Demand and resource planning for emergency services using atificial neural networks

Emergency Department (ED) must be managed effectively since it is the first point of care in hospitals for urgent and critically ill patients. ED can be effective, only by using resources efficiently. Generally, simulation model is used in healthcare systems to optimize resources. In ED, the optimum number of bed resource is crucial, since most of the resources in ED can be planned according to the number of bed resource. This study has devoloped a simulation model to determine the optimum number of beds. This simulation model needs some inputs and the most important input for this model is patient's arrival rate. 10 different machine learning algorithms are utilized to predict the patient's arrival rate. These machine learning algorithms need optimum feature subsets and this optimum subset has been determined by using exhaustive feature selection method. The most significant feature is identified as mean arrival rate. The long short-term memory (LSTM) model has the best accuracy with a MAPE value of 46,7%, and by the help of the simulation method, the length of stay (LOS) has been minimized by 7% and the number of beds at the ED has been optimized.