非药物性干预措施对荆州市手足口病发病影响的间断时间序列方法综合评价  

Comprehensive evaluation of the impact of non-pharmaceutical interventions on the incidence of hand foot and mouth disease in Jingzhou City based on interrupted time series analysis

作　　者：刘天[1] 向泉 秦周 宋开发[1] 姚梦雷[1] 黄继贵[1] 赵婧[2] LIU Tian;XIANG Quan;QIN Zhou;SONG Kaifa;YAO Menglei;HUANG Jigui;ZHAO Jing(Department for Infectious Disease Control and Prevention,Jingzhou Center for Disease Control and Prevention,Jingzhou,Hubei 434000,China;不详)

机构地区：[1]荆州市疾病预防控制中心传染病防治所,湖北荆州434000 [2]中国疾病预防控制中心卫生应急中心传染病溯源预警与智能决策全国重点实验室

出　　处：《中国预防医学杂志》2024年第9期1178-1184,共7页Chinese Preventive Medicine

基　　金：中国疾病预防控制中心应急综合业务管理项目(23170101)。

摘　　要：目的基于间断时间序列分析方法综合评价非药物性干预措施对荆州市手足口病发病的影响,为手足口病防控提供依据。方法采用广义回归模型(generalized linear model,GLM)、Prais-Winsten回归、自回归移动平均法(autoregressive moving average,ARIMA)和广义最小二乘法(generalized least squares,GLS)4种间断时间序列方法分别分析手足口病,计算超额风险(excess risk,ER)及95%置信区间(confidence interval,CI);对4种不同方法分析结果采用Meta分析进行合并,计算综合ER及95%CI。评价4个模型残差自相关性和平均绝对百分比误差(mean absolute percentage error,MAPE),选择残差不存在自相关性、MAPE最小的方法进行亚组分析,采用反事实估计法计算相对下降(relative decrease,RD)。结果2011—2019年共报告57286例手足口病例,年均发病率9.33/10万;2020—2022年共报告7693例,年均发病率4.00/10万,两个时期发病率差异有统计学意义[发病率比(incidence rate ratio,IRR)为2.33,95%CI:2.28~2.39]。最优ARIMA模型为ARIMA(1,0,1)含回归项。GLM、Prais-Winsten回归、ARIMA、GLS 4种模型计算ER依次为66.04%(95%CI:33.11%~82.76%)、63.15%(95%CI:2.85%~86.02%)、59.26%(95%CI:9.24%~81.73%)、58.41%(95%CI:3.57%~82.06%)。Meta分析合并的ER为63.00%(95%CI:46.00%~79.00%)。ARIMA模型残差不存在自相关性,MAPE最小为85.02%,为最优模型。利用ARIMA进行亚组分析显示,2020年1月—2022年12月、2020年1—6月、2020年7月—2021年12月、2022年1—12月的RD依次为39.22%、94.94%、20.41%、41.62%;0~<1岁、1~<3岁、3~<6岁和≥6岁的ER依次为70.53%、57.58%、-11.47%、-86.52%。结论非药物性干预措施显著降低了荆州市手足口病发病,非药物性干预措施对不同时期、年龄人群的降低作用存在异质性。Objective Based on interrupted time series analysis,this study aims to comprehensively evaluate the impact of non-pharmaceutical interventions on the incidence of hand,foot,and mouth disease(HFMD)in Jingzhou City and provide a basis for the prevention and control of HFMD in the city.Methods Four interrupted time series methods,namely generalized linear model(GLM),Prais-Winsten regression,autoregressive integrated moving average(ARIMA),and generalized least squares(GLS),were used to analyze and calculate excess risk(ER)and 95%confidence interval(CI).Meta-analysis was conducted to combine the results of the four different methods and calculate the overall ER and 95%CI.The autocorrelation of residuals and the mean absolute percentage error(MAPE)of the four models were evaluated,and the method with no autocorrelation in residuals and the lowest MAPE was selected for subgroup analysis.Relative decline(RD)was calculated using the counterfactual estimation method.Results From 2011 to 2019,a total of 57286HFMD cases were reported,with an average annual incidence of 9.33 per 100000 population.From 2020 to 2022,a total of 7693 cases were reported,with an average annual incidence of 4.00 per 100000 population.There was a statistically significant difference in the incidence rates between the two periods[incidence rate ratio(IRR):2.33,95%CI:2.28-2.39].The optimal ARIMA model was ARIMA(1,0,1)with a regression term.The calculated ERs for GLM,Prais-Winsten regression,ARIMA,and GLS models were 66.04%(95%CI:33.11%-82.76%),63.15%(95%CI:2.85%-86.02%),59.26%(95%CI:9.24%-81.73%),and 58.41%(95%CI:3.57%-82.06%),respectively.The meta-analysis combined ER was 63.00%(95%CI:46.00%-79.00%).The ARIMA model showed no autocorrelation in residuals,and the lowest MAPE was 85.02%,indicating that it was the optimal model.Subgroup analysis using ARIMA revealed that the RDs from January 2020 to December 2022,January to June 2020,July 2020 to December 2021,and January to December 2022 were 39.22%,94.94%,20.41%,and 41.62%,respectively.The ERs for age

关 键 词：手足口病 非药物性干预措施 间断时间序列 META分析 

分 类 号：R183.4[医药卫生—流行病学]