There are ways to do analytics right
With apologies to Internet meme-makers everywhere, analytics experts have a message for healthcare providers trying to get their heads around business and clinical intelligence: "Big data, you're doing it wrong."
So much attention and energy have been put toward "big data" in the last couple of years, for perfectly understandable reasons. For example, health systems collectively have spent billions of dollars installing EHRs in recent years. "They want to get their value," says Cora Sharma, analytics analyst for Chilmark Research, a Cambridge, Mass.-based health IT research firm.
They certainly see a lot of potential in the data. A March poll from MeriTalk and EMC found that 63 percent of healthcare executives in the federal government believe that big data will improve population health management. Similar numbers show that advanced analytics would "significantly improve patient care" and make it easier to deliver preventive care in the Military Health System and Veterans Health Administration.
[See also: Deloitte taps the Zen of data analytics.]
But so few have proper goals and strategies for their data, according to Graham Hughes, MD, chief medical officer of business analytics firm SAS, based in Cary, N.C.
"They're looking to accumulate data and how to get data in," Hughes says. In his opinion, this is a faulty course of action. "It's not about the data. It's about how you're going to manage it."
The focus, according to Hughes, should be on information management, including data governance, stewardship and quality. "If you are just about grabbing data, you will be on a data grab forever."
Optum, the IT and analytics division of UnitedHealth Group, published a white paper in February that corroborated this belief, particularly when it comes to clinical analytics.
"It may sound impressive to say that your organization has access to terabytes of patient information, but without robust technology and smart people to manipulate it, that data is simply words and numbers without context," researchers point out in the white paper.
"Raw data from claims or from an EMR database are not suitable for analysis. Turning raw data into usable information requires preparation, including normalization and validation. Only then can an organization gain trustworthy insights from the information and put it to use in maximizing patient care, reducing risk and strengthening a business's bottom line,” they add.
Hughes says that organizations have been spending too much time and money on enterprise data warehouses, which he sometimes refers to as "data landfills." The repository is not as important as the location of the data, according to Hughes. "An EDW isn't where data goes to die. An EDW is a staging point for analytics," he says.
"An EDW needs to be easy for clinicians to understand and interpret, and also needs to interoperate with and push data back out to other systems," Hughes continues. Too many organizations wrongly assume that data should get moved to the analytics software, he says. "Modern analytics run directly from transactional systems," so there is no need to replicate the data in every situation.
"It's where the data is being moved to [that] makes it actionable," he says.
Indeed, Hughes notes that analytics historically have been seen as retrospective reviews, but data stores are so great now that predictive analytics are now possible, even if much of the data remains unstructured. "Analytics is now about providing actionable insights back into workflow," in close to real-time if necessary, he says.
"Sometimes this is done in too fragmented a fashion," Hughes says, a reflection of the "best-of-breed" strategy of years past. Organizations bring in pieces of analytics technology every time they see a new "shiny object," he says.
"To me, this feels where we were with clinical systems in the late '80s and '90s," Hughes says. He suggests thinking about it not as "niche buying," but rather as a strategic, enterprise-wide investment.
The MeriTalk-EMC study found that only a third of federal healthcare executives had invested in technology to optimize data processing, and less than 20 percent said their agency was "very prepared" to manage big data.
In the private sector, according to Sharma, early adopters such as Intermountain Healthcare, Kaiser Permanente and Partners HealthCare are farther along than most, but she worries about smaller, less-tightly-integrated organizations. "There's no kind of off-the-shelf software out there for them," Sharma says. So they turn to the best-of-breed strategy.
A severe shortage of analytics pros makes navigating this landscape all the more difficult, according to Hughes. "It's also a mistake to think you can staff up on this easily," he says.
Hughes suggests managing data in the cloud, through a vendor or looking for "self-service solutions" that provide expertise. "You need clinical analysts to create models, and let the system be smart enough to give good recommendations," Hughes says.
Sharma views the lack of qualified data engineers as an opportunity for payers; Aetna and United are among those insurance companies that have been beefing up their analytics divisions lately. "They're basically offering ACOs in a box," Sharma says. However, she adds, "Providers are still more reluctant to work with payers, for a lot of reasons."
Thus, big data in healthcare remains fraught with pitfalls.
[Learn more about analytics best practices: Healthcare Business Intelligence Forum]