A virtual hospital: How a model can test real-world innovations
A new hospital is an incredibly expensive endeavor. Building a good one can cost more than $1 million per bed, so it makes sense to test out new layout ideas before any actual construction gets underway.
Duane Steward, DVM, MSIE, Ph.D., assistant professor in the Texas A&M Health Science Center Biomedical Informatics Center, uses a software program to test different variables in the efficient functioning of a hospital emergency department.
There are several types of simulation commonly used in health care: Device-based, role-based and model-based. An example of device-based simulation would be using a manikin to learn how to properly perform CPR. In role-based simulations, future nurses and physicians act out scenarios with standardized patients. Model-based simulations—like the ones Steward works on—are a little different in that they don’t require a physical object or interaction, but instead can exist solely on a computer. However, they are extremely powerful tools for testing novelty and innovation in health care.
One example of this is when Steward worked on testing new ideas for the emergency department of the Nemours Children’s Hospital in central Florida several years before it opened on October 22, 2012.
“On opening day, the entire team would have to function like they knew what they were doing,” Steward said. “And hospital leadership wanted to move patients through the system in a way that had never been done before.”
The goal of any hospital is to get patients through the emergency department and either discharged or admitted to the hospital as quickly as possible—a measure called “length of visit”—while still caring for each effectively. The simulation can test how factors like number of nurses or how the physicians are allocated can change, for better or worse, this length of visit.
Instead of labeling incoming patients as those that were critical and those that could wait, Steward and his team tested out the system of dividing them into four groups he calls “streams.” In addition to the critical care stream and the “fast track” stream (perhaps echoing the system at nearby Disney World that gets certain riders through the lines quickly, this group doesn’t need much care and can be treated and discharged quickly), Todd F. Glass, M.D., the hospital’s chief of pediatric emergency medicine, devised two middle streams: therapeutic and diagnostic. These two streams are patients who likely do need emergency care but they are not likely to die if they’re not seen immediately. Therapeutic stream patients have an obvious ailment—a broken arm, for example—and can be treated immediately. The source of diagnostic stream patients’ complaint is less clear, so testing needs to be performed before they can be treated.
“We took ideas and embodied them into the model,” Steward said. “In doing so, we determined that we actually needed a fifth stream, a subset of the critical care stream, of ‘code’ patents.”
“Just taking ideas out of people’s heads and putting them into a computer program is illuminating,” Steward added.
Even something as seemingly simple as the size of the waiting room needed can be tested using these simulations. Because this is an emergency department for a children’s hospital, one assumption will be that each patient will come with one or more adult family members, and those parents might also bring other children who will also need to be comfortable while their sibling is being treated.
The hard work paid off when the design for the hospital won a citation in the 2011 Modern Healthcare Design Awards.
Modeling the flow of people through the hospital is not a new idea: The technology in some form has existed since the 1970s, and the hospital emergency room was the classic example. However, modern technology has allowed for greater complexity in the modeling process and faster processors make it easier to input details and run multiple scenarios for “centuries” to see what percentage of the time certain events occur. “I can give them centuries of experience with their emergency department before they ever stick a shovel in the ground,” Steward said.
Essentially what Steward and his team are able to do is feed a number of “building blocks,” each of which represents some aspect of the real world, into the computer program. The building block might be an “agent,” like a nurse or a patient, who has a specific goal: Tending to the patient or being treated, respectively. Those agents then interact with other building blocks, like the number of available treatment rooms, and these interactions lead to different outcomes.
[pullquote align=”right” cite=”” link=”” color=”” class=”” size=””]“I can give them centuries of experience with their emergency department before they ever stick a shovel in the ground,” Steward said.[/pullquote]
“It’s a matter of deciding what parts of the real world you want to explicitly present,” Steward said. He explained that you can almost always add new details to further refine the simulation, but although modern processors are extremely powerful, the human building the model still only has a set number of hours in the day. “It really comes down to a negotiation of time—how soon you need the answer and what level of detail you are willing to live with. It’s a matter of who is using it to make what decisions about actual hospital design or operations.”
Of course, those decisions will only be as good as the assumptions put into the model in the first place. “As scientists, we want to go find some record of history to make the assumptions, but you don’t have to throw away the possibilities just because you don’t have historical data,” Steward said. “An estimate is better than a wild guess.” And even if there is no good estimate, the scenario can be run with several different values for that variable, which allows testing of how sensitive the ultimate outcome is to that particular feature.
Because the world really does have some degree of randomness, the computer can factor that natural variability in as well. “We set up the assumptions, we turn it loose, and we sit back and see what it tells us,” Steward said. Because the same scenario can be run multiple times with different values for the random variables each time, these simulations can help the designers anticipate—and thus prepare for—the unusual.
Ultimately, though, it’s what happens in the real world that matters.
“Until we run further tests, we can’t know for sure how much we can take credit for,” Steward said. “All we know is that since opening, the emergency department has had an average length of visit that is only half that of the rest of the industry.”