Market changes, labor shortages, and the introduction of a form of activity-based costing (diagnosis treatment combination (DTC) financing policy) in the Netherlands have provided an impetus for hospitals to reorganize care processes to improve efficiency. Establishing a form of capacity management with regard to agreed service levels of quality of care is essential to gaining insight into the balance between available and required resources, like staff and equipment. For inpatient care facilities in a hospital, this requires information on bed capacity and nursing staff capacity, on a daily as well as annual basis. Quantitative models can be used to calculate capacity needs for different planning purposes and for short, medium and long term planning issues. Although several useful models are described in the international literature [1–3] many of them are difficult to apply in practice because they require a great deal of data and clerical work [1].

To be able to apply capacity management in practice, models must fulfill different functions: "annual staff planning," "roster scheduling support," and "daily assignment of nurses to wards [4–6]." In addition, "strategic decisions" are sometimes mentioned as a separate planning level [3, 7, 8]. Models based on mathematical optimization techniques from operations research are generally focused on short-term scheduling [3, 9–11]. Models that do integrate different planning horizons (daily, periodical (1–2 months), and annual) are for example described by Abernathy et al.[7] and Wright et al.[11]. These models contain connected models for periodical staff planning and daily scheduling. However, models incorporating operational planning issues with tactical and strategic decisions or operational scheduling support with annual staff planning were not found in the literature. For accurate capacity management, operational planning issues (like the number of nurses on each shift) should be handled together with tactical planning issues (like annual nurse staffing).

Ideally, capacity management would support discussion between management and staff on all these issues and promote improvements.

Three cases will be used to demonstrate how this model can be applied to support issues at operational, tactical, and strategic levels.

A comprehensive model for capacity planning was developed. In many Dutch hospitals, as is the case in other countries, nurse staffing is historically-based. The model offers the opportunity to calculate staffing requirements for wards based, on recent ward-specific data. With the model, required capacity can be calculated on an annual basis and for each shift. Decision support for operational, tactical, and strategic levels is possible with the model: efficiency of working processes can be evaluated, wards can be compared, capacity needed can be compared with staff budgets, consequences of changing LOS and production targets (number of admissions) can be calculated, and ward sizes can be evaluated in terms of required nursing staff capacity. It appeared to be a useful tool for supporting discussions regarding capacity management issues between wards and hospital management by giving objective and quantitative insight into staff and bed requirements.

In three cases, it was shown that the model offered different ideas for the best way to improve capacity utilization. For pediatrics, opportunities for improvement were found at the operational level in the organization of working processes. With more efficient physician rounds and standardized admission and discharge procedures, the nurse-patient ratios agreed with the wards were attainable. This calculated annual nursing staff was adopted in the 2007 budget, and the existing inexplicable differences between wards were removed. For the cardiology department, the key to reducing workload was to reduce LOS, by keeping the same number of admissions. Finally, for two internal medicine wards, possible savings were calculated in nursing staff capacity by merging the wards, a strategic level decision. In all cases, the model was applied as part of a larger project.

The model goes beyond efficiency and cost reduction [17]. It offers insight into opportunities to improve working processes and reduce workload, which makes the nurses' work easier. Employing an adequate number of nurses is beneficial for patients and for nurses themselves [18, 19]. One of the strong points of the model is the use of historical utilization data rather than available beds. This means both annual production and fluctuation in utilization are taken into account and provides a more fair comparison between wards and between current and calculated numbers of staff. In the AMC, the hourly bed utilization data are regularly recorded and are easily accessible for use in the model. The model is also applicable with less detailed data like maximum or average bed utilization per shift. For the observation at the pediatric wards, structured observations were used. Limitation was that no inter-observer consistency was calculated.

The model has not the aim to replace integrated hospital and patient planning methods (see for example [20, 21]). Our model had the aim to support capacity management decisions on ward level, by calculating staffing needs for different planning levels. Apart form hospital wide planning, a lot of capacity improvements can be attained by carefully analyzing capacity utilization on single hospital units. With our model alternative possibilities for improvement can be determined by what-if analyses.

Daily staff scheduling is not incorporated in the model. For the scheduling of nurses, several solutions are available in literature (see for example [11]). Other models concern the short term adaptation of schedules (see for example [22, 23]). Our model adds a possibility to provide a quick insight in capacity needs and opportunities to improve capacity usage.

The model could potentially be criticized for not considering patient acuity. In the literature, there is some criticism of the application of nurse-patient ratios [24, 25]. However, alternative models are much more complicated and many aspects of patient acuity and organization of care can be incorporated into nurse-patient ratios. Differences in patient care needs between wards can be taken into account by adjusting the nurse-patient ratios. We used the US and Australian ratios as guideline in our discussions with department staff. However, in all case studies, we applied ward-specific ratios that take the Dutch situation and ward specific patient needs into account. For example, as can be seen in Table 3, we applied different ratios for each pediatric ward. These ratios were set by mutual agreement between hospital management and department staff of the pediatric wards. For the levels of analysis in this model, it is sufficient to apply an average nurse patient ratio per ward. This will be sufficient to calculate the number of nurses needed in each shift and yearly, and to perform the what-if analyses. To measure the acuity for patients on a specific ward for a period of time the AUKUH Acuity/Dependency tool developed for the NHS may be useful [26]. This tool supports the measurement of acuity and dependency across a range of wards and specialties. For detailed adaptation of daily nursing staff, daily measurement of patient acuity could be required, due to variety in admitted patients at a moment [22, 23, 27]. However, in our cases, we checked the acuity mix of patients to investigate the stability of the number of needed nurses in each shift. The patient mix for a sample of weekdays was analyzed for the pediatric and cardiology wards. Each day on each ward there appeared to be a balanced patient mix of mostly "average" patients, a few patients with lower care needs, and even fewer patients with relatively high care needs. Therefore, we are convinced that the calculated number of nurses for each shift is appropriate for safe and adequate patient care. When this is not the case, for example for very small wards, daily adjustment of staffing numbers may be needed. The smaller the number of patients, the larger the variety in number of patients as well as acuity mix will be [28]. These short term adaptation are, however, beyond the scope of this model, that is developed for planning purposes on operational, tactical and strategic level, and not for daily scheduling.

Nursing mix by training levels is not an explicit part of the model. The literature shows that deploying highly educated nurses is cost-effective compared with lower educated nurses [3, 29, 30]. Due to increasing patient complexity, there is an increasing need for Registered Nurses compared to lower educated staff [3, 17]. The model calculates the total number of nurses needed during each shift. In Dutch hospitals, the majority of those in nursing teams are certified nurses. It was assumed that in each shift a sufficient number of qualified nurses would be scheduled. The actual allocation of nurses, and therefore the skill-mix available in a shift is more a scheduling issue which is outside the scope of the model. A future model extension could be the incorporation of different skills in the staffing needs to overcome this limitation.

Two research areas were explored to extend research with the model. First, the application of flexible resource pools will be studied. A model option is to separate average capacity needed per shift and the extra capacity needed in peak periods. This peak capacity could possibly be shared by two or more related wards. Second, the model appears to be suitable for benchmarking resource utilization between comparable wards in different hospitals. Benchmarking projects were started with two other academic hospitals in the Netherlands.

A comprehensive model could be developed that covers both capacity planning for nursing staff and improving capacity utilization in hospital wards. The model was applied successfully in supporting capacity decisions on operational, tactical, and strategic levels. For two pediatric wards, improvements were made by improving working processes so that the agreed nurse-patient ratios could be attained. In a second case, for a cardiology ward, what-if analyses with the model showed that the workload could be substantially decreased by reducing LOS. The third case demonstrated possible savings in capacity by merging two small internal medicine wards. The model appeared to be useful in calculating capacity needed and in initiating organizational improvements, resulting in more efficient capacity utilization.