Framing the Issue
- Information technology will play an essential role in helping health systems manage population health.
- Population-level analysis requires hospitals to know what's going on outside of their system.
- Hospitals face a skills gap when it comes to having staff who can produce population-level intelligence.
- Health care organizations are scouring their own ranks and luring professionals from the insurance industry to get their analytical capabilities up to speed.
The notion of managing and being accountable for the health status of defined populations has gone from unfamiliar concept to new reality across the health care field in the space of a year or two. Health system leaders now know what accountable care is, how it's dependent on coordination-of-care activities on a wide scale, and the essential role of information technology in that coordination. Discussions that may have lingered over whether to plan for population-based management have moved on to when and how to proceed.
All the purposeful planning in the world, however, won't by itself produce the very different technological capabilities and expertise required for this new mission, likely at a significant cost in high-priced personnel and data analytics services. Traditional clinical enterprises, staffed and organized to treat people who show up sick, now have to unravel the medical mysteries of whole groups of people who may not mind their personal health, to the detriment of their providers' financial health. The key skill set for today? How to keep people from showing up sick.
The ability to do prospective, pre-emptive care management requires a much more sophisticated and expansive level of data collection on covered populations than most health care organizations have now, according to experts in this emerging field. Even those enjoying the greatest progress with electronic health records have only a portion of the data foundation necessary for population-based care, says Matthew Cinque, managing director for performance technologies with the Advisory Board Co.
One prevailing presumption surrounding the push for EHR adoption is that as soon as all providers are plugged in, "they'll have all the information they need to manage the patient," says Cinque. But a driving force behind their design has been to capture data for individual encounters and better billing, not pull up and combine data for a better understanding of a lot of patients at once. "The data are all there — just not necessarily accessible in aggregate when you're going to better manage populations rather than persons," Cinque adds.
And that only covers patient information entered into the enterprise EHR. To get a grip on people's health status, managers must know at what other places they've received care and what services or goods they received. Receipts for care, when amassed and analyzed, can identify overall health problems in greater detail. A key source of these population health clues is none other than insurance claims, the clinically thin source of data that EHRs were supposed to replace and surpass. "Today, that's really the only source of data that truly reflects 100 percent of the care received by the individual patient," says Cinque.
Claims and EHR data working in tandem will supply the foundation to size up health status, stratify groups of people into categories according to their ills and financial consequences, and come up with action plans, he says. Those are the means, but they're of little use without the skills to make the most of them.
"There is no ready-set population of health care data analysts sitting out there waiting to be employed," says James Gaston, senior director of business and clinical intelligence with Healthcare Information and Management Systems Society Analytics. "They're very rare, and they don't necessarily teach this in college — it's not like you can get a degree in health data management like you can in accounting or marketing. So these people have to be grown, they have to be developed and they have to evolve." The pivotal skill, he says, is an understanding of clinical and claims data and how they work together with social and census data to put together a comprehensive picture.
Tools of a new trade
The starting point for many organizations that plan to take risk for a population's health is to identify a group with the highest utilization, much of it the result of uncontrolled health complications, and get that group under control with intensive case management, thereby saving money by avoiding hospital admissions, readmissions and emergency department visits. Lower-risk groups receive preventive services to keep them from losing control of their health. The process amounts to an informational two-step: assigning people to high-, medium- and low-risk categories; and using clinical information to guide courses of action.
Providers with strong EHRs can closely manage people targeted as high-risk once identified, but the ability to identify them is limited by what is captured in the EHR, says David Abelson, M.D., president of Park Nicollet HealthPartners Care Group, which operates an enterprisewide EHR and is a participant in Medicare's Pioneer ACO demonstration project. Population-level analysis requires a wider view, including hospital stays outside the system and prescriptions filled elsewhere. "Those data don't exist within delivery systems; those data are with payers," says Abelson. "And so it's important to have a partnership with payers to actually have a complete data set to do population management."
In addition to getting claims data from Medicare, Park Nicollet has negotiated contracts with the three largest commercial payers in Minnesota with "very significant dollars" in play around sharing savings through population management, he says. Central to those arrangements are feeds of claims data on patients both inside and outside of Park Nicollet's information reach.
"Most of the contracts that are either being put forth by the insurance companies or being negotiated by providers include language guaranteeing access to this type of data," says Cinque. "The advice we give to providers is essentially, 'Don't sign a contract unless it includes access to this type of information.'"
It's telling, he says, that the Centers for Medicare & Medi-caid Services have placed such a big emphasis on sharing this data with providers approved for Medicare's shared-savings program and other demonstrations. It's a recognition that "the data are necessary at least as a starting point to get your hands around the population, what's wrong with them, how much do they cost, what the trends are."
But all this intelligence on the population at large doesn't extend to what to do about it. Accountable care organizations need rich clinical detail at ground level to see into the medical situations of their targets, and that's the province of the EHR, says Gaston. "Once you have comprehensive information at a patient level that can be rolled up to a population level, then you can begin to grasp [population health] — with the proper analytical environment, infrastructure and tools." Success with that investment requires "staff that are competent and capable of using the data, the infrastructure and the tools to drive that understanding."
Rare new skills
If health care is short of the skills to produce population-level intelligence, it's partly because it didn't need those skills until the business model was marked for change. "When you talked about understanding populations from a health care perspective, that was something that some researcher did in a back room and was published in a paper that nobody read," Gaston says. "People weren't necessarily very focused or even interested in that." The type of data needed to do such analysis never existed in large enough quantities, "and, therefore, there's been no reason to invest in these specific competencies," adds Cinque.
There are reasons now. Health care organizations are scouring their own ranks and luring professionals from the insurance industry to get their analytic capabilities up to speed. "There's plenty of expertise across all of health care in order to approach this," says Stephen Moore, M.D., senior vice president and chief medical officer of Catholic Health Initiatives, Englewood, Colo.
"It's hiding in silos: care-management organizations, payer organizations, traditional hospital-system organizations, and physician enterprises," he says. "I think the challenge for us is [that] in moving from a hospital-centric organization to a community-centric organization, we're going to have to reach out and beg, borrow and steal those competencies and skills from those other areas in order for us to be successful."
At Norton Healthcare, Louisville, Ky., a lead analyst with doctorate-level training selected a team over the years with expertise in other sectors of business besides health care that manages large data volume, says Steven Hester, M.D., senior vice president and chief medical officer. It's not an information technology team but rather a data analysis group capable of understanding population sets, registries and the impact of data on operations. One expert, for example, reports on incidence of infections and looks for opportunities to improve, says Hester. Another is focused on cardiac disease, and so on.
At CHI, an executive from United HealthCare of Colorado was hired to be vice president for care management throughout a care continuum spanning 18 states, says Moore, and managed care experts from payers were added to set up and execute networking and contracting strategies. Other positions included insurance actuaries and "dozens and dozens of people" tasked with creating a data analytics capability in conjunction with a systemwide data warehouse, he says.
Health care organizations won't have to round out the whole range of expertise at once, says Abelson. In the Pioneer ACO program, for example, the initial step of improving results for shared savings doesn't call for the actuarial heft that goes with predicting patient utilization and costs. "Once you go beyond shared savings and actually are accepting risk, you blur the lines between delivery and the typical health care insurer functions," he says. "And in order to reliably take risk, you have to understand what the population is likely to cost over the next year. And that requires actuarial analysis."
When Park Nicollet last year made the decision to merge with Health Partners, a hybrid health plan and integrated delivery system in the Minnesota market, part of the reasoning was that it would provide instant expertise from its health plan side. "Delivery systems that want to move in that direction of increasingly taking risk need to get that actuarial talent, and there are three ways to do it: you make it, you buy it or you partner," says Abelson.
Not for the small enterprise
Park Nicollet and other health systems in Minnesota had gone through a previous round of population risk and case management in the 1990s, when managed care was at its height, and "in many ways I'm going back to my early years," says Abelson. "Even though we didn't have the richness of the data, we still knew our care model had to produce predictable results in terms of achieving quality and making it more affordable."
"Any system with an affiliated health plan gets this," Abelson adds. "[For] organizations that either haven't been involved [in] or are not part of a health plan, there will be a very steep learning curve." Providers need to get on a similar footing with insurance companies, says Cinque, regarding the sense of a population that those companies have maintained, primarily through analyzing available claims data.
The smaller the hospital or health system, the less likely it will be able to accomplish that, says Teresa Koenig, M.D., senior vice president with the Camden Group. "This is a foreign language for hospitals, let alone rural hospitals that have the pressures of declining reimbursement," says Koenig. "They have to start by spending all this money just to become able to transmit data; they have to get and work with everybody's data; they have to find people who can process it in a way to find the areas of population to address first; they have to build new care models; and then they have to measure if they're successful. And they have to do it with no increases in revenue."
At the CHI network, which includes 21 critical-access hospitals and six other rural hospitals that provide the sole source of health care in their communities, the system provides some of the data management capabilities centrally at its headquarters and is locating other duties out in the field, says Moore. The data warehouse takes in information from 2 million data points across the system daily, then manages the data through analysis and reporting, pushing those reports back out to the organization as a national back-office service.
The investment will cost between $30 million and $60 million, but "if we were to replicate the same warehousing and specialty analytics necessary in every region that we're in, we'd be spending somewhere between $150 million and $200 million," Moore says.
People who are competent on both the payer side and care management, are in high demand, says Moore.
CHI is in good shape because of its size and ability to rationalize what to do nationally and locally, but "the small to medium regional hospital systems, as well as the stand-alone hospitals, are going to be very hard-pressed to be able to get into this in a way in which they will have the expertise to manage," he says.
Moore points to four systems in the eastern half of Iowa that formed the University of Iowa Health Alliance in mid-2012 to gain the scale needed to become an ACO. CHI is providing all the data warehousing. The component health systems found that the technology project would be four times as expensive to set up by themselves compared with getting it from CHI, which already had made the investments and could extend its infrastructure to them. "This is going to drive consolidation as much as the financial environment drives it: the need for expertise, competency."
— John Morrissey is a freelance writer in Mount Prospect, Ill.
Setting the target: Getting useful data to the front lines
Arriving at a target for analysis is as important as amassing the analytics capability for population health management. Whether it's a percentage of the sickest people, a top 50 to 100 cases or other starting point, a health care organization has to zero in on a goal everyone can rally around.
Norton Healthcare has a focus on the top 100 users of health services and how to manage them more effectively, says CMO Steven Hester, M.D. "They're folks who have legitimate chronic illness or even legitimate acute illness. But how can you look at that differently?" The target group affects how the system's analysts look at data and feed information to clinicians, he says.
Park Nicollet looks at the top 5 percent of its patient population in terms of cost and sickness, many needing case management, says CEO David Abelson, M.D. Once armed with the identifying data, "that's a very simple thing to do with an Excel spreadsheet."
A general focus on the sickest people, however, might not prove the best use of management resources depending on the odds of improving their health status, says Joseph Kimura, M.D., medical director of analytics and clinical reporting systems at Atrius Health, a federation of six medical groups in eastern Massachusetts. A claims-based scoring of people according to financial risk has to be combined with information from an EHR to make judgments on the best risks, not just the highest, he says.
"If we threw all our really high-end resources at the high-end, high-risk population, I think we may be sorely disappointed — patients will love it, families will love it, but the cost curve may not bend a heck of a lot, because, honestly, you can't do a lot clinically for a lot of these particular patients," says Kimura. "You can save those kinds of resources for a slightly less morbid, or high-risk but potentially more actionable, population." Ultimately, a care team has to put its heads together with EHR information to make those judgments, he says.
Atrius, a participant in the Medicare Pioneer ACO demonstration, has had a data warehouse in operation for about seven years, taking in all-payer claims data as well as clinical information from all six medical groups, which use the same EHR vendor.
"Any organization with a semicomplex, semirobust data warehouse can use its own tools, commercially available tools, etc., to help highlight patients with any number of characteristics," says Kimura. "The true challenge of being a data-driven organization is to try to take that analytic capability and help decision-making on the front lines. … We as an organization are really trying to emphasize how we take this wealth of data and analytics that we have and try to bring it down to the spot that actually changes patient care."
Determine if the organization can afford the investment.
Population-level analysis is a multimillion-dollar expense for a reasonably robust data warehouse, the personnel to manage it, special talent to turn the data into courses of action and, ultimately, actuarial expertise to predict costs of at-risk patients. Providers may have to consider becoming part of a larger group if they can't commit to this outlay and ongoing cost.
Determine how to obtain the necessary analysis skills.
Large organizations may be able to lure analysts from health plans and from outside the health care field and be able to pay the going rate for such talent. For others, "quite honestly, you're just going to be relying on a lot of vendors and external products for at least the near future to do a lot of this," says Matthew Cinque of the Advisory Board Co.
Bring the EHR into the population-management age.
Greater emphasis has to be placed on what the organization is getting out of its EHR, says Cinque — not just tracking patient encounters and recording the results, but actual analysis of trends, areas of concern for large groups of people, and the guidelines and prompts that change behaviors of providers and patients.
Arrange to acquire and use population-level data on patients.
Medicare has information on its elderly beneficiaries; commercial payers can supply claims data on a sub-65 population mix. That's just the first step. Are organizations capable of grouping the information coming in and making it meaningful so reporting can be shared with the relevant people? "The answer to that question is quickly 'No,' because that's not a competency or a data set they've ever had to deal with," says Cinque. "But certainly the question needs to be asked."