中国媒介生物学及控制杂志 ›› 2023, Vol. 34 ›› Issue (5): 654-663.DOI: 10.11853/j.issn.1003.8280.2023.05.013

• 预测预警 • 上一篇    下一篇

基于多源地理数据的广州市精细尺度登革热传播风险预测

张梦真1,2, 任周鹏2, 范俊甫1, 肖建鹏3, 张应涛4   

  1. 1. 山东理工大学建筑工程与空间信息学院, 山东 淄博 255000;
    2. 中国科学院地理科学与资源研究所资源与环境信息系统国家重点实验室, 北京 100101;
    3. 广东省疾病预防控制中心广东省公共卫生研究院, 广东 广州 511430;
    4. 广东省疾病预防控制中心, 广东 广州 511430
  • 收稿日期:2023-04-06 出版日期:2023-10-20 发布日期:2023-10-27
  • 通讯作者: 任周鹏,E-mail:renzp@lreis.ac.cn
  • 作者简介:张梦真,女,在读硕士,主要从事GIS与公共健康研究,E-mail:zhangmz@lreis.ac.cn
  • 基金资助:
    国家自然科学基金(42071377,42171413)

Fine-scale dengue transmission risk prediction based on multi-source geographic data in Guangzhou,China

ZHANG Meng-zhen1,2, REN Zhou-peng2, FAN Jun-fu1, XIAO Jian-peng3, ZHANG Ying-tao4   

  1. 1. School of Civil Engineering and Geomatics, Shandong University of Technology, Zibo, Shandong 255000, China;
    2. State Key Laboratory of Resources and Environmental Information Systems, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Science, Beijing 100101, China;
    3. Guangdong Provincial Institute of Public Health, Guangdong Provincial Center for Disease Control and Prevention, Guangzhou, Guangdong 511430, China;
    4. Guangdong Provincial Center for Disease Control and Prevention, Guangzhou, Guangdong 511430, China
  • Received:2023-04-06 Online:2023-10-20 Published:2023-10-27
  • Supported by:
    National Natural Science Foundation of China (No. 42071377,42171413)

摘要: 目的开展精细尺度登革热传播风险预测研究,为满足精细化预防和控制的现实需求和疾控部门制定更加精准的登革热疫情应对方案提供参考。方法收集广州市2017-2019年的登革热病例数据,结合降水、地表温度、人口密度、道路密度、归一化植被指数、医院可达性、公交站点密度和土地利用香农均匀度指数等自然和社会经济数据,采用随机森林模型实现1 km×1 km精细尺度登革热传播风险预测。结果基于过采样方法的预测模型精度优于欠采样和组合采样,检验受试者工作特征曲线的曲线下面积(AUC)值为0.999,准确率为0.978,精确率为0.999,查全率为0.959,F1分数值为0.979。分析单一因素对登革热预测的重要性程度发现,人口密度的重要性程度远高于其他变量,其均方误差增加量平均值为63.76。医院可达性为第2重要特征变量,平均地表温度在所选变量中重要性程度最低,其均方误差增加量平均值为35.42。广州市登革热传播风险分布与人口区分布一致,高风险区面积占总面积的6.18%,位于高风险区内的风险人口占总人口的39.13%。越秀、荔湾、海珠和天河区4个区均有80.00%以上的人口处于高风险区。结论广州市登革热传播风险区主要分布于广州市中心城区,以越秀、荔湾和海珠区为中心,向北延伸至白云区中部,向南延伸至番禺和南沙区交界处,向东延伸至黄埔区东部。预测结果中的风险区域与病例分布高度吻合,表明该研究提出的方法能够较为准确描述登革热传播风险地理分布。

关键词: 登革热, 传播风险预测, 随机森林, 重采样, 广州市

Abstract: Objective This study predicted the risk of dengue fever transmission on a fine scale, aiming to meet the practical needs of meticulous prevention and control and to provide a reference for relevant departments to formulate more precise response plans against dengue fever.Methods A Random Forest model was constructed to predict the risk of dengue fever transmission at a fine resolution of 1 km×1 km based on the data on dengue fever cases as well as natural and socio-economic factors including precipitation, land surface temperature, population density, road density, the normalized difference vegetation index, hospital accessibility, bus stop density,and the Shannon evenness index of land use in Guangzhou, China from 2017 to 2019.Results Compared with the models based on undersampling or combined sampling, the oversampling-based prediction model had better precision, with the area under the curve (AUC) being 0.999, accuracy being 0.978, precision being 0.999, recall being 0.959, and F1 value being 0.979. The analysis of the importance of single factors in dengue fever prediction revealed that the importance of population density was much higher than those of the other variables, with an average increase in mean squared error of 63.76; hospital accessibility was the second important feature;the average land surface temperature had the lowest importance among the selected variables, with an average increase in mean squared error of 35.42. The distribution of dengue fever transmission risk was consistent with the distribution of populated areas in Guangzhou. The high-risk areas accounted for 6.18% of the total area, and the at-risk populations in the high-risk areas accounted for 39.13% of the total population. More than 80.00% of the population in Yuexiu, Liwan, Haizhu, and Tianhe districts were in the high-risk areas.Conclusions The risk of dengue transmission in Guangzhou was mainly distributed in the central urban areas of Guangzhou, with Yuexiu, Liwan, and Haizhu districts as the center, extending northward to central Baiyun District, southward to the junction of Panyu and Nansha districts, and eastward to eastern Huangpu District. The predicted risk areas were highly consistent with case distributions, indicating that the method proposed in this study can more accurately depict the geographical distribution of dengue transmission risk.

Key words: Dengue fever, Transmission risk prediction, Random forest, Resampling, Guangzhou

中图分类号: