An Integer Programming-Based Nurse Rostering System

This paper considers a real-world rostering problem at the Gold Coast Hospital, Queensland. Nurse rostering is a constraint satisfaction problem (CSP) [3]. The task is to find a consistent allocation of shift values, for a group of nurses, over a fixed period of time, that satisfy a set of rostering constraints. These constraints include i) acceptable shift combinations (or schedules) for individual nurses, and ii) acceptable overall staffing levels for each shift. Previous work in the area has looked at optimally solving small or simplified problems [5] or using non-optimal heuristics to solve larger, more realistic problems [1]. The current research uses an optimising integer programming (IP) approach to solve a large rostering problem within reasonable time and computing resource bounds. The paper shows how the nurse rostering problem can be simplified by combining different modelling and decomposition techniques. Finally, the research demonstrates an IP model can capture “difficult” constraints and remain flexible in changing circumstances.