IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

Predictive Analytics Used to Determine Hospital Bed Capacity

Hospital bed availability changes daily, especially during COVID-19, as patients are admitted and discharged. The software gives staff a real-time look at data as it evolves, so they can develop models from different variables.

Louisiana doesn’t enjoy the same level of hospital capacity of say, Texas, nor does it have the capacity to handle a surge of patients or evacuees. Those often have to be moved out of state.

That’s why accurate pre-disaster forecasts of the number of hospital beds or ICU beds, or the number of people who will have to be evacuated to another state is so critical. And often, coming up with a single number of beds or evacuees, for instance, leads to miscalculation.

That’s why modeling that finds a more accurate range is such a benefit.

“When you rely on one number, you tend to overcommit resources or undercommit resources,” said Henry Yennie, a program manager for the Louisiana Department of Health and Hospitals. “So large-scale hospital evacuations are different in Louisiana than they are in Texas, for example, which has a lot of health-care resources so they can move [hospitalization] from their coastal areas inland and they have tremendous hospital capacity. Louisiana doesn’t.”

When the coronavirus pandemic hit this spring, the department deployed a predictive analytical software tool, Palisade’s @RISK decision support software, to support policy decisions and data probabilities more accurately instead of trying to arrive at one single number of available beds or how many people might become infected with the virus.

The “deterministic” models that provide that one number tend to be based on uncertain variables and often produce overestimates, Yennie said.  The probabilistic or stochastic models, such as @RISK, examine a wider range of possibilities and how likely each one is to occur. “I just quit paying attention to those other models,” Yennie said.

“We started off with that @RISK in the early days of COVID trying to predict the number of fatalities, and from my perspective, that’s important because we’re responsible for any sort of mass fatality response,” he said.

The department used the software to do simulations to try to get ballpark ranges of numbers for fatalities and how many ICU beds might be needed. They used Palisade software to run charts on a variety of different levels based on data availability for ICU bed availability in different areas of the state over time to get an idea of where a shortage might reveal itself.

Bed availability changes by day as patients are admitted and discharged. The software allowed Yennie and staff to get a real-time look at the data as it evolved and develop models based on different variables.

“The simulations we were able to do weren’t exact— they never are — but they got us in the ballpark and we were able to avoid committing major resources for something that was never going to happen,” Yennie said. 

The models also allowed hospitals to determine when to open up for “normal business” — such as everyday surgeries, elective surgeries and trauma surgeries — and keep the health-care system going.