Topics
More on Operations

UMMC analytics hub helps save system big money, improves patient safety

The $1.6 billion health system has made significant gains, including a four-fold return on its substantial investments in data analytics.

Mike Miliard, Editor, Healthcare IT News

John Showalter, chief health information officer at University of Mississippi Medical Center, with members of his analytics team: Clinical Intelligence Manager Keith Hodges (left) and Research Informatics Manager Alex Castillo. (Photo courtesy UMMC).

University of Mississippi Medical Center's John Showalter, MD, has led some serious gains at the $1.6 billion health system – not least a four-fold return on its substantial investments in data analytics.

In building its Center for Informatics and Analytics and relentlessly reinventing itself as a knowledge-driven health system, UMMC has made some significant strides with regard to patient safety and cost savings. Part of that transformation included Showalter's title change, from chief medical information officer to chief health information officer.

While not every hospital may be as analytically-advanced, many of the lessons learned at UMMC can be applicable even to smaller hospitals looking to make the most of their data projects, Showalter said, adding that robust data governance, a well-considered strategy and leadership engagement are essential to gaining value from any analytics initiatives.

Showalter's CHIO title is somewhat unique – at least when compared with the much more ubiquitous CMIO.

[Also: Hospitals hope 'Moneyball' analytics approach can help bend the cost curve]

"The chief health information officer position here is really much more focused on analytics and driving institutional return on investment from our clinical IT," he said. "When I was the CMIO, I was much more focused on adoption and usability for the clinicians."

The new role came about two years ago, after UMMC completed a large Epic implementation and ascended to Level 6 on the HIMSS Analytics EMR Adoption Model.

Having achieved those benchmarks, the next step was to "tackle the return on investment side – the quality improvement, the revenue optimization – and really bring predictive and descriptive analytics to the table to get that done," said Showalter.

The aim was to become a "learning health system" – which he defined as "having the ability to generate a clinical evidence base, looking internally to see what your opportunities for improvement are and what you're doing well – as opposed to looking externally toward evidence-based medicine."

Evidence-based medicine has a role to play, for sure, but the real goal is the ability "to do continual quality and process improvement," Showalter added.

[Also: MultiCare Health System to save $25 million in Health Catalyst analytics deal]

To get there, UMMC made use of an associative data lake – an analytics strategy that depends on "keeping data in its natural state and putting it together when we need it," he said.

"We don't do a lot of transformation of data into standardized data, as in a typical data warehouse model. We're using much more of a logical data warehouse model that combines some internal data lake structure with external Hadoop high processing computing, machine learning, and then combining those together. We do a combination of outsourcing and homegrown analytics that focus on use cases as opposed to comprehensive data management."

To get the most out that methodology, UMMC also employed so-called "honest brokers."

"The honest broker concept is that if you're going to have your data in a more natural state you have to have a group that can go access all the data and pull it together when you need it," Showalter said. "That level of access brings a lot of privacy and security concerns."

UMMC's honest broker team "has access to pretty much every data set in the entire institution," he explained. "So they're a group that is specially trained in security and privacy and has responsibility for ensuring compliance with all data releases and all data aggregations. It's a highly audited, highly monitored group of about five people that have the responsibility to pull all of our data together – but in response they have access to all of it without having to pull it into a single environment and do security around it."

That elite tactical squad has helped the health system find success on an array of analytics initiatives.

[Also: IBM Watson will acquire Truven Health Analytics for $2.6 billion]

"We've focused on improving our documentation, which required us to look into multiple systems to see whether physicians are answering queries, to see whether we're at benchmarks with our case mix index and to go into our billing and coding systems to identify additional opportunities to turn those into data visualizations that allowed us to work directly with our physicians and then pull up our CMI for several million dollars in return," he said.

A similar data visualization initiative focused on physician engagement and making more robust use of problem lists, he added. "More recently, we've integrated predictive analytics into our treatment of pressure ulcers and are about to go live with descriptive analytics approach around treating those ulcers. That's projected to be between a $500,000 and $1 million savings to the institution."

How do those specific initiatives get identified and prioritized? What goes into determining what gets attention and resources and when?

"It comes from two directions," said Showalter. "The first is we have a five-year strategic plan with top tier priorities, and those are the ones we're building our analytics suites around, where there will be a whole set of scorecards and discovery applications for each of our top five priorities.

"We also do do ad hoc requests that get prioritized by our senior leadership in conjunction with those strategic plans," he said. "Every request goes to a senior director or above that scores it from one to 10 and tells us whether or not we should do it."

As what he's learned as UMMC has attained such a high level of analytic maturity, Showalter said he's surprised somewhat at how the technology has greatly improved: "It has taken leaps and bounds over the past couple years," he said. "The tools to do data visualization and predictive analytics have really jumped ahead to where I thought they would be at this point."

But just as important as the technology, if not more so, is having a good team of people to staff these projects, said Showalter.

"The primary thing is to recruit a team that works well together, has a good work ethic and is very mission-driven. My team completely realizes that their work reaches the bedside, and reaches the patient, and that changes how they do their jobs. We work very hard for them to understand the impact," Showalter explained. "I would also suggest you recruit a team that's interested in learning, because the technology we have in three years is not going to be the same as the technology we have today."

Twitter: @MikeMiliardHITN