Deep Learning Techniques for Forecasting Emergency Department Patient Wait Times in Healthcare Queue Systems

Abstract

A lot of hospitals make use of the duration of patients' stays in queue as a gauge for overcrowding in the emergency room (ER). Many emergency rooms have lengthy wait times, which make it more challenging to provide patients with appropriate care and increases overall expenses. In queuing system applications, Innovative techniques like machine learning and deep learning (DL) have become crucial. In order to forecast Waiting periods for patients in a system, this research will use deep learning techniques for historical queuing variables, either in addition to or instead of queuing theory .SGD, Adam, RMSprop, and AdaGrad were the four optimization algorithms that were applied. To determine which model has the minimum absolute mean error (MAE), there was an algorithmic comparison. To facilitate more comparisons, a traditional mathematical simulation was utilized.. The findings demonstrated that the DL model may be used to estimate patients' waiting times utilizing the SGD algorithm, with the lowest MAE of 09.60 minutes (23% reduction of errors) activated. In order to better priorities patients in a queue, this study contributes theoretically to the field of patient waiting time prediction using alternative methodologies by establishing the highest performing model. This study also makes a useful addition by utilizing actual data from emergency rooms. In addition, we suggested models that, compared to a conventional mathematical approach, produced more accurate predictions of patients' waiting times. Our method can be readily applied to the healthcare sector's queue system by utilizing data from electronic health records (EHRs). Since over 40% of people who are admitted to hospitals do so through the emergency rooms (ER), most hospitals suffer from extreme patient overcrowding. Since most ER departments in hospitals have lengthy patient wait times, they are an important component of healthcare facilities.