As patient care organizations engage in more rigorous and challenging risk-based contracts, the need to leverage data and analytics to support value-based care delivery operations is becoming paramount
Industry sages have been predicting for years now the time when all the processes around the leveraging of data analytics would become fundamental to operational and financial success among the leaders of U.S. patient care organizations. Well, with the emergence of more and more risk-based contracting, including two-sided risk, that moment has clearly arrived.
Indeed, the August 9 announcement on the part of Seema Verma, Administrator of the federal Centers for Medicare and Medicaid Services (CMS), of a proposal dubbed “Pathways For Success,” appeared to mark an inflection point in the U.S. healthcare system’s journey around risk-based contracting. That proposal, if finalized, would remove the traditional three tracks in the voluntary Medicare Shared Savings Program (MSSP), and replace them with two tracks that eligible ACOs would enter into for an agreement period of no less than five years: the BASIC track, which would allow eligible ACOs to begin under a one-sided model and incrementally phase in higher levels of risk; and the ENHANCED track, which is based on the program’s existing Track 3, providing additional tools and flexibility for ACOs that take on the highest level of risk and potential rewards. That proposal has received very mixed reviews among patient care organization and ACO leaders, with some ACO leaders pushing back on CMS, asserting that two-sided risk is still too difficult and challenging for many provider organizations.
But regardless of the near-term and middle-term fate of “Pathways For Success,” industry leaders fully recognize that the U.S. healthcare system is inevitably headed further into risk-based contracting, including into two-sided risk-based contracting—however long it might take to reach a tipping point, in which the majority of patient care organizations are receiving significant percentages of their reimbursement via two-sided risk.
A Lot of Basic Blocking-and-Tackling Obstacles Remain
Of course, the urgency on the part of the purchasers and payers of healthcare to push providers into two-sided risk makes absolute economic sense, says Don Crane, president and CEO of America’s Physician Groups (APG), a Los Angeles-based, nationwide association of physician groups involved in risk-based contracts. “Upside risk makes sense for a while, baby steps,” Crane says. “But it’s weak tea in terms of driving real change. If you get lucky with the right benchmarks, you can do well. But it doesn’t induce real structural change. But when an organization faces downside risk, also known as bankruptcy—that really forces change. When you take that higher risk, you have to have the data. You have to risk-stratify your population, and treble down on the resources you’re using on the patients at highest risk.” And, he adds, “All of a sudden, you move into the big leagues.”
Still, despite the fact that the purchasers and payers of healthcare are pushing providers forward to take on more risk and downside risk in their contracts, the leaders of patient care organizations remain challenged by many rather basic blocking-and-tackling types of data management and analysis challenges. As Liam Bouchier, principal advisor at the Naperville, Ill.-based Impact Advisors consulting firm, puts it, “The data challenge is not a new challenge; the biggest challenge is still getting a complete and comprehensive view of the patient. It’s only been in the past five years that information has really begun to flow. When you look at a longitudinal view of the patient’s care, it’s important to see what’s happened between hospital stays,” he adds.
Fred Horton, president of AMGA Consulting, the consulting arm of the Alexandria, Va.-based American Medical Group Association (AMGA), a nationwide association of large medical groups, says that “There are a couple of different areas” of fundamental concern, as provider organizations move to leverage data analytics in order to manage risk. “Number one is the investment in the EHR [electronic health record]. A lot of health systems are roll-ups of multiple organizations, so you may have multiple EHRs to gather data from. There are EHR-agnostic tools that can analyze data coming from multiple EHRs. That trend will continue into the future. But then there’s the piece involving how much a particular patient is seen by a particular health system, and that’s a challenge. Let’s say some of the urgent care visits and some of the primary care visits are happening outside your health system. How do you ultimately obtain all the data, so you can analyze what’s going on?”
What’s more, he notes, the amount of relevant data in any organization’s EHR “could be less than 25 percent of all the data would be necessary to understand what the potential for risk is. The EHR might capture two or three medical visits a year, but that might actually be a minor element in what you need to truly manage under a risk-based or value-based environment.”
Towards “Generation 2.0” with Data: Predicting First Hospitalizations
The good news in all this is that the leaders of some of the most advanced multispecialty physician groups have steadily been learning how to leverage data, says John Cuddeback, M.D., Ph.D., chief medical informatics officer at AMGA Analytics, the analytics-focused division of AMGA. What have they learned? “I think the fundamental strategy for managing population health depends on risk stratification and matching what you’re doing for each individual person within the population, with their needs. And if you can use data to figure out where they fall on the risk continuum and proactively address primary prevention, secondary prevention, or monitoring of chronic conditions needs, that’s the big opportunity,” Cuddeback says. “And in the very first generation of population health involved clinical decision support in the EHR,” medical group leaders gathered together data on the diseases of their patients with chronic illnesses, and used clinical guidelines to manage their care.
John Cuddeback, M.D., Ph.D.
“The next thing,” Cuddeback says, “was registries, finding the patients who are not coming in, and need a screening or preventive service or something, that’s something like Generation 1.1. Generation 2.0 is using longitudinal data on populations to create predictive models. We know what’s going on for a patient at a particular time, and we know what will probably be going on in two or more years. So let’s take advantage of that and create good predictive models. We need to find the rising-risk patients and address their needs.”
Very importantly, Cuddeback says, “One of the big advantages of having clinical data is that you can predict an initial hospitalization. Historically, people have used claims data well to build predictive models. But the weakness with claims data is that you’re building models based on historical data, meaning you’re looking towards potential readmissions. So predicting initial admissions is the new thing that people are working on. But the implementational aspect of this is that you have to become much more proactive in your care design. Because if you’re just waiting for people to arrive, the predictive model does you no good. So the predictive model will help you figure out who your rising risk populations are. But we have to change our care delivery and management models.”
Moving Forward in Massachusetts: MACIPA’s Path
The senior executives of physician groups who have been working in the trenches in value-based healthcare delivery and payment are working through a series of challenges. One medical group that exemplifies intensive effort in the face of those challenges is the Mt. Auburn Cambridge Independent Practice Association, or MACIPA, a Boston-area, 500-physician independent practice association that participated in the Pioneer ACO Program for three years from 2012 through 2014, and is now in the Track 3 section of the MSSP program. MACIPA president and CEO Barbara Spivak, M.D., a primary care physician, says flatly, “I don’t think you can do risk-based contracting without data. I’m not even talking about the business side; on the clinical side, you need to know who your sick patients are, and we all know that costs go up when there are gaps in care. So you need to know who your sick patients are, where they’re going, and where the gaps in care are. And that’s even before quality management.”
What’s more, Spivak says, “The physicians need actionable data. It doesn’t help me that my colon cancer screening rate is 50 percent and needs to be 60 percent; as a PCP, I need to have the patients identified and to bring them in. The issue is not what we need; what we need is similar to what we needed when we started with electronic records and population health 15 years ago. The challenge is that the data is exponentially more.”
Barbara Spivak, M.D.
Spivak says that she and her colleagues have had no regrets whatsoever about participating in the Medicare ACO programs, despite the frustrations—including that, as of mid-summer of this year, they still had not received key performance data from CMS on their 2017 performance, for example. Indeed, she says, the reality of the complexity of the MIPS (Merit-based Incentive Performance System) program that now governs physician payment under Medicare for those physicians in practice not involved in ACOs and other authorized alternative payment models, is forcing physicians forward. “I believe the way they’ve structured MIPS and MACRA [the overarching law, the Medicare Access and CHIP Reauthorization Act of 2015] is way too complicated for anybody but someone who’s in a big system, where the system’s going to take care of it. If you’re in a small private practice on your own, it’s going to be virtually impossible. It is probably succeeding in pushing people into alternative payment models.”
Moving into Social Determinants Data Management
For years, it’s been a truism that one of the key challenges in harnessing data to support value-based healthcare delivery, has been that of marrying clinical and claims data—a challenge that continues to vex many, if not most, patient care organizations moving into risk. At the same time, one of the important frontier areas emerging around the leveraging of data analytics for value-based healthcare, is the pursuit of data around the social determinants of health. All those who are moving into that area agree that it remains challenging. As MACIPA’s Spivak says, “The EHRs are not always set up to help you with the social determinants. And we are trying to do some of that. I’ll give an example,” she offers. “We have social workers and health coaches. We don’t have a lot, because we’re small. But last year, they identified all the patients 90 and over in our senior products, and called them to make sure that they were being taken care of medically, and had transportation, weren’t living on the second floor and isolated, so we tried to reach out to them.”
Meanwhile, Spivak continues, “This year, we looked at whether they could identify family members—in patients who had dementia, we tried reaching out to those family members of patients. We were not doing it based on people’s insurance, we were doing it based on other factors. So we’ve been trying to do that”—gather and use more social determinants data. “Our care managers of course look at people who are in the hospital two or three times or use the EDs more.”
Of course, it really helps to be working in a payer-provider environment. That’s what the folks at the UPMC health system are doing in western Pennsylvania. The umbrella UPMC organization includes the UPMC Health Plan, and the health plan and provider divisions of the umbrella organization have been collaborating around data for years, in order to improve the health status of covered populations.
Pamela Peele, Ph.D., chief analytics officer at UPMC Insurance Services and UPMC Enterprises, says, “We absolutely have a great advantage in being an integrated health system. You can’t manage what you can’t measure. And providers can only see the patients in front of them. Payers, we get everything. We’re asking providers to manage risks they can’t even see. So we’re trying to give providers a broad view of the risks they’re trying to manage.”
Pamela Peele, Ph.D.
Peele notes that the UPMC Insurance Services and UPMC health system professionals are working forward in a few distinct areas right now. “The first one,” she says, “is around opioid abuse. In conjunction with providers, we’ve been identifying worrisome prescriptions that are ordered in EDs, etc., and we’ve been giving them the data and are identifying those worrisome situations, and are pushing that information right into the EHR so the provider can see it at the point of care. That,” she said, “is a great example of how a payer and provider can come together for the benefit of our membership. And per readmissions, the provider can only see readmission rates in their own facilities. We can see all readmissions via our claims, and can provide information to our providers about true readmission rates.”
And the results? “We’ve definitely seen a decrease in opioid prescribing; and our providers love the collaboration,” Peele reports. “We’ve also seen a decrease in readmission rates over time. If you’re asking a provider to take risk on readmissions, you have to give them accurate data,” she adds. “So when a payer and provider come together, you can accomplish that. And once again, you can’t change what you can’t measure."
And in doing so, Peele and her colleagues are using “very advanced techniques in machine learning,” she says, though she adds, “I don’t like the term ‘artificial intelligence.’” But, she notes, one exciting proect has been the use of a data model that “we’ve built with natural language processing and trained on our clinical care notes. It assigns the probability that somebody is homeless or has housing insecurity. And that’s incredibly important,” she notes. "If you’re homeless, how are you refrigerating your insulin?”
Meanwhile, “We’re still in the early stages of physicians and physician groups moving into social determinants of health data work,” APG’s Crane says. “If you talk to a fee-for-service doctor about social determinants of health, he’ll say, ‘Nice idea, but are you kidding? I didn’t go into healthcare to be a social worker.’ If you talk to APG members, though, you’ll see that they totally understand it.” Indeed, he says, more advanced medical groups, those taking on more risk, are trying to work forward, even as their leaders come quickly to see the challenges of connecting the data side of the SDoH phenomenon with the process side. “You so often need to get into the home, and into transportation, and nutritional support,” in order to make meaningful use of any data gathered for such purposes, Crane says. “If the patients can’t get to the doctor’s office, and aren’t eating and are living in a high-crime area, no amount of good diagnosis and prescription will produce a good outcome.”
Things are moving forward on multiple fronts, Crane notes. “In Medicare Advantage, given the latest rate note and the bipartisan Balanced Budget Act, Medicare is basically beginning to cover social determinants of health stuff; that’s in a nascent stage, but people are gearing up for it,” he says. “I think I saw that Humana and Ascension Health have created a new venture around social determinants. And we’ve entered into a partnership with Partners in Care, a foundation headquartered in Los Angeles. They’re available for hire to do home visits and other similar sorts of social work items. And my members are hiring them to do that kind of outreach into patient’s homes. It’s really helpful with the frail elderly and such. So seeing where the puck is headed there, we entered into a partnership with Partners in Care. And that’s a whole new frontier.”
Advice for Health I.T. Leaders
As the leaders of patient care organizations move their organizations further into risk-based contracts, inevitably, those interviewed for this article agree, the data analytics processes to support value-based care delivery will become more sophisticated, nuanced—and successful. In the meantime, what should the CIOs, CMIOs, and other healthcare IT leaders of patient care organizations, be thinking and doing, in this important area?
“Figuring out how to improve predictive modeling is one thing, and that means getting additional information that has predictive value,” says the AMGA’s Cuddeback. “And, in that context, everyone is recognizing that gathering social determinants of health information is important. For example, Lehigh Valley Health Network is actually working at the level of the census block group. Most of the socioeconomic measures are available from the Census Bureau at the census block group level, roughly 1,500 people. So if you actually know the resident’s address and are able to map it at that level—then you’re able to get the census data for that group of about 1,500 people,” for example. That granular level of data collection and analysis, he says, will be one key to moving forward successfully in the rapidly evolving world of risk-based contracting, he says.
Meanwhile, speaking of the foundational importance of data analytics to the entire value-based healthcare operations venture, APG’s Crane says bluntly, “There’s no future around fee-for-service; it’s eroding out from under doctors,” says Crane, whose association represents more than 300 physician groups operating in 45 states, the District of Columbia, and Puerto Rico. “You look at the Medicare fee schedule and increases slated for the future,” he says. “What are they? The anticipated increases to physician payment under Medicare are going to be 0.5 percent, 0.25 percent, from here out to as far as the eye can see, they’ll be nearly flat; and the increases in costs of running practices will be increasing 2, 3, 4, 5, 6 percent. So you’re quickly on the way to the poorhouse if you’re trying to stay in a fee-for-service world. So how will we make a living? To make a living doing what you want to do, you’re going to need to find a different way to make a profit under flat revenue. How do you do that? You keep the population healthier. You stare into the data and figure out who will get sick next, by using predictive analytics.”