流行性感冒发病态势感知与风险预警模型的构建与应用  

Construction and Application of Situation Awareness and Early Risk Warning Model for Influenza

作　　者：刘星航 郑雅旭[1] 崔欣 钱晨嗣 夏天[1] LIU Xinghang;ZHENG Yaxu;CUI Xin;QIAN Chensi;XIA Tian(Shanghai Municipal Center for Disease Control&Prevention,Shanghai 200336,China)

机构地区：[1]上海市疾病预防控制中心,上海市200336 [2]上海市卫生健康统计中心,上海市200040

出　　处：《中国卫生信息管理杂志》2022年第6期792-797,共6页Chinese Journal of Health Informatics and Management

基　　金：科技部科技创新2030-“新一代人工智能”重大项目《新冠肺炎疫情等公共卫生事件的智能流调研究》(项目编号:2021ZD0114005);上海市科委2020年“科技创新行动计划”技术标准项目《基于医疗健康大数据的新发重大传染病监测预警标准》(项目编号:20DZ2200400)。

摘　　要：目的研究基于多源数据的流感样病例数预测模型,为呼吸道传染病流行态势感知与风险预警探索方法。方法以面向多源数据的不同特征组合为基础,分别构建流感样病例数预测模型,评估不同特征对预测准确率的影响,并选择最优模型进行应用。结果历史流感样病例数、流感病原阳性率和节假日特征对模型预测准确率具有正向影响,诊疗和气象特征对模型预测准确率无明显正向影响。最优模型预测准确率79.5%,实际应用效果良好。结论基于多源数据、特征分组的流感发病态势感知与风险预警方法具有较高可行性,可为其他呼吸道传染病流行态势感知和风险预警模型的构建提供借鉴。Objective A prediction model for numbers of influenza-like illness(ILI)cases based on multisource data was studied to explore a method for epidemic situation awareness and early risk warning of respiratory infectious diseases.Methods Using different feature combinations of multi-source data,the prediction models of influenza-like cases were constructed respectively.The impact of each feature on prediction accuracy was estimated,and the optimal model was selected and brought into practice.Results Numbers of historical ILI cases,positive rate of influenza pathogens and holiday features have positive effects on the accuracy of prediction,while weather and clinical data showed no obvious positive impact.The accuracy of the optimal prediction model is 79.5%and has a good practical effect.Conclusion The influenza disease situation awareness and risk warning methods based on multi-source data and feature grouping is highly feasible.This could provide a reference for model construction of epidemic situation awareness and early risk warning on other respiratory infectious diseases.

关 键 词：多源数据 流感样病例 态势感知 风险预警 呼吸道传染病 

分 类 号：R-058[医药卫生] R319