As Sanford Health grew through mergers and acquisitions, people were going everywhere to get data, IT was decentralized, and there was a lack of governance with very little standardization.
“We had data anarchy, everybody was getting their data, so we were going to quickly move into a data dictatorship,” said Doug Nowak, Senior Executive Director of Enterprise Data and Analytics at Sanford Health. “We took access away, we took licenses to Crystal away, it ticked off people.”
[Also: Change management is IT's greatest challenge in delivering needed innovation]
While Sanford Health built an enterprise data warehouse, Montefiore Medical Center created what Parsa Mirhaji, Director of Clinical Informatics called its approach a network tapestry.
Analytics as foundation
In its journey to become an information-driven health system, Sanford had to make even more major changes.
“To get to one source of the truth, we had to build a foundation, have a team, centralize the data, and agree on a common language,” Nowak said.
Nowak and Sanford wanted to get to a point of less data with more analytics, then put advanced analytics in place to ultimately deliver descriptive data, and not just more of it, back to users.
[Also: Go figure: The key to big data is actually small data]
And they did just that with a self-service option. The enterprise data warehouse that Sanford built comprises technologies from Cisco, Hadoop, CDW and others, as well as the Epic EHR.
“We eliminated over 25,000 reports,” Nowak added. “Most of them were never used, they were just junking up the system and we were maintaining them.”
Leveraging ‘Tapestry’
Montefiore collaborated with Cisco, Franz and Intel to build a data lake to track which patients are touched by what analytics and how that data being is used.
“Tapestry is the link between everything that goes back and gets injected into the data lake so we know every score being generated by every analytic and by-product for every single patient,” Mirhaji said.
The overarching goal is to relentlessly understand patients, Mirhaji added, especially as they continue generating data with smartwatches, fitness trackers and social networking.
Mirhaji and his team built predictive models to determine which respiratory failure patients would deteriorate in 48 hours and who might need life support.
The next project they’re working on is tackling readmissions, starting with heart failure. Rather than creating a prediction for readmission at time of discharge to look at the next 5, 30 or 90 days, Mirhaji said they want to span the entire lifecycle of patients.
The hospital wants to take every opportunity to predict readmission every time a patient touches the health system or it touches a patient. Users can visualize whether they’re increasing or decreasing the chance of readmissions down the road, Mirhaji said.
“The beauty is the simplicity of the query. The system knows so we don’t have to know the back-end infrastructure, the system does it, which is the key.”
The bottom line
Becoming an information-driven hospital is going to require a culture shift and robust technologies.
To harvest value from investments in IT, Mirhaji recommended creating a data culture by optimizing communication internally and externally and implementing a mature repository to hold the data and advanced analytics to make it actionable and operational.
And it’s likely you will have to sell it to your own users at some point.
When Nowak presented the self-service reporting functionality to Sanford clinicians, for instance, they at first they pushed back because they thought it was just shifting the work from IT and analytics onto them.
“One of our team members said ‘my bank calls it on-demand,’” Nowak said. “That was more palatable.”
Twitter: SullyHIT
Email the writer: tom.sullivan@himssmedia.com
