Computer Science Points to Holy Grail of Medical Scheduling
Most researchers develop collaborative projects with colleagues across the campus or across the hall, but the collaboration between Janet Jokela, MD, and Sheldon Jacobson, PhD, is unique. Their collaboration was created across the kitchen table.
Jokela is the head of medicine in the University of Illinois College of Medicine at Urbana, and Jacobson is a professor in the Department of Computer Science, but they also happen to be married. They worked together to create a new computer program that solves one of the toughest scheduling problems in medicine. In fact, the marriage of their two minds stands as a fitting metaphor for the marriage of research efforts that is increasingly taking place between the Colleges of Medicine and Engineering on the Urbana campus.
When Jokela became director of the University of Illinois Internal Medicine Residency Program in 2010, she found herself grappling with the nightmare task that faces every residency director across the country each spring. She had to balance a multitude of needs and requirements as she forged the annual schedule for 48 medical residents working in two Champaign-Urbana hospitals and the Veteran’s Administration hospital in Danville.
Jacobson was astounded by the massive number of hours that Jokela was putting in to develop this complex schedule, so he proposed a solution. “I think this can be solved algorithmically,” he told her. “I believe we can build a model that would enable you to press a button and get your schedule automatically.”
Several years later, they have succeeded.
Jacobson and two students worked with Jokela and others in the College of Medicine to create a user-friendly computer program that transforms a six-week task into something that takes 45 seconds. Their effort took top honors in the research poster competition for the Association of Program Directors in Internal Medicine this April in Nashville.
The new program, ReSchedule | Med, is in the initial testing stage—the “alpha stage” when they work out the most glaring bugs. The program has successfully produced schedules for the U of I residency program, and they are aiming to do the beta testing in early 2015; other residency programs around the country are already lining up to participate.
When Jacobson first told Jokela that he might be able to create a program to create schedules automatically, she was skeptical. “It was difficult for me to envision how this could possibly be translated into a mathematical formula or algorithm.”
What makes the task so difficult is that the schedules first must ensure that every one of the 48 medical residents meets all of the educational requirements during 13 four-week rotation blocks at the hospitals. For instance, every resident must do at least one emergency medicine rotation before they graduate, and they must do at least three intensive care unit rotations, but no more than six.
On top of that, the schedule must meet the many needs of the hospitals, and a hospital’s requirements differ for first-, second-, and third-year residents. Making it even more complex are other requirements set by the national residency accreditation program. For example, residents cannot do two consecutive night-shift rotations, working 7 p.m. to 7 a.m.
“The complexity is daunting,” Jokela said.
Jacobson compared it to playing a high-dimensional game of chess. “When you make a move in one dimension, it affects every other dimension.”
Jokela said it takes roughly four to six weeks for several people to hammer out the initial draft of a residency schedule the old way by hand, and then they start getting feedback from residents. If a resident cannot do a rotation during a particular week, that means tinkering with the schedule, setting off a cascade of changes.
“The schedule explodes in multiple dimensions,” Jacobson said. The new RESIDENT program has been designed to handle these changes with ease.
Today, Jokela is head of the Department of Internal Medicine and no longer has to set residency schedules. But she continues to advise Jacobson’s team as they refine the computer program. She also raises awareness of the new tool. When she told Tom Cooney, MD, the governor of the Oregon chapter of the American College of Physicians, about the computer program, “His eyes got very big. He told me, ‘Do you realize that this is the Holy Grail of internal medicine residency directors? If you could solve this problem, that would be huge.’”
Cooney, who is also the former long-time director of the internal medicine residency program at Oregon, told her that a number of years ago a private company poured millions of dollars into a similar effort to solve the scheduling dilemma, but they never succeeded.
To create the new program, Jacobson worked with David Morrison, a PhD student in computer science, and two computer science undergraduates: Jiayi Guo and J. Taylor Fairbank. The Illinois team used integer programming, a modeling technique in the field of operations research—Jacobson’s area of expertise.
“You define a set of variables that have integer values,” he said. “Every single resident has a large number of variables, which are like switches that are either on or off. If residents are in a particular rotation at a particular time period, the switch is on; and if they are not in a rotation, the switch is off. What we want to do is make sure all of the right switches are on at the right time to satisfy all of the educational requirements and all of the hospital requirements.
“It’s computational voodoo,” he added with a smile. “And it works.”
According to Jacobson, “This collaboration exemplifies exactly what the campus is trying to establish. The physical proximity of medicine and engineering on campus makes it possible to find solutions—and not just for local issues. These are national and international issues that people are grappling with.”
The RESIDENT program has been successful so far with the University of Illinois residency program, which has 48 people. But what about larger programs?
“Can it handle 70 residents or 90 or 120? The only way to find out is by trying it,” Jacobson said. “We look forward to the challenge.”