IV. Nursing Model Components

Published 2022

In this module:

This module describes the data, methods, and assumptions used in the Health Workforce Simulation Model (HWSM) to model supply and demand for registered nurses (RNs) and licensed practical/vocational nurses (LPNs). The latest year for which reliable supply data are available is the “base year.” The base year is currently 2020. The period from the base year through the last year for which projections are made is the “projection period.” The projection period is 2020-2035. The data and methods presented here update those in HRSA’s previous nursing workforce report published in 2017.1

Modeling Supply

The microsimulation approach to modeling supply used in HWSM begins by creating a database with information on each nurse in the base year supply of nurses. Moving through the projection period is simulated by:

  • adding new entrants to the nursing workforce each year,
  • removing those who leave the RN/LPN workforce during the year,
  • adjusting hours worked based on age of nurses in the new year, and
  • adjusting for nurses moving between states during the year.

Details of the supply modeling process are found in the Supply Modeling Overview module. The data, methods, and assumptions specific to nurses are found in the following subsections.

Estimating Base Year Workforce Supply

State-level estimates of RN supply in 2020 start with National Council of State Boards of Nursing (NCSBN) counts of active RN licenses.2 The number of active licenses overstates available nurse supply in a state for several reasons.

  1. Some nurses have a license as both an RN and an advanced practice registered nurse (APRN). To avoid double counting, these nurses are modeled as APRNs. (A separate module describes modeling of APRN supply and demand.)
  2. Some nurses may have retired from practice, yet their license remains active until it expires.
  3. Some nurses are licensed in multiple states.
  4. Similarly, some nurses may be working in states other than the state in which they are licensed under the Nurse Licensure Compact (NLC). The NLC allows nurses to work in other compact nursing states without additional licensing.
  5. A separate issue, discussed later, is that many licensed nurses choose not to be active in the workforce.

To account for differences between active license counts from NCSBN and available RN supply, we adjusted the NCSBN state-level estimates of active licenses by the ratio of final to initial sample frame used for the 2018 National Sample Survey of Registered Nurses (NSSRN). When the Census Bureau conducted the NSSRN, licenses were removed from the survey sample frame:

  • for RNs who were APRNs,
  • to leave a single observation when RNs were licensed in more than one state, and
  • for other reasons.3

Nurses working in compact nursing states are required to be licensed in the state they reside and not the state they practice. However, NSSRN data includes a nurse’s state of practice. Thus, generally, we use the NSSRN data instead of licensing data to estimate the number of active and available nurses in each state for the model.

The base year supply of nurses is sampled from the NSSRN using numbers derived from NCSBN for each state that approximates the states’ nursing workforce on characteristics such as age, sex, highest education level, and employment status. Employment status indicates whether a nurse is (1) employed or actively looking for employment, or (2) not employed and not looking for employment. Each states’ sample is built by drawing a random sample, the size of the estimated number of available nurses in the state, from the NSSRN data described above. For example, if NCSBN indicates 100,000 RNs are licensed in a particular state then we draw 100,000 records (with replacement) from the NSSRN respondents from that state. Each nurse in the NSSRN has a sample weight that reflects the amount of nurses in the state with whom they share age, sex, education level, and employment status characteristics. A nurse with popular characteristics might have double the sample weight of a nurse with a unique set of characteristics. The nurse with double the sample weight has double the probability of being drawn for the sample. Nurses drawn from the NSSRN sample frame are included in the starting supply sample then placed back in the NSSRN data. Thus, nurses in the NSSRN data may be chosen as many times as they are selected at random. The sample size of nurses in the NSSRN from RI, VT, ND, SD, MT, WY, AK, and HI was too small for the Census Bureau to identify their state in the NSSRN public use file. Therefore, the samples of RNs drawn from NSSRN for these states were drawn from nurses with unknown states but in their states’ Census Division. For example, nurses sampled for the Vermont HWSM starting supply came from either Vermont or Rhode Island, the two states in the New England Census Division for which state name was not available in the NSSRN data. Nurses in a state’s starting supply sample with the characteristic “employed” were drawn from all nurses in NSSRN employed in that state. However, nurses with the characteristic “unemployed” were drawn from all nurses in NSSRN living in that state.

Data to create the starting supply samples of LPNs by state come from the 2020 Occupational Employment and Wage Statistics (OEWS) survey and 2016-2020 American Community Survey (ACS). State-level estimates of the size of LPN supply in 2020 come from the OEWS. The ACS sample size in the 2020 file was too small to accurately estimate the size of the LPN supply by state. The approach described above for RNs also was used to create a starting sample of LPNs in each state by sampling the state’s LPNs in the 2016-2020 ACS data. We used ACS sample weights in the sampling process to match the characteristics of LPNs in the starting sample to those of LPNs in the state. The state supply samples include each LPN’s race/ethnicity, sex, age, and employment status.

For five states (Ohio, Oklahoma, Oregon, New Jersey, and Texas) the modeling team gained access to licensing data with RN and LPN counts and demographics after conducting an outreach campaign to relevant entities in each state. However, we did not use the counts from this data to determine starting supply. Using licensure data could place some nurses in nurse compact states in their home state even though they work in a different state, which would create less accurate estimates of nurse supply by state. However, we did use the licensure data to sample both RNs and LPNs into the starting supply instead of the NSSRN and ACS data for these states. This allows the nurses’ characteristics in each of these states to come from a database of all the state’s nurses instead of surveys that only contact about 1% of nurses in the state.

Not all nurses with an active license are active in nursing. HWSM contains prediction equations, estimated using the NSSRN (for RNs) and the ACS (for LPNs), that return the probability that nurses are employed in a nursing position or are actively seeking employment as a nurse. Nurses not in the workforce are still included in the starting supply microsimulation file but not counted as part of the starting year supply total. As described later, nurses that are inactive in the starting year have a chance to become active again in later years.

Modeling New Entrants

New entrants reflect the number of nurses entering the workforce for the first time upon completion of a nursing program and certification. These graduates must pass the National Council Licensure Examination (NCLEX) to practice as a nurse. The NCSBN administers the NCLEX and reports the number of RNs who are first time takers and the number who passed on the first try by nurse education level:

  • Bachelor of Science in Nursing (BSN): an undergraduate-level degree from an accredited college or university
  • Associate Degree in Nursing (ADN): a 2-year degree from an accredited college or technical program
  • Diploma: typically, a hospital-based nursing school requiring 2-3 years of training

NCSBN also reports the number of people taking and passing the National Council Licensure Examination for Practical Nurses (NCLEX-PN) exam to become an LPN.

For RNs who fail the exam on the first try (often around 10-15% depending on education level) and retake the exam, information is not provided on their education level. Hence, relying on annual pass numbers (rather than calculating eventual passing rates) would not provide exact number of RNs entering the workforce by education level. Education level is important in HWSM and is correlated with propensity to leave RN work to become an APRN, hours worked patterns, and geographic mobility patterns.

Modeling the annual number of new entrants starts with state-level numbers of first time, U.S. educated candidates taking the NCLEX and the pass rates by education level. In 2020, there were 177,394 first-time, U.S.-educated takers of the NCLEX-RN across the 50 states and District of Columbia.4 Of these, 88,635 RNs had completed a baccalaureate degree. Another 86,508 RNs had completed a diploma, associate degree, or special program. (Because of small numbers for diploma and special program graduates, we combine them with the associate degree graduates for modeling). We assume that nurses who initially fail the NCLEX will retake the test up to two more times and as such calculated initial state-level estimates of the eventual NCLEX passers, by education level. We compare the initial state-level estimates of passers by education level with NCSBM reports of aggregate numbers and scale our estimates by education level to equal the reported total. For LPNs, we use reported numbers of total NCLEX-PN passers in each state.

Pass rates for the initial and subsequent tests differ by state and by nurse education level. At the national level, this modeling assumption gives an eventual pass rate of:

  • 98.8% of RNs trained at the baccalaureate level
  • 96.9% of RNs trained at the associate or diploma level
  • 93.0% of LPNs

Historically HWSM has excluded data of international students who pass NCLEX when calculating new entrants. The primary reason is the lack of data on how many of these students actually practice in the U.S. The number of internationally educated RNs who passed the NCLEX-RN decreased from 14,931 in 2019 to 9,700 in 2020 (Exhibit IV-1). About half of these RNs came from the Philippines. Other geographic areas with larger numbers of NCLEX-RN passers include India, Puerto Rico, Nigeria, Kenya, Canada, Nepal, and South Korea. The number of internationally-trained LPNs passing the NCLEX-PN is relatively small (245 in 2019 and 485 in 2020), representing less than 1% of new LPNs trained.

Exhibit IV-1 summarizes the number of internationally-trained RNs who passed the NCLEX-RN and the proportion of passers who are U.S.-trained over the past two decades. Available data over the past 10 years (2011-2020), indicate that 95% of NCLEX-RN passers are U.S.- trained and the number of internationally-trained passers averaged 8,063 RNs per year.

The Status Quo supply scenario models that annually the number of nurses eventually passing the NCLEX is 92,456 RNs (baccalaureate level), 83,979 RNs (associate or diploma level), and 42,444 LPNs. Our analysis of the 2018 NSSRN indicates that about 16,000 LPNs will further their education and become RNs each year. The counts of nurses taking the NCLEX-RN include the 16,000 LPNs. They are also included in the attrition from the LPN workforce, as described later.

Alternative supply scenarios modeled include training 10% more or 10% fewer nurses relative to current numbers. These scenarios illustrate the sensitivity of supply projections to the number of nurses being educated each year and the uncertainty that the annual number of graduates passing the NCLEX will change over the projection horizon.

Demographic information for the new entrants to the workforce came from the 2018 NSSRN for RNs and the 2016-2020 ACS for LPNs. Due to concerns regarding the race and ethnicity estimates of the NSSRN, we compared the new nurse race and ethnicity distributions in the model to published estimates from the American Association of Colleges of Nursing (AACN).5 The RN subset consists of those nurses who received their degree in nursing in 2000 or later. The LPN subset contains only those LPNs under age 35 for the sex and race distribution, as graduation year is unavailable in ACS. Exhibit IV-2 summarizes the demographic characteristics of new nurses.

Modeling Workforce Attrition

In this section, we describe analyses and assumptions regarding nurses who permanently leave working as an RN or LPN. A permanent departure from the nursing workforce differs from a temporary departure (discussed later) such as for child rearing, illness, or other reasons where the nurse might return to employment. We model three types of nurse workforce attrition: (1) Nurses under age 50 who leave the workforce, often to change occupations (2) Nurses aged 50 and older who leave the workforce, presumably retiring and (3) Nurses who transition from LPN to RN, or from RN to APRN.

Attrition of Nurses Under Age 50

As part of our analysis of whether using NCLEX and NCLEX-PN pass statistics provide a good estimate of the number of new nurses entering the workforce, we conducted analyzes comparing number of nurses entering in each year cohort to the number of nurses still working in each cohort. This analysis explored whether some nurses who pass the exam might decide not to enter the workforce as a nurse, or might leave nursing shortly after entering. For other occupations modeled in HWSM, such as behavioral health counselors, there appear to be a substantial number of people trained who (a) never make it into the workforce in which they were trained, or (b) leave the occupation within a short time of entering the workforce. This observation comes from comparing year-over-year growth in published statistics such as the OEWS or ACS. In this research we explored the issue for nurses.

Of the three main data sources that offer a historical count of nurse employment (OEWS, ACS, and NSSRN), we focused on the ACS and the NSSRN. Each data source has limitations for estimating how many newly trained nurses remain in the workforce in subsequent years. The OEWS counts nurse employment, but a substantial number of nurses work part time or for multiple employers. Because OEWS counts positions rather than unique nurses, this data source was not used. The ACS includes RNs, who can be identified by education level, as well as LPNs. However, ACS does not ask respondents when they received their nursing degrees, so one cannot identify recent entrants to the workforce. The NSSRN does ask respondents when they received their first nursing degree but does not survey LPNs.

The NSSRN data regarding when RNs received their first nursing degree is important because nurses who complete additional training to become APRNs need to be included in this analysis for comparison to NCLEX-RN annual numbers, just as they are included in the model. The goal here is to track nurses at the point of entry into the nursing workforce. The NSSRN includes first licensing year for nurses from 1963 to 2014, with all years before 1963 grouped together and all years after 2014 grouped together. Using the NSSRN weights, we created a table of licensed RNs by first license year for their initial degree type for the years 2000 to 2014 (Exhibit IV‑3). For each year, we compared the number of active nurses of each education level in 2018 from the NSSRN with estimates of the number of nurses who passed the NCLEX exam during that year.

The expectation was that the difference between number of nurses in NSSRN who reported receiving their first RN license in a particular year and the number passing the NCLEX-RN in that year would show the proportion of nurses who left the workforce between year of graduation and 2018. Any attrition would likely not be due to traditional retirement because most new nurses are unlikely to reach traditional retirement age within 4-18 years of entering the workforce. (Presumably, a nurse who completes a nursing program in his/her 40s might retire within that timeframe). Additionally, any difference between NCLEX-RN exam completers and nurses in the NSSRN should not be caused by RNs becoming APRNs because the first license of APRNs is also tracked in the NSSRN.

As shown in Exhibit IV‑3, we did not find the expected increase in the ratio of original NCLEX completers to weighted 2018 NSSRN participants as nurse education year becomes more recent for BSNs, though we do observe a small trend for ADN and Diploma nurses. This suggests that much of the attrition that happens for nurses under age 50 occurs in the first years of their career. For BSNs, the number in the NSSRN who report graduating in a particular year often exceeds the estimated number of NCLEX passers in that year. This finding could reflect potential overestimation of the NSSRN sample as RNs holding multiple state licenses may have residual duplicates in forming the weights. There might be slight differences in the year in which RNs completed their nursing program and the year they completed the NCLEX. Given that the NSSRN figures are estimates based on weighted survey responses, we suspect that the NSSRN weights are overestimating the number of nurses at the baccalaureate level. This difference is only 8,000 nurses in the most recent analysis.

There were fewer ADNs in the NSSRN weighted respondents than NCLEX exam completers for all but one (2009) graduation year. Recognizing that some ADNs transition to BSNs, and that NSSRN responders could have misinterpreted questions about initial education level, we combined the BSN, ADN, and Diploma nurses. For each graduation year, we divided the percent difference between weighted NSSRN respondents in 2018 and NCLEX completers for the specified year. We then divided that result by the number of years between the graduation year and 2018 to obtain an annual attrition rate. Across the years analyzed, the average annual attrition rate is 0.552%. When applied to new entrants in 2020 (151,334 who pass the NCLEX), this equates to 830-840 RNs leaving the workforce annually from an entering cohort. This number is dwarfed by the approximately 3,600 RNs (and 2,100 LPNs) that graduate from a nursing program but never join the nurse workforce because they failed to pass the NCLEX despite repeated attempts. Still, over 10 years, this equates to approximately 8,350 RNs leaving the workforce from the 2020 graduation cohort.

Exhibit IV‑4 summarizes a similar analysis for internationally-trained RNs who take and pass the NCLEX. We did not conduct a year-to-year analysis because of the small sample size of internationally-trained RNs in the NSSRN. However, of the 150,889 internationally-trained RNs who passed the NCLEX between 2001 and 2014, only 91,254 (60%) appear to be represented in the NSSRN compared to 95% of U.S.-trained RNs during this period who passed the NCLEX and also appear in the NSSRN. HWSM does not separately track whether RNs are U.S.- or internationally-trained. To incorporate internationally-trained RNs into the pipeline numbers, we apply the 0.552% annual attrition rate calculated based on U.S.-trained RNs but assume that only 65% of international-trained NCLEX-RN passers actually enter the workforce in the 50 states + District of Columbia. This combination of 65% of NCLEX-RN passers entering the workforce and 0.552% annual attrition would equate to 60% of NCLEX-RN passers from 2001 to 2014 remaining in the U.S. workforce (as suggested by comparison to NSSRN numbers). Applying this 65% estimate to the number of internationally-trained NCLEX-RN passers in recent years years (Exhibit IV‑4) suggests approximately 5,873 additional RNs entering the U.S. workforce in 2020 down from 9,040 in 2019.

In addition to the NSSRN analysis, we performed similar analysis using the 2015-2019 ACS Public Use Microdata Sample. (We excluded the 2020 ACS from this analysis as it is unclear whether the COVID-19 pandemic affected data collection and/or nurse labor force decisions). A person is identified as a nurse if the person is or has been a nurse within the past 5 years and has not changed to a non-nursing occupation, regardless of their current active or inactive status. If an RN is in a teaching role in academia, it is unclear whether that nurse will self-report as a nurse or as an educator for their occupation. Likewise, if an RN is in a management role, it is unclear whether that nurse will self-report as a nurse or as an executive/manager. For modeling, we assume that nurses in non-patient care roles are still part of the nursing workforce, as nurse educators and nurse managers presumably rely on their nursing experience in the performance of their role.

ACS does not include questions about licensure status nor graduation year. Therefore, we used multiple ACS years to determine the change in the number of BSNs, ADNs, and LPNs between years. If all nurses remained in the workforce and no nurses entered, we would expect to see, for example, the same number of 31-year-old LPNs in 2019 as we saw 30-year-old LPNs in 2018. By comparing across all ages between 20 and 49 for the time periods of 2015-2016, 2016-2017, 2017-2018, and 2018-2019 for all three nurse types, we can determine the number of nurses entering the workforce. After accounting for new entrants and nurse career progression we arrive at a residual attrition rate. The annual attrition rates we found for nurses under age 50 were 2.9% for BSNs, 4.6% for ADNs, and 6.4% for LPNs. Data limitations in ACS likely overstate calculated attrition rates. For example, we cannot identify nurses who stopped working as a nurse, but still retained their nursing license. Hence, the above rates cannot be applied to estimates of licensed nurses. Because of these data limitations with ACS, we relied on the NSSRN analysis described previously to model RN attrition rather than the ACS analysis. For LPNs, we used the RN ratio of the NSSRN-estimated attrition rate to the ACS-estimated attrition rate and multiplied this by the LPN ACS-estimated attrition rate. This equates to 1% of LPNs under age 50 permanently leaving nursing each year (in addition to LPNs becoming RNs and LPNs age 50 and older retiring).

Attrition of Nurses Aged 50 or Older

For nurses aged 50 and older, we model retirement intention which includes nurses intending to leave the workforce for chronic illness to care for family members and for other reasons. Attrition does not account for unexpected deaths or other unexpected reasons for leaving the workforce. There is insufficient data to model mortality rates for nurses. Using national average mortality will likely overstate mortality rates for nurses and is presumably correlated with nurses’ stated intentions to retire.

Multiple approaches have been explored and used to estimate nurse attrition patterns. Prior to 2016, we used ACS-derived attrition rates by age and sex for RNs age 50 and younger. A challenge with ACS data is that ACS does not collect occupation data if a respondent has been out of the workforce for five or more years. However, if a trained nurse respondent remains in the workforce but changes to a non-nursing occupation, their occupation will indicate the current occupation instead of nurse. The approach to model attrition patterns changed in 2016 to use refined estimates of nurse attrition patterns based on licensure data from Oregon, South Carolina, and Texas. In this latest 2022 update, the attrition patterns are based on the most recent data available—which is the 2018 NSSRN for RNs and the 2016-2020 ACS for LPNs.

Retirement for RNs is based on a question in the 2018 NSSRN that asks nurses when they plan to retire. Those who say that they plan to retire within a year are considered as retiring at their current age. RNs who say they plan to retire in 1-2 years are considered as retiring at the age they would be in 1-2 years and with half the weight due to the two-year time span covered. Finally, RNs who say they plan to retire in 3-5 years are considered as retiring at the age they would be in 3-5 years and with a third of the weight due to the three-year time span covered. We assume that all other nurses remain in the workforce at their current age. For LPNs, the ACS identifies individuals who are not in the labor force at the time of the survey but were in the labor force one year prior to taking the survey. For individuals aged 50 and older, we assume that combination of responses represents a permanent departure from the workforce (retirement). Retirement patterns in the model differ by age, nurse type, and education level (RN with baccalaureate degree, RN with diploma or associate degree, or LPN). Samples sizes are insufficient to estimate retirement patterns by nurse sex, race/ethnicity, or other factors such as region. For each age and nurse type combination, the number of nurses retiring in the next year (in the NSSRN) or in the past year (in the ACS) is divided by the total number of nurses in that combination, which results in a probability of retirement at that age for that type of nurse. For nurses aged 70 and older, the sample sizes are small and estimates of retirement patterns fluctuate accordingly.

Career progression: LPN-to-RN and RN-to-APRN

The other modeled reason for nurses leaving the nursing workforce of their current nurse type is career progression. HWSM changes education level based on the probability that an LPN will become an RN or an RN will become an APRN (nurse practitioner, nurse midwife, or nurse anesthetist). The NCLEX examination data used to determine the number of new RN entrants includes those who formerly practiced as LPNs, so changing LPNs to RNs would introduce double counting issues. Instead, career progression from LPN to RN is treated as attrition from the LPN workforce. Similarly, RNs who become APRNs are also included in the attrition count because APRNs are not part of the nursing model component.6

To determine the probability of an LPN becoming an RN, the supply model uses a similar process to the retirement pattern described previously. LPN responses from the 2015 ACS are combined with 2018 NSSRN data on registered nurses whose highest degree was awarded in 2015 and who reported having an LPN license. An attrition pattern is then created for LPNs by dividing the number of LPNs who become RNs in 2015 by the total number of responses in the combined NSSRN and ACS data for each age from 20 to 49. For RNs who become APRNs, the process is similar, except that there is no need to use ACS data. We created a subset of responses from the NSSRN including RNs active in 2015 and APRNs who achieved their highest education level in 2015. We created an attrition pattern by each individual age using the nurses’ age in 2015. Though ACS data later than 2015 are available, we used 2015 data due to anomalies in the NSSRN data on the number of LPNs becoming RNs in later years.

We combined the career progression and retirement patterns to form a single attrition pattern that the model uses the same way regardless of attrition reason. The model generates a random number between 0 and 1 every year of the simulation for each individual in the simulation. If this number is less than the attrition probability for that age and nurse type, the model removes the nurse from the supply output for that year and every subsequent year.

Exhibit IV-5 shows the attrition pattern for RNs with a baccalaureate. These nurses have a small annual probability of leaving the workforce due to becoming APRNs, with retirement probability accelerating after about age 55.

  • 6APRNs are modeled in other HRSA reports with physicians and physician assistants.

Exhibit IV-6 shows the attrition pattern for RNs with a diploma or associate degree. Each year some RNs trained at the associate or diploma level complete RN-to-BSN programs and transition to a different education level in HWSM. This is counted as attrition for the purposes of this exhibit, and explains the steep drop in RNs at this education level among the under age 40 workforce. However, these nurses are switched to the baccalaureate degree level in the model and do not leave the nursing workforce.

Exhibit IV-7 shows the attrition pattern for LPNs. The steep drop in remaining active in the LPN workforce indicates that about half of new LPNs at age 20 will eventually become an RN. This does not mean that half of the current supply of LPNs will become RNs, as LPNs who stay in that career for longer will be overrepresented compared to those who only briefly practice as an LPN.

Modeling Workforce Participation

Activity status for nurses is modeled using prediction equations derived from the 2018 NSSRN for RNs and ACS (2016-2020) for RNs and LPNs. This analysis focuses on nurse clinicians under age 50 as the activity status for clinicians aged 50 and over is modeled as attrition. The dependent variable is whether the nurse is active in the nursing workforce (employed or actively seeking employment) or not active. Explanatory variables are the same as those used to model hours worked: sex, race/ethnicity, and 5-year age groups.

The overall activity rates for RNs and LPNs under age 50 were 85% and 87%, respectively. The odds of being employed vary by nurse characteristics, in particular age (Exhibit IV-8). Nurses are more likely to be active in the workforce as they age, with RNs at the baccalaureate level having a 13% higher odds of being active at ages 45-49 compared to those under 30 and under. Non-Hispanic Black LPNs have 59% higher odds of being active in nursing compared to non-Hispanic White LPNs.

Modeling Hours Worked

Forecasting equations model the correlation between weekly hours worked to nurse age, sex, and race/ethnicity by nurse type and education level. Data for all variables came from the 2018 NSSRN for RNs and the 2016-2020 ACS for RNs and LPNs. Year was included in the regression because multiple years of ACS data were analyzed. Total projected nurse hours worked were converted to FTEs by dividing by 40, as starting in 2017 workforce projections for all the health occupations modeled using HWSM have defined an FTE as 40 hours per week for the full year.

Ordinary Least Squares regression coefficients reveal the following (Exhibit IV-9):

  • Average weekly hours worked decline among older nurses, especially from age 65 onward.
  • On average, male RNs work about 3 more hours and male LPNs work about 2 more hours than their female peers, controlling for any differences in age, race/ethnicity and education level.
  • Non-Hispanic Black nurses work more hours than nurses in other race/ethnicity groups.

These prediction equations have low predictive power for explaining weekly hours worked for specific nurses, as illustrated by the low R-squared values. Other predictors of labor force participation that are not included in HWSM, but are likely correlated with nurse demographics are family composition (including marital status, presence of young children in the family, and older family members who need caregivers), nurse health, wages, and local economic conditions.7 Still, at the aggregate level the prediction equations allow HWSM to capture the implications of changing nurse demographics and education level on FTE supply.

Modeling Cross-state Migration

We model that nurses initially enter the workforce in the state where they took the NCLEX exam. We then model cross-state migration based on prediction equations estimated using logistic regression with the 5-year (2016-2020) ACS file combined with the 5-year (2011-2015) ACS file. Multiple 5-year files were used to increase sample size, as the number of nurses observed changing states is low as a proportion of the total number of nurses. Cross-state migration models whether a person moves out of a state. It then models whether a person moves into a state. Of the 66,640 baccalaureate-level RNs in the 5-year file (with different nurses surveyed each year), 2,220 (3.3%) indicated working in a different state compared to a year ago. Of the 32,662 RNs at the diploma or associate degree level in this file, there were 640 (2.0%) who indicated working in a different state compared to a year ago. Of the 23,916 LPNs in this file, there were 479 (2.0%) who indicated working in a different state compared to a year ago.

Analysis of nurse cross-state migration patterns found in Exhibit IV-10 suggests that:

  • The probability of migration declines with age. Nurses aged 30 and below have the highest probability of migrating to another state.
  • Male nurses are more likely to move than female nurses.
  • Non-Hispanic RNs with a baccalaureate and LPNs who do not identify as either White or Black are more likely to relocate compared to other race/ethnicity groups in the same nurse type (Exhibit IV-10).

Using the ACS sample weights, the findings suggest that approximately 44,700 RNs with a baccalaureate, 14,200 RNs with a diploma or associate degree, and 9,500 LPNs change states annually. When modeling cross-state migration patterns, HWSM uses the above equations to generate a probability that each nurse will migrate out of the state. The model then compares this probability to a random number between 0 and 1 using a uniform distribution. If the random number is below the estimated probability of moving, then the nurse moves out of that state.


We ensure that the national number and characteristics of nurses moving out of states matches the number and characteristics of nurses moving into states. When the model moves a nurse out of state, it generates a random number. It compares that number against a cumulative distribution created from the national distribution column in Exhibit IV-11. This gives each nurse a chance to move to a specific state based on how frequently nurses in the past have been observed moving to that state. For example, between 2016 and 2020, of the estimated 9,500 LPNs who moved to another state each year, approximately 2.1% moved to Alabama and 4.7% moved to California. Over time, projections of the number of nurses exiting a state change based on the characteristics of nurses in that state and the overall number of nurses. The variation across states and years reflects the modeling of migration determinants. It also reflects the use of the random number generator to move nurses across the various states based on the geographic distributions described previously. 

Supply Scenarios Modeled

Nurse supply is modeled under a Status Quo scenario that models the continuation of recent numbers and characteristics of nurses completing their nursing education, and recent patterns of labor force participation. Labor force participation decisions include attrition (retirement, career change out of nursing, or career advancement from LPN to RN or from RN to APRN), being temporarily out of the workforce, and hours worked patterns. Labor force participation varies by nurse demographics and education level. The Status Quo scenario models the continuation of these patterns taking into account the changing demographic and changing education levels of the nursing workforce.

Alternative supply scenarios modeled include the impacts of:

  • retiring two years earlier or delaying retirement by two years, on average
  • graduating 10% more or 10% fewer nurses annually than the Status Quo

The early or delayed retirement scenarios simply shift workforce attrition patterns for nurses aged 50 and older by ±2 years. For example, a nurse who would have retired at age 62 under the Status Quo scenario would now retire at age 60 under the Early Retirement scenario and at age 64 under the Delayed Retirement scenario.

Modeling Demand

Demand modeling for nurses follows the overall HWSM demand modeling approach described in the other modules. HWSM applies prediction equations to the simulated U.S. population data to estimate use of health care services in the settings where nurses work. Projected demand for health care services is the driver of projected demand for nurses. For example, projected growth in hospital inpatient days and emergency visits are the drivers of project demand for nurses employed in hospital inpatient and emergency department settings, respectively. For work settings outside the traditional health care system, growth in the population most likely to use the services is the driver of growth in demand for nurses (Exhibit IV-12). For example, projected growth in demand for school-based nurses is based on projected growth in the population of children ages 5 to 17.

As illustrated in Exhibit IV-12, nurses are found in almost all care delivery settings. Data from the NSSRN and ACS estimate the portion of national FTE nurses providing care in each setting. We multiply the national starting supply of nurses by these proportions to estimate the number of FTEs in each work setting. We calculated a staffing measure for each setting by dividing the number of FTE nurses working in that setting in the base year by the base year estimate of the workload measure. Workload measures include office and outpatient visits in ambulatory settings, inpatient days in hospital inpatient settings, emergency visits, home health visits, and population size metrics. The workload measures for 2020 are modeled expected values in the absence of COVID-19, as the health care use patterns are based on 2015-2019 data that predates COVID-19. Applying these patterns to the 2020 population provides estimates of what health care use would have been in 2020 absent COVID-19.

In the Status Quo demand scenario, these ratios are constant over time. For example, demand for RNs under the Status Quo scenario is based on the 2020 ratio of inpatient days to RNs for hospital inpatient settings for every year of the projection period. Demand for nurses in academia is based on the estimated population of college graduates and applies the ratio of nurse educators to students in 2020 across the projection period. Nurses working in adult day service centers are isolated from nurses working in the ‘other’ setting via the 2015-2016 National Center for Health Statistics report estimates of long-term care providers.8 Demand for nurses in adult day service center is based on the national ratio of day service center users to nurses. In the HWSM framework, probabilities weighted by age group are assigned to estimate how many people use adult day care services daily.

Data from the 2018 NSSRN provide the distribution of RNs across employment settings.9 Estimates of the distribution of LPNs across employment settings came from analysis of the detailed industry classification of the combined (2019 and 2020) ACS.9 10

National staffing ratios by employment setting at baseline were applied to the projected service use to generate staffing requirements by setting. Demand projections were calculated at the county level and summed to produce state and national estimates for reporting.

Demand Scenarios Modeled

As with other health occupations modeled, HWSM models demand for nurses under two scenarios as described in the other modules:

  1. The Status Quo scenario models a continuation of recent (2015-2019) national patterns of care use extrapolated to the future population. This scenario captures geographic variation in demographics, health risk factors, disease prevalence, insurance coverage, level of rurality, and household income that can affect demand for nursing services. Similarly, the scenario captures population growth and aging—and the associated implications for disease prevalence and other health risk factors—over the projection horizon. The scenario assumes national demand equal to national supply in 2020. The scenario evaluates whether the nation’s future nursing workforce is sufficient to provide at least the current level of care.
  2. The Reduced Barriers scenario estimates the number of nurse FTEs required if populations that historically faced barriers to accessing health care services demonstrated care use patterns comparable to populations perceived to have fewer barriers to accessing care. The scenario modifies the health care use patterns of nonmetropolitan county residents, racial and ethnic minority populations, and people without health insurance to the health care use patterns of their peers living in metropolitan counties, who are non-Hispanic White, and who have health insurance. Each of the three components of this scenario (non-metropolitan versus metropolitan county of residence, minority versus non-Hispanic White, and uninsured versus insured) is modeled in isolation and together to quantify the magnitude of each factor on demand for services. This hypothetical scenario describes the implications on nurse demand if policies and programs reduced access-based disparities to health care services. The impact of reducing barriers to accessing care is modeled only for care provided in ambulatory settings and hospital settings. Reduced Barrier Demand projections for nursing services delivered in nursing homes, residential care facilities, school-based settings, nurse education, and public health settings equal those of the Status Quo scenario.
Date Last Reviewed: