For years now, large integrated health systems such as Oakland, Calif.-based Kaiser Permanente and Intermountain Healthcare in Salt Lake City have been combining clinical and claims data for population health analytics. In today’s era of health reform, with many providers taking on more risk, even health systems without provider-sponsored health plans are starting to integrate payer data.
One example is Carolinas HealthCare System in Charlotte, N.C., which is taking advantage of one of Verisk Health’s population management solutions in Premier’s PopulationAdvisor suite, which manages and analyzes both claims and clinical information to help providers make better financial, operational and clinical decisions.
It is still unusual for a provider that is not part of an integrated system to be able to combine claims data and clinical data, notes Michael Dulin, M.D., Ph.D., medical officer for analytics and outcomes at Carolinas. “We are trying to understand the total picture with our patients across a spectrum of care,” he explains. “As a large health system, we do capture a lot of that in our network from primary care to the hospital setting. But still, we want to know what happens when our patients go to emergency rooms outside our network or while traveling. Some provider organizations are payers themselves, but our thinking is that through a partnership with the payers, we can get a better picture of the total cost of care and get a 360-degree view of the patient.”
Michael Dulin, M.D., Ph.D.
Carolinas negotiates with commercial payers such as Aetna to share this data, Dulin says, and his team is working on a grant proposal to request Medicare data for these same purposes.
Dulin says providers often get regular reminder letters from multiple payers. “They might say you have five patients covered by our company with reminders about screenings or other best practices. That is helpful, but these paper notices from multiple companies are difficult to coordinate with workflows,” he says. “What we are doing is getting all that information together in one place and at the point of care in a decision support tool. We are taking our own clinical data and matching it with the claims data under a single interface. The hope is that it will drive down unnecessary tests, such as mammograms, if the provider can see that one has been ordered outside the network. And they can see whether a patient has actually filled a prescription.”
The most important aspect to the payers, Dulin explains, is to help Carolinas bring costs down and improve care. “I think it is a win-win if we can avoid redundant testing and variability in care,” he adds.
Keith Figlioli, senior vice president of healthcare informatics for the Charlotte, N.C.-based Premier Healthcare Alliance, says it is critical to pull claims and clinical data together. “The data the payers have is a mile wide but an inch deep. The providers’ data is deep but narrow. You need the nexus of the two.”
CHALLENGES PERSIST AMID PROGRESS
Although data-sharing between payers and providers is on the increase, there is still tension in the relationship. “The single biggest challenge in negotiating agreements on new types of contracts is who has access to what data, and there is something of a data grab under way,” Figlioli says.
“We have concerns about giving the data to payers, because then the payers are controlling the care processes. I think that for the population health component, where you are getting paid for outcomes, you should control the data repository.” He says his organization counsels its provider organization members that they should own their own data warehouse and analysis infrastructure. (Premier has an enterprise data warehouse offering in the cloud for members.)
One physician executive who agrees with Figlioli is Scott Hines, M.D., co-chief clinical transformation officer for Crystal Run Healthcare, a multispecialty physician organization in the mid-Hudson Valley of New York that is one of the fastest growing accountable care organizations (ACOs) in the country. “You don’t want to be at the mercy of trusting data that is provided to you. You need the ability to do the analyses yourself,” he says. Crystal Run has a business intelligence team of six and has created its own tools to gather data on 70 quality measures.
SIDEBAR: A Joint Effort Focused on Big Data, Research
Most of the efforts to join clinical and claims data involve pay-for-performance contracts between payers and providers. But the health services organization Optum, a subsidiary of UnitedHealthcare Group which is also the parent of payer UnitedHealthcare, founded Optum Labs in 2013 as an open collaboration center. Initially working with Mayo Clinic, Optum Labs, based in Cambridge, Mass., is expanding to include other stakeholders across the ecosystem of payers, providers, academics, life science companies, and consumer organizations. “All have an interest in improving the delivery and value of patient care,” says Paul Bleicher, M.D., Ph.D., CEO of Optum Labs.
Both Mayo and Optum realized that they had substantial amounts of data, but many participants in the healthcare system don’t have access to this type of data. “By combining this data and involving other providers in the ecosystem, such as medical device companies, we can create cross-industry organizations that wouldn’t normally work together,” says Bleicher, who was formerly chief medical officer for Humedica, a clinical informatics company that was acquired by Optum in January 2013.
Paul Bleicher, M.D., Ph.D., CEO of Optum Labs.
It took Optum Labs a while to get up and running, but it is starting to see interesting insights, with two peer-reviewed papers already accepted for publication. And because Mayo is an early and active participant, it can quickly bring to its clinical setting new insights about issues such as anticoagulant usage and hip and knee replacements. An example of the work being done is a focus on better predictions of which patients with congestive heart failure will be readmitted to the hospital.
“You also have to realize that putting data together is just the first step,” Bleicher says. “Ideas about causalities and pathways are actually just ways to generate hypotheses. You have to do the big science work to know what you are looking at, because there are many errors you can make going on assumptions.”
Crystal Run has spent several years looking at quality and cost data in its own outpatient offices. “We have been able to lower overall per-patient cost by 7 percent over three years by doing that,” Hines says. “But to do an even better job, we need that claims data to get information on things such as hospital lengths of stay and pharmacy data.”
Crystal Run’s quality measures in risk-based contracts with payers include whether a patient has had a mammogram in the last two years. “We need access to all that claims data in order to accurately report on the quality measures we are being held to,” Hines says. The largest payer his group has been working with is the Centers for Medicare and Medicaid Services (CMS) and it has been very good at sharing the data in a timely fashion, Hines says; however, as it converts contracts with commercial payers to risk-based contracts, getting the data Crystal Run needs can be a challenge. “We started one risk-based contract July 1 but didn’t receive the claims data about patients attributed to us until Nov. 1, so we have only two months to move the needle,” he says. He believes there is some reluctance on the part of some payers to share the data. “There is some thinking on payers’ part that if they are not in charge of this data, what are the insurers for? I think they are nervous that if they share too much data they risk becoming irrelevant,” Hines says. “Of course, there are others that are eager to collaborate and reshape how care is delivered.”
PROGRESS IN MICHIGAN
One place where the spirit of collaboration seems to be flourishing is Michigan, where the Michigan Value Collaborative (MVC) is providing decision support for hospital leaders using both claims and clinical data.
Blue Cross Blue Shield of Michigan, the largest private payer in the state, has funded a longstanding program called Value Partnerships, which encompasses several different initiatives, including collaborative quality initiatives. Continuous quality improvement (CQI) addresses some of the most common, costly areas of surgical and medical care by facilitating collaboration between Michigan providers. CQI registries permit a more robust analysis of the link between processes and outcomes than can be achieved by examining one group or institution alone. These efforts have done a remarkable job of moving the needle on technical quality, says John Birkmeyer, M.D., a professor of surgery and director of the Center for Healthcare Outcomes & Policy at the University of Michigan and director of the MVC.
The MVC is going one step further to look at cost efficiency and to integrate private payer claims data. “We started with performance measurement using exclusively Blue Cross claims, but we are in discussions with the other biggest private payers and Medicaid next year,” Birkmeyer says.
The early work of integrating granular clinical data with claims-based data has already led to some interesting insights. “There is a remarkable and as yet unrecognized variation in how efficiently hospitals deliver care,” Birkmeyer says. For instance, of the 29 Michigan hospitals that do coronary artery bypass, there is a $20,000 spread between the most and least efficient hospitals for episodes of care. “The hospitals didn’t have an appreciation of that,” he says.
Launched in March 2013, the MVC is helping to look at entire episodes of care. What is done before a surgery and after? How does it correlate with outcomes? “The goal is to help providers yield better outcomes at lower cost,” says David Share, M.D., M.P.H., senior vice president of value partnerships at Blue Cross Blue Shield of Michigan. It is too soon to say exactly how the hospitals will use the results, he says, but as hospitals are being held accountable for population health performance in value-based contracts, they are getting interested in the overall value of an episode of care and what works to optimize value. “For instance, after joint repair, there is a wide variety in how long people stay in the hospital, whether they move to a sub-acute care center, have home care, how much rehabilitation they do,” Share says. “There is no evidence that more intensive care in those cases have better outcomes.”
Birkmeyer has found that providing claims-based performance feedback is much more meaningful to hospitals when it is tied to clinically rigorous registry data. “If it is exclusively a claims and cost analysis, it is difficult for conveners to get past sources of pushback,” he says. First, they say ‘your data is crappy, and my patients are much sicker than those you are comparing us to.’ The clinical registry data is essential to debunk that argument. Second, hospitals will say, ‘Yeah, but our quality is so much better than that of other hospitals.’ That is why the comparative registry data is so important.”
Birkmeyer co-founded a company called ArborMetrix that is helping MVC with its analytics. Its CEO, Brett Furst, explains that with more risk being put on providers in pay for performance contracts, there is an underserved need for analysis of episodic care in the acute setting. “What we go after is that clinically granular data in acute and specialty settings,” he says. “With just billing data, it doesn’t do much good to determine that one health system charges 10 percent more than another for a specific type of care,” he explains. “Without the clinical data, you can’t address the why. Maybe the outcome is better. But for some ACOs, we are seeing two providers that are $25,000 apart on some episodes of care and the lower cost provider is reporting better outcomes.”
Keith Blankenship, vice president of technical solutions at St. Louis-based Lumeris, a healthcare consulting and software services firm, agrees that some efforts at clinical quality improvement that involve only claims data struggle. “We worked with one group that wanted to highlight uncontrolled hypertension. But without the clinical data, they couldn’t identify uncontrolled hypertension,” he says, “so they had to wait until they had both types of data in the system.”
Lumeris offers a cloud-based delivery system that pulls together EHR and claims data, and has a master patient index and analytics tools.
Blankenship adds that now that payers and provides are sharing data more frequently, they are gaining a better understanding of how powerful the combination can be. “Our platform supports multi-payer models driven by providers,” he says. “Then it becomes like a matrix, where the provider sees his or her own whole panel, but each payer sees just its customers’ data.”
Baptist Memorial Health Care in Memphis, Tenn., is working on a coordinated care initiative that involves incorporating claims data from Aetna with the data in its Epic EHR.
Previously, physicians had localized data on the patient: whether they kept an appointment or were admitted to the hospital, but they had no view beyond that outside their four walls, explains Paul DePriest, M.D., chief medical officer for Baptist Memorial. “The data from Aetna can give us a global view for a population.”
In the past, Aetna would send out letters to physician offices with reminders about their members, and so would other insurers, but physicians might not have been able to deal with that information in a structured way, explains Bob Kropp, M.D., regional medical director, accountable care solutions, for Aetna. “Now with EHRs, re-engineered work flows and care coordinators in physician offices, doctors are finding efficient ways to use the data, and they are seeing the value and validity of it.”
Clinicians have traditionally chafed at having data sets presented to them, DePriest adds, “but there is more potential value in having the data in clinical decision support tools sitting on top of Epic to remind them about high-risk patients.”
SIDEBAR: Analytics Tools to Collaborate on Claims Administration
Analytics can help improve care delivery, but it can also help with the main point of contention between payers and providers: claims administration.
When Juan Davila, executive vice president, healthcare quality and affordability, first came to Blue Shield of California in 2006, he saw right away that relationships with providers were terrible. “It is an industry-wide problem,” he says. “There is a disconnect and it is an adversarial relationship with lots of finger pointing. We would go into meetings and providers would say, ‘We have $10 million in claims you owe us and haven’t paid.’”
To address this problem, Blue Shield worked with performance management company MedeAnalytics (Emeryville, Calif.) to create a web-based tool where providers could look at the raw claims data. “This opened the kimono and gave them both the data and slice-and-dice capabilities to get at the root of the problems,” Davila says.
Today, every hospital in the state is participating in what Blue Shield dubbed the Partnership in Operational Excellence and Transparency (POET). Quite often, Davila says, claims are denied or delayed because data is coded the wrong way. “It is mistakes that both the hospitals and insurers make. I consider that waste, and it is fixable,” Davila says. Using the tool, Mission Hospital in Mission Viejo went from 30 percent denials to the low double digits, he adds.
“This is about transparency,” Davila says. “We can both look at and improve our processes. And it is not just based on anecdotes. It gets the emotions out of the way. These partnerships help lay the groundwork for ACO work.
Davila says the POET team’s work is never really done. “We have ICD-10 going in and the POET team will have to work on the coding differences for that. “This is a tool that allows for dialogue,” he says, “and we have to keep that dialogue going.”