基于地理加权广义线性模型探索山东省新型冠状病毒肺炎的影响因素

Influence factors of COVID-19 in Shandong Province based on geographically weighted generalized linear model

目的探讨山东省新型冠状病毒肺炎(COVID-19)的空间分布与相关影响因素,进一步了解山东省疫情的区域分布特征,为指导防控策略提供科学依据。方法收集2020年1月21日至3月1日山东省COVID-19确诊病例数及相关影响因素数据,采用地理加权广义线性模型(GWGLM)分析COVID-19确诊病例数及各影响因素间的空间异质性及其相关关系。结果对558例确诊病例的空间分布进行分析,广义线性模型(GLM)分析结果显示,人口密度、人均可支配收入、公共预算支出、湖北迁入规模占比和距武汉的空间距离均有统计学意义。人口越密集、人均可支配收入越高、公共预算支出越多,则确诊病例数越多;绝大多数县区的湖北迁入人口规模和距武汉的空间距离与确诊病例数呈负相关。GWGLM的R2为0.363,模型可解释COVID-19确诊病例数总变异的36.3%。结论GWGLM能够揭示COVID-19及其影响因素的空间异质性,有助于局域精准施策,应根据各因素的空间分布特征及其与确诊病例数的局域关系制定不同区域的分级防控措施。

Objective To explore the related influence factors of coronavirus diseases 2019(COVID-19)in Shandong Province and understand the regional distribution characteristics of the epidemic situation,and to provide a scientific basis for guiding prevention and control strategies.Methods The number of confirmed cases of COVID-19 and related influence factors in Shandong Province from January 21 to March 1,2020 were collected.The geographic weighted generalized linear model(GWGLM)was used to analyze the number of confirmed cases and the spatial heterogeneity among various influence factors.Results We analyzed spatial distribution of 558 confirmed cases.The results of GLM analysis showed that the population density,per capita disposable income,public budget expenditure,the proportion of Hubei immigrations and the spatial distance from Wuhan were statistically significant.The denser the population,the higher the per capita disposable income,and the higher the public budget expenditure,the greater the number of confirmed cases;the size of the Hubei immigrants and the spatial distance from Wuhan were inversely related to the number of confirmed cases in most counties and districts.In this study,the R^2 of GWGLM was 0.363,and the model could explain 36.3%of the total variation of COVID-19 confirmed cases.Conclusion GWGLM reveals the spatial heterogeneity of COVID-19 and its influence factors,and helps the local area to apply the policy precisely;the hierarchical prevention and control measures of different regions should be developed according to the spatial distribution characteristics of each factor and its local relationship with the number of confirmed cases.