Prediction of SARS-CoV-2 transmission dynamics based on population-level cycle threshold values: An epidemic transmission and machine learning modeling study

Polymerase chain reaction (PCR) cycle threshold (Ct) values can be used to estimate the viral burden of Severe Acute Respiratory Syndrome Coronavirus type 2 (SARS-CoV-2) and predict population-level epidemic trends. We investigated the use of epidemic transmission modeling and machine learning (ML) based on Ct value distribution for SARS-CoV-2 incidence prediction in British Columbia, Canada during an Omicron subvariant BA.1-predominant period from November 2021 to January 2022. Using real-world data, we developed an epidemic transmission model that was first validated on outbreak data, and subsequently fitted to province-level data to predict incidence. Using simulated data, we developed a ML pipeline including five models to predict the reproductive number as a measure of transmission potential based on Ct value distribution, and validated it on out-of-sample province-level data. The epidemic transmission model demonstrated accurate prediction with the real incidence falling within the 95% credible interval of the predicted MCMC chains for both the long-term care facility outbreak, and province-level data. The ML models demonstrated good performance with a median mean squared error (MSE) lower than 0.17 across all models, and improved performance with increasing sample size. The variability of the Ct distribution around the mean was the strongest predictor of the reproductive number. These modeling approaches demonstrated utility for incidence and reproductive number prediction, and have potential to complement traditional surveillance in real time to guide public health interventions.