An empirical and theoretical study of outpatient scheduling problems employing simulation and genetic algorithm methodologies

• Authors: Marie E. Matta;Sarah Stock Patterson;Salah E. Elmaghraby
• Venue: An empirical and theoretical study of outpatient scheduling problems employing simulation and genetic algorithm methodologies
• Year: 2004

Operationally, hospitals are complex systems. Despite their prevalence, a void exists in the literature on patient flows that exhibit multiple stages of activities in multiple facilities. Research has failed to address the operation of hospital outpatient facilities from a system-wide and common scheduling perspective. This dissertation assists in filling this void by developing approaches to schedule and improve the operations of two multi-facility departments in a hospital—an oncology center and a testing center. The first study focuses on the operations of a real oncology center. A simulation model of the center is constructed for experimentation. Past simulation analyses of outpatient facilities have been limited in that they have been predominantly steady-state models and focus on only one response variable. This dissertation overcomes these limitations by formulating a terminating simulation model with multiple response variables. It derives a method for handling the multiple response problem imposed by performance measures being dependent upon patient classes, patient routings, and days of the week. A new scoring methodology aids in selecting the “best” alternative from among several possible. Changes in scheduling, process flow, and resources are predicted to reduce patient wait time and resource overtime. This research has already been successful as the oncology center has made several improvements. The second study focuses on hospital diagnostic testing facilities (MRI, X-ray, CT Scan, etc.). It proposes a theoretical way to design and schedule these facilities as a centralized multiprocessor open shop (MPOS). The MPOS scheduling problem is NP-complete. Despite its complexity, two restricted cases are found to be solvable in polynomial time. For the general MPOS problem, however, a three-phase heuristic is developed to design, load, and schedule the center. A novel genetic algorithm is presented as the method for scheduling the MPOS. A clever chromosome is developed that succinctly encodes a schedule of patients across multiple testing facilities. The innovative design of this chromosome enables any permutation of its genes to yield a feasible solution and enables the genetic operators to easily manipulate a schedule. Computational results reveal that the algorithm performs extremely well.