2004-2015年中国狂犬病发病数据ARIMA乘积季节模型的建立及预测

Human Rabies Incidence in China:Trends and Predictions from a Time Series Analysis from 2004 Through 2015

目的了解我国大陆地区2004-2015年狂犬病的发病情况,建立狂犬病发病的时间序列模型,利用模型进行短期预测,为狂犬病的预防和控制提供参考。方法通过查阅2004-2015年每月的《中华人民共和国卫生和计划生育委员会公报》,获得狂犬病发病的月统计数据,利用2004-2014年的数据建立ARIMA乘积季节模型,并利用建立的模型预测2015年数据,与实际发病数据比较。结果中国2004-2015年总计报告狂犬病25561例,年平均发病率为0.1592/10万,总计报告死亡病例22196例,年平均死亡率为0.1383/10万,2004年-2007年,狂犬病的发病人数和死亡人数逐年上升,2008年至2015年,持续下降。狂犬病具有一定的季节趋势,其中夏秋季节报告发病人数较多,而冬春季节发病人数较少。根据2004-2014年发病资料建立的最优时间序列模型为ARIMA(0,1,1)(0,1,1)12,模型预测2015年发病人数为764,相对误差7.73%。结论我国大陆地区狂犬病发病在2007年达到峰值之后,之后年发病率持续降低。ARIMA乘积季节模型能很好地拟合狂犬病发病的长期趋势和季节趋势,回代拟合和短期预测效果较理想。

Objective With an increasing population of dogs and low vaccination coverage,rabies is remerging and becoming a serious public health problem in China. Control and prevention of rabies requires know ledge of recent and future incidence trends. Methods M onthly information on the incidence of human rabies from January 2004 through December 2015 w as collected from the Gazette of the M inistry of Health of the People＇s Republic of China. Autoregressive integrated moving average（ ARIMA） multiple seasonal models were established to fit and predict rabies incidence using these data. The dataset was divided into tw o parts： data reported from January 2004 to December 2014,w hich w ere used to develop the time series model,and data from 2015,w hich w ere used to validate the established model. Results Human rabies remains a serious infectious disease in China. During the 12 years that this study investigated,25561 rabies cases w ere reported in China. The annual average number of cases w as 2 130 and the incidence w as 0. 1592 /105 people. The seasonal ARIM A（ 0,1,1）（ 0,1,1）12model w as the best to fit the data from the year 2004 to 2014. Using this model,w e predicted 764 rabies cases in all of China in 2015. The actual number of reported cases w as higher than the prediction for all of 2015（ 828 vs. 764）,for a relative predictive error of 7. 73%,w hich w as statistically acceptable. Conclusion the number of rabies cases reported reached its peak in the year of 2007,then it has been show ing a dow nw ard trend since then. The seasonal ARIM A（ 0,1,1）（ 0,1,1）12model w as significant,w ith an acceptable estimative and predictive accuracy. The time series model can be used to analyze and predict rabies cases in China,and w ill help frame strategies to control and prevent the disease.