【摘要】 目的 基于2005―2019年湖北省结核病月发病率数据，分析其结核病流行特征，采用X12季节调整结合自回归移动平均模型(X12-adjusted autoregressive integrated moving average model,X12-ARIMA)、X12季节调整结合季节性自回归移动平均模型(X12-adjusted seasonal autoregressive integrated moving average model,X12-SARIMA)和季节性自回归移动平均模型(seasonal autoregressive integrated moving average model,SARIMA)，比较3种模型的预测性能，并为湖北省结核病防控资源的科学调配提供方法学依据。方法 基于2005―2019年湖北省结核病月发病率数据，利用EViews 13软件进行X12季节分解并分别构建SARIMA、X12-SARIMA与X12-ARIMA模型，以绝对误差、相对误差、平均绝对误差(mean absolute error,MAE)和平均绝对百分比误差(mean absolute percentage error,MAPE)评估2019年2―12月预测效果。结果 湖北省结核病年均发病率为7.07/10万(95%CI:6.92/10万～7.22/10万)，呈持续下降趋势(年均降幅3.83%),3月为发病高峰(较年均高出20.03%)。SARIMA预测性能最优(MAE=0.16,MAPE=3.06%)，优于X12-SARIMA(MAE=0.18,MAPE=3.68%)和X12-ARIMA(MAE=0.27,MAPE=5.49%)。结论SARIMA预测精度最高; X12-SARIMA可解析季节成分，为防控资源季节性调配提供依据。更多还原

【Abstract】 Objective Based on the monthly incidence data of tuberculosis in Hubei Province from 2005 to 2019,analyze the epidemiological characteristics of tuberculosis,using X12-adjusted autoregressive integrated moving average model( X12-ARIMA),X12-adjusted seasonal autoregressive integrated moving average model( X12-SARIMA),and seasonal autoregressive integrated moving average model( SARIMA) to compare the predictive performance of the three models, and provide a methodological basis for the scientific allocation of tuberculosis prevention and control resources in Hubei Province. Methods Based on the monthly incidence data of tuberculosis in Hubei Province from 2005 to2019,the time series was decomposed using the X12 method in EViews 13, and SARIMA,X12-SARIMA,and X12-ARIMA models were constructed. The predictive performance for February to December 2019 was evaluated using absolute error,relative error,mean absolute error( MAE) and mean absolute percentage error( MAPE). Results The average annual incidence rate of TB in Hubei Province was 7. 07 per 100 000( 95% CI: 6. 92/100 000-7. 22/100 000),showing a declining trend with an average annual decrease of 3. 83%. The peak incidence occurred in March,which was 20. 03% higher than the annual average. The SARIMA model demonstrated the best predictive performance for February to December 2019( MAE = 0. 16,MAPE = 3. 06%),outperforming the X12-SARIMA model( MAE = 0. 18,MAPE = 3. 68%) and the X12-ARIMA model( MAE = 0. 27,MAPE = 5. 49%). Conclusions The SARIMA achieves the highest prediction accuracy. The X12-SARIMA model,while enabling the analysis of seasonal components,provides a basis for the seasonal allocation of prevention and control resources.更多还原