Adapting duration categorical value to accommodate duration variability in a next-day operating room scheduling

Abstract

Operating room scheduling is a complex problem since it involves the synchronization of multiple resources and uncertainty factors. The duration data is the significant input for scheduling, but it is naturally uncertain. We assume that the variability of duration in operating room scheduling can be represented by the categorical value. We use a time bin, a group of adjacent numerical values of duration, to represent a category value of duration. This study evaluates the impact of duration prediction using the categorical value on the operating room scheduling performance. For the scheduling model, we discuss a next-day operating room scheduling model considering surgeons' preferences for surgery starting time with limited resources and solve it using a priority dispatching rule-based heuristic algorithm. The numerical experiments using the data from a middle-scale university hospital in Indonesia show that the scheduling performance of the heuristic algorithm outperforms the actual schedule for most cases. The underestimated and overestimated duration predictions have no impact on increasing PACU beds, but underestimated surgery duration predictions have a significant impact on increasing surgeons' and patients’ waiting time. However, our results show the opportunity to use the categorical value of duration to accommodate the duration variability in the operating room scheduling problem.

Keywords

• Duration categorical value prediction
• Machine learning
• Operating room scheduling
• Recovery bed
• Waiting time