基于互联网搜索数据研究全球高致病性禽流感病毒H5N1的暴发监测  

Surveillance of HPAI H5N1 global outbreaks based on internet searching data

作　　者：李锐[1] 孙利谦 熊成龙[1] 胡艺 林燧恒[1] 张志杰[1] 姜庆五[1] 

机构地区：[1]复旦大学公共卫生学院流行病学教研室教育部公共卫生安全重点实验室,上海200032

出　　处：《中华疾病控制杂志》2015年第8期773-777,共5页Chinese Journal of Disease Control & Prevention

基　　金：国家自然科学基金(81102167);上海市新优青计划(XYQ2013071);复旦大学自主科研项目(20520133105);全国优秀博士学位论文作者专项资金(201186)

摘　　要：目的 分析互联网搜索数据与高致病性禽流感病毒H5N1的关系,探讨利用网络搜索工具对其监测和预警的可能性。方法 基于联合国粮食及农业组织（Food and Agriculture Organization of the United Nations,FAO）和世界动物卫生组织（World Organization for Animal Health,OIE）收集整合了2004-2009年全球高致病性禽流感病毒H5N1在家禽中的暴发数据,从GoogleTrends获取同时期的相关关键词数据。在对二者的数据进行描述分析与对比的基础上,计算Spearman等级相关系数评价其相关程度,并通过平移技术以相关系数为指标分析互联网数据预测H5N1暴发的时间提前期。结果 以2004-2009年为整体,互联网数据与H5N1的相关性并不高（r=0.276,t=3.57,P〈0.001）,但年度数据则均表现出了较强的相关性,2004-2009年相关性系数分别为r2004=0.718（t=3.64,P〈0.001）,r2005=0.576（t=3.58,P〈0.001）,r2006=0.760（t=3.62,P〈0.001）,r2007=0.474（t=3.45,P〈0.001）,r2008=0.750（t=3.47,P〈0.001）,r2009=-0.442（t=3.32,P=0.001）。且在各个年份中,将OIE/FAO流感监测数据提前1-4周后出现与GoogleTrends监测数据的相关系数最大值,可认为GoogleTrends数据可提前1-4周预测H5N1暴发的趋势。结论 利用互联网搜索数据预测高致病性禽流感病毒H5N1的暴发可作为传统方法的补充手段,但由于搜索技术等的限制该方法仍需进行进一步的改进以提高准确性,其思路可推广至其他传染性疾病暴发的监测与预警。Objective To analyze the relationship between the prevalence of high pathogenic avian influenza（ HPAI） H5N1 virus and the Internet searching data,and to explore the possibility of monitoring H5N1 by using the internet searching tools. Methods Reported outbreaks of HPAI H5N1 in the poultry of 2004- 2009 were obtained from Food and Agriculture Organization（ FAO） and International Epizootic Office（ OIE）,and the Internet searching data with relative keywords were retrieved from Google Trends（ GT）. Combined with the descriptive analysis and comparison of the two kinds of data source,spearman correlation coefficient（ SCC） was employed to measure the association between the traditional surveillance data and GT based internet search data. The ahead of time that could be obtained by the internet search data was also evaluated using lagging analysis with the correlation coefficient as an index. Results The SCC for combined data from 2004 to 2009 was only 0. 276（ t = 3. 57,P〈0. 001）,whereas SSCS for data at each year were higher,with0. 718（ t = 3. 64,P〈0. 001） in 2004,0. 576（ t = 3. 58,P〈0. 001） in 2005,0. 760（ t = 3. 62,P〈0. 001） in 2006,0. 474（ t = 3. 45,P〈0. 001） in 2007,0. 750（ t = 3. 47,P〈0. 001） in 2008,and- 0. 442 in 2009（ t = 3. 32,P =0. 001）,which indicated that GT data had a strong yearly correlation with the traditional surveillance data. In the lagging correlation analysis,the best-lagged time for the highest correlations was about 1-4 weeks. Conclusions Google trends can be used to monitor the epidemic of H5N1 though more work is needed to improve the accuracy of the method. The idea of such analysis is helpful in monitoring and forecasting other infectious diseases.

关 键 词：禽流感 流感病毒A型 H5N1亚型 人群监测 计算机通信网络 

分 类 号：R181.8[医药卫生—流行病学]