Dynamic Estimation of Epidemiological Parameters of COVID-19 Outbreak and Effects of Interventions on Its Spread

Hongzhe Zhang; Xiaohang Zhao; Kexin Yin; Yiren Yan; Wei Qian; Bintong Chen; Xiao Fang

Our study aims to characterize the evolution of the COVID-19 outbreak in Wuhan and reveal the effects of major government interventions on its spread. The underlying challenge in studying the pandemic dynamics lies in the discrepancy between the officially reported number of infected cases in Wuhan and the actual number of infections, in together with the lack of reliable data sources after the city's complete lockdown (e.g., most existing work focuses on static estimation before the lockdown on January 23, 2020 and often relies on exported case numbers 4,5,10 ) . To address the data discrepancy issue, we employ a straightforward yet effective data transformation approach under a Bayesian dynamic epidemic modeling framework, which leads to important implications in understanding the evolution of Wuhan's outbreak. First, using prior literature knowledge on COVID-19, we adjust for the reported data to estimate and gauge the actual outbreak sizes, which is shown to be substantially larger than those from official reports particularly in early periods. Second, taking into account the adjusted numbers, the resulting trajectory for effective reproduction numbers serves as more accurate reflection of disease spread trends and the temporal changes in response to official intervention policies. Third, our study results are crucially equipped with under-reporting factors that, to some extent, reflect the difficulty level in recording the actual infective numbers and the stress of COVID-19 on medical resources. In particular, by comparison with two other major cities in China, our results from the under-reporting factors are in alignment with the reality that Wuhan as the epicenter experienced the longest periods of high stress on health care system while the numbers outside Wuhan tend to be generally trustworthy at smaller outbreak scale with better medical preparedness.