Modeling allows those in power to rely on science to make critical decisions
Computer simulations have been a critical part of the world’s response to the COVID-19 pandemic, as millions of people closely follow projections about infection rates. Some computer models in mid-March predicted more than 2 million deaths if the U.S. didn’t act fast to contain the spread.
But simulation models can do much more than project death rates. One international group of computer modelers has called on government decision-makers and their own academic community to be more proactive in using simulations for a pandemic. The Operational Research Society, based in England, published an article recently titled, “How simulation modelling can help reduce the impact of COVID-19” in the Journal of Simulation. One of the co-authors is John Fowler, the Motorola Professor of Supply Chain Management in the W. P. Carey School of Business at Arizona State University.
“We wanted to make people aware that there are models that could have helped with a lot of the elements that are trying to minimize the impact of the pandemic,” Fowler said.
“It really is important for policymakers to get heavily engaged with the modelers during a time like this so they can make more informed decisions.
“This is one of those cases where science should inform policy decision-making.”
John Fowler, the Motorola Professor of Supply Chain Management in the W. P. Carey School of Business.
The article is a “call to arms,” the authors said, meant to elaborate on the different types and uses of models and the need for the modeling community to take advantage of the momentum created by the COVID-19 pandemic.
Fowler said that the public now has seen the effectiveness of computer simulations.
“We have a team at ASU that are building those models and they were advocating not opening back up (after the shutdown in April) because they predicted this influx of cases,” he said. “And lo and behold, their models pretty accurately predicted what would happen.”
Fowler and his co-authors compared computer modelers working in the throes of the pandemic to Bletchley Park, home of a transdisciplinary community that worked feverishly to break enemy codes during World War II. “What this shows is that good teams help to solve difficult problems,” they wrote.
The article explains that computer modeling is essentially a way to run experiments “in silico,” or in a computer, rather than on real people. It can answer a question such as, “What happens if three people who are infected with the virus get on a plane for three hours?”
“Modeling refers to building computer software that will in some sense mimic what’s happening in the real world by manipulating bits inside the computer,” Fowler said.
“This falls under operations research, which is a discipline within applied mathematics. It started most formally during World War II to try to do things as effectively as possible.”
Modeling doesn’t have to be done on a computer, but doing so greatly increases the speed of the calculations, he said.
There are two main types of models.
“One would be simulation models, where you don’t try to make any judgments about what comes out of the model — whether it’s good or not.
“It doesn’t drive to a good answer. You play with the model and change the inputs to see what happens to the outputs,” he said.
The second kind is an optimization model.
“That would be a model that says, ‘How can I set the inputs to get the best output for a given criteria for some measure, such as the least cost?’
“It’s, 'I want to set the inputs to give the best cost solution, or minimal deaths or the most number of healthy people.'"
The public has become familiar with epidemiological models, which simulate the spread of the disease itself, Fowler said.
“There are models that will predict what will happen partly based on what has happened so far but are also dynamic in the spread of disease,” he said. “The model doesn’t tell you an exact number but it will tell you a direction.
“You’ve seen those graphs with lines going up and down. Those are epidemiological models, which would allow you to look at the impact of social distancing and masks.”
The article describes an array of other pandemic decisions that can be managed with different modeling techniques, such as delivery of vaccines, hospital bed capacity, staffing, thresholds for patient admission and discharge, and mental health services. For example, one type of modeling, called “agent-based modeling,” takes into account the variability of human behavior and is useful for situations such as how to deploy a limited number of testing kits.
The authors recommend combining several types of models for some complicated issues, such as how to manage the end of social distancing.
Fowler, who is an expert in supply-chain modeling, said that simulations can improve the way resources are shared.
“You could build models of what happens if you share resources within a region, such as with PPE or ventilators. You would try to match the supply of services to demand that you see,” he said.
Fowler and his co-authors cite some of the obstacles to wide use of computer simulations, including the fact that different modeling teams that are currently siloed would need to work together, like at Bletchley Park.
Another critical hurdle is that decision-makers would need to be involved in creating the models. The problem with that is not only are they too busy during a pandemic, but that kind of collaboration has rarely happened. So the authors call on the modeling community to think about creating “templates” that could be quickly deployed.
“You still would have to supply data, transmission rates, etc. But if you have the platform already built, you could plug in the data rather than build new models.”
Models can be adjusted continuously along the way, based on things like the actual percentage of people wearing a mask, he said.
“There’s a lot we learned from the 1918 flu that influences the way we build models,” Fowler said.
“Part of the call to arms would be to have these previous pandemics in the memory banks of these models. And very early on, you try to determine which of the previous pandemics does this new one mimic most closely? You could use that as a guide.”
The authors ask for government agencies to provide resources now to create better computer simulation in the future.
“Because there will be another pandemic,” Fowler said. “It’s a fact of life.”
Top image by Pixabay
