is important for that encounter.

But there may be a lot that goes unsaid, that doesn’t get recorded. And that gets back to one of the reasons why there is such a challenge with Artificial Intelligence and its potential use in health care, because the data, by definition, are incomplete.

ConvergenceRI: How important do you believe the work of the study commission on rate reimbursements is?
SARKAR: What Sen. Miller has put out in front of us is really just the very beginning of what we together hope is an important dialogue and discussion around what needs to be done.

What we are trying to take on in our state is really trying to revolutionize how health care operates. And, frankly, it’s because we don’t have a choice. We all know that. Not just in our state but in our nation.

I’ll say this with the risk of getting in a lot of trouble. If we can’t make some significant advances in health care in health care in Rhode Island, the rest of the country is in really, really deep trouble.

ConvergenceRI: Is that because Rhode Island has a relatively stable population of 1 million residents?
SARKAR: That’s part of it. We’re 1 million and we’re a nice snapshot, not a perfect snapshot, but a nice snapshot of diversity. Also, we have a limited number of places where a person get care, one medical school, a limited number of schools of nursing, one accredited school of public health, and a true sense of community.

It’s what makes Rhode Island unique. There’s a community in Rhode Island. We are in this together, and we are going to figure this out together.

 

View article on ConvergenceRI