When patients arrive at the hospital for a surgery, they probably don’t think about what goes into scheduling the operating room (OR) where the procedure will happen.
But OR scheduling is something that Chaithanya Bandi, associate professor of operations at the Kellogg School, thinks about—a lot.
“How best to schedule ORs is an important puzzle hospitals need to solve,” Bandi says. “They need to allocate rooms while both honoring the requests of surgeons and keeping costs in check.”
ORs are indeed costly to operate—$1000 per hour is typical—but they also make money. Surgical operations and associated hospitalizations typically generate about 70% of total hospital revenues. So optimizing OR scheduling is a top priority for administrators.
Bandi and collaborator Diwakar Gupta of the University of Texas, McCombs School of Business studied surgical-request data at a large community hospital. They found that hospitals generally keep more ORs at the ready than might be needed at any given time in order to accommodate unpredictable surgeon requests.
In hopes of reducing this cost, the researchers developed an innovative algorithm to improve OR operations, minimizing the number of ORs hospitals had to keep open, while still honoring surgeon requests.
“Our goal was to help hospitals improve on current OR operations while still satisfying surgeon needs.”
Application of the algorithm can drive savings of about 20% in OR costs—a figure with significant impact on hospitals’ bottom lines. And Bandi sees applications for this algorithm-driven approach in other fields that have high costs and unpredictable scheduling requests.
“We designed the algorithm for a specific medical context,” he says, “but its applications stretch well beyond that.”
The Challenge of Hospital Scheduling
Hospital administrators generally use a scheduling system to determine which OR will be assigned to meet a surgeon’s request for a particular surgery.
This process is deceptively complex because in order to optimize OR scheduling, hospitals must take three main considerations into account.
First, the cost: a fully staffed OR costs an estimated $15-20 per minute to run, or about $1,000 or more per hour. That means the fewer ORs the hospital can keep open at any given time, the greater the savings.
Second, hospital administrators try their best to honor surgeons’ requests for specific OR slots.
“Surgeons are very important from a hospital’s point of view,” Bandi says, “so the administrators work hard to respect their preferences.”
The third factor is the “noisy” nature of OR requests, which arrive at irregular, unpredictable times given that patients’ health problems do not occur on a regular schedule.
“The challenge is that the hospital doesn’t know when the OR requests will arrive,” Bandi says. “It’s just a call that the surgeon makes, and the hospital needs to find the space—they can’t move the patient to a different hospital or surgeon.”
Another unknown is the exact amount of time an upcoming surgery will take. “Surgeons have a general sense of how long a given procedure takes,” Bandi says, “but there is a very wide range—it could be half the requested OR time or double, and hospitals need to find a way to accommodate this.”
Because so many factors in OR scheduling are unpredictable and unknown, hospitals tend to prep and staff an excess number of ORs to accommodate all requests. Bandi says, “It’s a problem of excess capacity—and a costly one.”
Better OR Scheduling, Lower Costs
The factors above suggest an intriguing research opportunity.
“Our goal was to help hospitals improve on current OR operations while still satisfying surgeon needs,” Bandi says.
Using 18 months of surgical scheduling data shared by a large hospital, the researchers developed an algorithm to do exactly that. Their model used a novel approach to analyze OR request data and determine the minimum number of ORs to keep open at any given time.
“But better utilization of OR capacity helps the hospital see more patients in a shorter period of time, with lower rejections or delays for surgery requests.
The results confirmed that the hospital was needlessly operating with excess OR capacity.
A typical hospital, like the one in the study, keeps about 10 ORs open, Bandi says. “But in Phase 1 of our study, we show that keeping only eight open would allow the hospital to handle the requested volume of surgeries while achieving the same service levels.”
The researchers knew, however, that this first analysis did not take into account more specific surgeon requests, such as wanting the same surgery times each week. So the second phase of the research took this into account, accommodating specific requests by either shuffling surgeries among available ORs, adding another OR, or a combination of both.
Importantly, the model enables hospital administrators to place greater weight on specific factors as needed. For example, if a particularly important surgeon wants a specific OR at a specific time, the hospital can give this request top priority, making adjustments to the remaining surgeons’ schedules as needed.
“We enable the hospital to control trade-offs related to scheduling while still saving a lot in overall OR costs,” Bandi says.
Indeed, the reduction in the number of ORs the hospital must keep open means large savings—roughly 20% of OR costs, on average. “We simulated multiple OR request scenarios that resulted in anywhere from 10% to 25% in total savings,” Bandi says.
This is great news for hospitals. But it does not change the financial equation much for insurers or the patients who occupy those ORs.
“In healthcare settings, these kinds of savings don’t translate immediately to consumers,” Bandi says. “But better utilization of OR capacity helps the hospital see more patients in a shorter period of time, with lower rejections or delays for surgery requests.” In the long run, however, he expects that these savings will lead to lower healthcare costs and thus lower insurance premiums.
Other Uses for a Scheduling Algorithm
This optimization algorithm has potential uses beyond the hospital setting, Bandi says.
For example, cloud-computing data centers must optimize how they schedule “big data” analytics projects, such as large simulations requested by researchers.
“Projects like these can take hours to run and incur very large energy costs,” Bandi says. “By some estimates, as much as 12% of the world’s total energy goes toward these projects.”
As with ORs, data centers receive analytics job requests at unpredictable times, and it’s not clear upfront exactly how long a given project will take. As a result, the centers tend to overcommit, retaining excess computing capacity to accommodate all requests. The result: wasted energy and extra costs.
Bandi is exploring how to apply an algorithm similar to the OR-optimization model to help data centers improve project efficiency. “There’s a lot of discussion about how to ‘green’ these data centers to help them use less energy,” he says. “Our algorithm can play a big role in this effort.”