Mitigating overtime risk in tactical surgical scheduling. (June 2020)

• Record Type: Journal Article
• Authors: Zhang, Yu; Wang, Yu; Tang, Jiafu; Lim, Andrew
• Is Part Of: Omega. Volume 93(2020)
• Publication Date: 2020-06
• Subjects: Surgery scheduling -- Overtime -- Risk aversion -- Exact algorithm -- Healthcare; Management -- Periodicals; 658.4005
• DOI: 10.1016/j.omega.2019.01.002
• Languages: English
• ISSNs: 0305-0483

Abstract

Highlights: We formulate a stochastic programming model for a tactical surgery scheduling problem. Our decision criterion is able to mitigate both the probability of overtime and its magnitude. We show the computational complexities of several models for the problem. We develop an efficient hill-climbing algorithm to solve our model. We use real data simulation to illustrate the effectiveness of our approach against existing ones. Abstract: Overtime is a common phenomenon in surgery departments, causing stress to physicians, dissatisfaction to patients, and financial loss to hospitals. We help risk-averse managers of operating rooms (ORs) to mitigate overtime in tactical surgery scheduling, which determines the assignment of elective patients to available ORs in upcoming time periods. We model the uncertain surgical durations via partial, full, or empirical distributions. To mitigate overtime, our model maximizes the risk aversion level of the OR manager (and thus the risk-hedging ability of the solution) while ensuring that the certainty equivalent of surgery duration in each OR at each time period does not exceed the stipulated working hours. The corresponding decision criterion, termed the maximized risk aversion level, is demonstrated in theory and in numerical experiments to be able to mitigate both the overtime probability and the expected overtime duration. To solve the problem, we develop an exact hill-climbing algorithm and demonstrate its convergence and correctness. Numerical experiments based on real-life surgery data show that our method outperforms the existing methods in several indicators that of concern to OR managers. In particular, this method is computationally amiable and hence is applicable to larger-scale instances.