A stochastic shortest-path MDP model with dead ends for operating rooms planning

This paper addresses the problem of operating rooms (ORs) planning with different demands from both elective patients and non-elective ones. Two types of uncertainties are incorporated: new arrivals of patients and surgery durations. A time-dependent policy to manage the waiting lists of patients is applied to determine the patient priorities in accordance with urgency levels and waiting times. In order to reduce the waiting times of patients as well as control the over-utilization of ORs, sequential decisions should be made by selecting some patients from the waiting lists and serving them every day. This problem is formulated as a stochastic shortest-path MDP (Markov decision process) with dead ends, and solved by the method of asynchronous value iteration. Results of numerical experiment show that, compared with the regular MDP model, the proposed model with time-dependent policy is more efficient in reducing the waiting times of patients, and does not lead to significant increase in over-utilization of ORs.