基于ARIMA乘积季节模型的中国流行性腮腺炎发病趋势预测分析  

Analysis and prediction of mumps incidence in China using a multiplicative seasonal ARIMA modeI

作　　者：李平[1] 黄澳迪 包黎明 程立雪 王富珍[1] 杨宏[1] 马超[1] 尹遵栋[1] Li Ping;Huang Aodi;Bao Liming;Cheng Lixue;Wang Fuzhen;Yang Hong;Ma Chao;Yin Zundong(National Immunization Program,Chinese Center for Disease Control and Prevention,Beijing 100050,China)

机构地区：[1]中国疾病预防控制中心免疫规划中心,北京100050

出　　处：《中国疫苗和免疫》2023年第2期174-179,共6页Chinese Journal of Vaccines and Immunization

基　　金：中国CDC公共卫生应急反应机制的运行(131031001000150001)。

摘　　要：目的构建自回归求和移动平均(Auto-regressive integrated moving average,ARIMA)乘积季节模型,预测分析新型冠状病毒感染(Coronavirus disease 2019,COVID-19)疫情前后中国流行性腮腺炎(流腮)发病趋势。方法收集2008-2021年中国流腮月报告发病数据,基于2008-2018年数据拟合流腮发病ARIMA乘积季节模型;利用拟合模型预测2019-2021年流腮月发病数,评价预测效果。结果2008-2018年中国流腮发病呈3-5年一次流行高峰,夏季和冬季高发。流腮发病的最优拟合模型为ARIMA(2,1,2)(0,1,1)12,模型的相关参数估计值均具有显著性,其残差序列白噪声检验显示均为白噪声序列。2019年、2020年、2021年流腮月发病数的模型预测值与真实值的相对误差范围分别为1.56%-19.30%、41.24%-360.66%、64.46%-267.61%,平均相对误差分别为6.65%、159.08%、177.39%。结论拟合模型可准确预测COVID-19疫情前中国流腮发病,但对疫情期间的发病预测结果偏差较大;需要补充COVID-19疫情后流腮发病数据以拟合更优的预测模型。Objective To construct a multiplicative seasonal auto-regressive integrated moving average(ARIMA)model and predict mumps incidence before and after the coronavirus disease 2019(COVID-19)pandemic in China.Methods We obtained monthly incidences of reported mumps in China during 2008-2021.We used the 2008-2018 data to fit a seasonal product ARIMA model of mumps incidence and used the optimal fitted model to predict monthly incidence of mumps in 2019-2021 and evaluate model prediction accuracy.Results Mumps incidence peaked every 3-5 years and in summers and winters of each year during 2008-2018.The optimal fitted model for mumps incidence was ARIMA(2,1,2)(0,1,1)12;estimated values of model parameters were all significant and their residual series were all white noise by the Ljung-Box test.Relative error between model-predicted values and actual values of monthly mumps incidence in 2019,2020,and 2021 ranged 1.56%to 19.30%,41.24%to 360.66%,and 64.46%to 267.61%,with average relative errors of 6.65%,159.08%,and 177.39%,respectively.Conclusions The fitted model could accurately predict mumps incidence in China before the COVID-19 pandemic,but the predictions during the pandemic had large error rates.We should incorporate mumps incidence data after the pandemic to further optimize the ARIMA model.

关 键 词：流行性腮腺炎 发病 自回归求和移动平均(ARIMA)乘积季节模型 新型冠状病毒感染 预测 

分 类 号：R186[医药卫生—流行病学] R512.1[医药卫生—公共卫生与预防医学]