Special Report: Faster Analytics at University of Michigan Health System
Editor’s Note: Healthcare Informatics has compiled together a five-story Special Report section on data analytics for its July/August print issue. This story, and one posted online yesterday about provider organizations making strides in their data analytics work, are part of that special section.
As Healthcare Informatics reported as part of its Up and Coming health IT vendors’ section last issue, many data analytics companies have recently gained momentum because they address a pain point in the provider space. But other vendors in this segment, such as Seattle-based Tableau, featured in last month’s Healthcare Informatics 100 issue, offer data visualization products with a focus on business intelligence (BI). One of Tableau’s health system clients is the University of Michigan (U-M) Health System, which has a Fast Analytics team whose job is to crunch data and assist more than 30 groups across the health system—which has three hospitals, 40 outpatient hospitals, and more than 140 clinics—with their dashboards.
Jonathan Greenberg, director of the five-member Fast Analytics team at U-M Health System, says the team’s core question that it asks itself over and over is, “How do we do reporting and analytics here better?” To that end, the Fast Analytics team has a key philosophy around three major groups: skills, process, and tools, Greenberg says. “By fine tuning that triangle, based on the staff you have, you can come up with a successful environment to improve reporting and information decision making, which was our main goal.”
Jonathan Greenberg
About five years ago, when the 990-bed U-M Health System was looking for analytics tools, Greenberg’s analytics team was being carved out from the organization’s central IT team; it was mostly focused on professional billing, he notes. “We were the red-headed stepchild in a medical school IT shop. We were managing the billing system and things like RVU [relative value unit] recording, and all recordings surrounding billing.”
Indeed, at that time, the analytics team at the U-M Health System managed a large and outdated billing system. To produce reports, they either had to do the coding themselves or pay large fees to an outside vendor, and the final output was often late or incorrect. “We had to fight our way through all kinds of vendor interactions in order to have the right to pay huge sums of money for report alterations, and it wasn’t a cost effective way of reporting. Also, it wouldn’t let us see the next layer of reporting in our organization,” Greenberg notes. At the same time, the health system was looking at Epic, he adds. “I was very involved in those initial stages, and I realized there was a huge missing reporting and analytics component there too. And so everything kept coming back to Tableau,” he says, specifically noting his organization’s capability to leverage the vendor’s automation and data visualization techniques. Greenberg said that throughout the whole vendor selection process, U-M Health System looked at 14 different packages spanned across an 18-month period.
Now, Greenberg points to having the ability to find outliers and being able to do multi-dimensional outlier searchers so the analytics team can zoom in on characteristics. He gives an example of such characteristics that are causing doctors to be margin negative rather than margin positive on the same procedure. “It’s about helping to answer questions like that by finding trends and outliers, but more important than that, it’s giving people a simple and intuitive way to drill down into the data,” he says. “If we build the dashboards right, the menu structures and everything else are intuitive, and people know how to use the tool without much training. That is so incredible,” he says. “Doctors out there don’t have time; if they wait more than seven seconds for a page to open, they’re gone.”
Andrew Rosenberg, M.D., CIO at U-M Health System, adds that problems that are being solved with Tableau are concentrated. He says visual analytics help solve the complexity of the interaction of financial and billing data as it relates to operational needs that go beyond revenue cycle, such as the ambulatory space, and the health system’s cancer center. “It’s our ability to extract new meaning, new information, at the minimum, from what one might consider as traditional finance and revenue cycle billing data,” he says. “Our [head] of the entire revenue cycle recently kept making it clear to me, ‘Remember that you’re not just helping me and the revenue cycle, but you’re helping me become a provider to many other business customers in our health system who always want to see financial and revenue cycle data to meet a variety of different needs.’”
Andrew Rosenberg, M.D.
For Rosenberg, the dashboards that stick out to him are when the analytics team is, in a refined manner, look for the work productivity by type of work of the health system’s actual IT employees—those who are building reports and analytic extracts for the entire medical center’s mission. “Our ability to dive into details across a large team of 40 people to enhance productivity with Tableau has improved our efficiency by between 10 and 15 percent, just by having a tool to look across a group’s complex work,” he says. Also, Rosenberg notes, the RVU dashboards are used to demonstrate a variety of RVU-based measures to pretty much every type of division in the entire academic medical center. And there isn’t one dashboard, but rather some 80, that could stem off a common view, he says.
What’s more, at HIMSS16 in Las Vegas, Tableau announced an agreement with Epic in which client-created Tableau analytics workbooks and dashboards will be integrated with Epic’s electronic health record (EHR), enabling direct access from EHR users’ workflows. On the partnership, Rosenberg notes, “I don’t mind, if when I am logging into Epic, I then have to click on a hyperlink to go into Tableau natively and see things. I would prefer if it were fully and deeply integrated, but what I don’t want to do is keep having to log in and back out of everything. What I have started to see is the ability to start integrating Tableau dashboards into Epic further.”
Regarding results, Greenberg adds the team was able to create a charge estimate tool for customer service staff that went from taking 4.5 hours to complete to 4.5 seconds with Tableau. “And 17 people could do work that only two specialists could do before. Also, we eliminated 4,000 hours of work out of their schedule for next year, and for future years,” he says. What’s more, Greenberg points to some 10,000 hours work eliminated, or four to five full-time equivalents (FTEs). He also notes a recovery of $3 million in RAC money from Blue Cross as part of a one-time amnesty program the payer had. While Greenberg’s analytics work is limited to revenue cycle, where he says he has made 105 workbooks, the Fast Analytics team has helped 26 other departments spin off their own sites, sharing their visualizations and reports. “They’re really doing all kinds of marvelous things,” he says.
Greenberg additionally explains the team’s “one-hour challenge” strategy in which it would get a customer’s data and do a quick brainstorming session. “Then, we go off and spend one hour programming a dashboard. And we’d shoot for an 80/20 on one of their problems [finding an 80 percent solution]. And we’d go back and we’d hand it to them. And we would hear, ‘This solves 80 percent of my problems.’”
