What proportion of the population has been infected by the COVID-19? Answers to this questions are important for knowing the true morbidity, case-infection ratio, herd immunity, and vaccination policies of COVID-19. Rigorous estimation of COVID-19 infected population has been elusive. At the early stage of the pandemic, due to lack of data, models projected that infected proportion could reach 50% or higher. Seroprevalence tests (antibody tests) suggested that a much much lower percentage of the population has been infected with COVID-19.

The main challenge for mathematical models to accurately estimate the proportion of infected population is the nonidentifiability issue. Most of the models can be calibrated to fit the reported case data with good precisions. However, they could predict widely different estimations of the size and proportion of the infected population. Simply put, an SIR model fitting to reported case data is nonidentifiable.

To overcome nonidentifiability, data independent of the reported cases is needed for model calibration. Such data was typically not available at the early stage of the COVID-19 pandemic, and became available after the first wave. In this talk, I show that how administrative data related to COVID-19 testingcan be used to improve an SIR model and mitigate the nonidentifiability issue in the model. The estimated proportion of infected population is comparable to the level indicated by serosurveillance tests.

Dr. Michael Li is a Professor of Mathematics at the University of Alberta. He had his BSc and MSc degrees from Jilin University, and his PhD from the University of Alberta in Canada. He did his postdoctoral training at University of Montreal and Georgia Institute of Technology. His research interests are mathematical investigation and modeling of population dynamics of disease transmission, and in vivo dynamics of viral infections and immune responses. His research group collaborates with Alberta Ministry of Health on modeling HIV, TB and Influenza, and with China CDC on estimation of HIV incidence using mathematical modeling. He also collaborates with virologists and immunologists at the Li Ka-Shing Institute of Virology at the UofA on modeling viral dynamics. During the COVID-19 pandemic, his group has provided modeling support for public-health policies in Alberta.

Related publication: Why is it difficult to accurately predict the COVID-19 epidemic? https://pubmed.ncbi.nlm.nih.gov/32289100/