Bayesian Analysis on the Spatial Difference of Input Risk of Overseas Cases of COVID-19 in China

Abstract

To analyze the spatial difference of COVID-19 import risk is helpful for scientific prevention and control. On the basis of clustering 25 provinces and cities with epidemic input in study time, a multinomial distribution model was established under the Bayesian framework. All parameters Bayesian estimation was obtained by MCMC method. 25 provinces and cities with overseas input were divided into 9 categories from March 3 to April 23, 2020. 468 overseas input risk values are regarded as parameters, and the maximum MC-error estimated by Bayesian is only 0.677% of the standard deviation. During the study period, 25 provinces and cities have input risk. The highest risk areas of overseas import are 12 provinces and cities in the first category represented by Beijing, Shanghai and Guangdong Province, including 10 provinces and cities along the coast / border. The lowest risk areas are the eighth category (Henan Province) and the ninth category (Anhui Province); the fourth category (Heilongjiang Province and Shanxi Province) risk is higher than the first category in 7 days and it has the largest input vary fluctuation. Taking 2020-3-22, 4-7 and 4-18 as time nodes, the overseas input risk is divided into four stages. In the first stages, the highest risk of overseas import is the first category (59.613%); in the second and third stages are the first category (decline from 60.505% to 37.056%), the fourth category (increase from 16.071% to 33.852%); in the fourth stage, the first category (42.622%), the third category (Shaanxi Province and Jilin Province, 17.556%) and the fourth category (10.056%).