时间序列模型在肿瘤患者医院感染发病率预测中的应用  

Application of time series model in prediction of nosocomial infection for cancer patients

作　　者：夏聪聪 王丽娟 蔡黎霞 张树敬[1] 王袁[1] 张燕[1] Xia Congcong;Wang Lijuan;Cai Lixia;Zhang Shujing;Wang Yuan;Zhang Yan(Department of Hospital Infection Management,Fudan University Shanghai Cancer Center,Shanghai 200032,China;Department of Head and Neck Surgery,Fudan University Shanghai Cancer Center,Shanghai 200032,China)

机构地区：[1]复旦大学附属肿瘤医院医院感染管理科,上海200032 [2]复旦大学附属肿瘤医院头颈外科,上海200032

出　　处：《中华传染病杂志》2021年第4期199-203,共5页Chinese Journal of Infectious Diseases

摘　　要：目的探讨时间序列模型在上海市肿瘤患者医院感染发病率预测中的应用,为医院感染的早期预警和防控提供参考。方法基于2013年至2018年上海市某三级甲等肿瘤专科医院住院患者的医院感染发病率数据,应用SPSS 22.0统计软件构建自回归求和移动平均(autoregressive integrated moving average,ARIMA)模型和指数平滑模型,通过比较两种模型的拟合精度,确定最优拟合模型。以2019年1月至6月的医院感染发病率数据作为验证数据集,评价模型的预测效果。结果2013年至2018年共监测379477例住院患者,其中3170例发生医院感染,总感染率为0.84%。霍尔特-温特相加法模型为拟合最优模型(R^(2)=0.82)。应用霍尔特-温特相加法模型预测2019年1月至6月的医院感染发病率,预测值和实际值动态趋势较一致,平均相对误差为15.22%。结论霍尔特-温特相加法模型能较好地拟合和预测肿瘤患者医院感染的发生趋势,可为肿瘤专科医院感染的监测预警提供依据。Objective To investigate the applicability of time series model in predicting incidence of nosocomial infection in a cancer center in Shanghai,and to provide the references for early warning and prevention.Methods The nosocomial infection data of inpatients of a tertiary oncology hospital in Shanghai from 2013 to 2018 were collected.The autoregressive integrated moving average(ARIMA)model and the exponential smoothing model were established by SPSS 22.0 expert modeler.The fitting predictions were compared between these two time series models to select the optimal one.The nosocomial infection data from January 2019 to June 2019 were used to test the predictive effect of the model.Results A total of 379477 cancer inpatients were studied,3170 of which acquired nosocomial infection and the incidence was 0.84%from 2013 to 2018.Additive Holt-Winters method exponential smoothing model was the better model with R^(2)of 0.82.Using this model,the predicted value fitted well with observed value from January 2019 to June 2019,and the mean relative percentage error was 15.22%.Conclusion Additive Holt-Winters method exponential smoothing model could be used to fit and predict the tendency of nosocomial infection among cancer patients,which can provide reference for surveillance of nosocomial infection in oncology hospitals.

关 键 词：交叉感染 预测 肿瘤专科医院 自回归求和移动平均模型 指数平滑模型 

分 类 号：R51[医药卫生—内科学]