Abstract：Objective To construct a model based on the number of past dischared patients by month and to predict the number of discharged patients in a tertiary referral hospital. Methods We predicted and analyzed time series of discharged patients and monthly discharged patients per bed (monthly discharged patients/the bed number at the beginning of the year) respectively. Then, we validated the results by true value and predictive value of discharged patients in 2014. Results In model ARIMA(0,1,1)(1,1,1)12, the mean absolute percentage error of monthly discharged patients was 14.49% and the mean absolute percentage error of monthly discharged patients per bed was 7.75%. We selected the final model on the basis of mean absolute percentage error from true value and predictive value. So the autoregressive moving smoothing model was superior to the exponential smoothing model. The best fitting model was Winters′s multiplicative model which directly predicted monthly discharged patients with the mean absolute percentage error of 2.96%, better than the Winters′s additive model of monthly discharged patients per bed with that of 6.62%. We predicted the number of yearly discharged patients in 2015 would be 104 248 with Winters′s multiplicative model of monthly discharged patients and 95% CI between 89 109 and 119 386. Conclusion Using the expert modeler of SPSS software to construct Winters′s multiplicative predictive model is easy to operate and the predictive effect is relatively better.
刘晓钰，陈莉雅，陈平雁. 用指数平滑模型预测广州某三甲医院出院量[J]. 中国医院统计, 2016, 23(2): 92-95.
Liu Xiaoyu, Chen Liya, Chen Pingyan. Prediction of the number of discharged patients in a tertiary referral hospital in Guangzhou with exponential smoothing model. journal1, 2016, 23(2): 92-95.
