Predicting Patient Waiting Time in the Queue System Using Deep Learning Algorithms in the Emergency Room

Abstract

Many hospitals consider the length of time waiting in queue to be a measure of emergency room (ER) overcrowding. Long waiting times plague many ER departments, hindering the ability to effectively provide medical attention to those in need and increasing overall costs. Advanced techniques such as machine learning and deep learning (DL) have played a central role in queuing system applications. This study aims to apply DL algorithms for historical queueing variables to predict patient waiting time in a system alongside, or in place of, queueing theory (QT). We applied four optimization algorithms, including SGD, Adam, RMSprop, and AdaGrad. The algorithms were compared to find the best model with the lowest mean absolute error (MAE). A traditional mathematical simulation was used for additional comparisons. The results showed that the DL model is applicable using the SGD algorithm by activating a lowest MAE of 10.80 minutes (24% error reduction) to predict patients' waiting times. This work presents a theoretical contribution of predicting patients’ waiting time with alternative techniques by achieving the highest performing model to better prioritize patients waiting in the queue. Also, this study offers a practical contribution by using real-life data from ERs. Furthermore, we proposed models to predict patients' waiting time with more accurate results than a traditional mathematical method. Our approach can be easily implemented for the queue system in the healthcare sector using electronic health records (EHR) data.

Performance Evaluation on Resolution Time Prediction Using Machine Learning Techniques

“…Hijry and Olawoyin [27] proposed a solution that studies deep learning techniques for historical queueing variables that will be used in addition to, or instead of, queueing theory to anticipate patient waiting times in a system (QT). They employed four optimization strategies, including Stochastic Gradient Descent (SGD) [28], adaptive moment estimation (Adam) [29], [30] Root Mean Square Propagation (RMSprop), and Adaptive Gradient (AdaGrad) [31], The model is evaluated using the mean absolute error (MAE) method.…”

An Optimized Deep Learning Approach for Improving Airline Services

“…It improves the accuracy of the deep neural network, speeds up the training, and improves performance. It is used to update the neural network's attributes and leads to minimizing the loss function and maximizing the accuracy [30]. The optimized approach achieved the highest accuracy compared with previous studies, which applied data mining, machine learning, and deep learning to airlines [31].…”

“…Different combinations of hyperparameters that include different numbers of epochs (150,200), hidden layers (5,6,7), and neurons (30,200) are addressed. In addition, several activation functions (ReLU, Tanh) were applied as shown in Table 3.…”

Prediction of Waiting Times in A&E

“…The nature of the A&E department is complex and multifactorial [1]. Several methods have been used to predict the waiting time in emergency departments; however, even though recent research has accomplished better prediction models, the need for more precise results is still present [3][4][5][6][7][8][9][10][11][12]. The current project implemented and compared two supervised tree-based ensemble machine learning algorithms, random forest and XGBoost [13,14].…”