Objective: To explore the time series characteristics of 5 types of viral hepatitis in China and predict their incidence through effective models. Methods: The monthly incidence data of 5 types of viral hepatitis (A, B, C, D and unspecified) in China from 2009 to 2018 were collected for descriptive and time series analyses, decomposition methods were used to explore the seasonality in the form of seasonal indices and the long-term trend in the form of a linear regression model. Autoregressive integrated moving average (ARIMA) models were established for each type of viral hepatitis. Results: From 2009 to 2018, a total of 14 856 990 cases of viral hepatitis were reported, the seasonal index range of 5 types of viral hepatitis were all lower than 1, the seasonality of hepatitis E was significant, and its incidence was unimodal, but no obvious seasonality characteristics were observed for other four types of viral hepatitis. The incidences of hepatitis A, hepatitis E and unspecified hepatitis remained at lower levels, showing slow declines. Although the cases of hepatitis B accounted for the highest proportion (79.59%, 11 824 262/14 856 990) among 5 types of viral hepatitis, the decline was fastest (-0.01/100 000). The incidence of hepatitis C was on rise, and the rate of increase remained stable (0.005/100 000). The predicted incidences of 5 types of viral hepatitis in China from January 2009 to December 2018 fitted by ARIMA model were consistent with the actual incidences, and the mean absolute error percentage (MAPE) ranged from 3.756 8 to 8.068 3. Conclusions: Time series analysis on surveillance data is useful for better understanding the incidence of the viral hepatitis. The ARIMA model has good application value in the short-term prediction of viral hepatitis incidence in China.
目的： 探讨我国5种病毒性肝炎（肝炎）的时间序列特征，并通过有效的模型预测其发病率。 方法： 按照甲型肝炎、乙型肝炎、丙型肝炎、戊型肝炎和未分型肝炎5种不同类型肝炎分类方式收集2009-2018年的月度发病数据，进行描述性和时间序列研究，采用趋势分解法以季节指数形式表示时间序列中的季节性，以线性回归模型表示其长期趋势，为每种肝炎建立差分自回归移动平均（ARIMA）模型。 结果： 2009-2018年报告肝炎14 856 990例，5种肝炎季节指数的极差均<1，戊型肝炎的季节性特征较为显著，其发病呈单峰型，其余4种肝炎的季节性特征一般。甲型肝炎、戊型肝炎和未分型肝炎的发病基本趋于平稳，在一个较低的水平上呈缓慢下降趋势，乙型肝炎发病数在5种肝炎中占比最高（79.59%，11 824 262/14 856 990），但其下降趋势也为各型肝炎中最快（-0.01/10万）。丙型肝炎发病呈不断上升的趋势，上升速率一直保持稳定（0.005/10万）。ARIMA模型拟合的2009年1月至2018年12月的预测值与实际值较一致，平均绝对误差百分比范围为3.756 8~8.068 3。 结论： 对于法定报告传染病监测数据的时间序列分析有助于更好地了解我国肝炎的发病特征，ARIMA模型可用于我国肝炎的短期预测，具有较好的应用价值。.
Keywords: Decomposition methods; Long-term trend; Prediction; Seasonality; Viral hepatitis.