III.
Nursing Demand Model
The
NDM projects Statelevel demand for FTE
RNs, LPNs and vocational nurses, and nurse
aides/auxiliaries and home health aides
(NA) through 2020. Moreover, the NDM projects
demand for RNs, the focus of this paper,
in 12 employment settings. Nurse demand
is defined as the number of FTE RNs whom
employers are willing to hire given population
needs, economic considerations, the healthcare
operating environment, and other factors.
Changing
demographics constitute a key determinant
of projected demand for FTE RNs in the
baseline scenario. The U.S. Census Bureau
projects a rapid increase in the elderly
population starting around 2010 when the
leading edge of the baby boom generation
approaches age 65 (Exhibit 15). Because
the elderly have much greater per capita
healthcare needs compared with the nonelderly,
the rapid growth in demand for nursing
services is especially pronounced for
longterm care settings that predominantly
provide care to the elderly.
Exhibit
15. Population Growth, 2000 to 2020
[D]
In
addition to Statelevel U.S. Census Bureau
projections of changing demographics,
the NDM projects nurse demand as a function
of changing patient acuity, economic factors,
and various characteristics of the healthcare
operating environment.
The
NDM (Exhibit 16), which combines input
databases and projection equations to
project demand, contains two major components:
(1) the data and equations to project
future demand for healthcare services
and (2) the data and equations to project
future nurse staffing intensity. It first
extrapolates expected use of healthcare
services by combining national healthcare
use patterns and State population projections
by age and gender. Then, the model adjusts
the healthcare use extrapolations for
each State to account for factors that
cause healthcare use to deviate from expected
levels (e.g., Statelevel variation in
managed care enrollment rates).
Exhibit
16: Overview of the Nursing Demand Model
[D]
The
model next projects nurse staffing intensity
(e.g., FTE RNs per hospital inpatient
days) as a function of current staffing
intensity and trends in major determinants
of nurse staffing intensity (e.g., average
patient acuity). Combining projected healthcare
use (e.g., inpatient days) with projected
nurse staffing intensity (e.g., FTE RNs
per inpatient day) produces projections
of demand for FTE RNs by setting, State,
and year. We describe the data, assumptions,
and methods used to estimate demand for
healthcare services and nurse staffing
intensity, and we present our findings.
A more complete description of the NDM
is available in other reports. ^{[5]}
A. Modeling
Demand for Healthcare Services
The
demand for nurses derives from the demand
for healthcare services. To accurately
project the demand for nurses, therefore,
one must first project the demand for
healthcare services. The NDM projects
demand for healthcare services for half
of the 12 employment settings in the NDM
(Exhibit 17). (For five settings, demand
for RNs is projected using RNperpopulation
ratios. Demand for nurse educators is
projected assuming that nurse educators
remain a fixed proportion of total RN
demand in each State). Measures of demand
for NDMprojected healthcare services
include inpatient days, outpatient visits,
and emergency visits to shortterm hospitals;
inpatient days at longterm hospitals
(e.g., psychiatric, rehabilitation, and
all other hospitals); nursing facility
residents; and home health visits.
Exhibit
17. Overview of the Nursing Demand Model
Setting 
Healthcare
Use Measure Projected 
Staffing
Intensity Measure Projected 
Shortterm
hospitals:
Inpatient
Outpatient
Emergency 
Inpatient
days
Outpatient
visits
Emergency
visits 
FTE
RNs/1,000 inpatient days
FTE
RNs/1,000 outpatient visits
FTE
RNs/1,000 emergency visits 
Longterm
hospitals 
Inpatient
days 
FTE
RNs/1,000 inpatient days 
Nursing
facilities 
Residents 
FTE
RNs/resident 
Physician
offices 
NA 
FTE
RNs/10,000 population 
Home
health 
Home
health visits 
FTE
RNs/1,000 home health visits 
Occupational
health 
NA 
FTE
RNs/10,000 population ages 18–64 
School
health 
NA 
FTE
RNs/10,000 population ages 5–17 
Public
health 
NA 
FTE
RNs/10,000 population 
Nurse
education 
NA 
FTE
RN educators/total FTE RNs 
Other
healthcare 
NA 
FTE
RNs/10,000 population 
The
NDM employs a twostep process to make
Statelevel projections of demand for
healthcare services for each of the six
settings modeled. Step 1 applies national
per capita use rates for 32 population
subgroups to U.S. Census Bureau population
projections for each State and year. ^{[6]}
The 32 population subgroups are defined
by eight age categories (ages 0–4, 5–17,
18–24, 25–44, 45–64, 65–74, 75–84, and
85 and older), gender, and metropolitan
or nonmetropolitan location.
Multiplying
each per capita use rate by its respective
Statelevel population projection creates
a Statelevel extrapolation of the expected
demand for healthcare services that controls
for differences across States and over
time in demographics. (Step 2 adjusts
these extrapolations based on trends in
the healthcare operating system and other
factors.)
The
following equation describes this step,
where EU_{S,H,Y} is the
expected level of healthcare use in State
S in healthcare setting H
in year Y. The variables P
and R are, respectively, the size
of the population in State S and
the national per capita healthcare use
for each age category (a), by gender
(s) and by metropolitan or nonmetropolitan
location (l). The first component
of this equation is a calibration factor
to ensure that base year estimates of
expected healthcare use equal estimates
of actual use. ^{[7]}
Step
2 adjusts up or down these initial extrapolations
of healthcare use in each State and year
based on projected changes in the healthcare
operating environment, economic considerations,
and other factors. The use adjustment
factor differs by State, year, and setting
and is calculated using projection equations
whose parameters describe the relationship
between healthcare use and exogenous variables.
[D]
We
estimated the parameters in the projection
equations (bs) (Exhibit 18) using multiple
regression analysis and a panel data set
consisting of Statelevel data for the
period 1996 to 2000. The dependent variable
in the regression equations, measures
the degree to which actual use (AU) of
healthcare services deviate from expected
use (EU) in a given State and setting
during the period included in the regression
analysis as described in Step 1.
[D]
The
actual regression equations contain the
logged form of the dependent and many
of the exogenous variables. Taking the
logged form of these variables has two
major advantages over the unlogged form.
One, using a logged form ensures that
the model will not project a negative
value of the dependent variable. Two,
the coefficients of logged exogenous variables
can be interpreted as elasticities that
represent the percentage change in the
dependent variable for each 1% change
in the exogenous variables (holding constant
the other variables in the model). Having
the coefficients in a common metric (e.g.,
elasticities) allows easier comparison
of the magnitude and precision of coefficients
between variables, across regression equations,
and with empirical findings in the literature.
The health maintenance organization (HMO)
variable and the region dummy variables
are the only variables not in log form.
Selection
of the exogenous variables employed in
the healthcare use regressions, as well
as those employed in the staffing intensity
regressions, was based on both theory
and empirical analysis. We considered
three criteria when determining which
variables to include in the regression
equations.
 Theorybased
model specification. A logical relationship
should exist between the exogenous variable
and the dependent variable. That is,
there should be a priori expectations
of the direction of the relationship
between the exogenous variable and the
dependent variable based on theory and
prior empirical evidence.
 Identification
of major determinants. We used stepwise
regression to identify factors that
exert a statistically significant effect
on either demand for healthcare services
or nurse staffing intensity. Stepwise
regression considers the pool of potential
exogenous variables—the pool consisted
of only exogenous variables that logically
would affect the dependent variable—and
adds or subtracts variables based on
the predictive power of each variable.
One result of using this approach is
that nearly all the exogenous variables
in the final regression equations are
statistically significant. Unfortunately,
another result of using stepwise regression
is that the statistical significance
of the regression equations and the
predictive power of the equation are
overstated.
 Reliable
extrapolations of future values.
We considered for inclusion in the final
regression equations only variables
whose future values can be extrapolated
with some degree of reliability or that
are important for policy modeling.
Several
factors complicated the selection of exogenous
variables in the regressions. First, in
a few cases an exogenous variable is not
statistically significant, though the
factor that this variable reflects is
presumed essential for developing a dynamic
model (e.g., the HMO variable in the equation
to estimate RN staffing patterns in hospital
inpatient settings). We had to determine
whether to include these variables with
low statistical significance. In a few
cases, variables deemed important that
had a level of statistical significance
between 0.05 and 0.2 were included in
the final regressions. The coefficients
on these variables are unbiased, despite
the lack of precision. We closely scrutinized
these coefficients and compared them with
other findings from this analysis and
from the literature to help ensure their
reasonableness.
A
second complication is that some of the
exogenous variables that theory suggests
are determinants of the dependent variable—and
thus should be considered for inclusion
in the equation—are correlated. For example,
HMO enrollment rate is correlated with
population density, and both HMO enrollment
rates and population density might affect
healthcare use and staffing intensity.
(An example of how population density
might affect nursestaffing patterns is
that healthcare providers in metropolitan
areas might benefit from economies of
scale that rural areas might not realize.)
Multicollinearity among the exogenous
variables means that their independent
effects might not be precisely estimated
even though the estimated effects are
unbiased. Also, the stepwise regression
approach might result in one variable
forcing a correlated variable from the
equation. Preliminary regressions were
estimated to test the robustness of the
regressions with respect to the inclusion
or exclusion of correlated variables,
and the results helped determine which
variables to include or exclude from the
final regression specifications.
A
third complicating factor is that some
regressions contain data from multiple
years, and observations from the same
State are not completely independent,
meaning some heteroskedasticity occurs
in the data. Heteroskedasticity can result
in underestimates of the coefficient standard
errors, which in turn overstates the statistical
significance of the coefficients. ^{[8]}
The
dependent and exogenous variables in the
equations are estimates based on hospital
census data and surveys of patients and
healthcare providers. The concern that
estimates for smaller States are less
precise than estimates for larger States
led to the decision to weight each observation
in the regression by the square root of
the State’s population.
Multiple
regression analysis provides estimates
of the relationship between healthcare
use and its determinants. Note that the
regressions predict the relationship between
healthcare use and its determinants after
adjusting for differences in the
demographic composition by age category,
gender, and urban or rural location.
Consistent
with other studies, this analysis finds
that HMOs decrease the number of inpatient
days at shortterm hospitals (Exhibit
18). The number of emergency department
visits and nursing facility residents
also decline as HMO enrollment rates rise.
The baseline scenario assumes a 0.5 percentage
point increase annually in enrollment
rates, which equates to a 10 percentage
point increase between 2000 and 2020.
^{[9]}
Consequently, the NDM projects that, in
2020, inpatient days at shortterm hospitals
will decline by 3 percent, emergency department
visits will decline by 2.8 percent, and
the number of nursing facility residents
will decline by 3.6 percent relative to
the levels that would exist if no change
in HMO enrollment rates occurred. Statelevel
estimates of HMO enrollment rates for
1996 through 2000 come from the Interstudy
Competitive Edge.
As
improvements in technology and cost pressures
shift more surgeries from an inpatient
to an outpatient setting, the number of
inpatient days at shortterm hospitals
will fall and the number of outpatient
visits and home health visits is expected
to rise. The baseline scenario assumes
that per capita inpatient surgeries will
decline by 2 percent annually from 2000
to 2020 and that these surgeries will
instead be performed on an outpatient
basis. For every 1 percent increase in
the proportion of hospitalbased surgeries
performed on an outpatient basis, the
regression findings suggest that inpatient
days will decline by 0.47 percent, outpatient
visits will increase by 1.66 percent,
and home health visits will increase by
0.86 percent. Statelevel estimates of
the proportion of hospital surgeries performed
on an outpatient basis were obtained from
American Hospital Association (AHA) annual
Hospital Statistics publications.
An
increase in the percentage of population
uninsured decreases demand for healthcare
services in longterm hospitals and nursing
facilities. The baseline scenario assumes
a modest decline in the percentage of
population uninsured due to changing demographics.
The variable was primarily included to
increase the NDM’s policy analysis capabilities.
A 1 percent increase in the proportion
of the population that is uninsured decreases
inpatient days at longterm hospitals
by 0.38 percent and decreases nursing
facility residents by 0.16 percent.
The
percentage of population enrolled in Medicaid
is positively correlated with higher use
of healthcare services in five settings.
Given that Medicaid enrollment is generally
associated with higher need for healthcare
services, access to medical services,
and lower income (which some studies have
found to be correlated with greater healthcare
needs), this positive relationship is
not surprising. The baseline scenario
assumes a modest change in the percentage
of population enrolled in Medicaid due
to changing demographics. A 1 percent
increase in the proportion of the population
enrolled in Medicaid increases demand
for inpatient days, outpatient visits,
and emergency department visits at shortterm
hospitals by 0.26 percent, 0.17 percent,
and 0.29 percent, respectively; increases
demand for inpatient days at longterm
hospitals by 0.26 percent; and increases
demand for home health services by 0.34
percent.
An
increase in the proportion of the population
that is nonwhite is associated with a
slight increase in the use of shortterm
hospital outpatient services and longterm
hospital inpatient days. An increase in
the proportion of the population that
is Hispanic is associated with a slight
decrease in emergency department visits.
These demographic variables might be capturing
differences across racial and ethnic groups
in healthcare needs, behavior that affects
healthcare use, or access to care via
insurance and local availability of services.
Population
density, as measured by percentage of
population living in an urban area, is
negatively correlated with use of inpatient
services at shortterm hospitals and nursing
facilities. The reader will recall that
the approach already controls for urban
or rural location of the States’ population
before estimating the regressions. Consequently,
these findings are difficult to interpret.
Population density is also correlated
with HMO enrollment rates. When the population
density variable is omitted from the shortterm
hospital inpatient day and nursing facility
regressions, the coefficients on the HMO
variable grow more negative.
The
inclusion of regional dummy variables
in the regressions improves the overall
fit of many of the equations and helps
estimate more precisely the relationship
between the dependent and exogenous variables
in the model. Over time, the values of
these dummy variables remain constant.
After controlling for differences in demographics
and the exogenous variables in the model,
the regressions show significant regional
variation in demand for healthcare services.
Exhibit
18. Healthcare Use Regression Results

ShortTerm
Hospitals 
LongTerm/
Psych/Other Hospital Inpatient
Days 
Nursing
Facility Residents 
Home
Health Visits 
Inpatient
Days 
Outpatient
Visits 
Emergency
Department Visits 
Intercept 
0.30^{a}
(0.127) 
1.39
(0.162) 
0.50
(0.080) 
0.24
(0.173) 
4.62
(1.151) 
0.85
(0.267) 
Healthcare
Operating Environment 
Percentage
of population in an HMO 
0.30
(0.105) 

0.28
(0.075) 

0.36
(0.138) 

Percentage
of hospitalbased surgeries performed
in an outpatient setting 
0.47
(0.143) 
1.66
(0.206) 



0.86
(0.345) 
Economic
Conditions 
Percentage
of population uninsured 



0.38
(0.069) 
0.16
(0.051) 

Percentage
of population Medicaid eligible 
0.26
(0.040) 
0.17
(0.054) 
0.29
(0.032) 
0.26
(0.073) 

0.34
(0.098) 
Per
capita personal income 




0.40
(0.116) 

Demographics 
Percent
of population nonwhite 

0.06
(0.023) 

0.27
(0.029) 


Percentage
of population Hispanic 


0.05
(0.008) 



Geographic
Location 
Percentage
of population in urban area 
0.25
(0.062) 



0.17
(0.089) 

EastNorthCentral
Region 



0.35
(0.054) 


EastSouthCentral
Region 
0.09
(0.038) 
0.25
(0.054) 



0.58
(0.095) 
MidAtlantic
Region 
0.24
(0.031) 
0.15
(0.045) 


0.35
(0.051) 
0.26
(0.077) 
Pacific
Region 
0.35
(0.033) 

0.17
(0.028) 
0.54
(0.057) 

0.56
(0.079) 
New
England Region 
0.19
(0.034) 

0.10
(0.030) 
0.30
(0.072) 
0.45
(0.055) 
0.79
(0.085) 
SouthAtlantic
Region 

0.26
(0.038) 




WestNorthCentral
Region 


0.16
(0.027) 



WestSouthCentral
Region 

0.17
(0.047) 



0.83
(0.080) 
Mountain
Region 
0.27
(0.031) 





Central
Regions 




0.39
(0.032) 

RSquared 
0.7659 
0.4679 
0.6299 
0.5559 
0.6061 
0.7125 
Years
Included in Regression 
1996–1999 
1996–1999 
1996–1999 
1996–1999 
1996–2000 
1996–1998 
^{a}
Regression coefficients with standard
errors in parentheses.
Note:
The projection method already controlled
for population age, gender, and urban
or rural location distribution before
estimating the regression equations. Also,
the use of stepwise regression to determine
which exogenous variables to include inflates
the statistical significance of the results.
Modeling
Nurse Staffing Intensity
Nurse
staffing intensity is defined as the number
of FTE RNs divided by some measure of
workload specific to the setting being
modeled (e.g., FTE RNs per 1,000 inpatient
days at shortterm hospitals). The NDM
calculates base year values of nurse staffing
intensity for each State and setting by
dividing estimates of RN employment by
estimates of healthcare use. Thus, in
nursing facilities, base year estimates
of employed FTE RNs per resident are used
as the staffing intensity measures.
We
use 1996 as the base year for several
reasons. First, the importance of the
SSRN in estimating baseyear RN supply
and demand limits the base year to a year
in which the SSRN was conducted (e.g.,
1992, 1996, 2000). Second, indications
that the nurse shortage has grown more
severe in recent years suggests that an
earlier year (e.g., 1996 versus 2000)
might produce nurse staffing intensity
estimates that reflect a market where
a relative equilibrium existed between
nurse supply and demand. We make one exception
to the argument that nurse employment
in a setting is the best measure of nurse
requirements. In hospitals, we estimate
that RN demand was approximately 7 percent
higher than RN employment in 1996. The
lowerthandemanded number of RNs employed
in hospitals reflects the rapid and significant
changes taking place in the hospital sector
during the early and middle 1990s when
hospitals were downsizing in response
to the rapid rise in managed care and
hospital consolidations. We arrive at
this 7 percent estimate by comparing RN
staffing intensity in hospitals using
SSRN and AHA data for 1992, 1996, and
2000.
After
establishing base year nurse staffing
intensity, the NDM then projects future
nurse staffing intensity. For four employment
settings, nurse staffing intensity is
measured as a nursetopopulation ratio
(because of data limitations) that is
assumed constant over time. Demand for
nurse educators is calculated as a constant
fraction of total demand for RNs. For
7 of the 12 employment settings modeled,
future nurse staffing intensity is projected
as a function of changes in exogenous
variables (X) such as average patient
acuity levels, economic considerations,
and characteristics of the healthcare
operating environment. The projection
formula is specified as
_{
} [D]
where
the parameters b represent the estimated
relationship between nurse staffing intensity
and its determinants and δ is an
adjustment factor so the base year projections
equal actual nurse staffing intensity
in the base year. We estimated the parameters
using multiple regression analysis with
Statelevel data from 1996 through 2000
(although most regression equations were
estimated using a subset of these years
based on data availability).
Both
theory and empirical analysis helped determine
the exogenous variables to employ in the
projection equations. As with the healthcare
use regressions, the dependent variable
and most of the exogenous variables enter
into the regression equation in a log
form. Also, we estimated the equations
using a stepwise regression that results
in a parsimonious model but that overstates
the significance statistics often used
to assess how robust the regression findings
are.
1.
Nurse Wages
The
ratio of RN to LPN wages is used to estimate
the degree to which employers substitute
lowercost LPNs for highercost RNs as
RN wages rise relative to LPN wages. ^{[10]}
In the baseline projections, we assume
that this ratio stays constant over time.
The regressions do not simultaneously
control for nursing supply, which could
bias the wage elasticities (e) towards
zero. The size of the estimated elasticities,
however, appears reasonable based on a
priori expectations and a comparison with
the literature. Demand for RNs is less
responsive to changing relative wages
in physicians’ offices (e=0.64) and inpatient
settings at shortterm hospitals (e=0.65)
compared with home health (e=1.06) and
longterm hospitals (e=1.20).
The
wages elasticity estimates from this analysis
are comparable to the few studies in recent
literature that report wage elasticities.
Lane and Gohmann (1995), in their analysis
of nurse shortages, estimate the wage
elasticity of nurse demand by simultaneously
estimating a supply and demand equation.
^{[11]}
The authors combine both RNs and LPNs
in their analyses. They estimate nurse
ownwage elasticity in shortterm hospitals
to be approximately 0.9.
Spetz
(1999) estimates a demand equation for
RNs using hospitallevel data for shortterm,
general hospitals in California during
the period 1976 to 1994. To control for
the endogeneity of nurse wages, Spetz
uses an instrumental variables approach
to estimate the RN demand curve, which
she compares to a demand curve estimated
using the ordinary least squares (OLS)
regression. As expected, her estimate
of wage elasticity from the OLS regression
(e=0.194) is less elastic than the estimate
obtained using the instrumental variables
approach (e=2.778) when she models the
daily services units of California hospitals.
Similarly, when she estimates demand equations
for the medicalsurgical units of California
hospitals, the wage elasticity estimates
are less elastic from the OLS regression
(e=0.342) than from the instrumental
variables regression (e=3.653). Spetz
also finds that an increase in LPN wages
is associated with a statistically significant
rise in RN employment in daily services
units of hospitals, but the converse is
untrue.
As
discussed previously in the context of
RN supply, the shortterm wage demand
elasticities are typically smaller than
longterm wage elasticities. In the short
term, employers might have few options
to replace RNs as they become relatively
more expensive. In the long term, employers
can change nurse staffing practices and
adopt new technologies that alter how
RNs are used.
2.
HMO Enrollment Rates
An
increase in HMO enrollment rates produces
mixed effects on staffing intensity. The
HMO variables in the regressions are not
logged, so the interpretation of the coefficients
is different from the other variables.
An increase in the HMO enrollment rate
by one percentage point increases RN staffing
intensity in shortterm hospital inpatient,
shortterm hospital outpatient, and home
health by 0.30 percent, 0.67 percent,
and 0.97 percent, respectively. An increase
in the HMO enrollment rate by one percentage
point decreases RN staffing intensity
in physician offices by 0.51 percent.
HMO
enrollment rates affect nursestaffing
patterns for two possible reasons. One,
HMOs decrease inpatient days in shortterm
hospitals through efforts at preventive
care and efforts to channel patients with
lesssevere problems to lessexpensive
settings. This reduction in inpatient
days might be raising the average acuity
level of patients admitted to the hospital,
which results in higher RN staffing per
1,000 inpatient days. Two, the efforts
of HMOs to reduce costs could contribute
to their adopting technologies or substituting
between different types of healthcare
professionals. As discussed previously,
HMO enrollment rates are correlated with
other variables such as percentage of
population in urban area. Consequently,
the coefficient on the HMO enrollment
rate variable could be capturing some
of the relationship between staffing intensity
and other factors correlated with HMO
enrollment rates. In both regressions
where HMO enrollment rate affects staffing
intensity, the variable percentage of
population in urban area is also included.
3.
Hospital Inpatient and Outpatient Surgeries
Changes
in technology can exert a mixed effect
on the demand for healthcare services
and staffing intensity. One measure used
in the NDM that reflects, in part, technological
advances is the percentage of hospitalbased
surgical procedures performed on an outpatient
basis. Improvements in technology and
medical procedures that shift some surgical
procedures from an inpatient to an outpatient
setting could affect nursestaffing intensity
in both inpatient and outpatient settings.
If patients with lesssevere health problems
are shifted from an inpatient to an outpatient
setting, then average patient acuity in
both settings could rise. This situation
could result in greater staffing intensity
per inpatient day and per outpatient visit
while decreasing overall nurse demand.
Each 1 percent increase in the proportion
of hospital surgeries performed in an
outpatient setting increases staffing
intensity for FTE RNs per 1,000 shortterm
hospital inpatient days by 0.64 percent.
As discussed previously, a 1 percent increase
in the proportion of hospitalbased surgeries
performed on an outpatient basis reduces
shortterm hospital inpatient days by
0.47 percent, increases outpatient visits
by 1.64 percent, and increases home health
visits by 1.86 percent. Surprisingly,
a 1 percent increase in this surgery variable
causes virtually no change in overall
demand for RNs—it just shifts where the
RNs are providing services.
4.
Healthcare Reimbursement Rates
A
rise in average Medicare and Medicaid
payments for services is associated with
greater staffing intensity. Part of this
increase might be due to greater patient
acuity, and part might be due to the ability
of healthcare providers to purchase nursing
services. A 1 percent increase in average
Medicare payments per home health visit
increases demand for RNs by 1 percent.
A 1 percent increase in average Medicaid
daily rates for nursing facilities increases
staffing intensity of RNs in nursing facilities
by 0.34 percent.
5.
Percentage of Population Uninsured
The
rate of uninsured in the population could
increase the level of uncompensated care
provided by healthcare providers. A 1
percent increase in the proportion of
the population that is uninsured decreases
RNs per 1,000 shortterm hospital inpatient
days by 0.37 percent and decreases RNs
per 1,000 visits to physician offices
by 0.21 percent. RN per 1,000 inpatient
days in longterm hospitals rises by 0.3
percent for each 1 percent increase in
the rate of uninsured, although the reason
for this positive relationship is not
readily surmised.
6.
Percentage of Population Medicaid Eligible
A
1 percent rise in the proportion of population
that is Medicaid eligible decreases RN
staffing per 1,000 emergency department
visits by 0.19 percent. As discussed in
the previous section, a 1 percent rise
in percentage of population that is Medicaid
eligible increases demand for emergency
department services by 0.29 percent, so
the net effect of a 1 percent rise in
this variable is to increase demand for
RNs in emergency departments by 0.05 percent.
7.
Per Capita Personal Income
As
the population grows wealthier, the demand
for higherquality healthcare services
likely will rise. A 1 percent rise in
per capita income increases RN staffing
intensity in physician offices by 0.33
percent.
8.
Patient Acuity Levels
A
population with greater healthcare needs
requires greater levels of services as
measured by both the quantity of services
provided and staffing intensity per unit
of service provided. The NDM contains
two measures that are proxies of population
health status: (1) population mean age
and (2) average number of activities of
daily living (ADL) limitations of nursing
facility residents. (In addition, the
Medicare and Medicaid reimbursement rate
variables discussed previously might also
be capturing variation in average patient
acuity across States and over time.) A
1 percent increase in population mean
age increases RN staffing intensity in
physician offices by 1.52 percent. A 1
percent increase in average number of
ADL limitations of nursing facility residents
increases demand for RNs per nursing facility
resident by 0.63 percent.
9.
Geographic Location
The
percentage of population living in urban
areas exerts a mixed impact on nurse staffing
intensity. A 1 percent increase in this
variable decreases RN staffing per 1,000
inpatient days at longterm hospitals
by 0.60 percent. In shortterm hospitals,
a 1 percent increase in this variable
increases RN staffing intensity in inpatient
settings and outpatient settings by 0.16
percent and 0.39 percent, respectively.
As discussed previously, this variable
is correlated with HMO enrollment rates;
consequently, the precise relationship
among HMO enrollment rate, percentage
of population living in urban areas, and
nurse staffing intensity is unclear. Significant
regional variation occurs in nurse staffing
intensity, but few visible patterns emerge
in the findings (Exhibit 19). Changes
in staffing intensity will vary by State
depending on the projected values for
exogenous variables and changing demographics.
Between
2000 and 2020, staffing intensity is projected
to increase 34 percent in home health,
from approximately 2.8 FTE RNs per 1,000
home health visits to approximately 3.8
FTE RNs per 1,000 visits (Exhibit 20).
In shortterm hospital inpatient settings,
FTE RNs per 1,000 inpatient days is projected
to increase by 18 percent at the national
level (from 6.5 to 7.7). For nursing facilities
and physician offices, we project a 13
percent increase in staffing intensity,
while for shortterm hospital outpatient
settings we project a 6 percent increase
in staffing intensity. In shortterm hospital
emergency settings and in longterm hospitals,
we project virtually no change in staffing
intensity. The staffing intensity measures
for RNs in occupational health, school
health, public health, nurse education,
and “other” healthcare settings is assumed
constant over time at their 1996 levels.
To fully comprehend the magnitude of additional
FTE RNs required, the overall impact of
staffing intensity must be considered
in conjunction with healthcare use projections.
Exhibit
19. Nurse Staffing Intensity Regressions

ShortTerm
Hospitals 
LongTerm
Hospitals 
Nursing
Facilities 
Home
Health 
Physician
Offices 

Inpatient 
Outpatient 
ED 




Intercept 
1.62^{a} 
1.7 
0.53 
2.69 
5.15 
5.16 
7.13 
(0.247) 
(0.122) 
(0.177) 
(0.462) 
(0.922) 
(0.787) 
(3.593) 
Healthcare
Operating Environment 
Ratio
of RN to LPN hourly wage 
0.65 


1.20 

1.06 
0.64 
(0.258) 


(0.671) 

(0.537) 
(0.391) 
Percentage
of population in an HMO (variable
not logged) 
0.30 
0.67 



0.97 
0.51 
(0.202) 
(0.389) 



(0.316) 
(0.230) 
Percentage
of hospital surgeries performed
in outpatient setting 
0.64 






(0.255) 






Average
Medicare payment per home health
visit 





1.00 






(0.198) 

Average
Medicaid NF daily rate 




0.34 






(0.153) 


Economic
Conditions 
Percentage
of population uninsured 
0.37 


0.30 


0.21 
(0.069) 


(0.147) 


(0.091) 
Percentage
of population Medicaid eligible 


0.19 

0.19 




(0.091) 

(0.103) 


Per
capita personal income 






0.33 






(0.202) 
Population
Health/Patient Acuity 
Population
mean age 






1.52 






(0.761) 
Average
number of ADL limitations of nursing
facility residents 




0.63 






(0.444) 


Geographic
Location 
Percentage
of population in urban area 
0.16 
0.39 

0.60 



(0.114) 
(0.201) 

(0.206) 



EastSouthCentral
region 
0.11 



0.5 
0.22 

(0.066) 



(0.098) 
(0.139) 

EastNorthCentral
region 
0.23 






(0.054) 






MidAtlantic
region 
0.34 

0.15 
0.43 

0.23 

(0.057) 

(0.077) 
(0.138) 

(0.119) 

SouthAtlantic
region 




0.24 






(0.067) 


New
England region 



0.41 






(0.166) 



WestSouthCentral
region 

0.19 


0.91 
0.62 


(0.111) 


(0.091) 
(0.123) 

Western
regions 
0.20 
0.40 

0.26 


0.16 
(0.045) 
(0.076) 

(0.103) 


(0.072) 
Coastal
regions 

0.40 






(0.076) 





Rsquared 
0.7988 
0.4544 
0.1365 
0.5217 
0.5664 
0.7121 
0.3869 
Years
included in regression 
1996 
1996 
1996 
1996 
1996,
1999, 2000 
1996 
1996 
^{a}
Regression coefficients with standard
errors in parentheses.
Exhibit
20. National Measures of Projected Nurse
Staffing Intensity


Baseline 
Projected 
Setting 
Staffing
Intensity Measure 
1996 
2000 
2005 
2010 
2015 
2020 
Increase
from 2000–2020 
Shortterm
hospitals:
Inpatient
Outpatient
Emergency 
FTE
RNs/1,000 inpatient days
FTE
RNs/1,000 outpatient visits
FTE
RNs/1,000 emergency visits 
6.16
0.18
0.93 
6.54
0.19
0.94 
6.81
0.19
0.94 
7.12
0.19
0.94 
7.42
0.20
0.95 
7.69
0.20
0.94 
18%
6%
0% 
Longterm
hospitals 
FTE
RNs/1,000 inpatient days 
5.31 
5.25 
5.28 
5.29 
5.28 
5.27 
0% 
Nursing
facilities 
FTE
RNs/resident 
0.09 
0.10 
0.10 
0.11 
0.11 
0.11 
13% 
Physician
offices 
FTE
RNs/10,000 population 
5.50 
5.51 
5.69 
5.88 
6.04 
6.20 
13% 
Home
health 
FTE
RNs/1,000 home health visits 
2.59 
2.87 
3.08 
3.31 
3.57 
3.84 
34% 
Occupational
health 
FTE
RNs/10,000 population ages 18–64 
Constant
at 1996 levels 
School
health 
FTE
RNs/10,000 population ages 5–17 
Constant
at 1996 levels 
Public
health 
FTE
RNs/10,000 population 
Constant
at 1996 levels 
Nurse
education 
FTE
RN educators/total
FTE RN demand 
Constant
at 1996 levels 
Other
healthcare 
FTE
RNs/10,000 population 
Constant
at 1996 levels 
B.
Nursing Demand Projections
Below,
we present projections from the NDM. We
present projections for alternative scenarios
that use different assumptions about the
trends in the major demand determinants.
1.
Baseline Projections
Under
the baseline scenario, demand for FTE
RNs is projected to increase 41 percent
between 2000 and 2020 at the national
level (Exhibit 21). As shown in the appendix,
the projected change in demand varies
substantially by State. In percentage
terms, the fastest growth will occur in
settings that predominantly serve the
elderly (e.g., home health and nursing
facilities) and in hospital outpatient
settings.
Exhibit
21. Baseline Projections of Demand for
FTE RNs
Setting 
2000 
2005 
2010 
2015 
2020 
Increase
from
2000–2020 
Total^{
a} 
2,001,500 
2,161,300 
2,347,100 
2,569,800 
2,824,900 
41% 
Hospitals^{
a} 
1,239,500 
1,324,800 
1,427,900 
1,555,600 
1,698,900 
37% 
Shortterm
hospital, inpatient 
874,700 
930,200 
999,100 
1,086,800 
1,187,000 
36% 
Shortterm
hospital, outpatient 
83,500 
95,900 
110,400 
126,400 
142,000 
70% 
Shortterm
hospital, emergency 
90,300 
92,200 
94,500 
97,300 
100,400 
11% 
Longterm
hospitals 
191,000 
206,500 
223,900 
245,100 
269,400 
41% 
Nursing
facilities 
172,800 
197,200 
224,500 
253,600 
287,300 
66% 
Physician
offices 
155,000 
166,400 
178,800 
191,600 
204,700 
32% 
Home
health 
132,000 
157,300 
187,500 
226,200 
275,600 
109% 
Occupational
health 
20,200 
21,000 
22,000 
23,100 
23,900 
18% 
School
health 
57,600 
59,700 
60,400 
61,100 
62,200 
8% 
Public
health 
99,800 
103,500 
107,300 
111,500 
115,800 
16% 
Nurse
education 
45,900 
49,600 
53,800 
58,800 
64,500 
41% 
Other
healthcare 
78,500 
81,700 
84,900 
88,400 
92,000 
17% 
^{a}
Due to rounding, category totals might
fail to equal the sum across component
settings.
2.
Alternative Scenarios
Nurse
demand will be determined, in part, by
political decisions, changes in technology,
changes in the healthcare operating environment,
and changes in other factors difficult
to predict. In addition, projection models
such as the NDM are relatively simplistic
simulations of a complex healthcare system
that try to capture the major trends affecting
demand for nurses, so the RN demand projections
are made with some level of imprecision.
The degree of imprecision is difficult
to determine. A sensitivity analysis shows
how the projections change as we change
key assumptions in the model. We present
projections under four alternative scenarios
(Exhibit 22):
(1) Scenario 1 assumes no changes in managed
care enrollment rates (compared to the
baseline that assumes an annual 0.5 percentage
point increase). At the national level
across all settings, this modest change
in the growth rate of managed care enrollment
has virtually no effect on demand for
RNs. However, substantial changes occur
at the setting level. Managed care growth
simply shifts care from inpatient to outpatient
settings, and the decline in projected
inpatient days is offset by a likely increase
in staffing intensity as the average level
of patient acuity increases.
(2) Scenario 2 assumes that RN wages increase
1 percent annually compared to LPN wages.
(The baseline assumes that RN and LPN
wages grow at the same rate.) Under this
scenario, a rise in RN wages gives employers
greater financial incentive to substitute
lowercost LPNs for highercost RNs, where
possible. Between 2000 and 2020, the compounding
effect of a 1 percent annual growth in
relative wages for RNs results in a real
increase of 22 percent relative to wages
of LPNs. By 2020, demand for FTE RNs would
be approximately 10 percent lower (or
285,000 FTE RNs) relative to the baseline.
(3) Scenario 3 assumes that the U.S. population
grows 20 percent faster than projected
by the U.S. Census Bureau. By 2020, this
accelerated growth results in demand for
88,000 additional FTE RNs (or 3 percent)
relative to the baseline.
(4) Scenario 4 assumes that the U.S. population
grows 20 percent slower than projected
by the U.S. Census Bureau. By 2020, this
decelerated growth results in the demand
for 85,000 fewer FTE RNs (or 3 percent)
relative to the baseline.
[D]
