The NSM and NDM are built on a theoretical foundation supported by empirical research. Still, efforts to update and enhance both models faced numerous challenges—many due to data limitations. Below, we describe limitations of the two models and suggest areas for research that could address these limitations. Such research could improve the theoretical underpinnings of the models and improve the precision of key parameters in the model.

The NSM and the NDM are independent models. The NDM makes projections without considering the potential supply of nurses and vice versa. The future nurse workforce, in reality, will be influenced by the combination of supply and demand. A rising demand for nursing services at a time when supply is flat or falling will place upward pressures on nurse wages. This rise in wages would increase the number of new graduates, increase employment participation rates, and delay retirement for some nurses—all actions that will increase supply. Local shortages, on the other hand, could increase nurse wages locally contributing to local increases in the number of nurse graduates and an increase in the number of nurses migrating to that locality. Rising nurse wages will also place downward pressures on demand for nurses.

Both models use the SSRN to estimate the number of RNs employed in the base year. The NSM uses the 2000 SSRN to estimate supply of RNs by age, education level, and State. The NDM uses the 1996 SSRN to estimate number of FTE RNs by setting and State. Because the precision of estimates is proportional to sample size, the RN supply and employment estimates for the base year become less precise the smaller the unit of aggregation. Consequently, the base year starting values and projections for future years are less precise the smaller the unit of analysis. For example, estimates of demand for RNs in a particular setting within a State likely will be less precise than the State-level estimates, which in turn likely will be less precise than the national-level estimates.

One criticism of many attempts to model nurse demand is the limited consideration of important determinants of nursing demand (e.g., see Dumpe, Hermon, and Young [1998] and Prescott [2000]). Projections models such as the NDM and NSM are scaled-down versions of complex systems. Data and resource limitations prevented building models that include a wider array of determinants to better model the complexities of RN supply and demand. Consequently, many determinants of RN supply and demand are excluded from these models. Still, these models attempt to account for the major trends affecting RN supply and demand and project future supply and demand under a set of assumptions that constitutes an educated guess at whether current trends will continue.

Regarding the NDM, we use State-level data to estimate the relationship between demand for RNs and its determinants. One consequence of using State-level data is that relatively few degrees of freedom exist for estimating the regression equations. Future efforts might investigate the use of alternative approaches or lower levels of data aggregation to estimate the relationship between healthcare use and its determinants and between staffing intensity and its determinants.

Additional research could provide estimates of key parameters that improve the accuracy of the models and make the models more flexible policy tools. The NSM, for example, was built with the capacity to model the RN supply implications of changes in nurse wages, working conditions, tuition costs, and number of nursing school faculty. The empirical research has yet to be conducted to estimate the parameters necessary to use these features.

The NSM models only the supply of RNs and, unlike the NDM, fails to consider LPNs and nurse aides. The adequacy of the LPN supply holds implications for both the supply of and demand for RNs. On the demand side, employers have some ability to substitute between RNs and LPNs—taking into consideration legal and practical constraints. On the supply side, some LPNs seek further training to become RNs. Using the 2000 SSRN, we estimate that approximately 9.5 percent of the RN workforce, or 257,784 RNs, were employed as LPNs before starting their basic nurse education. The RN and LPN workforces are competing for the same candidates, many of whom could become either RNs or LPNs. Consequently, policies designed to recruit more RNs could have the unintended consequence of reducing the LPN supply.

Parts of both models are static. In the NSM, for example, the probability of cross-State migration is based on historical patterns that fail to consider the current shortage of RNs in each State. The NDM has limited ability to model substitution between types of nurses and between nurses and other healthcare workers. The NDM does model substitution between RNs and LPNs if their relative wages change, but future research might look at other ways to incorporate substitution effects. Similarly, the NDM has limited ability to capture the interaction of healthcare settings. For example, some settings might be viable substitutes (e.g., home health versus nursing facilities), while other settings might be complementary (e.g., increased use of outpatient services leading to increased use of home health services).

In summary, the NSM and NDM constitute powerful tools for projecting RN supply and demand under alternative sets of assumptions. The models help quantify the growing shortage of RNs as an aging population increases demand for nursing services at the same time an aging RN workforce and difficulties attracting new entrants to the nursing profession portend relatively little growth in the national RN supply.