住院患儿医院感染发生率ARIMA时间序列模型

目的初步探索和评价儿童患者医院感染发生率的自回归滑动平均混合模型(ARIMA)预测模型。方法以某院2011年1月—2014年12月4年医院感染发生率数据建立ARIMA模型,依据信息量准则,确定最优模型;以2015年医院感染发生率数据作为验证样本,评价模型的可行性。结果 ARIMA(0,1,1)为医院感染率最优预测模型,其最小信息量准则(AIC)、贝叶斯信息准则(BIC)值分别为66.61、70.76,模型残差序列的LjungBox统计量Q=14.14,差异无统计学意义(P=0.658),提示残差为白噪声序列,模型拟合良好。医院感染率实际值与预测值的平均绝对百分误差值(MAPE)为22.4,实际值均在预测值95%可信区间内,未见超出点。结论 ARIMA时间序列模型拟合医院感染率效果良好,具有预测住院患儿医院感染发生情况的实际价值。

Construction of ARIMA time series model for healthcare associated infection in hospitalized children

ObjectiveTo investigate the applicability of autoregressive integrated moving average (ARIMA) model in predicting healthcare associated infection(HAI) in children. MethodsThe ARIMA model was constructed according to the incidence of HAI in a hospital from January 2011 to December 2014. With the use of information criterion, optimal model was determined; HAI data in 2015 was as test samples, the feasibility of the model was evaluated. ResultsARIMA (0,1,1) was the optimal prediction model for HAI rate, the Akaike Information Criterion（AIC）and Bayesian Information Criterion(BIC）of the ARIMA(0,1,1) were 66.61 and 70.76, respectively. The Ljung Box statistics value Q= 14.14 was not significantly different (P= 0.658), suggesting a white noise sequence of residuals with a good model fitting. The mean absolute percent error(MAPE) between actual and fitting value of HAI was 22.4, the actual values were within the 95% confidence interval. ConclusionARIMA model fits the time series data, and can achieve satisfactory effect on predicting the incidence of HAI in hospitalized children.