基于哨点医院电子病历计算机自动识别技术的流感样病例试点监测评价  

Pilot surveillance and evaluation of influenza-like illness based on automatic computer analysis of electronic medical record in sentinel hospital

作　　者：朱爱琴 刘建华[2] 徐承中[2] 张皓[2] 杨孝坤 赵宏婷 李志丽 王丽萍[1] 冯录召[1] 郑亚明[1] 秦颖[1] 李中杰[1] Zhu Aiqin;Liu Jianhua;Xu Chengzhong;Zhang Hao;Yang Xiaokun;Zhao Hongting;Li Zhili;Wang Liping;Feng Luzhao;Zheng Yaming;Qin Ying;Li Zhongjie(Division of Infectious Disease,Key Laboratory of Infectious Disease Surveillance and Early Warning,Chinese Center for Disease Control and Prevention,Beijing 102206,China;Yichang Center for Disease Control and Prevention,Yichang 443003,Hubei Province,China)

机构地区：[1]中国疾病预防控制中心传染病监测预警重点实验室传染病管理处,北京102206 [2]湖北省宜昌市疾病预防控制中心,443003

出　　处：《中华预防医学杂志》2020年第6期691-695,共5页Chinese Journal of Preventive Medicine

基　　金：国家科技重大专项(2018ZX10713001-005);国家自然科学基金(91846302);中美新发和再发传染病合作项目;中国疾病预防控制中心应急反应机制运行项目(131031001000015001)。

摘　　要：目的比较基于病历资料的医生报告与计算机自动识别方式对门急诊流感样病例(ILI)监测的准确性。方法2019年4—10月在宜昌市中心医院发热门诊、呼吸内科门诊及急诊内科门诊就诊的患者中,选取国际疾病分类第10版诊断编码属于J00-J22范围且病历症状信息完整的病例,共3542例,以流感监测专业人员根据ILI病例定义的判断结果为金标准,绘制受试者工作特征(ROC)曲线,计算灵敏度、特异度、诊断一致率,分析计算机自动识别与医生报告两种方式对ILI病例监测的准确性。结果病例年龄M(P25,P75)为30(24,38)岁;其中符合定义标准ILI病例为1179例(33.29%),医生报告ILI为1306例(36.87%),计算机自动识别ILI为1150例(32.47%);男性为1391例(39.27%)。计算机自动识别和医生报告与金标准判断的结果一致性Kappa值分别为0.97和0.66,ROC曲线下面积分别为0.98和0.84。计算机自动识别方式的灵敏度和特异度(分别为96.95%和99.70%)均高于医生报告方式(灵敏度和特异度分别为82.27%和85.78%)(P值均<0.001)。结论基于电子病历的计算机自动识别方式开展ILI病例监测,具有良好的灵敏度和特异度。Objective To compare the accuracy of influenza-like illness(ILI)surveillance by automatic computer analysis based on electronic medical records and by doctor’s report.Methods A total of 3542 patients who presented to Yichang Central Hospital fever clinic,respiratory department or emergency department between April to October 2019 with an ICD-10 code for acute respiratory illness(J00-J22)and complete electronic medical information of ILI related syndromes were drawn as the study sample.Taking the classification of the study sample according to the ILI case definition by influenza surveillance professionals as the gold standard,draw the receiver operating characteristic(ROC)curve and calculate sensitivity,specificity,diagnostic consistency to compared the accuracy of ILI surveillance by automatic computer analysis and by doctor's report.Results Median age of 3542 cases was 30(24,38)years old;1179 cases(33.29%)compliance with the case definition,ILI reported by doctors was 1306 cases(36.87%),and computer automatic identification ILI were 1150 cases(32.47%);1391(39.27%)cases were men.The results of automatic computer analysis and doctor report consistency of kappa values with gold standard judgment were 0.97 and 0.66 respectively;area under the ROC curve was 0.98 and 0.84,respectively.And the sensitivity and specificity of automatic computer analysis were higher than that of doctor's report(all P values were<0.001),the sensitivity was 96.95%and 82.27%,and the specificity was 99.70%and 85.78%,respectively.Conclusion The automatic computer analysis based on electronic medical records can identified ILI cases with good sensitivity and specificity in ILI case surveillance.

关 键 词：流感样病例 计算机自动识别 电子病历 监测方法评价 

分 类 号：R511.7[医药卫生—内科学]