Published 2024
In this module:
This module describes the data and assumptions used in the Health Workforce Simulation Model (HWSM) to project the future supply and demand for physician assistants (PAs). The overall modeling approach is described in other modules. PAs work closely with physicians, and PA specialty is generally defined by the specialty of the physician(s) with whom the PA works. Hence, the modeled specialties for PAs overlap the modeled physician specialties.
Modeling supply
Supply modeling consists of estimating the number and characteristics of starting year supply, modeling the number and characteristics of new entrants to supply, attrition, and workforce behavior such as patterns of hours worked and geographic mobility. Although some PAs will change specialties over their career, the HWSM does not currently have sufficient information to model that change. This is a potential area of improvement to the model.
Estimating the base year workforce supply
Estimates of the number, characteristics, practice location, and specialty of PAs used to construct a starting year supply database combine published information from the National Commission on Certification of Physician Assistants (NCCPA) and de-identified individual-level PA data from the American Academy of Physician Assistants (AAPA). NCCPA produces annual reports on the number of certified PAs in each state and the national number of certified PAs by specialty in 2022.1 2 We use estimates of certified PAs by state as the starting point for developing the HWSM supply database for PAs.
HWSM simulates the career choices of PAs, and as such needs individual-level data. While AAPA’s database contains data on most PAs (both AAPA members and some non-members), the file is not as complete as the NCCPA licensure database. AAPA survey data combined with membership files provide flexibility to categorize PAs as working in a metropolitan or nonmetropolitan area and provide demographic information on PAs. Consequently, we use NCCPA published estimates of the number of licensed PAs in each state and draw a random sample (with replacement) equal to the number from the NCCPA database. For example, if NCCPA indicated that a state had 5,000 licensed PAs then we would draw a sample of 5,000 records from the AAPA database for that state. The AAPA data includes PA specialty, age, and gender which are necessary for supply modeling. We compare aggregate statistics on the number of PAs by specialty in the constructed starting supply file and add and subtract sampled PAs until the sample matches published NCCPA estimates by specialty.
Taking into consideration the response rate to NCCPA surveys and the percentage of certified PAs who are in non-clinical positions, retired, or working outside the U.S., we estimate the number of active PAs in 2022 (which is not the same as full time equivalents [FTEs]) at 157,121. This includes 37,657 PAs in primary care, 24,578 in medical specialties, 39,792 in surgery specialties, 5,760 in hospital medicine, and 49,334 in other specialties.
Modeling new entrants
The mechanism for modeling new PA entrants to the workforce each year is the creation of a “synthetic” population of the profession, based on the number and characteristics of recent graduates in each specialty. Each new PA is assigned an age, sex, and specialty that reflects the distribution seen in recent years.
NCCPA reports 11,092 physician assistants were newly certified in 2022.3 NCCPA surveys new graduates to ascertain the specialty area for PAs who have accepted a position and targeted specialty area for PAs who have not yet accepted a position. Specialty is defined as the specialty of the physician(s) with whom the PA primarily interacts. Supply projections assume that the number, age distribution, and sex distribution of new entrants (Exhibit VII-1) remain the same over time at the most recently observed numbers. The annual number of PAs completing training has grown over time. While the Status Quo supply scenario models the number of new PAs entering the workforce remaining constant, alternative scenarios model different assumptions of higher or lower growth in the number of new graduates.
HWSM simulates the geographic distribution of new PAs by practice based on probabilities summarized in Exhibit VII-2. The state distribution for newly trained primary care physicians and NPs is based the AAPA Masterfile state distribution of PAs who have completed training since 2000.
Modeling workforce attrition
PAs can leave the PA workforce of their current specialty permanently by changing careers. Each year, PAs under age 50 have a low chance of being removed from the workforce based on observed career change using the Current Population Survey (CPS) Annual Social and Economic Supplement survey data. This simulates PAs leaving the PA profession entirely, and not PAs changing specialty, which is much more common. While these PAs may join other health care occupations, we assume that they are captured in the new graduate data for that occupation.
HWSM simulates provider attrition probabilities based on the provider’s age, sex, and occupation. Exhibit VII-3 summarizes the cumulative probability that male and female PAs will be active in the workforce after age 50. Attrition probabilities for PAs are based on responses to the question in AAPA’s 2023 National Survey of PAs asking if the PA plans to retire in the next three years. The retirement age distribution of PAs from age 50 to 74 is based on the answers to this question. We assume that all PAs will have retired by age 75. Comparison of results from the 2023 survey to the prior (2015) survey show a substantial shift in PA retirement expectations. Prior data suggested that many PAs would start to retire around age 65. The updated retirement probabilities suggest that many PAs are leaving the workforce prior to age 65. The sample of PAs in many specialties in this survey is insufficient to obtain reliable estimates of retirement intention by age, so the retirement patterns used for modeling are based on all PAs—with separate patterns for male and female PAs.
Modeling hours worked
Physician assistant hours worked in professional activities differ systematically by provider age and sex. The changing demographics of the PA workforce has implications for future FTE supply. We modeled hours worked patterns using Ordinary Least Squares regression with self-reported total weekly hours worked as the dependent variable. The explanatory variables were age group, sex, and age group by sex interaction.
AAPA’s 2023 Salary Survey (n=4,209) collected data on hours worked per week for the primary employer. Using these survey responses, we created hours worked patterns by specialty category for each of the four broad specialty categories (primary care, medical specialties, surgery, and other specialties) as the number of responses is insufficient to create hours worked patterns for each of the 35 PA specialties in the model. Average hours worked per week were 42.96. PAs worked similar hours in the Midwest, South, and West. They worked about 1.9 fewer hours in the Northeast (though not statistically significant). There is no statistically significant difference in hours worked by PAs in metropolitan versus nonmetropolitan areas.
PA hours worked remain relatively steady through age 65, at which time average hours declined for those PAs remaining in the workforce. Hours are lower for female PAs compared to male PAs. Defining a full time equivalent (FTE) as 40 hours per week, total FTE PA supply in 2022 is approximately 7% higher than total active supply.
Modeling state-level supply and migration
The model accounts for annual movement of physician assistants between states in two steps. First, logistic regression4 using 2018-2022 American Community Survey (ACS) data estimates the probability of a PA under the age of 50 migrating to another state based on their age group, sex, the population of the state from which the person moved, and a year indicator. Comparing each person’s annual movement probability to a random number between 0 and 1 determines which PAs move each year. The likelihood that each person who moves will relocate to a specific state is based on the proportion of PAs moving to that state as observed in ACS data. For example, if 10% of PAs who relocated according to ACS moved to California, then each PA who moves in the simulation has a 10% probability of moving to California. This approach ensures that cross-state migration has no impact at the national level in terms of the number and characteristics of PAs.
Second, when a PA moves to a specific state, the model then assigns them a value of metropolitan or nonmetropolitan location designation equal to the proportion of PAs practicing in metropolitan and non-metropolitan areas of the new state. For example, if 70% of PAs in a moving PA’s new state work in a metropolitan area (per the latest AAPA Masterfile), then the moving PA has a 70% probability of being assigned a value of metropolitan in the simulation because of the move.
Supply scenarios modeled
The Status Quo supply scenario models the continuation of current numbers and characteristics of new graduates from PA training programs. The scenario extrapolates current patterns of hours worked, attrition, and cross-state migration to PA supply in future years.
Alternative supply scenarios model the sensitivity of future supply to changes in key assumptions. These alternative scenarios illustrate the uncertainty of how PA supply determinants might change over time. Two scenarios model a 10% increase and a 10% decrease in number of PAs trained each year, respectively. Two other scenarios model PAs retiring 2 years earlier or 2 years later relative to current patterns, respectively. For example, under the early retirement scenario, a PA who would have retired at age 65 instead retires at age 63. Under the delayed retirement scenario, a PA who would have retired at age 65 instead retires at age 67.
Modeling demand
The approach to model demand for PAs follows the common model outlined in the other modules. The main demand components are (1) a constructed population file for each county in the U.S. with a representative sample of the population from 2022 through 2037, (2) prediction equations of the demand (i.e., annual expected use) for health care services, and (3) PA staffing ratios. This section focuses on construction of the staffing ratios for each PA specialty for the setting where they provide services.
The staffing ratios are calculated as the national expected health care use in 2022 divided by estimated national FTE demand for PAs. FTE demand is estimated as starting supply plus estimated shortfalls for select PA specialties to address the unexpected demand increase from COVID-19 becoming endemic (as described in prior modules about COVID-19 implications for hospital-based care and increased outpatient visits to deal with acute and long COVID-19. Altogether, these adjustments imply PA supply adequacy of 99.3% (or slightly over 1,200 FTE shortfall) in 2022. Hence, HWSM extrapolates the national patterns of health care use and PA staffing to the local level as well as into the future. Consequently, demand projections are interpreted as the number of FTE PAs that would be required in the future to maintain current national average levels of PA care. To the extent that a larger proportion of future health care services will be provided by PAs (and advanced practice nurses [APRNs]), comparison of future PA supply adequacy should be interpreted in the context of projected adequacy for physicians and APRNs.
Projecting the quantity of PAs demanded has been a historically difficult task because of the overlapping roles that PAs and physicians can play in the administration of health care services. Accordingly, past analyses using HWSM have used physician setting distributions as stand-ins for PA setting distributions when PA-specific data was unavailable. Rather than rely purely upon the assumption that PA setting distributions mirror those of their physician counterparts, we have broken out survey response data from the NCCPA’s “2022 Statistical Profile of Certified Physician Assistants by Specialty” report to formulate a more accurate picture of where PAs work (Exhibit VII-5). Throughout 2022, the NCCPA collected survey responses and published summary statistics concerning the specialties and settings of a cohort that reflect roughly 84% of the total NCCPA-licensed PA population.
Employment setting distributions for some PA specialties—mainly specialties with few PAs—are unavailable in the NCCPA report. In those cases, physician setting distribution is used as a proxy to categorize PAs into their respective employment settings. Specifically, we assume that PA setting distribution mirrors the distribution of their physician colleagues in Allergy & Immunology, Endocrinology, Infectious Diseases, Nephrology, Pulmonology, Rheumatology, Anesthesiology, Radiation Oncology, Radiology, Colorectal Surgery, Ophthalmology, Thoracic Surgery, and Other Specialties.
Multiplying total licensed PAs in each specialty by the setting distribution from the NCCPA survey, and adjusting for hours worked to define FTE as 40 hours per week, produced estimates of FTE PAs by setting and specialty (Exhibit VII-5).
The workload demand drivers for PAs are the same as the demand drivers for physicians within the same specialty. For example, an estimated 33% (1,569) of PAs in cardiology work in an office setting, and the demand driver for office-based care is projected visits to a cardiology office (which is estimated as nearly 31.5 million visits in 2022 (Exhibit VII-6). Likewise, an estimated 62% (2,952) of PAs in cardiology worked in hospital settings. In 2022, an estimated 12.5 million cardiology inpatient days were recorded.
Dividing estimates of health care use by estimates of FTE PAs in each specialty and setting (plus starting year shortages) yields a ratio of demand for services per PA (Exhibit VII-7). For example, dividing the estimated number of office visits to family medicine (303,421,000) by the number of office-based FTE PAs in family medicine (14,054) yields a ratio of 22,230. This ratio does not mean that PAs provided services for 22,230 patients’ visits, on average, as many patients to a family medicine practice might not have seen a PA. Under the Status Quo demand scenario, if this ratio held constant over time, then demand for office-based PAs in family medicine would grow at the same rate as number of visits to a family medicine practice. For settings where the number of PAs in a specialty is small, that setting category is combined with the largest setting for modeling purposes. The ratios in Exhibit VII-7 are based on unrounded numbers, so they might differ slightly from ratios calculated by dividing the rounded numbers in Exhibit VII-6 by the FTE numbers in Exhibit VII-5.
Demand scenarios modeled
As with other health occupations modeled, HWSM models demand for PA under two scenarios as described in the other modules:
- The Status Quo scenario models a continuation of national patterns of care use (2017-2021) 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 physician 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 evaluates whether the nation’s future PA workforce is sufficient to provide at least the current level of care. Interpretation of projections should be within the context that physicians and APRNs overlap in scope of practice with PAs. Hence, imbalances between supply and demand for PAs should consider supply adequacy for physicians and APRNs.
- The Reduced Barriers scenario estimates the number of PA 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 PA 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 Barriers demand projections for PA services delivered in nursing homes, residential care facilities, and other settings (e.g., academia, public health) equal those of the Status Quo scenario.