Efficient scheduling of integrated operating rooms and post-anesthesia care units under uncertain surgery and recovery times: an artificial neural network-metaheuristic framework Efficient scheduling of integrated operating rooms and post-anesthesia care units under

• Published in: Soft computing (Berlin, Germany) Vol. 29; no. 8; pp. 3909 - 3941
• Main Authors: Ahmadian, Mohammad Amin, Varmazyar, Mohsen, Fallahi, Ali
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.04.2025
• Subjects: Artificial Intelligence; Computational Intelligence; Control; Engineering; Mathematical Logic and Foundations; Mechatronics; Robotics; Soft Computing in Decision Making and in Modeling in Economics; Computer simulation; Machine learning; Artificial neural network; Evolutionary algorithms; Healthcare scheduling
• ISSN: 1432-7643; 1433-7479
• DOI: 10.1007/s00500-025-10620-0

Hospital managers must plan and schedule their operating rooms (ORs) and related units to increase productivity and deliver high-quality care. In this situation, some surgery parameters, like surgery and recovery times, are subject to uncertainty due to the unstable environment of hospital processes. Therefore, efficient methodologies should be designed to address these uncertainty sources and provide acceptable solutions in a reasonable time. Typically, solving the problem involves simulation-optimization approaches known to be time-consuming. This work presents a new data mining-optimization approach for integrated ORs and post-anesthesia care unit (PACU) scheduling, known as the operating theater room (OTR) scheduling problem, under uncertainty in surgery and PACU times. Concerning the complexity of the problem, we design four metaheuristics, including genetic algorithm (GA), particle swarm optimization (PSO), simulated annealing (SA), and tabu search (TS) based on the problem features to search the solution space. Our presented approach aims to improve the traditional simulation-optimization frameworks, and utilizes simulation experiment results to train an artificial neural network (ANN), which evaluates the generated sequences of metaheuristic algorithms. The goal is to optimize the patients’ scheduling to minimize the problems’ single objective function. Two cases of the problem are defined based on makespan and total tardiness objectives. Extensive analysis of the results is established by providing three different sizes of example categories from the literature with several probability distributions for surgery and PACU times of patients. The input parameters of metaheuristic algorithms are tunned using the Taguchi design of experiments method, and the results are analyzed using several computational measures. The analysis of results reveals that GA performs better than the other algorithms regarding the solution quality measure for both cases. For the makespan objective case, the average relative percentage deviation (RPD) of GA is zero, while for PSO, SA, and TS, the deviations are 0.03, 0.01, and 0.04, respectively. This difference becomes even more pronounced in the total tardiness objective case, where the average RPD of GA is 0.04, compared to 0.36, 0.41, and 0.84 for PSO, SA, and TS, respectively. Finally, analysis of variance (ANOVA) and Kruskal–Wallis tests are utilized as parametric and non-parametric tests for statistical comparisons. The p value of the conducted tests is less than 0.05 at the 0.05 significance level, confirming a significant difference in the performance of the algorithms. Some managerial insights and practical implications are presented based on obtained results.