摘要目的:分析医疗机构报告学生流感个案和学校因病缺勤数据的时空关联性，探讨贝叶斯结构时间序列模型（BSTS）在学校流感疫情预测中的应用。方法:选取深圳市大鹏新区13所学校，收集2015年1月1日至2019年12月31日在中国疾病预防控制信息系统的流感报告数据和深圳市学生健康监测系统的因病缺勤数据，分析比较2个系统数据间的时空关联性；采用BSTS，使用2个系统并联数据，对2019年学生流感月发病数进行长期预测，并分别对2019年第1~8周和第45~52周的周发病数进行短期预测。结果:中国疾病预防控制信息系统数据和深圳市学生健康监测系统具有时间关联性（ r=0.93， P<0.001），前者滞后1 d（ r=0.73， P<0.001）。不同学校间的流感疫情呈随机分布，无空间关联性。长期预测的均方根误差（ RMSE）为0.35，平均绝对误差（ MAE）为0.28；在2019年第1~8周和第45~52周的短期预测中， RMSE分别为0.33和0.34， MAE分别为0.26和0.28，均显示良好的预测精度与拟合效果。 结论:通过并联中国疾病预防控制信息系统和深圳市学生健康监测系统的数据，结合BSTS，可准确预测学校流感疫情动态，为流感预警与防控应对提供有效技术支持。

abstractsObjective:To analyze the spatiotemporal correlation between the surveillance data of influenza in students reported by medical institutions and school absenteeism due to illness, and evaluate the application of Bayesian Structural Time Series model (BSTS) in the prediction of school influenza epidemic.Methods:A total of 13 schools in Dapeng new district of Shenzhen were selected. The incidence data of influenza in schools in Shenzhen from January 1, 2015 to December 31, 2019 were collected from China Disease Control and Prevention Information System and the illness related school absentence data during this period were collected from Shenzhen Student Health Surveillance System, and the spatiotemporal correlation between the data from two systems was analyzed and compared. BSTS was used to make long-term predictions of the monthly incidence of influenza in students in 2019 and short-term predictions of the weekly incidence of influenza in week 1-8 and week 45-52 of 2019 by using the data from two systems.Results:There was a temporal correlation between the data from China Disease Control and Prevention Information System and the data from Shenzhen Student Health Surveillance System ( r=0.93, P<0.001), and the lag of the former one was 1 day ( r=0.73, P<0.001). Influenza outbreaks were randomly distributed in different schools in Shenzhen, and there was no spatial correlation. The root mean square error ( RMSE) and mean absolute error ( MAE) were 0.35 and 0.28, respectively, in the long-term prediction, and the RMSE was 0.33 and 0.34, and the MAE was 0.26 and 0.28, respectively, in the short-term predictions of week 1-8 and week 45-52 of 2019, respectively, showing good prediction accuracy and fitting effect. Conclusion:By analyzing the data from China Disease Control and Prevention Information System and Shenzhen Student Health Surveillance System with BSTS, the dynamics of the school influenza epidemic can be accurately predicted, and effective technical support can be provided for the early warning and prevention and control of influenza epidemic.