Like everyone else in the space, Clarify Health President and Co-Founder Todd Gottula is excited about how advances in artificial intelligence are reshaping the healthcare analytics sector. In a recent interview with Healthcare Innovation, Gottula spoke about how his healthcare analytics software company is beginning to incorporate generative AI into its solutions.
HCI: Todd, before we dive into the impact of AI, could you talk a bit about your background and how you came to co-found Clarify?
Gottula: My background has been in enterprise software development for all of my career. Prior to founding Clarify, I was the chief technology officer of what was at the time the world's largest independent provider of financial services technology, a company called Advent Software, where we were delivering portfolio accounting, trading and network connectivity tools to about 3,500 of the world's financial services institutions. I have had an extensive career building enterprise-scale applications, but outside of the healthcare industry.
After selling Advent, a group of engineers and I were looking for something with really significant social impact. We had the opportunity to found Clarify with a physician coming out of McKinsey as a senior partner, Dr. Jean Drouin. He had the vision to say there's going to be more and more data assets becoming available. Data liquidity is going to go up in the healthcare industry. There's an opportunity to take technologies that had been developed in other industries and bring them into healthcare to finally unlock the promise of having national-scale insights on historical care delivery, for the purpose of identifying opportunity and then making recommendations for those who are responsible for care journeys or the provisioning of care plans for better benefit. Ultimately, we exist to help organizations manage overall medical spend and ensure that providers are connecting patients with the most appropriate care provider that can give them efficient and appropriate healthcare outcomes.
HCI: You work with health plans and provider organizations. Could you give examples of how you would work with them?
Gottula: On the provider side, we're really focused on connecting patients to the most appropriate provider. That's helping organizations understand market share; it's helping them understand the referral patterns, and it's helping them understand the rate data — what services cost so that either a primary care physician or a specialist who's looking to refer on into a site of care for a surgery or subspecialty knows exactly what the landscape of available providers looks like, what the historical quality has been of those providers for patients like the one that they're making a recommendation to, and ultimately, they can make an informed decision about what's the most appropriate next step in a patient's journey. We've worked with health plans and large physician groups to give them this access and transparency into the rest of the care providers and sites of service in their geography.
For plans and health systems that are also taking risks, what we effectively do is give them a view of where their opportunities are to drive improvement in the context that matters to them. It could be quality incentive programs; it could be where they're at risk and they need to drive down medical spend. It could be where they've got utilization patterns that they're trying to adjust to other patterns that they've determined are more appropriate for their members or for specific patient cohorts.
HCI: As you work with accountable care organizations, what are some of the biggest challenges they tend to have in working with their data or getting access to data? And what kind of insights does Clarify help unlock that they couldn't have seen before?
Gottula: This is a gross oversimplification, but the first thing that organizations struggle with is how do they unlock the insights that are latent in the data they already have. Accountable care organizations are typically going to get claims experience in addition to the clinical data they have for the care that they're providing to their patients, but that's a lot of data. And these are health systems, right? Their business is providing healthcare; it's not to build systems that can unlock insights. That's my job. They have really talented IT organizations and development organizations, but they can't bring the scale of what an organization like Clarify can to that problem.
The first order of business is: how do you identify what's happening in your health system with the data that you already have? And the same thing goes for health plans. They sit on a tremendous amount of information, and getting the insights out of that data is a significant challenge for them. Step two is saying, how do I identify opportunity? Where is there positive and negative variation in my underlying observations that I've gleaned from the data that I have? Identifying healthcare variation is driven by having enough information to be able to say with confidence that this cohort of patients to which this specific member is a participant typically performs like this. Well, even the largest health systems don't have enough experience to truly identify variation. That's where nation-scale data is the second pillar of how you deliver value.
The third is where generative AI is unlocking significant value for us, which is the needle in the haystack problem. I've got my observations for my data. I've got national benchmarks. But how do I go figure out what matters most to me? That is the key, because some people are focused on driving quality improvement; some people are looking for reducing certain types. of high-dollar therapy use. Others want to drive people to use ASCs [ambulatory surgery centers] instead of acute-care hospitals. When you have the context of the question, then you have to dig into the data, and be able to understand what are the practice patterns that seem to be having success that I can then drive others to model towards. Generative AI is the missing link to then be able to take these two very large data assets — observations and national benchmarks — and drive value or unlock value that you can then drive through your system.
HCI: Clarify just announced a beta program for a generative AI tool called Clara. What does it enable that other types of analytics wouldn’t?
Gottula: The way to think about what we've developed is an accelerant for the existing analyst who is tasked with going into large data assets and using traditional SQL-type tools to be able to make recommendations to their organization about where they can improve. That task is often one of exploration. You test the hypothesis and dig into the data asset and then validate that hypothesis against other opportunities, data assets, and so forth to finally make a recommendation and then link that recommendation to fiscal benefit for whomever it is that we're ultimately trying to drive better outcomes for. That process can be reduced by over 90 percent by putting a generative AI interface on top of these tabular datasets. It's really incredible. An example use case is: I'm going into a rate negotiation as a provider with a payer. Tell me in my market where I am being most underpaid relative to my peers, but the quality of my care is superior. And you will literally get out of our interface the list of service codes, DRGs, and so forth, that you should go in and renegotiate with your payer because you're delivering higher quality care for and you're not getting rewarded for the work that you're doing.
HCI: And if that provider group were asking that same question without the AI tools, say a year and a half ago, what would it take to come up with the same answer?
Gottula: Hours of lots of digging into the underlying data. Well, a year and a half ago, without the No Surprises Act, the rate data wouldn't be available. That data is available, which has taken significant computing and engineering time to aggregate, which we've done, but now I need to dig into all the places where there is service code variation on price. I’d need to build a query that says, ‘Where am I the lowest price by over 10 percent relative to this peer group? Okay, great. I've got the service codes; now I have to link that to performance data, to be able to ask: where am I a high performer on a subset of those codes where I've got this underpricing relative to what is apparently the market price for this payer? You're talking hours of work, if not days of work, digging into large datasets. With Clara’s generative AI experience on top of the same data assets, it's literally minutes to be able to answer these questions. And it makes the analysts so much more productive because now instead of spending hours and hours digging into the data, they can spend that time saying: this is what the proposal should look like for us and why we should justify this side of the rate negotiation in order for us to ensure that we're getting paid fairly for the work that we're doing.
HCI: Do you have a group of customers beta testing this already?
Gottula: We have a small set of providers and a small set of payers who are using this data for their respective purposes, to be able to interrogate both provider performance data and this aggregated rate data that we've pulled together under the No Surprises Act data availability.
HCI: Clarify is a CMS Qualified Entity. Does that grant you access to certain datasets, but also confer some responsibility to publish reports on it?
Gottula: The Qualified Entity program is a really incredible program where CMS entitles organizations like Clarify that qualified through a series of rigorous security reviews, and a validation of our intent to use the data for essentially public good. It can be used for commercial purposes, but those commercial purposes need to be driving better outcomes, reduce medical spend, better access, things like that. They grant us access to the entire Medicare claims repository and essentially the Medicare membership roster. In return, we need to produce periodic reports that show interesting insights that then become publicly available around utilization of healthcare. And the unique requirement is that it has to be a combined Medicare and commercial data analysis. So in order to even receive the Qualified Entity data, you also have to have sufficient commercial claims data in order to be able to do these Qualified Entity reports that we publish.
HCI: I saw that a couple of examples were reports on patient safety measures and mental health utilization among children and adults. Who are intended audiences? Policymakers?
Gottula: Exactly. Part of the legislation was to help inform overall healthcare policy. We're incredibly proud of the work, although it is sobering: the insights that the Clarify Health Institute surfaced from looking at the under-18 incidents of depression and the various forms of depression and the utilization or lack thereof of healthcare services for those individuals. It is really an epidemic-level crisis that we're facing, particularly during the COVID period where children faced that isolation, and it's continued. What that report is intended to do is to signal for policymakers that this needs to be prioritized, that we need to put effort behind ensuring that the youth in this country have access to mental health services, because they desperately need them. And that is not an area that is typically prioritized right now. It's not something that your typical pediatrician is going to readily recommend. There's a lot of continued stigma in this country around access to mental health services.
HCI: Now that you've been working in healthcare space for several years, are you optimistic that the focus on the value-based care movement is going to bend the cost curve dramatically?
Gottula: I love what you said there, which is bend the cost curve, because I incredibly naively came into this industry and said the words that I often hear other people say, which is that we're going to reduce the amount of money that's spent in the U.S. healthcare industry, and we're going to make those dollars available for other things. And I have learned through a lot of blood, sweat and tears that that's wrong. That's not the way we should be prioritizing. Because behind every dollar, if you follow those dollars far enough, there's a job. Somebody is being employed through the distribution of these healthcare dollars, whether it's a doctor or it's somebody who's building the MRI that eventually gets sold to the hospital, and then gets used for services. To say that you're going to reduce the medical spend means you're going to put a lot of people out of work, and what are those people going to go do? That's the wrong way to think about this problem. We need to slow the growth of healthcare spending.
Our tools that help settle value-based care contracts and therefore implement them at scale are very valuable tools, but they have to be taken i as part of a complete system to say that we need to make sure that referrals are optimized. We need to make sure that people have access in the first place, which means their holistic condition needs to be understood, not just medically, but also social determinants. And we have to be willing to say that physicians need proportionate income relative to their outcomes, and that last piece is the area where we have to spend much more time as a country and as a healthcare system, focusing on. That's ultimately why I believe in value-based care programs because most other industries are not just based on volume, right? It's also the quality of service I'm offering. Moving to a quality-based incentive program in the healthcare industry is something we absolutely have to do and that's how we’ll eventually keep medical spend under control.
