Analytics or BI? Centralized or federated data? Geisinger's CDO shares insights
Healthcare jargon is filled with terms that are often used interchangeably but have different meanings: electronic medical records and electronic health records, for instance; or public health and population health. Another such pair? Business intelligence and analytics.
They're similar, but not the same. Nonetheless, the two "absolutely have to exist in parallel," says Geisinger Health System Chief Data Officer Nicholas Marko, MD.
"People like to say, 'Advanced analytics, that sounds really cool, really high-power,'" says Marko. "But the fact of the matter is, most clinical care decisions, most business decisions, most operational decisions, are driven by much more basic analyses – reporting of numbers on a consistent basis, display of data in dashboards."
At Geisinger – a longtime analytics leader that's been able to recognize big gains in quality and efficiency from its targeted number-crunching – "most of the business processes that we inform with data are really BI-centric," he says. "Reporting and dashboarding are the workhorses of what we do: 80 percent of the problems are perfectly suited-to and very well-addressed by those tools."
So at the Danville, Pennsylvania-based health system, "we always need to maintain a strong, consistent and robust BI environment," says Marko. "It's a must."
But advanced analytics is the exciting technology upon which Geisinger has built its reputation for quality – to the point where it's able to offer money-back guarantees to dissatisfied patients.
Those initiatives go "two or three steps beyond dashboarding – which is just presenting and summarizing data – to a place where we're actually using that data to drive computations that we use to then inform more complex decisions: predictive analytics, decision modeling, these sorts of things."
Any provider organization hoping to make the most of its data should keep both BI and analytics programs in top shape.
"It's important, if you're going to be on the leading edge, to have systems in place that can deal with those kind of complex things," says Marko, speaking of advanced analytics. "But common things are common, and most business runs on BI. We've got lots of BI, we focus on it, we nurture it right along with the more advanced stuff. In fact we probably dedicate more resources to the BI than to the advanced analytics because there's more demand."
Another this-or-that question many data strategies need to answer has to do with centralization versus federation. What factors might shift the balance one way or the other? That can depend on the organization, said Marko, and definitely depends on a key question: "centralization vs federation of what?"
Efficiency and quality improvement depends on data management, data analysis and data governance, he says. At Geisinger, "all three of those things have strategies for us around how we manage them and to what degree they're centralized vs federated."
When it comes to data management – warehousing, big data operations, etc., "bringing all the pieces of information together, putting them somewhere we can use them and making sure people have access" – Geisinger's strategy is to centralize the raw data.
"It makes sense to have one copy of everything, rather than 50 different copies that change all the time and no two people are looking at the same thing," says Marko.
But when it comes to data analysis, Geisinger is much more federated.
"You just can't scale a centralized analysis process. You can't hire enough people, you can't buy enough computers to keep up with everybody wanting everything they do to be data driven. We try to federate as much as we can, so it's just a care provider or a business leader and their data: They've got tools, guidance and support, but don't need to go through another centralized layer to get most of their questions answered."
Finally, for all-important task of data governance, Geisinger pursues a balance between centralized and federated, says Marko.
"There are certain things we do at a core level – certain data features and data elements that we govern very closely because we want to make sure everyone in the institution is using the same definition for 'readmission' or 'length of stay,'" he says.
"There are other things that are governed and maintained out of the point of origin – lots of places where data enters our system, and at most of them there's some sort of data steward who makes sure it's well-maintained and clean and orderly," says Marko. "Similarly, our policies – who can access data, how we use it – are pretty centralized because we want it to be consistent. We want there to be a predictable, transparent and level playing field for all things data-related."
But when it comes to the question of centralization vs. federation, your mileage may vary: "It's really balance between the two," Marko says. "What defines that balance – the set point in the middle of the fulcrum – is your strategy, and your organization's goals."
This article is part of a series focusing on big data and analytics. Other pieces include: Geisinger's guiding principles for moving away from one-off analytics projects toward a data-driven culture and Do you need an enterprise analytics strategy? It depends. (But probably, yes.)