Dynamic appointment rescheduling with patient preferences

This study examines patient-initiated appointment rescheduling with consideration of patient preferences. Online rescheduling policies are investigated for the selection and sequential offering of new appointments upon the arrival of a rescheduling request via a telephone call. Appointments are offered until the patient accepts one or the maximum number of offers is reached. The aim is to reschedule appointments using a weighted function to maximise the patients’ satisfaction, optimise the operational performance, and minimise the number of patients deferred to a future time horizon. Different patient types are taken into account characterised by their uncertainties in rescheduling, cancellation, no-show, and service duration. The rescheduling process is formulated as a stochastic dynamic scheduling problem and approximated using a Markov Decision Process (MDP). Two heuristic policies are proposed, referred to as the myopic stochastic and the MDP-based algorithms. Both policies apply a simulation-optimisation approach that considers patient preferences and expected operational performance. To determine the set of offered appointments, the MDP-based algorithm additionally accounts for expected future rescheduling requests. Computational experiments are performed on real-life instances. The results demonstrate that the two proposed policies yield solutions of high quality. The myopic stochastic policy outperforms the MDP-based policy when it is challenging to offer suitable slots due to high capacity utilisation or a lack of clear patient preferences. Conversely, the MDP-based algorithm delivers better results when capacity utilisation is lower and there is some variation in preferences across days and patients.