The Booth Healthcare Analytics Laboratory is giving new solutions to students—and new life to hospitals.
- January 10, 2016
- Health Care
It’s April 15, two weeks into the course. Adelman has told the student UCM heart-failure readmissions team that they will be giving a presentation next week on data visualization. He starts with the basics: “Tell us about your dataset. How big? How many patients?”
Then he ups the ante to ensure that the team engages their audience: “Draw eight to ten visualizations of data. Use histograms. Use color. Kick it up a notch. Make your graphics the most interesting graphics they can be.” He shows a YouTube video of a TED talk by Swedish doctor and statistician Hans Rosling. It’s a dynamic collage of facts and motion graphics that tells the story of Asia’s economic ascendancy over the West. This is the standard Adelman sets for his students. The team has their work cut out for them: distilling several years of discharge data and thousands of patient records, each containing hundreds of clinical variables, into a compelling 15-minute visual narrative.
Pat Ward, a PhD student in economics and a former Air Force officer, selected the students who would address the UCM heart-failure readmissions issue and served as their mentor. He chose these six students (of 24 in the course) to contribute "a huge gamut of skills," ranging from a PhD statistician to a statistics neophyte, and offering experience in industries from medicine to marketing:
Adam Vohra, a third-year student at the UC School of Medicine pursuing a joint MD/MBA degree
Ai Nguyen, BS '05, a pharmacy manager at Walgreens
Hyuna Yang, a statistician at a pharmaceutical company in the Chicago suburbs
Katerina Steele, AB '03, an associate director at a pharmaceutical-marketing agency
Kristin Francoz, a former health-policy researcher at a think tank
Yizhen Dong, formerly in strategic planning at a biotech company
“I’ve got the only class I’ve heard of that does big data and brings in the teaming skills as well.”
“It’s a pretty large change we’ve initiated,” Dr. Corey Tabit, a cardiology fellow and the project sponsor at UCM, had told Adelman back in February. “We’ve focused on standardizing the heart-failure care at the hospital, at home, and also in the emergency department.”
When patients are admitted to the hospital with a diagnosis of acute heart failure, they are evaluated by nurses and are assigned either to the preexisting course of treatment (the control group) or to the new intervention. Patients in the new pathway are given an intensive education program, specifically designed by the nurses to be easy to follow after discharge from the hospital. Pharmacists, nutritionists, occupational and physical therapists, and social workers also spend time with the patients and offer management strategies for their condition. All efforts aim at keeping the patient healthy enough to live comfortably at home.
“It isn’t one specific intervention; it’s the bundle, the whole program,” said UCM cardiologist Rupa Mehta Sanghani, who worked with Tabit on the project. “The idea is that the more of them you do, the lower your readmission rate.”
In Tabit’s early evaluation, the new protocol appeared to show an 80 percent reduction in 30-day readmissions compared with the control group. But he recognized the need for a more sophisticated analysis—which is where Booth came in.
A data analytics lab course could be designed around any industry: technology or manufacturing, media or telecoms. But healthcare generates so much information and is so resource intensive, that “there is this massive interest in using big data to improve health-care delivery,” Adelman said. Healthcare is such an economic disaster area in the United States that cost-effective solutions in the sector need to be found. It’s a growing area of interest at Booth. “More and more students want to get into healthcare,” Adelman said.
When Steele, on the UCM student team, was originally choosing among MBA programs, Adelman’s class “sealed the deal for me at Booth,” she said. “This was a professor offering a class in exactly what I want to be doing.”
Matthew L. Primack, ’13, (XP-82), vice president of clinical institutes and business development at Advocate Condell Medical Center in Libertyville, Illinois, sponsored a project several years ago. “How you analyze data is fairly straightforward,” he said. “What you do and how you implement the changes based on the data is a very, very different art form. What Dan Adelman did very well was that he made sense of the data analytics and turned it into actionable opportunities.”
By early May, the Healthcare Analytics Laboratory heart-failure readmissions team zeroes in on determining actionable opportunities out of 36 variables—in demographics, clinical data, and treatment information—for thousands of patients. The students have run about 1,000 iterations of their statistical program in an attempt to discern a small subset of variables that would be most likely to predict which patients entering the hospital are at highest risk of returning after discharge.
“We are better able to predict low risk than high risk,” Nguyen said. “We can predict 80 percent of those who won’t be readmitted, but only 60 percent of those to be readmitted.” The students need to try more variables to get a more precise picture of what issues predict future readmissions.
“We know from the literature that the time of greatest risk is right after discharge from the hospital; then the risk goes down,” Vohra said. “Then it goes back up at some point later in time. The problem is, we don’t have a good predictor of when that comes.”
The team won’t get their final data dump from the hospital until mid-May, which forms the basis of the analysis they have to present at the final class on June 3. Only a week later, they have to present their findings to the hospital sponsors.
“If you have the right data, and you know what this data is about, the statistical analysis can be done very smoothly. You know what you have to do,” said Yang, the statistician. “The difficult part is identifying the right data and doing the quality control.”
“How you analyze data is fairly straightforward. How you implement changes based on the data is a very different art form.”
At some schools, Adelman said, the students come to class, complete a project, and they’re done. “They aren’t held accountable for the discipline-based learning they receive.” He’s developing what he considers “more of a Chicago version of experience-based learning, which leverages our unique strengths.” Students in his course not only must master content-anchored statistics and analytics work, but as part of the soft-skills training they have to draft team contracts, write personal reflections, and submit peer feedback. Communication skills are key. Every two weeks, each team makes a presentation to update the rest of the class on their research. Adelman critiques the content, flow, and persuasiveness, as well as the student’s speaking style and body language.
For student Yizhen Dong, an early presentation, when Adelman had a public-speaking coach on site, was an eye opener. “I say a lot of ‘ums,’ and I fill in words that don’t add any value to the presentation,” Dong realized. “It’s something I’m trying to work on.”
The chance to combine data analytics with the soft skills—presentation, team dynamics, and storytelling—is unique, said Harry L. Davis, Roger L. and Rachel M. Goetz Distinguished Service Professor of Creative Management, who pioneered experience-based learning at Booth in the 1980s. “This is, in a sense, a start-up, with really interesting opportunities for you here at Booth that you cannot get anywhere else,” Davis told students in the first class session.
By the June 3 class presentation, the UCM team has identified that, in the roughly six months the hospital’s new protocol has been in place, the intervention patients each received an extra 90 minutes of care. The cost savings from the intervention would be $385,000 a year—the equivalent of three new nurse practitioner hires.
The students have also distilled the predictive algorithm to eight yes-or-no questions for the hospital staff to assess on each patient’s admission. They have found the questionnaire to achieve the same predictive effect as the far more complicated medical models cited in the published literature on this topic.
Adelman appreciates their data insights—and he also knows this is a dress rehearsal of sorts for the team’s presentation next week at the hospital. The professor offers some directorial advice: “Start thinking how the different components fit together, and you can build some drama.”
It’s 8 a.m. on Wednesday, June 10. In a small lecture hall at UCM, the cardiology team, as well as the hospital’s C-suite executives, awaits answers—from Booth students. The students wear jackets and ties, skirts and heels; the vascular surgeons are still in their scrubs.
Francoz sets the stage: “UCM has an above-average rate of readmissions. But even if you control for poor case mix, it’s still in the middle of the pack.” Each student takes the floor in turn and advances the narrative. They look sharp; they sound knowledgeable and well rehearsed. The entire presentation has momentum and confidence, poise, éclat.
Yang unveils the key finding around the intervention protocol: “It was effective . . . but not that effective. OK, so now we can all go home and relax.” She has the cardiologists laughing. The students lay out the numbers, explain the analysis, and give recommendations to the senior executives. They recommend that to achieve statistical significance, the program would need 420 participants in the intervention group and 420 in the control group. That would take about a year.
Tabit listens, rapt, and says he couldn’t be more pleased. “The quality of analysis is just superb. It’s easily the equivalent of what we would have gotten from a consulting firm. The capabilities of these students are amazing.”
Sanghani says she can’t believe how much was accomplished between January and June. “I would love to have you guys keep going with the data.”
“Is this class every quarter?” interjects cardiologist Kirk T. Spencer. “Because we have lots of problems.”
Indeed, Adelman is currently seeking the kinds of problems that will challenge his next cohort of students when the course begins again this spring. Meanwhile, the UCM cardiology team continues to build on the insights from the Booth report. “The model they came up with was very impressive,” Tabit said. “It has performed remarkably well. It has only eight variables, so it’s easy to work with. We have put it into a larger risk-stratification scheme. It’s by far the most important piece of the model.”
The Booth student findings led the hospital to streamline their process and better predict patient risk. In the July through October period, readmissions fell to between 8 and 9 percent. UCM Medicare payment penalties are down to 0.38 percent for fiscal 2016 from 0.54 percent in 2015. “For $100 per patient, I can reduce readmissions by 50 percent,” Tabit said. Assume each readmission costs the hospital around $10,000. “That’s a tremendous cost savings to the health system. The Booth project let us figure out which patients need these aggressive interventions and which don’t.”
Adelman is rightfully proud. “The analyses were just impressive,” he said. “They certainly lived up to the school’s reputation. Any sponsors sitting there had to be thinking, ‘Damn, these guys are good.’”
We'd love to hear your Booth memories, stories, connections...everything.