Reinforcement Learning for Patient Scheduling with Combinatorial Optimisation

Patient scheduling is a complex task that plays a crucial role in the quality of care. Effective scheduling management mitigates dissatisfaction among patients and physicians, serving as a crucial indicator. Traditionally, the approach to patient scheduling has been ad hoc, often overlooking key factors that may influence scheduling.

In this paper, we propose a reinforcement learning approach that utilises an early stopping mechanism which balances exploration and exploitation to provide combinatorial optimisation from both theoretical and experimental perspectives. Our study utilised datasets from NHS Scotland and The First Affiliated Hospital of Anhui Medical University to evaluate patient scheduling. Our results demonstrate that our Reinforcement Learning (RL) method with early stopping can successfully conduct preliminary practice on realistic examples of the General Practitioner (GP) Scheduling Problem and hospital scheduling issues.