基于环比增长率和空间约束聚类算法分析中国COVID-19疫情的风险现状

Current risk status and spatial distribution of COVID-19 in China based on the day-on-day growth rate and spatial constrained clustering algorithm

目的分析中国31个省、自治区、直辖市和新疆生产建设兵团(以下简称"31个省(区、市、兵团)")新型冠状病毒肺炎(coronavirus disease 2019,COVID-19)疫情的近期防控态势,基于各省份的疫情防控现状提出分区管理的设计建议。方法基于2020年1月25-2月8日中国31个省(区、市、兵团)COVID-19累计确诊病例数,计算累计和新增确诊病例数的环比增长率,并拟合随时间的线性函数得到环比增长率的线性趋势。进一步结合环比增长率的中位数(median,M)、标准差(standarddeviation,S)和累计确诊病例数进行AZP-SA(auto zone procedure with stimulate annealing,AZP-SA)的空间约束聚类,对各省份疫情现状进行合理分区。结果中国31个省(区、市、兵团)累计确诊病例数的环比增长率在近期均呈现下降趋势。新增确诊病例数的环比增长率则没有呈现出明显的趋势方向。选择AZP-SA方法将中国31个省(区、市、兵团)分成8个区,其中湖北及湖北周边省份属于疫情下降相对较慢的地区,西北和北方省份疫情控制良好,东部和东南省份的疫情控制效果处于两者之间。结论累计确诊病例数的环比增长率是衡量疫情短期趋势的良好指标,基于累计确诊病例数及其环比增长率应用空间约束聚类可以对中国各省份的疫情现状进行划分,为下一步的复工复产提供建议。

Objective Analyze the current situation of COVID-19 epidemic prevention and control in various provinces in China, and propose a regionalized management method based on current status of epidemic prevention and control. Methods We calculated day-of-day growth rate of cumulative and new confirmed cases based on the reported confirmed cases between Jan 25 to Feb 8, and fitted it with the linear regression model accompanied with date. Furthermore, we applied auto zone procedure with simulator annealing(AZP-SA) spatial constrained clustering in cumulative confirmed cases and the linear trend, median, standard deviation of the growth rate, and provided a regionalized management suggestion based on previous result. Results The growth rate of cumulative confirmed cases in all provinces has shown a downward trend in recent. The growth rate of new confirmed cases did not show a significant trend. AZP-SA method partitions the China's Mainland into 8 regions. Region surrounding Hubei with relatively slow decline trend, and regions of the northwest and north provinces have good control of epidemic situation, and the eastern and Southeast provinces have a control effect between them. Conclusions The growth rate of the cumulative confirmed cases is a good indicator to measure the short-term trend of the epidemic. Based on the cumulative confirmed cases and its growth rate, the application of spatial constraint clustering can partitions the epidemic status of China, and provide a regionalized management for work resumption.