Optimising Elective Theatre Lists: Machine Learning and Mathematical Optimisation for Patient Scheduling

Current elective surgical scheduling often depends on planners’ judgement and local rules of thumb. While machine learning and optimisation have advanced rapidly and show clear promise for improving such decisions, they are still not widely used in routine theatre list planning. This study introduces an integrated approach to elective surgical scheduling that reflects the uncertainty and day-to-day pressures of real theatre planning. It uses pre-trained machine learning model(s) to predict how long operations are likely to take, and a statistical component to capture how variable those durations can be. Together, these estimates feed into the scheduling problem so that the lists can be built to be more reliable and less prone to disruption. To solve the core scheduling problem, two optimisation methodologies are employed: (1) an Integer Linear Programming (ILP) solver for exact mathematical optimisation, and (2) a custom-developed meta-heuristic algorithm designed for scalability and flexibility in complex scheduling environments. Evaluating both methods across different operational conditions enables decision-makers to select the most suitable approach based on usability, solution quality, and runtime requirements (e.g., exact optimal schedules when time permits versus fast, scalable schedules when responsiveness is critical). The ultimate model(s) are designed in line with an NHS Trust setting— scheduling patients from the waiting list while taking into account the Trust's objectives in optimal resource allocation and patient satisfaction. The proposed approaches demonstrate the potential for improved scheduling efficiency, better resource utilisation, and enhanced patient satisfaction, highlighting the value of integrating AI-driven techniques in healthcare operations.