During a disaster, looking at a range of possibilities, rather than a target number, gives emergency planners a better way to gauge risk.
The basis of a solid risk assessment is understanding that probability exists in a range, not a specific number, says Henry Yennie, a program manager at the Louisiana Department of Health.
In fact, deterministic -- or number-based -- models have been off by a factor of 10 or more, he said. “It’s better to ask for a probability,” Yennie said. In an emergency, “it gives you a way to gauge what the likely risk is that some intervention might have to happen, and it gives you enough time to consider those responses that might have to happen, rather than just reacting.”
For instance, as Hurricanes Laura and Sally threatened the state in recent weeks, Yennie was able to assess the risk that each of the 1,494 health care facilities the department works with would lose power or be flooded, which helped prioritize response plans. To do that, he used Palisade’s @RISK decision support software, an add-on to Microsoft Excel that performs Monte Carlo simulations, a type of analysis that provides a range of possible outcomes and the probabilities they will occur.
The software is part of Palisade’s DecisionTools Suite, which can be used for decision trees and optimization. The department used another of the company’s products – StatTools – for statistical analysis and forecasting related to the coronavirus pandemic.
On March 13, the department began using the technology to forecast the availability of hospital beds, including those in intensive care units. Hospitals report data on bed availability daily, and an alert feature automated by the model color codes forecasts to indicate when help may be necessary.
“As we saw the pandemic ebb and flow, we needed some sort of early warning system about our hospitals and whether or not to have conversations … about suspending certain kinds of procedures that weren’t life-threatening or urgently needed,” Yennie said.
The department had to modify its models to collect the data and train hospital personnel to provide the information before it could perform simulations, but once the data structure was in place, “we quickly developed a template upload system where the hospitals could populate a template and upload it to the system so they didn’t have to manually input numbers as they had been,” Yennie said. “That has made the data we feed the models much more accurate.”
A key to shifting from hard numbers to ranges is getting leaders to see the value of probability, he said. Relying on a specific number can easily lead to over- or under-preparation.
“I think it’s educating the leadership that has to evaluate risk,” said Yennie, who’s been using @RISK since Hurricane Katrina devastated the state in 2005. “From my perspective, it’s educating them about the dangers of looking for a number.”
In 2008, Hurricane Gustav made the need for probabilistic modeling apparent after it left several large hospitals on generator power for weeks. “After about Week Two, they started making these frantic requests for fuel because their normal fuel distributor was blown away,” Yennie said. “We had no way to predict who might need fuel, and we were just reacting to these last-minute requests,” he said. “It was really, really nerve-wracking.”
Now, all health care facilities the department works with use applications in the ESF 8 Portal – developed after Gustav -- to store and share data on every generator they have. The health department extracts data from the portal applications to feed the @RISK models.
“We’ve got data on thousands of generators, including their type, model, type of fuel they use, where the fuel tank is, how big the tank is, their burn rate,” Yennie said. “Because we have that data in the background, we can start to spit out the data that feeds the Palisade model that’s based on that risk. It’s a very simple process of loading the data that we get and then running those fuel simulations,” he said. “It generates kind of a calendar for us – the top 20 hospitals or nursing homes or group homes or assisted living facilities that are in jeopardy of running out of fuel first.”
The department uses the probabilities to help determine response timelines, working with a concept called an H-hour. For storms, H-0 is the time at which tropical storm-force winds are expected to hit the coast -- and the point when ambulances can’t travel and flying becomes dangerous, Yennie said. Preparations are backed up to H-minus 72 -- about three days before a storm might hit and when risk assessments begin in earnest. At H-minus 60, the department decides whether hospitals need to decide whether to shelter-in-place or evacuate patients.
“So far it’s worked really well,” he said.
Editor's note: This article was changed Sept. 23 to correct one of the H-minus numbers.
NEXT STORY: How transit data can drive urban recovery