Constructing COVID-19 Pandemic Prediction Models Using Positivity Rate

Data science-based techniques have been widely applied in studies related to COVID-19 spread prediction. In these studies, different modeling techniques have been deployed to estimate the current and future trajectories of the pandemic. The estimation models for these trajectories mainly consider the change in the daily number of confirmed cases as a key modeling input variable. The enforcement of precautionary measures related to COVID-19, in several countries, relied on the estimation results obtained using these prediction models. This paper demonstrates the use of the positivity rate, instead of the daily number of confirmed cases, to obtain more accurate estimates for future pandemic trajectories. We used COVID-19 datasets for eight countries to obtain the daily positivity rate. Susceptible–Infected–Recovered (SIR) modeling was applied for the two compared cases, the case of using the positivity rate and the case of using the daily number of confirmed cases. For each case, we obtained estimated dates for the end of the pandemic. Pairs of results are statistically compared. The results for the predicted pandemic end dates obtained using the positivity rate were found to be statistically different from those obtained using daily confirmed cases. Based on these results, health authorities are advised to consider the pandemic prediction results based on the positivity rates because these rates consider both the daily number of confirmed cases and the number of performed tests.