基于季节性指数平滑法和灰色GM(1,1)模型的小儿重症监护病房医院感染率的预测比较

Prediction and comparison of nosocomial infection rates in Pediatric Intensive Care Units based on seasonal exponential smoothing method and gray GM(1,1)model

目的预测某院小儿重症监护病房(PICU)医院感染发病率,为医院感染监测提供参考。方法收集某院PICU 2016年1月—2018年12月医院感染监测数据,分别采用季节性指数平滑法和灰色GM(1,1)模型进行预测。结果季节性指数平滑法的Holt-Winters相乘模型为最佳预测模型,R^(2)值为0.727,标准化的BIC为1.591,RMSE为1.96,该模型在α水平、γ(趋势)、δ(季节)上均有统计学意义(P<0.05);灰色GM(1,1)模型方程:X(k+1)=170.5784e^(0.017k)-164.9684,a=-0.0179,u=2.9488,检验统计量C=0.49,P=0.81。结论季节性指数平滑法的Holt-Winters相乘模型能够较好地预测某院PICU医院感染情况,可为医院感染监测提供参考依据。

Objective To predict the incidence of nosocomial infection in Pediatric Intensive Care Unit(PICU)of a hospital,and to provide reference for nosocomial infection monitoring.Methods The data of nosocomial infection monitoring in PICU in a hospital from January 2016 to December 2018 were collected,and the seasonal exponential smoothing method and the gray GM(1,1)model were used to make predictions.Results The Holt-Winters multiplication model of the seasonal exponential smoothing method was the best prediction model.The R^(2) value was 0.727,the normalized BIC was 1.591,and the RMSE was 1.96.The differences of this model were all statistically significant onαlevel,γ(trend),andδ(season)(P<0.05);the gray GM(1,1)model equation:X(k+1)=170.5784^(e0.017 k)-164.9684,a=-0.0179,u=2.9488,test statistic C=0.49,P=0.81.Conclusion The Holt-Winters multiplication model based on seasonal exponential smoothing method can better predict the nosocomial infection of PICU in a hospital,and can provide a reference for nosocomial infection monitoring.