Generación automática de una metaheurística para el problema de asignación de cirugías electivas considerando afinidad y preferencias en el equipo quirúrgico.
Loading...
Date
2025
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad de Concepción
Abstract
La programación de cirugías electivas implica asignar pacientes a quirófanos en días y bloques horarios junto con un equipo quirúrgico. Aunque este problema ha sido ampliamente estudiado, la afinidad humana y las preferencias de recursos de los cirujanos, factores que mejoran el bienestar, acortan la duración de los procedimientos y fomentan un mejor ambiente laboral, rara vez han sido incluidos. En enfoques tradicionales, dichos factores se tratan como restricciones duras en los modelos, exigiendo un nivel mínimo de afinidad en cada equipo. Sin embargo, esta rigidez puede reducir el número de cirugías agendadas, una decisión que podría afectar la salud de un paciente. Para mitigarlo, surgieron modelos con sistemas de penalización por puntajes: cada cirujano dispone de un presupuesto diario que le permite integrarse en equipos mejor afinados y acceder a recursos preferidos, manteniendo el equilibrio en las asignaciones. Estos sistemas son incorporados a los modelos matemáticos. Sin embargo, suelen ser muy pesados computacionalmente, requiriendo largos tiempos de ejecución, realidades que pocos hospitales pueden afrontar sin afectar la práctica clínica. Así, esta tesis propone un framework de generación de un algoritmo metaheurístico, capaz de resolver la asignación de cirugías electivas en plazos cortos y de ofrecer soluciones competitivas. La generación automática explora todo el espacio de parámetros sin necesidad de calibración manual, garantizando una calidad difícil de alcanzar por ajustes convencionales. Para la calibración, se diseñan doce operadores de diversificación, diez de intensificación, dos de destrucción, tres constructores de soluciones iniciales y tres criterios de aceptación. Este conjunto de operadores equilibra búsqueda global y local, aumentando la robustez de la metaheurística en diversos escenarios. Se implementó un framework para llevar a cabo la calibración automática, seleccionando iterativamente las mejores combinaciones de operadores y parámetros. La metaheurística resultante se comparó con modelos matemáticos de la literatura y con metaheurísticas clásicas en tres tiempos limites definidos. Los resultados muestran que el algoritmo generado automáticamente produce mejores soluciones que los modelos matemáticos en el mismo tiempo y se sitúa en primer lugar frente a las metaheurísticas clásicas. Su implementación en un hospital ofrece menores tiempos de cómputo y mayor eficiencia operativa ante la llegada constante de pacientes, además de formar equipos quirúrgicos con mejor afinidad, repercutiendo positivamente en el bienestar del personal y de los pacientes.
Elective surgery scheduling involves assigning patients to operating rooms on specific days and time blocks, together with a surgical team. Although this problem has been widely studied, human affinity and surgeons’ resource preferences—factors that improve well-being, shorten procedure durations and foster a better working environment—are rarely included. In traditional approaches, these factors are treated as hard constraints in the models, requiring a minimum level of affinity in each team. However, this rigidity can reduce the number of scheduled surgeries, a decision that could impact a patient’s health. To mitigate this, scoring-based penalty models have emerged: each surgeon is given a daily budget that allows them to join better matched teams and access preferred resources while maintaining balance in the assignments. These systems are incorporated into mathematical models but are often computationally heavy, requiring long execution times—realities that few hospitals can afford without affecting clinical practice. Thus, this thesis proposes a framework for generating a metaheuristic algorithm capable of solving the elective surgery assignment problem within short time frames and offering competitive solutions. The automatic generation explores the entire parameter space without the need for manual calibration, guaranteeing a quality difficult to reach with conventional tuning. For calibration, twelve diversification operators, ten intensification operators, two destruction operators, three initial-solution constructors and three acceptance criteria are designed. This set of operators balances global and local search, increasing the metaheuristic’s robustness across diverse scenarios. A framework was implemented to perform automatic calibration, iteratively selecting the best combinations of operators and parameters. The resulting metaheuristic was compared with mathematical models from the literature and with classical metaheuristics under three defined time limits. The results show that the automatically generated algorithm produces better solutions than the mathematical models in the same time and ranks first compared to classical metaheuristics. Its implementation in a hospital offers shorter computation times and greater operational efficiency in the face of a constant influx of patients, as well as forming surgical teams with better affinity, positively impacting the well-being of both staff and patients.
Elective surgery scheduling involves assigning patients to operating rooms on specific days and time blocks, together with a surgical team. Although this problem has been widely studied, human affinity and surgeons’ resource preferences—factors that improve well-being, shorten procedure durations and foster a better working environment—are rarely included. In traditional approaches, these factors are treated as hard constraints in the models, requiring a minimum level of affinity in each team. However, this rigidity can reduce the number of scheduled surgeries, a decision that could impact a patient’s health. To mitigate this, scoring-based penalty models have emerged: each surgeon is given a daily budget that allows them to join better matched teams and access preferred resources while maintaining balance in the assignments. These systems are incorporated into mathematical models but are often computationally heavy, requiring long execution times—realities that few hospitals can afford without affecting clinical practice. Thus, this thesis proposes a framework for generating a metaheuristic algorithm capable of solving the elective surgery assignment problem within short time frames and offering competitive solutions. The automatic generation explores the entire parameter space without the need for manual calibration, guaranteeing a quality difficult to reach with conventional tuning. For calibration, twelve diversification operators, ten intensification operators, two destruction operators, three initial-solution constructors and three acceptance criteria are designed. This set of operators balances global and local search, increasing the metaheuristic’s robustness across diverse scenarios. A framework was implemented to perform automatic calibration, iteratively selecting the best combinations of operators and parameters. The resulting metaheuristic was compared with mathematical models from the literature and with classical metaheuristics under three defined time limits. The results show that the automatically generated algorithm produces better solutions than the mathematical models in the same time and ranks first compared to classical metaheuristics. Its implementation in a hospital offers shorter computation times and greater operational efficiency in the face of a constant influx of patients, as well as forming surgical teams with better affinity, positively impacting the well-being of both staff and patients.
Description
Tesis presentada para optar al grado de Magíster en Ingeniería Industrial.
Keywords
Metaheuristics, Algorithms, Surgery