The journey into population health management is a long, complex, and challenging one, regardless of the specific advantages any patient care organization might bring to it: that was the key message delivered on Sep. 17 by Rick Skinner, the chief information and technology officer of the University of Virginia Health System, which encompasses an academic medical center, medical school, nursing school, and other entities, all based in Charlottesville, Virginia. Skinner delivered a plenary session address entitled “Managing the Health of Populations: Many Approaches,” at the Health IT Summit in Vancouver, being sponsored by the Institution for Health Technology Transformation (iHT2, a sister organization of Healthcare Informatics), and offered his audience a nuanced view of the development of accountable care organizations (ACOs) in the United States, from the perspective of the senior healthcare IT leader at a health system that has developed one.
In his initial comments to the binational audience at the Rosewood Hotel Georgia in downtown Vancouver, Skinner began by asking a question that lies at the heart of ACO development in the United States. “With so much talk about population health management in the U.S., Canada, and across the world,” he asked, “are we really talking about managing the health of a population, or is what we’re really addressing the utilization of healthcare services, by the members of that population?” In fact, Skinner offered, utilization management really is at the core of U.S. efforts in this important area. “Oh, yes,” he said, “we want everybody to be healthy; but what we really want to do is to reduce their expenditures of expensive healthcare services, not because we’re cheap or mean, but because we have to reduce the overall cost of healthcare. And every jurisdiction I know of, including Canada, is facing that problem. No one has a model of the cost, quality and service needed” from any overall national healthcare system.
Describing his organization’s plunge into the ACO world, through its creation of Well Virginia ACO, which was created under the aegis of the Medicare Shared Savings Program (MSSP) for ACOs, and which takes care of 20,000 Medicare beneficiaries, Skinner noted that there are complex reimbursement strategies at play for an organization like his, “a traditional academic health center whose main bread and butter is very sick patients needing very specialized care. However, against the backdrop of the evolution of the U.S. healthcare system and payment model,” he quickly added, “we knew we had to gain some experience managing the health of an entire population. We had some advantages having an employed physician group and house staff; we don’t work with a large number of payer organizations; and we’ve had a comprehensive electronic health record with wide-scale adoption, for over five years, so we had the tools to do this,” he noted.
The strategy, Skinner explained to his audience, was to reduce fee-for-service payments from the Medicare program by somewhere between 2.0 percent and 3.5 percent, knowing that if successful, the University of Virginia Health System could lose up to $5.3 million in Medicare fee-for-service reimbursement, but hoping that, “Along with an overall reduction in fee-for-service Medicare volume, if there’s unmet need” in the communities the health system served, a gain in “some specialized patients” being able to be treated within the health system’s capacity, “would offset that.”
Of course, he noted, “The challenge with all these different payment models is aligning the incentives of the entire organization when going from a volume-based model to a model that has more to do with the longer-term provision of healthcare to members of a population.”
An early analysis of Well Virginia’s ACO population revealed some interesting findings, Skinner told his audience. That was particularly true of his and his colleagues’ findings around chronic illness. Looking at three of the most prevalent chronic diseases, they found that over 30 percent of the ACO’s population had been identified as having diabetes; 18 percent had identified congestive heart failure (CHF); and 68 percent had identified hypertension. Meanwhile, 25 percent had both diabetes and hypertension; 15 percent had CHF and hypertension; 9 percent had diabetes and CHF: and 8-9 percent had all three major chronic illnesses—a daunting prospect indeed.
Not surprisingly, Skinner and his colleagues found that “The top 10 percent of the people in that cohort” of those ACO members identified as having one or more of those three chronic diseases “account between 70 and 80 percent of the total costs of the cohort.” Interestingly, he noted that when he worked for an Ontario heath system, Skinner and his colleagues had found similar results with that health system’s population (and the Well Virginia ACO’s statistics generally match those of other health systems around the U.S., regarding the percentage of healthcare costs for the top 10 percent of patients.
So what does the treatment of patients with those diseases cost Well Virginia ACO on an annual basis? The following, according to Skinner: diabetes: $47.4 million; CHF: $59.8 million; hypertension: $95.2 million. Treating ACO enrollees with both diabetes and CHF costs Well Virginia ACO $27.8 million annually; treating those with CHF, $59.8 million; those with hypertension, $95.2 million; those with both diabetes and CHF, $27.8 million; those with both diabetes and hypertension, $41.2 million; those with both CHF and hypertension, $51.8 million; and those with all three chronic illnesses, $25.3 million annually.
Plunging into ACO development
Skinner told his audience that he and his colleagues made the decision early on to “jump off the deep end and get into this business,” meaning ACO development. That decision was followed up by nine months of strategic planning and analysis, he noted.
What essential IT needs were uncovered in the development of Well Virginia ACO? Skinner told his audience that there were five key needs: data acquisition; data analysis; the design of clinical interventions; the ability to apply interventions through their unified electronic health record (EHR); and the ability to measure the outcomes based on those interventions. “It wasn’t just enough to describe the population last year or the year before,” he noted, “but instead that population this year, and in particular how an individual in the cohort might behave in the future. And once you know that, you want to design some interventions to prevent that ED visit, that inpatient stay, that expensive specialty drug, from being necessary. Nobody wants to go to the ED who doesn’t have to, nobody wants to be an inpatient who doesn’t have to, and nobody wants to pay for an expensive drug they don’t need,” he said.
And, he said, “The EHR is the tool set by which to apply whatever those interventions are that need to be designed. And we heard loud and clear from all our practitioners, if it isn’t in the EHR, I’m not using it. If it’s not in my face while I’m treating a patient, I’m not going to link out to access information; it has to be in the tool set I’m using at the point of care. And we needed the ability and will to assess the effectiveness of interventions—the doctors needed that.”
Very importantly, Skinner told his audience, “Our opinion about the tools we would need changed dramatically as we proceeded. And I was one of the people who had thought we had to go out and buy a population health management solution, for a bezillion dollars.” In fact, he reported, he and his colleagues discovered that “we had he data we needed, and we already had an EHR to help us message our caregivers. So it turns out we had what we needed, tool-wise.” And, he said, in terms of being able to act on the data that they’ve collected and analyzed, is a work in progress. “We’re now in the end of year two,” he said. “We’re not really good at it yet, but we’re getting good enough.”
Among the key advances he and his colleagues have made around leveraging data has been to move forward to marry claims and EHR-derived clinical data, while at the same time pushing data out to the physicians closer and closer to real-time, so that doctors could act on it. “You want to identify your people in your population; the stratify them with regard to risk,” he said. “But then what you really want to do is to be able to predict which people will get sicker and which will require more interventions,” and then additionally to empower teams of care managers to intervene to avert more costly clinical interventions from taking place.
Among the most exciting innovations so far has included the development of what Skinner and his colleagues call a “Relative Readmission Risk Monitor,” which leverages a statistical learning algorithm to predict which plan members are “most likely to be seen in an ED or be hospitalized in the next so many days,” identifying the risk of inpatient readmission within the next six months, the risk of an ED visit within the next six months, and predicting the costs of care utilization.
That analytics-driven work has also been initiated along with the implementation of a new-ish “Care Coordination at Home” program, which Skinner described as being “designed for patients with discharge from the hospital with AMI [acute myocardial infarction/heart attack], CHF, COPD [chronic obstructive pulmonary disease], or pneumonia identified prior to or at discharge.” That program has involved the installation of such devices as scales, blood pressure equipment, glucometers, and other devices, along with close monitoring and potential intervention by care managers. Through such work, Well Virginia ACO has reduced the readmission rates for four common diseases by 12.4 percent altogether.
Doing all this important work, the University of Virginia Health System remains at risk financially, Skinner noted, as organizations like his, as they plunge into accountable care work, have to carefully balance progress in population health-driven care management with maintaining financial survival in its remaining fee-for-service reimbursement. “Our hospital system’s largest group of patients includes those who are very sick and who take a lot of money to care for,” he emphasized. “And that’s the calculus. In the U.S., that’s the valley of death we have to get across. You’re going to lose significant business in your current model for some period of time, unless you can become self-sustained to bridge over to the new model. That’s the challenge,” even with the best data analytics tools, deployed in the best way in order to make gains on the at-risk side, he emphasized to his audience.