A Data-driven Approach for Uncertainty in Multi-objective Operating Room Planning and Scheduling

Health care managers must deal with decisions across different hospital areas such as the emergency department, surgical services and postoperative care. Due to limited resources, hospitals seek efficient operation management while maximizing the satisfaction of patients, physicians, staff and other stakeholders. Healthcare systems’ dynamic and complex environment makes decisions difficult and sensitive to the uncertainty present in processes and patient conditions. Operating room management in surgical services is an attractive area for managers and researchers due to its influence on hospitals’ financial performance and quality of care. The operating room planning and scheduling problem refers to allocating surgeries in a suitable date and operating room and then sequence the allocated surgeries while optimizing objectives such as the number of late surgeries, tardiness, resource utilization, and stakeholder satisfaction. Moreover, in a real-world setting, planning and scheduling depend on usually unknown inputs such as surgery duration, patient arrival and unplanned emergency events. To describe uncertainty, ML provides alternatives through supervised and unsupervised learning. Once uncertainty is adequately described, stochastic programming, optimization-simulation or combined OR/ML approaches could be used to obtain a robust plan and schedule. OR/ML integration is a relatively new field and there is even less literature related to how to combine those fields to tackle problems under uncertainty. In this sense, three of the most relevant question that will be addressed during my Ph.D. studies are (1) how to solve the operating room planning and scheduling problem under uncertainty considering available data from hospitals (2) how sensitive the previous approach to the technique used to describe uncertainty is? and (3) what data-driven framework should be used by hospitals to choose the best method to solve the operating room planning and scheduling problem according to the available data?