基于ARIMA时间序列模型预测某三级甲等医院耐碳青霉烯类铜绿假单胞菌的感染率

Prediction of infection incidence of carbapenem-resistant pseudomonas aeruginosa by ARIMA time series model

目的探讨自回归移动平均模型(ARIMA)时间序列模型在耐碳青霉烯类铜绿假单胞菌(CRPA)感染率预测中的应用,为CRPA的预防和控制提供政策依据。方法以某三级甲等医院2016年1月至2019年12月的CRPA感染数据建立ARIMA时间序列模型,再利用模型对2020年1月至2020年9月的数据进行验证,评价模型的预测效果。结果经过建模、拟合,得出ARIMA(0,1,1)×(0,1,1)12为最佳模型。模型拟合正态化的BIC为3.461,决定系数R;为0.426,根据贝叶斯准则BIC值最小,R;最大为最优模型;Ljung-Box Q检验统计量为Q=16.02,P=0.38,可知残差属于白噪声值,说明本模型预测相对适合。建立模型之后,对2020年1月至2020年9月CRPA的感染率进行ARIMA预测分析,结果显示2020年1月至2020年9月实际发病趋势与预测曲线图较吻合,说明ARIMA模型拟合精度和预测效果均较好。结论 ARIMA模型能准确模拟和预测CRPA感染率,为预防和控制CRPA的感染率提供参考。

Objective To predict the infection incidence of carbapenem-resistant pseudomonas aeruginosa(CRPA) by multiple seasonal autoregressive integrated moving average model(ARIMA), so as to provide the basis for the development of control strategy of CRPA.Methods The time-series model of ARIMA was established using CRPA infection data from January 2016 to December 2019 in our hospital, and then the model was validated using the data from January 2020 to September 2020 to evaluate the prediction effect of the model.Results After modeling and fitting, it is concluded that ARIMA(0,1,1)×(0,1,1)12 is the best model. The normalized BIC of model fitting is 3.461, and the coefficient of determination R;is 0.426. According to the Bayes criterion, the model with the minimum BIC value and the maximum R;value is the optimal one;the Ljung-Box Q statistics are Q=16.02 and P=0.38, indicating that the residual belongs to the white noise value, and the prediction of this model is relatively suitable. After the model is established, the ARIMA prediction analysis is performed on the CRPA infection rate from January 2020 to September 2020. The results show that the actual incidence trend from January 2020 to September 2020 is relatively consistent with the predicted curve, indicating the accuracy and the forecast of the ARIMA model have a better result.Conclusion ARIMA model can accurately simulate and predict CRPA infection rate and provide a reference for the prevention and control of multi-drug resistant bacteria infection.