Simulated annealing for a multi-level nurse rostering problem in hemodialysis service

Hemodialysis service is provided with the cooperation of nurses taking on different roles: in-charged nurse, dispensing nurse, and treatment nurse. The goal of multi-level nurse rostering problem in hemodialysis service (MLHSNRP) is to assign multi-level nurses to satisfy demand of different roles which includes the number of nurses and required levels of qualification. Evaluation criteria consist of satisfying requirements of levels and preferences on shifts and roles of nurses. A 0-1 integer programming model is formulated with a synthetic objective and a simulated annealing (SA) based on a heuristic algorithm is developed. The heuristic algorithm rosters nurses to cover daily demands in order with some heuristic rules. Three neighborhood structures are embedded into SA to improve the solution obtained by the heuristic algorithm. A series of instances are generated based on real cases in Hemodialysis Service Center P in Wuhan, China. Computational experiments are conducted on different combinations of the neighborhood structures and the parameters of SA. The combination of neighborhood structures and parameters with the best optimization effect are found. A comparative evaluation of SA is carried out against a hybrid artificial bee colony (HABC). The results show that SA has a better performance as a whole, SA and HABC respectively have advantages on problems of different scales, and SA runs faster than HABC. (C) 2017 Elsevier B.V. All rights reserved.