Nurse rostering through linear programming and repair heuristics

Abstract

We consider a nurse scheduling problem in a large hospital in the center of The Netherlands. Approximately 50 nurses with different qualifications should obtain a work schedule for a period of 6 weeks. Every day is divided in three shifts (day, late and night) and we should make sure enough employees with the correct qualifications get assigned to each shift. While creating a solution, we should both take the general regulations and the personal roster preferences into account.

We present a three-stage solution approach. First, we create a set of suitable schedules for every employee using an individual roster generation scheme. Thereafter, all individual schedules are combined into a full schedule which satisfies all occupancy demands for the period. A Linear Program is used for this step. Finally, some remaining problems are resolved using a set of proposed repair heuristics. These heuristics can also be used to perform some schedule changes afterwards.

Our results look promising, although it is difficult to cope with all possible preferences and demands.