The need for a strong data governance structure has been talked to death in healthcare, but even for all the ink the topic has received, most healthcare organizations lack a sound plan, which could come back to haunt them as they embark on analytics and other complex data initiatives.
Without a firm foundation for assuring the overall management of the availability, usability, integrity and security of data in healthcare organizations, quality issues could limit the effectiveness of initiatives that rely on the trustworthiness of the underlying data.
John Moore, founder and managing partner at Boston-based analyst firm Chilmark Research, said the firm’s research finds that most healthcare organizations have “fairly rudimentary” governance structures. Only 15 percent to 20 percent have full-fledged data governance frameworks in place, he adds.
“You have to keep in mind that a decade ago, the industry had very few EHRs in place, so data governance structures to decide how to define and share data across systems wasn’t something people were working on,” he says. “The healthcare data value chain starts with strong governance and information management, but we don’t have a lot of good models in this industry, so you’re seeing a lot disparate point solutions instead of the integrated solutions really needed to move analytics forward.”
UPMC Presbyterian Hospital
"A strong data governance infrastructure means that we can ensure that our data privacy and security policies are applied consistently to all our data," says Rasu Shrestha, health systems CIO at UPMC Enterprises.The shift to shared saving and risk-based reimbursement has revealed the cracks in data governance and information management infrastructures at many organizations that are having significant problems getting the data pieces in place to keep their heads above water. Michael Hunt, M.D., chief population health officer at St. Vincent’s Health Partners, a 275-physician medical group in Bridgeport, Conn., cites research that shows 70 percent of accountable care organizations don’t make money.
“The data set for the Medicare Shared Savings Program requires reporting for 27 different quality measures, and many of those ACOs apparently couldn’t submit the appropriate quality data,” Hunt says. “Does anyone think they didn’t make a huge effort to hit their targets and qualify for incentive payments? Right now the industry is trying just to get an infrastructure in place to capture the utilization and quality metrics, and bring some visibility to costs. It’s difficult to start thinking about advanced analytics in an environment where you have to jump through so many hoops.”
Joe Kimura, M.D., chief medical officer at the Boston-based Atrius Health, says the 750-physician medical group alliance wouldn’t be able to launch analytics if it hadn’t done the extremely hard work—politically and technologically—of defining the high-level business concepts and clinical definitions that rule its data.
At Atrius Health, those concepts and definitions are guided by medical directors, with help from financial and operations staff. Most of the organization’s revenue is under full-risk contracts with the State of Massachusetts, so Atrius, like St. Vincent’s, poured resources into designing a governance infrastructure that could handle the rigors of quality reporting.
More than 90 percent of Atrius’ ACO reporting is captured in automated reports, but it still struggles capturing certain discrete information, such as details on follow-up measures, for its reporting.
But data governance, fundamentally, is getting together and defining the information on hand. For example, who is a patient of Atrius Health? “Marketing wants to count someone we haven’t see in four years; finance wants to say that if we haven’t seen them in 12 months, they are not on a roster and aren’t a patient; as a physician, I would say someone I’ve seen in the past three years is a patient,” Kimura says. “People have different ideas for different business purposes, but the bottom line is that you have to come together as an organization and decide. You can’t have three definitions, because if you try to go forward in that way, it takes a brutal amount of work to revise your data infrastructure.”
Another challenge is to ensure that an organization is defining data in way that’s clinically and financially valuable to its mission. For example, the Healthcare Effectiveness Data and Information Set (HEDIS) defines a diabetic patient for reporting purposes. But Atrius has a much more detailed definition of diabetes that includes additional claims, EHR and pharmacy data.
“The HEDIS definition is fine for reporting, but we feel our definition gives us a more accurate look at our diabetic patients and is more clinically valuable,” Kimura says. “Definitions are not necessarily universal, which is why governance is an enterprise responsibility, not just one person or department.”
The University of Pittsburgh Medical Center also has invested heavily in creating a data governance infrastructure years ago when it started having multiple information system go-lives and saw a need for those data streams to converge, says Rasu Shrestha, the health systems chief innovation officer and executive vice president at UPMC Enterprises, which funds incubators and leads commercialization efforts for UPMC technology products and services.
“We bet big on interoperability years ago, but to do so meant that right from the beginning, we realized that to converge data you had to have a common set of rules and definitions,” Shrestha says. “We operate more than 20 hospitals and a health plan with more than 2.5 million covered lives. Without a framework and rules around data ownership and stewardship, we couldn’t bring those streams together in a meaningful way. And just as importantly, a strong data governance infrastructure means that we can ensure that our data privacy and security policies are applied consistently to all our data.”
Shrestha adds that UPMC is discussing ways to commercialize its data governance best practices and information management models.