基于ARIMA模型的海南省流感样病例发病趋势预测研究  

Prediction of incidence trend of influenza-like illness(ILI)in Hainan Province based on ARIMA model

作　　者：李乔君 王南来 李卫霞 殷大鹏 靳妍 邱丽[2] 鲁英 LI Qiaojun;WANG Nanlai;LI Weixia;YIN Dapeng;JIN Yan;QIU Li;LU Ying(School of Public Health,Hainan Medical University,Haikou 571199,China;Hainan Provincial Center for Disease Control and Prevention,Haikou 570203,China)

机构地区：[1]海南医科大学公共卫生学院,海口571199 [2]海南省疾病预防控制中心,海口570203

出　　处：《新疆医科大学学报》2024年第11期1533-1538,共6页Journal of Xinjiang Medical University

基　　金：海南省重点研发项目(ZDYF2021GXJS018);海南医科大学高层次引进人才项目(XRC202028)。

摘　　要：目的应用自回归积分移动平均模型(ARIMA)对海南省流感样病例就诊百分比(ILI%)情况进行预测,以期为流感预防控制措施的制定和卫生资源的合理分配提供科学依据。方法收集2015-2019年海南省6家国家级哨点监测医院的流感样病例每周的数据,使用R软件基于2015年第1周-2019年第26周的流感样病例就诊百分比(ILI%)建立ARIMA模型,并用2019年第27周-第52周的数据作为验证集,对模型的预测能力进行评估和验证。结果海南省6家国家监测哨点医院累计报告门急诊就诊病例总数为6244159例,其中流感样病例数为188143例,流感样病例就诊百分比(ILI%)为3.01%。最优预测模型为ARIMA(0,1,1)(0,1,1)52模型,该模型的赤池信息量准则(AIC)和贝叶斯信息准则(BIC)分别为264.10、273.69,提示该模型拟合较好。运用该模型预测2019年第27周-第52周的ILI%,实际值均落在预测值的95%置信区间范围内。结论ARIMA(0,1,1)(0,1,1)52模型对海南省流感样病例就诊百分比的拟合效果较好,可用于海南省流感的短期预测和动态分析,具有较好预测、预警价值。Objective The incidence trend of influenza-like illness(ILI)in Hainan Province was predicted by autoregressive integrated moving average model,to provide a reference basis for influenza prevention and control and rational allocation of health resources.Methods The weekly data of influenza-like illness from 6 national sentinel surveillance hospitals in Hainan Province from 2015 to 2019 were collected.The ARIMA model was established based on percentage of influenza-like illness(ILI%)from the first week in 2015 to the 26th week in 2019 using R software.The data from the 27th week to the 52nd week in 2019 were used as validation set to evaluate and verify the predictive ability of the model.Results The cumulative total number of reported outpatient and emergency department visits to the 6 national surveillance sentinel hospitals in Hainan Province was 6244159 cases,of which the number of influenza-like illness was 188143,and the proportion of influenza-like cases visiting the hospitals(ILI%)was 3.01%.The optimal prediction model was ARIMA(0,1,1)(0,1,1)52 model.The Kaike information criterion(AIC)and Ayes-ian information criterion(BIC)of the model were 264.10 and 273.69,indicating that the model fitted well.The model was used to predict the ILI%of 27-52 weeks in 2019,and the actual values fell within the range of(95%CI)of the predicted values.Conclusion ARIMA(0,1,1)(0,1,1)52 model has a good fitting effect on the percentage of influenza like visits in Hainan Province,which can be used for short-term prediction and dynamic analysis in Hainan Province,and it has good prediction and early warning value.

关 键 词：ARIMA模型 流感样病例 预测 

分 类 号：R181.3[医药卫生—流行病学]