The complex optimisation problems arising in the scheduling of operating rooms have received considerable attention in recent scientific literature because of their impact on costs, revenues and patient health. For an important part, the complexity stems from the stochastic nature of the problem. In practice, this stochastic nature often leads to schedule adaptations on the day of schedule execution. While operating room performance is thus importantly affected by such adaptations, decision-making on adaptations is hardly addressed in scientific literature. Building on previous literature on adaptive scheduling, we develop adaptive operating room scheduling models and problems, and analyse the performance of corresponding adaptive scheduling policies. As previously proposed (fully) adaptive scheduling models and policies are infeasible in operating room scheduling practice, we extend adaptive scheduling theory by introducing the novel concept of committing. Moreover, the core of the proposed adaptive policies with committing is formed by a new, exact, pseudo-polynomial algorithm to solve a general class of stochastic knapsack problems. Using these theoretical advances, we present performance analysis on practical problems, using data from existing literature as well as real-life data from the largest academic medical centre in The Netherlands. The analysis shows that the practically feasible, basic, 1-level policy already brings substantial and statistically significant improvement over static policies. Moreover, as a rule of thumb, scheduling surgeries with large mean duration or standard deviation early appears good practice.

Additional Metadata
Keywords adaptive scheduling, operating room scheduling, operations research in healthcare, stochastic knapsack problem
Persistent URL,
Journal International Journal of Production Research
Xiao, G, van Jaarsveld, W.L, Dong, M, & van de Klundert, J.J. (2017). Models, algorithms and performance analysis for adaptive operating room scheduling. International Journal of Production Research, 1–25. doi:10.1080/00207543.2017.1328140