The centralized nurse scheduling model that we propose is a bicriteria integer scheduling model with objectives for schedule cost and schedule desirability. The schedule desirability objective accommodates each nurse's individual desirability (undesirability) for certain shifts (i.e. overtime, weekends). In addition to preferences, many nursing managers feel it is important to control the amount of work (patients) that each nurse is assigned. Therefore, the centralized scheduling model also includes constraints that control the service levels on each unit, based on nurse-to-patient ratios. The model accommodates both RNs and NAs, but could be expanded if necessary to accommodate additional nurse types. The centralized nurse scheduling model as formulated here is nonlinear.
The notation used in the model formulation is outlined below followed by the model. The model extends the one seen in Wright et al. [42] by presenting new problem instances and a new scheduling policy that schedules nurses across multiple units. This extension allows us to evaluate the benefits of centralized versus decentralized scheduling decisions and requires the decision variables to be expanded to include an additional subscript representing the units across which the scheduling occurs. In addition, we present a new linearization of the model that can be solved with integer programming software. This linearization (not provided in Wright et al. [42]) makes this large and difficult problem much more manageable. Finally, this model differs from Wright et al. [42] by adding new constraints that provide an analysis of cross-utilization policies (i.e. how cross-utilization impacts scheduling cost and desirability), and a more general treatment of the method that nurses can specify undesirable shifts.