中国2008-2016年手足口病月发病率时间序列分析及预测模型

目的  通过时间序列分析我国手足口病（hand-foot-mouth disease，HFMD）的发病趋势并构建时间序列预测模型，为制定防控策略提供科学依据。 方法  从公共卫生科学数据中心收集2008-2016年我国HFMD月发病数据，使用Excel 2007建立发病率数据库并进行图表绘制，通过SAS 9.1拟合自回归综合移动模型（autoregressive integrated moving average model，ARIMA model）。以2008年1月-2015年12月HFMD月发病率作为测试集构建时间序列模型，2016年发病率数据作为验证集检验预测效果，做出模型评价，利用该模型对2017年HFMD发病率做出预测。以P<0.05为差异有统计学意义。 结果  最终构建ARIMA（（12），2，0）疏系数模型，残差为白噪声序列，实际值均落在预测值95% CI内，模型回归系数有统计学意义，预测值与实际值总体吻合程度良好，均方误差平方根为3.6490，平均绝对误差=2.62，平均绝对百分比误差=28.24%。 结论  疏系数模型可较好的拟合我国HFMD发病率的时间序列趋势，对HFMD防控策略的制定有指导意义。

Application of the time series model in prediction of incidence of hand-foot-mouth disease from 2008 to 2016 in China

  Objective  To predict the monthly incidence of hand-foot-mouth disease (HFMD) in China by using autoregressive integrated moving average (ARIMA) model and provide evidence for prevention and control of HFMD.  Methods  The monthly incidence data of HFMD in China from 2008 to 2016 were collected from the Public Health Science data Center. The incidence database was established by Excel 2007 and graphed. SAS 9.1 was used to construct the ARIMA model, based on the data of the monthly reported incidence of HFMD in China from January 2008 to December 2015, and then the data in 2016 were used to verify the predicted results. The monthly incidence in 2017 was predicted in the same way.The difference was statistically significant when P<0.05.  Result  The model predicting monthly incidence of HFMD in China is ARIMA ((12), 2, 0) sparse coefficient and residuals is white noise. The parameters were as follows: moted mean squared error=3.6490, mean absolute error=2.62, mean absolute percentage error=28.24%.  Conclusion  The sparse coefficient model could well simulate the trend of HFMD case in time series, which has good reference of early warning and prevention of HFMD.