基于百度指数的ARDL模型在贵州省手足口病发病预测中的应用探讨  

To explore the application of ARDL model based on Baidu index in the prediction of hand-foot-and-mouth disease incidence in Guizhou Province

作　　者：黎倩 洪峰[1] 王丹[2] 平凯珂 方森洲 余春[2] 黄艳 LI Qian;HONG Feng;WANG Dan;PING kai-ke;FANG Sen-zhou;YU Chun;HUANG Yan(School of Public Health and Health,Guizhou Medical University,Key Laboratory of Environmental pollution and Disease Surveillance,Ministry of Education,Guiyang,Guizhou 550025,China;不详)

机构地区：[1]贵州医科大学公共卫生与健康学院,环境污染与疾病监控教育部重点实验室,贵州贵阳550025 [2]贵州省疾病预防控制中心传染病防治所,贵州贵阳550004

出　　处：《现代预防医学》2023年第3期416-421,共6页Modern Preventive Medicine

基　　金：2020年贵州省健康委科学技术基金项目(gzwjwkj2020-1-202);2019年贵州省健康委科学技术基金项目(gzwjwkj2019-1-236)。

摘　　要：目的 以贵州省手足口实际发病数和百度指数的相关数据为基础,利用自回归分布滞后模型分析关键词百度指数与实际发病数的关系,并对模型的预测效果进行评价。方法 通过中国疾病预防控制信息系统收集贵州省2011年1月1日—2021年12月31日每周手足口病发病数据,同时从百度指数官网收集同期预先定义的手足口病相关关键词在贵州省范围内的百度指数,选取与手足口病实际发病数相关系数r>0.5的关键词纳入分析。以贵州省手足口病实际发病数和百度指数的相关数据为基础建立自回归分布滞后模型(ARDL),利用ARDL模型对未来发病数进行预测,评价该模型预测效果指标在新型冠状病毒爆发前后是否有差异以及模型远期预测效率的衰减程度。结果 最终筛选出“儿童手足口病”“手足口病吃什么药”“如何预防手足口病”“手足口病有什么症状”4个关键词。建立ARDL(2,5,1,0,0)模型,利用该模型对未来4周发病数进行预测,与实际发病数相比,发现预测效果较好(RMSE=234.42±118.95,F=1.762 1,P=0.209),模型预测能力不受年度变化影响,且预测效果不受新型冠状病毒爆发影响。结论 基于百度指数的ARDL模型能够较为准确的预测贵州省手足口病发病数。Objective To investigate relationship between keyword Baidu index and actual incidence using autoregressive distributed lag model(ARDL) based on the data of actual incidence of hand-foot-and-mouth disease(HFMD) and Baidu index in Guizhou Province, and to evaluate the prediction effect of the model. Methods The weekly incidence data of HFMD in Guizhou Province from January 1, 2011 to December 31, 2021 were collected through the China Disease Prevention and Control Information system. The Baidu index of predefined keywords related to HFMD in Guizhou Province was collected from the website of Baidu Index, and the keywords with the correlation coefficient r > 0.5 were selected for analysis. Based on the data of actual incidence of hand-foot-and-mouth disease in Guizhou Province and Baidu index, an autoregressive distributed lag model was established. ARDL model was used to predict the number of future cases and to evaluate whether there was any difference in the prediction effect index of the model before and after the outbreak of COVID-19 and the attenuation degree of the long-term prediction efficiency of the model. Results Four key words were screened out: “HFMD in children”, “what medicine to take for HFMD”, “how to prevent HFMD”, and “what are the symptoms of HFMD”. An ARDL model was established to predict the number of cases in the next 4 weeks. The prediction accuracy was good with reference to the actual number of incidences(RMSE=234.42±118.95, F=1.762 1, P=0.209), model predictive power was not affected by the annual variation, and the prediction was not affected by the outbreak of COVID-19. Conclusion The ARDL model based on Baidu index can accurately predict the incidence of HFMD in Guizhou Province.

关 键 词：手足口病 百度指数 自回归分布滞后模型 

分 类 号：R725.1[医药卫生—儿科] R181.3[医药卫生—临床医学]