Base On Composite Models For Forecasting Emergency Department Visits

In our country,doctor-patient dispute problem is increasingly serious,and medical disturbance phenomenon is still often occur.The frequency of such problems is closely related to the crowded emergency room.Accurate prediction of emergency patient visits will directly affect and contribute to the allocation of medical resources.By realizing the reasonable allocation of medical resources,congestion in emergency department will be alleviated to a great extent.Based on the existing ARIMA-ANN and ARIMA-LR composite models,this paper puts forward the HW-ANN,HW-LR,ETSANN and ETS-LR composite models with higher prediction accuracy and stronger applicability,and realizes the accurate prediction of emergency patient visits in many hospitals.Data were collected from two hospitals in Dalian and one in Hong Kong.Dalian is a semi-humid continental monsoon climate with distinct seasons.Hong Kong has a subtropical climate with relatively high temperatures throughout the year.According to the different characteristics of the climate in the two places,the selection of factors affecting the visits of emergency patients in this paper is also different.This paper also compares and analyses the predictions of patient visits in three emergency departments in the two places.The purpose of this study is to explore the factors affecting the visits of emergency patients and what models are more suitable for the prediction of visits of patients in three emergency departments in the two places.The conclusions are as follows: In terms of influencing factors of emergency patient visits,weather factors,date factors and holiday factors are significant in predicting patient visits in large hospitals in Dalian.The emergency department of small hospitals in Dalian found that the date factor was significant in predicting patient visits,while the holiday factor and weather factor were not significant in predicting patient visits.The visits of patients with emergency in Hong Kong were significantly affected by weather and date factors,while the visits of patients with holiday factors were not significantly affected.As far as model selection is concerned,both large hospitals and small community hospitals in Dalian are suitable for ETS-LR model and ETS model.That is,the newly proposed ETS and its related composite model have the highest prediction accuracy.In addition,ARIMA-LR model also performs well in predicting patient visits in large hospitals.In the prediction of patient visits in small community hospitals,the accuracy of HW-LR model is also considerable.Therefore,GLM model is the most applicable one among the visits of patients in the emergency department of Hong Kong hospitals.At the end of the paper,by comparing the emergency department models of mainland hospitals with those of Hong Kong hospitals,the Enlightenment of using the emergency department management model of Hong Kong hospitals for reference in the emergency department of mainland hospitals from both macro and micro perspectives was obtained.