Estimation of the epidemic properties of the 2019 novel coronavirus: A mathematical modeling study

Jinghua Li; Yijing Wang; Stuart Gilmour; Mengying Wang; Daisuke Yoneoka; Ying Wang; Xinyi You; Jing Gu; Chun Hao; Liping Peng; Zhicheng Du; Dong Roman Xu; Yuantao Hao

Previous studies 5,7-9 found lower values for the basic reproduction number. This variation may have arisen for two reasons. First, the empirical data that previous studies used were collected before 25 th Jan, 2020. Testing protocols and diagnostic tools changed during the early period of the study 23 , and the number of diagnosed cases collected before 15 th January were considered underestimated and less reliable. This would flatten the epidemic curve in early studies, and the estimation of R0 based on these data may be underestimated and have larger confidence intervals. Second, previous studies only estimated the R0 based on a single method, and these estimates may have been affected by the implicit assumptions in these models. For example, a previous paper using the assumption of exponential growth found a value of " of 2.68 (95% Credible interval 2.47 -2.86) 5 using an SEIR model with Metropolis-Hastings MCMC estimates of uncertainty, but our modeling has shown that this method likely underestimated the basic reproduction number during the pre-closure period. Our model avoids the limitations of specific modeling choices by combining several methods with a Poisson Loss weight, using the most current and accurate case diagnosis. Through this approach we calculate a more robust estimate than previous studies, and find a higher value of " . Ours is also the first study to compare the pre-and post-closure periods in the data, and thus the first study to make a judgment about the effectiveness of this strategy. Given the high risk of epidemic from COVID-19, it is important to assess the value of this strategy before the disease takes hold in another global city.