Big data: Blind man with the elephant?
BOSTON — Big data means different things to different people. But as providers increasingly apply analytics tools to growing data sets, some realities are at least starting to become clearer. Among them: reasons why the technology need not be perfect to be effective now.
"We’re kind of like the blind man with the elephant," said Richard Finley, director for medical analytics at the University of Mississippi Medical Center, at the HIMSS Big Data and Healthcare Analytics Forum. "We’re all dealing with the same thing but we all see it a little differently."
Indeed, hospitals and networks are investing analytics for similar reasons, including value-based care, risk sharing, improving clinical quality, patient satisfaction, population health and overall performance.
[Also: How crowdsourcing can accelerate healthcare analytics]
"This big hope is that we can short-circuit and prevent high-cost problems from beginning," said Joseph Doyle, a health economist and associate professor at the MIT Sloan School of Management. "We’re using data sources in new ways and will propagate these as we get more and more data."
To make that happen, healthcare executives need to first understand what the business needs are, what to look for at the enterprise level and not to drown in all the details, according to Brian Doty, principal at ConvergeHealth by Deloitte.
In the case of Sanford Health, that meant taking a step back to create a data platform.
Benson Hsu, Sanford’s vice president of data and analytics said they created a "straight data universe" that became his team’s calling card.
The platform Hsu needed to create included three parts: a common data team, language and source. To assemble the team, Hsu, who reports to the CFO, pulled people from IT, enlisted clinicians and then they went hunting for our other stakeholder; only marketing and finance staffers were protected. The team eventually totaled some 60 people.
The team then agreed on a common data language by getting seven senior leaders together and agreeing on 300 discrete terms for what Hsu called quick and dirty governance. The common data source is going into effect right now.
The goal is not merely to pinpoint, say, four patients that will get sepsis tomorrow but, instead, to determine broader changes coming that Sanford can target at scale.
"The key is really centering on the event and intervention that will change that event," Hsu said.
UMMC’s Finley agreed.
"Prediction is not going to be perfect, but you don’t really need it to be," Finley said. "If you say such and such a patient will have a certain outcome on a certain date, we’ll never get there. Unless you have an actionable outcome that’s not going to make a difference."
Related articles from the HIMSS and Healthcare IT News Big Data & Analytics Forum in Boston, Oct. 24-25:
⇒ Charlotte hospitals analyze social determinants of health to cut ER visits
⇒ Big Data: Healthcare must move beyond the hype
⇒ Tips for reading Big Data results correctly
⇒ Small hospital makes minor investment in analytics and reaps big rewards
⇒ MIT professor's quick primer on two types of machine learning for healthcare
⇒ Must-haves for machine learning to thrive in healthcare
