摘要
目的了解我国1992—2022年乙型肝炎(乙肝)流行特征,探讨贝叶斯结构时间序列(BSTS)模型在预测疫情变化趋势中的应用价值。方法收集1992年1月—2022年9月我国乙肝发病监测数据,采用BSTS模型预测疫情变化趋势,并与自回归滑动平均混合(ARIMA)模型预测性能进行比较。结果1992年1月—2022年9月共报告乙肝27633674例(年均发病率为67.99/10万),年均上升百分比(AAPC)为4.64%(95%CI:3.25%~6.04%,P<0.01),每年1月和2月的季节指数(SI)最小(0.85、0.81),7月和8月的最大(1.10、1.11)。在向前93、69、33、9步预测中,模型产生的平均绝对误差(MAD)、平均绝对百分比误差(MAPE)、均方根误差(RMSE)、均方根百分比误差(RMSPE)和平均误差率(MER),BSTS模型均小于ARIMA模型,其中向前93、69、33步预测结果DM检验差异均有统计学意义(P值均<0.05)。使用BSTS预测的2023—2030年我国乙肝新发病为10023679(95%CI:5937860~13874521)例,年均发病1252960(95%CI:742232~1734315)例。结论我国乙肝疫情总体呈上升趋势,BSTS模型预测性能明显优于ARIMA,在乙肝疫情趋势预测方面有较高应用价值。
Objective To understand the epidemic characteristics in hepatitis B incidence in China from 1992 to 2022;to explore the application value of Bayesian structure time series(BSTS)model in predicting the changing trend ofthehepatitis B epidemic.Methods Monitoring data on the hepatitis B incidence in China from Jan 1992 to Sep 2022 were collected,and then the epidemic trend was predicted by use of the BSTS model and its prediction performance was compared with the autoregressive integrated moving average(ARIMA)model.Results A total of 27633674 hepatitis B cases were reported(mean annualized morbidity was 67.99/105)from Jan 1992 to Sep 2022,with an average annual percent change of 4.64%(95%CI:3.25%⁃6.04%,P<0.01).For the series,the minimum seasonal index(SI)values were observed in Jan(0.85)and Feb(0.81),whereas the maximum SI values in Jul(1.10)and Aug(1.11)every year.In the 93⁃,69⁃,33⁃,and 9⁃step ahead predictions,the predictive performance metrics[the mean absolute deviation(MAD),mean absolute percentage error(MAPE),root mean square error(RMSE),root mean square percentage error(RM⁃SPE),and mean error rate(MER)]from the BSTS model were less than those from the ARIMA model,and the DM statistics indicated a significantly statistical difference for the 93⁃,69⁃,and 33⁃step ahead predictions(all P<0.05).The forecasts under the BSTS model were 10023679(95%CI:5937860⁃13874521)hepatitis B cases in China,2023—2030,with the average annual 1252960(95%CI:742232⁃1734315)cases.Conclusions Collectively,the hepatitis B epidemic in China is on the rise.The prediction performance of the BSTS model significantly outperforms the ARIMA model,and thus it has a high application value in forecasting the morbidity trend ofhepatitis B.
作者
田珍榛
刘星言
赵蕊婷
王永斌
李言言
李继贞
TIAN Zhen-zhen;LIU Xing-yan;ZHAO Rui-ting;WANG Yong-bing;LI Yan-yan;LI Ji-zhen(The Ninth People's Hospital of Zheng zhou,Henan Zhengzhou 450052,China)
出处
《江苏预防医学》
CAS
2023年第3期285-288,359,共5页
Jiangsu Journal of Preventive Medicine
关键词
乙型肝炎
发病率
BSTS模型
ARIMA模型
发病趋势
预测
Hepatitis B
Morbidity
Bayesian structural time series
Autoregressive integrated moving average
Trend
Forecasting