Patients with COVID-19 have inundated hospitals in Italy, forcing doctors to make agonizing decisions about who should receive lifesaving care. Patient surges could soon demand distressing triage decisions in U.S. intensive care units (ICUs), too. As of Thursday, there weremore than 13,000 confirmed cases in the U.S., and the nationwide death toll had risen to 175.
In February a study in Operations Research used mathematical modeling to determine which kind of triage policy could be useful in an ICU during such a surge. The paper analyzed circumstances in which patients could be queued for admission to a hypothetical ICU with limited beds or transferred to a general ward as their condition changed. The goal was to find a heuristic, or rule of thumb, for clinicians that minimized the average mortality rate of all patients over time, which is the goal of triage in the real world.
“A lot of times, medical professionals are really focused on making this one decision for the patient who is right in front of them,” says Laura Albert, a systems engineer at the University of Wisconsin–Madison, who was not involved in the study. “It’s really hard when they have to ask the patient to wait because that will save many more lives across the system. These heuristics are really valuable for service providers, because otherwise it is so hard for them to make that call in the moment.”
On supporting science journalism
If you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
Using computer simulations, the researchers applied a heuristic that determined who should be admitted to an ICU bed by estimating how much each patient’s chances of survival increased by being there and then dividing the figure by the number of days that person would probably need to stay. Individuals whose ratio was highest were prioritized. The researchers also examined how the heuristic worked when additional patient health conditions were added.
The study compared the heuristic’s collective mortality rates with those of possible triage scenarios. One policy provided beds on a first come, first served basis. Another discharged patients who were the least likely to be worse off in a general ward to make room for new ones. A third approach randomly discharged people from the ICU when new patients arrived.
Overall, the ratio heuristic prioritized patients who were expected to receive the highest benefit per each day they occupied an ICU bed. Study co-author Nilay Argon, an operations researcher at the University of North Carolina at Chapel Hill, says it was particularly valuable when an individual’s state improved or worsened during his or her stay. “As soon as they change state, then their triage should be applied again,” Argon says. Previous operations models of triage decisions in ICUs have not considered a patient’s condition becoming different, she adds.
A first come, first served approach used in one hospital in Wuhan, China, when the coronavirus began to spread may not have been the best strategy. Shu-Yuan Xiao, a pathologist at the University of Chicago, was in Wuhan at that time and saw how health care workers responded. He even assisted them. “The hospitals were overwhelmed,” Xiao says. “They simply didn’t have that many ICUs, and the ICUs had a first come, first served” policyin the beginning, which may have contributed to the initial high mortality rates in the city.
“Health care is only as good as the resources that we have for it, and the resources available [for one patient] are actually a function of how you treat other patients,” Albert says. “You can’t always make these treatment decisions in isolation. And we really see this when there’s a big patient surge.”
Edieal Pinker, an operations researcher at Yale University, says reserving a bed for severely critical patients—a practice called “idling”—when less critical individuals are waiting brings yet another layer of complexity to triage. “Once you’ve tied up that bed, you’re blocking somebody,” he says. “It’s a tough decision to make, because you’re telling a patient who’s in front of you now [that] they can’t have this. That’s hard for people to do, so you need guidelines and discipline.” The new study only addressed nonidling policies.
Similar to what the new model indicated, when a patient with a low chance of recovery is tying up an ICU bed for many days, and multiple other patients could be stabilized in that bed, Pinker says, clinicians will have to make the decision to move that patient to palliative care. “The danger, though, when you move infectious COVID-19 patients, is that you need a place to move them where you’re not going to end up spreading the virus even further,” he adds.
Models are not necessarily the final answer, says Jennifer Horney, an epidemiologist at the University of Delaware’s Disaster Research Center. She cautions that built-in assumptions may not translate to real-world scenarios. “I think that we can consider [models] as part of a planning tool,” she says. But “it’s important to be judicious when using data from modeling to try and predict exactly what's going to occur in a real-life situation.” Horney says that “after-event” studies that collect data from health care facilities following a real outbreak, such as the 2009 H1N1 pandemic, and use them to predict what would occur in a similar event, may be preferable to models that make assumptions that may or may not play out.
Indeed, it may be too early for hospitals to apply the new study’s heuristic to a COVID-19 patient surge. One difficulty is a lack of data on the survivability rates of the disease, says Scott Levin, a biomedical engineer at the Johns Hopkins School of Medicine, who co-designed an electronic triage system for Johns Hopkins, a machine-learning program that uses health record data to help categorize emergency room patients. “We don’t really have a lot of historical data about who’s going to benefit from an intensive care unit,” he says. As data accumulates, Levin says,updated models can produce triage recommendations that are more attuned to what’s happening with COVID-19.
Without robust survivability data, flexibility will be key to dealing with a coronavirus patient surge, says Pinar Keskinocak, a systems engineer at the Georgia Institute of Technology. She says it is important for health system administrators and policy makers to think outside the box about how to modify workflow and processes.
One example comes from Demetrios Kyriacou, a physician at Northwestern Memorial Hospital’s emergency department—a front line for triage that has about 100 beds. Kyriacou says the hospital’s disaster committee has discussed expanding the triage area into other parts of the facility, even including an ambulance bay, should the need arise. “If we would have intervened earlier in terms of isolating people who are sick, I think we would have a much less problematic epidemic going on in this country,” he says.
Read more about the coronavirus outbreak here.