Author: “There are very few situations where a model perfectly corresponds with reality. By definition, models are simplified from reality. In some way or another, all models are wrong.”
BOULDER, Colo. — It’s proven incredibly difficult for scientists, organizations, and governments alike to accurately predict the spread of COVID-19 through countries and communities. Now, however, researchers from the University of Colorado, Boulder say they’ve developed a new mathematical tool that can help produce more accurate disease infection predictions.
The team behind this new study even went so far as to say that all prior disease models are somewhat wrong.
“I don’t think this tool is going to solve any epidemiologic crisis on its own,” explains study leader Rebecca Morrison, an assistant professor of computer science at CU Boulder, in a release. “But I hope it will be another tool in the arsenal of epidemiologists and modelers moving forward.”
Researchers used the 2016 Zika virus outbreak as a test case. While examining the spread of Zika, they created a new tool called an “embedded discrepancy operator.” They believe this tool is capable of fixing prior models that fail to correctly predict the spread of different diseases. So far, they’ve only achieved this feat regarding the Zika virus, but they believe the tool can also help with COVID-19.
“There are very few situations where a model perfectly corresponds with reality. By definition, models are simplified from reality,” Morrison says. “In some way or another, all models are wrong.”
Co-author Americo Cunha, a Brazilian mathematician and assistant professor at Rio de Janeiro State University, was trying to recreate the spread of Zika using a common disease model called SEIR (Susceptible, Exposed, Infected or Recovered). Unfortunately, Cunha’s work and estimates were never able to truly match up with recorded Zika data.
“The actions you take today will affect the course of the disease,” Cunha explains. “But you won’t see the results of that action for a week or even a month. This feedback effect is extremely difficult to capture in a model.”
Instead of just giving up, Cunha and Morrison collaborated to try and find a solution. It was clear that their current model couldn’t recreate real data all on its own, so they decided to use that data to create a better model in realtime.
That’s where this new tool came into play. They suggest imagining it as a “spy” that operates within a model. As more new data is added to a model, the tool constantly rewrites that model’s mathematical equations to better represent that new information.
“Sometimes, we don’t know the correct equations to use in a model,” Cunha notes. “The idea behind this tool is to add a correction to our equations.”
Regarding the spread of Zika, the tool was a big success; the newly revised model ended up being almost identical to real infection statistics.
The research team are already gearing up to use their new invention on coronavirus models that have proven inaccurate.
“This epidemic has revealed how difficult it is to model a real system,” Morrison concludes. “But I hope that people don’t take that to mean that we shouldn’t trust our scientists.”
The study is published in Chaos.