基于PCA-GTWR模型的中国肺结核发病率影响因素时空异质性研究  

Spatiotemporal heterogeneity of influencing factors of pulmonary tuberculosis incidence rate in China based on PCA-GTWR model

作　　者：戴萌萌 王雅怡 李佩吉 刘颖博 Dai Mengmeng;Wang Yayi;Li Peiji;Liu Yingbo(Department of Biostatistics,School of Science,China Pharmaceutical University,Nanjing 211198,Jiangsu Province,China)

机构地区：[1]中国药科大学理学院生物统计系,南京211198

出　　处：《中国基层医药》2024年第5期753-759,共7页Chinese Journal of Primary Medicine and Pharmacy

摘　　要：目的研究肺结核发病率及其宏观影响因素在中国的时空异质性,为相关部门制订防控政策提供参考。方法由于影响因素的复杂性,结合主成分分析(PCA)和时空地理加权回归(GTWR),提出一种新的方法分析肺结核影响因素的时空异质性。使用2010—2019年我国31个省、市、自治区(港澳台地区除外)肺结核发病率与地区生产总值(二级指标)等数据,建立宏观影响因素指标体系,21个二级指标PCA得分量化,确定综合经济、医疗保障、文教交运、资源环境4项一级指标。由此构建主成分-最小二乘(PCA-OLS)模型、主成分-地理加权回归(PCA-GWR)模型、主成分-时空地理加权(PCA-GTWR)模型。结果3个模型均通过F检验,F值分别为58.74、196.62、1202.90,均P<0.05,即一级指标对肺结核发病率的影响具有统计学意义。PCA-GTWR模型的均方误差(0.01)、平均绝对误差(0.08)、平均绝对百分比误差(0.02)、更正的赤池信息(-358.76)均低于PCA-OLS的0.13、0.28、0.07、258.38和PCA-GWR的0.06、0.15、0.03、23.41,确定系数(0.95)高于PCA-OLS的0.44和PCA-GWR的0.77,表明其拟合优度表现最好。同时PCA-GTWR模型回归系数分布图表明综合经济、医疗保障、文教交运、资源环境对肺结核的发病率有显著的时空异质性。结论应综合考虑多种因素因地制宜制定详细全面的防控措施,做好肺结核防治工作。Objective To study the spatiotemporal heterogeneity of the incidence rate of pulmonary tuberculosis and its macro influencing factors in China,so as to provide a reference for the relevant departments to formulate prevention and control policies.Methods In virtue of the complexity of influencing factors,a new method combining principal component analysis(PCA)with geographically and temporally weighted regression(GTWR)was proposed to analyze the spatiotemporal heterogeneity of influencing factors for pulmonary tuberculosis.Using the data of incidence rate of pulmonary tuberculosis and gross regional product(secondary indicators)of 31 provinces(excluding Hong Kong,Macao,and Taiwan)in China from 2010 to 2019,a macro influencing factor indicator system was established by the PCA scores of 21 secondary indicators quantified to determine four primary indicators:comprehensive economy,medical security,cultural and educational transportation,and resources and environment.Based on the indicator system,PCA-ordinary least squares(PCA-OLS)model,PCA-geographically weighted regression(PCA-GWR)model,and PCA-GTWR model were constructed.Results Three models passed F-test with F-values of 58.74,196.62,and 1202.90 respectively(all P<0.05),indicating that the impact of the primary indicators on the incidence of tuberculosis is statistically significant.The mean squared error(0.01),the mean absolute error(0.08),the mean absolute percentage error(0.02),and the corrected Akaike information criterion(-358.76)of PCA-GTWR were lower than those of PCA-OLS(0.13,0.28,0.07,258.38)and PCA-GWR(0.06,0.15,0.03,23.41).Meanwhile,the determination coefficient(0.95)of PCA-GTWR was higher than that of PCA-OLS(0.44)and PCA-GWR(0.77),indicating the goodness of fit of the model is the best.And the PCA-GTWR model showed that the comprehensive economy,medical security,cultural and educational transportation,and resources and environment had significant spatiotemporal heterogeneity on the incidence rate of pulmonary tuberculosis according to the distribution

关 键 词：结核 肺 发病率 影响因素分析 主成分分析 时空分析 

分 类 号：R181.3[医药卫生—流行病学] R521[医药卫生—公共卫生与预防医学]