Appointment scheduling and rescheduling are important problems in service systems. A good appointment schedule can tolerate uncertainties that occur frequently, such as no-shows and last-minute cancellations. Some other uncertainties are rare events and generally ignored in making an appointment schedule. However, if such uncertainty occurs, it will cause a large degree of disturbance on appointment schedules and force the service system to reschedule appointments. This thesis addresses the above two issues by two pieces of work, one on making an appointment schedule with random number of requests, the other on large-scale appointment rescheduling decisions caused by a pandemic.

We first study a dynamic appointment scheduling problem for a general service system in which a random numbe...[ Read more ]

Appointment scheduling and rescheduling are important problems in service systems. A good appointment schedule can tolerate uncertainties that occur frequently, such as no-shows and last-minute cancellations. Some other uncertainties are rare events and generally ignored in making an appointment schedule. However, if such uncertainty occurs, it will cause a large degree of disturbance on appointment schedules and force the service system to reschedule appointments. This thesis addresses the above two issues by two pieces of work, one on making an appointment schedule with random number of requests, the other on large-scale appointment rescheduling decisions caused by a pandemic.

We first study a dynamic appointment scheduling problem for a general service system in which a random number of homogeneous customers will call in sequentially to make appointments for service. Aiming to leverage the possibility of having a lower number of customers, we consider a non-sequential appointment scheduling policy, different from the conventional sequential (first-come-first-serve) scheduling policy, for more flexibility in managing appointment scheduling. The objective is to minimize the expected weighted sum of customer waiting time and server overtime. We develop a branch and bound algorithm to find the optimal policy, where the novelty of the algorithm lies in the integration of two different views to treat the problem. The first view is a dynamic assignment problem where each appointment request is assigned to one of the unoccupied time slots. Following this, we build up the general framework of the branch and bound algorithm. The second view is a shortest path problem on a directed network that has an exponential number of vertices. Such a formulation is integrated into the branch and bound algorithm by providing effective elimination rules and mechanisms of reducing repetitive computation. Computational results show the efficiency of our branch and bound algorithm and a number of managerial insights. In particular, the benefit of the non-sequential policy is more evident when the utilization rate is relatively low and the distribution of customer number is right-skewed.

In the second part of the thesis, we address a surgery rescheduling problem during a pandemic. The outbreak of an infectious disease causes a surge of infectious patients. The severity of the situation requires the involvement of multiple stakeholders at the strategic, tactical, and operational levels. First, at the strategic level, a government agency, such as Hospital Authority, needs to evaluate the impact on the healthcare system, in particular, the resource allocation to the hospitals. At the tactical level, each hospital needs to manage the scheduling of its operations to ensure the service level. At the operational level, detailed patient-by-patient schedule need to be made. We provide a decision-making framework for the three levels of the decision and conduct computational studies to validate the effectiveness of the framework.