Abstract

Physician scheduling is an important part of hospital operation management. Fatigue, nervousness, high levels of stress and depression are common negative effects of inappropriate work schedules on physicians. A robust and automated personnel scheduling system, which satisfies physicians' preferences, not only improves the quality of life for physicians but also helps to provide a better care for patients and potentially makes significant savings in time and cost for hospitals. Polyclinics reduce the burden on hospitals and help bridge the gap between primary and secondary care. They provide various hospital services such as X-rays, minor surgeries and out-patient treatment and gather several practices under one roof to cooperate, interact and share available resources. In addition, this structure provides an opportunity for physicians of different disciplines to work together and enables patients with chronic and complex conditions to visit multiple clinics at the same place during the same visit. Our problem of interest is mainly motivated by an extension of physician scheduling problems arising in ambulatory polyclinics, where the interaction of clinics and its consequences in terms of sharing their scarce resources introduce new constraints and add complexity to the problem.
In the first part of this thesis, we present an integrated physician and clinic scheduling problem arising in ambulatory cancer treatment polyclinics, where patients may be assessed by multiple physicians from different clinics in a single visit. The problem focuses on assigning clinic sessions and their associated physicians to shifts in a finite planning horizon. The complexity of this problem stems from the fact that several interdisciplinary clinics need to be clustered together, sharing limited resources. The problem is formulated as a multi-objective optimization problem. Given the inherent complexity for optimally solving this problem with a standard optimization software, we develop a hybrid algorithm based on iterated local search and variable neighborhood descent methods to obtain high quality solutions.
In the second part we propose a comprehensive bi-level physicians planning framework for polyclinics under uncertainty. The first level focuses on clinic scheduling and capacity planning decisions, whereas the second level deals with physicians scheduling and operational adjustments decisions. In order to protect the generated schedules against demand uncertainty, the first level is modeled as an adjustable robust scheduling problem which is solved using an ad-hoc cutting plane algorithm. To cope with variability in patients' treatment times, we formulate the second level as a two-stage stochastic problem and use a sample average approximation scheme to obtain solutions with small optimality gaps. Moreover, we use a Monte-Carlo simulation algorithm to demonstrate the potential benefits of using our planning framework.
In the last part of this thesis we investigate on the impact of physicians work schedules on patient wait-time under uncertain arrival pattern and treatment time of patients. We provide a methodology that combines discrete-event simulation with an optimization search routine to minimize patient wait-time and physician overtime subject to several scheduling/resource restrictions. We indicate the significant impact of adopting the proposed simulation optimization framework for physician scheduling on reducing the aforementioned key performance measures.