Illinois Tech Researchers Build Predictive Model That Warned Indian Officials About Omicron Wave
Long-term predictive models looking at how the COVID-19 virus spreads are lacking, but an Illinois Tech researcher’s work developing one of these models helped an Indian government think tank develop policy to help minimize the spread of the disease.
“” by Sanjiv Kapoor, professor of computer science, and Yi Zhang (Ph.D. CS ’25) was published by The Royal Society. It outlines a predictive model that accounts for relaxed versus strict lockdown policies, vaccine rates and time-lapsed efficacy, and new variants of the virus such as Delta or Omicron.
“We were able to use a model that we developed earlier, which incorporates human behavior into population interaction parameters, as well as changes in the susceptible population dependent on relaxed/stricter government policies,” Kapoor says. “This modifies the standard SEIR model developed many years ago. We were able to apply our model to a real-world situation, impacting a large population.”
The paper shows how their new Susceptible, Exposed, Infected, Asymptomatic/Undetected Symptomatic, Recovered with Linear population changes (SEIR-SD-L) model builds new compartments to account for those who have been treated in hospitals and at home, those who died, those under lockdown, and members of susceptible populations who are vaccinated. It also adds compartments for the virus variants.
Expanding the model by adding these new compartments provides more accurate predictions of how the COVID-19 virus could spread. According to the published data, predictions made using the new model on December 1, 2021, were 82 percent accurate, on average, 15 days later.
“The biggest challenge was the lack of predictability as to how the population would react to government policy changes, thus altering long-term predictions,” Kapoor says. “We would like to extend this work to improve the models of behavioral aspects of the population. In the current model, we have incorporated a pullback in the population’s interaction behavior when infection cases rise. Other parameters to include [down the road] should address exhaustion from restrictions and risk-taking behavior.”
The new model was used monthly to predict COVID-19 spread in India from June 2021 to March 2022. COVID-19’s second wave overwhelmed Indian hospitals. Results of the new model were supplied to an Indian public policy think tank, NITI Aayog, which was tasked with creating preventative policy to minimize the effects of COVID-19.
“I was connected to senior personnel at the government agency through an ex-colleague after he heard about my work on COVID modeling,” Kapoor says. “It is not clear to me as to exactly how the work was used for policy decisions, as long-term predictions are not accurate, but [can] serve as a warning about the severity of the infection spread. To our knowledge, the predictive reports were used for policy decisions and particularly welcomed was the pre-warning on Omicron."
Although the model did not pinpoint exact peaks in the spread of cases, the results were able to give policymakers useful forecasts to guide their work, according to the paper.
“We aim to apply this for modeling the spread of infectious processes, especially viruses in computer networks,” Kapoor says.
