Block-based Outpatient Clinic Appointments Scheduling Under Open-access Policy

Outpatient clinic appointment scheduling is an important topic in OR/IE studies. Open-access policy shows its strength in improving patient access and satisfaction, as well as reducing no-show rate. The traditional far-in-advance scheduling plays an important role in handling chronic and follow-up care. This dissertation discusses a hybrid policy under which a clinic deals with three types of patients. The first type of patients are those who request their appointments before the visit day. The second type of patients schedule their appointment on the visit day. The third type of patients are walk-in patients who go to the clinic without appointments and wait to see the physician in turn. In this dissertation, the online scheduling policy is addressed for the Type 2 and Type 3 patients, and the offline scheduling policy is used for the Type 1 patients. For the online scheduling policy, two stochastic integer programming (SIP) models are built under two different sets of assumptions. The first set of assumptions ignores the endogenous uncertainty in the problem. An aggregate assigning method is proposed with the deterministic equivalent problem (DEP) model. This method is demonstrated to be better than the traditional one-at-a-time assignment through both overestimation and underestimation numerical examples. The DEP formulations are solved using the proposed bound-based sampling method, which provides approximated solutions and reasonable sample size with the least gap between lower and upper bound of the original objective value. On the basis of the first set of assumptions and the SIP model, the second set of assumptions considers patient no-shows, preference, cancellations and lateness, which introduce endogenous uncertainty into the SIP model. A modified L-shaped method and aggregated multicut L-shaped method are designed to handle the model with decision dependent distribution parameter. Distinctive optimality cut generation schemes are proposed for three types of distribution for linked random variables. Computational experiments are conducted to compare performance and outputs of different methods. An alternative formulation of the problem with simple recourse function is provided, based on which, a mixed integer programming model is established as a convenient complementary method to evaluate results with expected value. The offline scheduling aims at assigning a certain number of Type 1 patients with deterministic service time and individual preferences into a limited number of blocks, where the sum of patients’ service time in a block does not exceed the block length. This problem is associated with bin packing problem with restrictions. Heuristic and metaheuristic methods are designed to adapt the added restrictions to the bin packing problem. Zigzag sorting is proposed for the algorithm and is shown to improve the performance significantly. A clique based construction method is designed for the Greedy Randomized Adaptive Search Procedure and Simulated Annealing. The proposed methods show higher efficiency than traditional ones. This dissertation offers a series of new and practical resolutions for the clinic scheduling problem. These methods can facilitate the clinic administrators who are practicing the open-access policy to handle different types of patients with deterministic or nondeterministic arrival pattern and system efficiency. The resolutions range from operations level to management level. From the operations aspect, the block-wise assignment and aggregated assignment with SIP model can be used for the same-day request scheduling. From the management level, better coordination of the assignment of the Type 1 patients and the same-day request patients will benefit the cost-saving control