Skip Navigation LinksHome > March 2014 - Volume 27 - Issue 3 > Supply of physician assistants: 2013-2026
Journal of the American Academy of Physician Assistants:
doi: 10.1097/01.JAA.0000443969.69352.4a
Original Research

Supply of physician assistants: 2013-2026

Hooker, Roderick S. PhD, PA; Muchow, Ashley N.

Free Access
Article Outline
Collapse Box

Author Information

Roderick S. Hooker is a retired PA. Ashley N. Muchow is a doctoral student in policy analysis at the Pardee RAND Graduate School in Santa Monica, Calif. The authors have disclosed that they were employees of the Lewin Group, a subsidiary of UnitedHealth. While at the Lewin Group, they were frequent users of the data and personally involved with upgrading and validating the PA component of Provider 360 Database.

Collapse Box

Abstract

ABSTRACT: As part of healthcare reform, physician assistants (PAs) are needed to help mitigate the physician shortage in the United States. This requires understanding the population of clinically active PAs for accurate prediction purposes. An inventory projection model of PAs drew on historical trends, the PA stock, graduation estimates, retirement trends, and PA intent to retire data. A new source of licensed health professionals, Provider 360 Database, was obtained to augment association information. Program growth and graduate projections indicated an annual 4.7% trend in new entrants to the workforce, offset by annual attrition estimates of 2.9%. As of 2013, there were 84,064 licensed PAs in the United States. The stock and flow equation conservatively predicts the supply of PAs to be 125,847 by 2026. Although the number of clinically active PAs is projected to increase at least by half by 2026, substantial gaps remain in understanding career trends and early attrition influences. Furthermore, education production could be constrained by inadequate clinical training sites and scarcity of faculty.

Around the turn of the century, physician assistants (PAs) and nurse practitioners (NPs) emerged as viable players in delivering clinical services and gained the notice of workforce planners.1 As attention turned to their role and outputs of care, PAs and NPs were found to be more productive than previously realized, and data sources for these providers were sought. Originally, data came from the American Academy of Physician Assistants (AAPA). The PA workforce was predicted to grow from 75,000 in 2010 to 127,000 in 2025 (a 72% increase).2 Since that estimate, the Patient Protection and Affordable Care Act was enacted, resulting in an increased demand for medical services. This demand is predicated on many factors—increasing the insured pool to 25 million, an expanding and aging population, sustainability of chronic diseases, rising number of visits per capita, and increasing demand for technology. At the same time, PA programs are growing.3 Validation of existing models, along with improved methods for workforce forecasting, is needed.

The shortcoming of many health workforce estimates is reliance on self-report survey data, membership rolls, graduation rates, certification attainment, and other methods of capturing individual characteristics and demographics. In the case of PAs, useful AAPA census surveys have been undertaken periodically since the PA Master File was developed in 1986.4 Unfortunately, age at graduation and retirement information also were missing from these estimates. In addition, limited numbers of observations and variance in response rates may have contributed to diminishing usefulness to health policy analysts.

A revised forecast was undertaken using a new data source. These data take into consideration the stock of active licensed PAs in the United States. The intent was to better-inform policy makers of the composition of the PA workforce with upgraded figures.

Back to Top | Article Outline

DATA SOURCES

Provider 360 Database (P360) is a health professional data source operated by OptumInsight, a business of UnitedHealth Group. One of several commercially available data sources for health providers, P360 is considered comprehensive due to its data collection strategy. The mission of P360 is to maintain an accurate database on all health providers in the United States who have a license or registration to treat patients. One of the key commercial uses of the data is to verify the identity and status of providers submitting claims for health services reimbursement. Accuracy of any provider holding a current license and without sanctions is critical for claims processing (fraudulent claims can be filed using deceased, retired, and inactive providers).

The data sources for P360 are numerous and include

* state licensure boards (state medical licenses and sanctions)

* the federal Office of Inspector General (federal sanctions)

* General Services Administration (federal sanctions)

* Drug Enforcement Administration (DEA numbers)

* National Plan and Provider Enumeration System

* Social Security Administration Death Master File (deceased individuals)

* Centers for Medicare and Medicaid Services (CMS) PECOS (Medicare enrollment)

* CMS MEDPAR (hospital data)

* licensees of US Postal Service (delivery point validated addresses, National Change of Address file)

* health plan provider network rosters.

Direct outreach to provider offices incorporates facsimile, phone, Internet websites, e-mails, and consensus modeling techniques for verification and resolving differences in data. A continuous focus on data integrity through monthly validation and quality improvement is core to the fidelity of the database operation.

P360 data are widely used for claims processing, database validation, correction, augmentation, public provider lookup sites, and research. Some examples include evaluating provider characteristics (such as age and sex of endocrinologists and oncologists) or supply and demand models (such as optometrists, dietitians, and nutritionists).5 The authors are experienced end users of P360.

Education programs The number of PA programs with students and annual graduates was obtained from the Accreditation Review Commission on PA Education, Inc. (ARC-PA) and the Physician Assistant Education Association (PAEA) through personal communication, reports, and website sources. Existing accredited PA programs and programs in development were identified using data reported by ARC-PA. PAEA annual reports were used to calculate the mean number of graduates per program. A National Commission on the Certification of Physician Assistants (NCCPA) report served as a source of data triangulation.6

Matriculate data Data on program matriculates provide the age and sex characteristics of projected graduates. This was obtained from the Central Application Service for Physician Assistants (CASPA), an application service for PA education similar to those for undergraduate, graduate, and medical schools. The system permits students to submit one application to multiple schools. Those enrolled are the matriculates of interest and provide the age and sex distribution of new PA stock each year.

Attrition Retirement rates were derived from a 2011 survey of certified PAs over age 55 years.7,8 Survey respondents were asked to indicate the age at which they retired or intended to retire. These responses were used to calculate retirement rates by age and were factored in the supply model. A set of career spans for each age band was used for forecasting purposes. The retirement studies also showed there was no difference in retirement rates or intended retirement rates by sex.

Back to Top | Article Outline

METHODS

Active stock Data extracted from P360 were used to establish the head count of clinically active (licensed) PAs for 2013. All variations of the National Provider Identifier (NPI) Taxonomy Codes, self-designated specialties, and degree titles for PAs that might occur across state license boards are included in the data for triangulation and validation purposes. After sanctioned and deceased providers were removed, the existing stock of PAs with licensure data as of 2013 was extracted. The data included age, sex, degree, NPI taxonomy, state, date of license, and license expiration dates. Active PAs were compared with inactive PAs to exclude double counting. Dual licenses in two or more states were reduced to one state using the date of the most recent license.

Entrants New entrants to the PA workforce were estimated by the number of programs with graduates (baseline), average number of graduates per program, and population (lower bound) projections. The age profile of graduates at the beginning of year 2013 used CASPA data on 2012 matriculates and projected this cohort profile for each subsequent year. The profile of those with complete information was used to impute missing information.

Sensitivity analysis An upper- and lower-bound sensitivity analysis was constructed to give a range of estimates based on different assumptions. The lower-bound growth of PAs uses a population-based method of analysis, and the upper bound uses a program-based method of analysis. Both methods are described below and illustrated in the results section.

Population-based method Census projections of men and women ages 20 to 59 years were used as a proxy for the size of the annual new entrant pool, and were marched forward each year, retiring older PAs and entering younger ones. The population-based method is an established technique of projecting workforce entrants, and is an alternative to the graduate cohort method. The equation assumes that the same proportion of the population in the base period will continue to enter the workforce as PAs. Mathematically, growth in graduates of sex (S) in year (Y) is estimated by multiplying the percent population growth by the baseline number of 2013 PA graduates:

Program-based method Population-based graduate projections are achieved by applying the 2013 average number of graduates per PA program to the anticipated number of annual programs with graduates. The ARC-PA provided estimates of programs that anticipate starting a class in subsequent years. The time from program inauguration to graduation was held constant at 28 months. The time from provisional accreditation to inauguration is 1.5 years, with 4 years elapsing between inauguration and the new program's first graduates. Five graduation classes are required before an application to expand the class size is accepted.

Figure. No caption a...
Image Tools

Attrition Any PA whose license expired in 2011 and was not renewed as a PA by 2013 was considered retired, deceased, emigrated, or not in the clinical designation of delivering domestic healthcare under a state PA or medical license. The retirement age distribution was applied to each successive year based on PA retirement trend data.7,8

Licensure Licensure enables a PA to legally manage patients clinically as a proviso of state law or regulation. The date a PA receives a license is the date the PA is considered clinically active. The date the license expires without renewal is used to identify attrition. P360 data is based on licensure count (which may explain why the number is lower than survey estimates by the AAPA and certification counts by the NCCPA).9,10 Furthermore, many PAs are certified but are not clinically active. The NCCPA estimates this unemployment was 5.2% of all 90,227 certified PAs at the end of 2012.6

Modeling the PA supply The projection technique used is an inventory or input-output model in which the input is annual entrants to the workforce and the output is the annual attrition of the workforce.11 The model estimates the annual supply (stock) of PAs by making assumptions about future population sizes and graduation (input) and retirement cohort effects (output). Cohort analysis shows the mean duration of a PA career to be 29 years with a range from 20 to 38 years.7,8 For comparison purposes, a population growth model and a program growth model were developed. All work was done in Microsoft Excel.

Back to Top | Article Outline

RESULTS

Clinically active PAs In 2013, there were 84,064 clinically active PAs (that is, each held a valid state-issued PA or medical license) in 50 states, 4 territories, and the District of Columbia. This 84,064 is the 2013 stock for the supply model. The mean age of PAs was 42 years and the percentage of women was 75%. Age and sex profiles are illustrated in Figure 1.

Figure 1
Figure 1
Image Tools

Graduates and entrants to the workforce As of 2013, there were 181 accredited PA programs; 146 had graduates. The 5,971 graduates in 2013 were based on an average of 40.9 graduates per PA program. Five years of PAEA data provide the numbers needed for projection purposes (Table 1).

Table 1
Table 1
Image Tools

As of 2013, ARC-PA had accredited 181 PA programs and 60 had applied for accreditation and were in the queue to be visited by ARC-PA by 2020. Graduates from these 60 anticipated programs were included in the upper-bound scenario and 54 (90%) were included in the baseline scenario. The working assumption for the baseline scenario is that accreditation and opening will vary among programs, and some will lose accreditation. As such, the estimated graduates were distributed in the upper-bound scenario over 6 years (10 programs per year over the years 2019 to 2024) and distributed 56 of the anticipated 60 in the baseline scenario over 6 years (9 programs per year over 2019-2024) (Table 2). The average number of graduates per program was held constant at 41 in the baseline scenario and grows at the historical 0.66% rate in the upper-bound scenario. Two projection methods were used: population-based and program-based. These projections illustrate the differences between status quo growth-–based on 2013 entrants-–and the growth anticipated by universities across the country, respectively.

Table 2
Table 2
Image Tools

Attrition For the purpose of this model, attrition is defined as anyone departing a clinical role from the United States, such as by death, retirement, or emigration. The age-specific attrition rate is the joint probability of retirement and mortality life tables based on reported estimates. Annual attrition was estimated by taking an age profile of those projected to die and those intending to retire in a given year from the CDC. The PA retirement rates were published in 2013.7,8,12 (Figure 2). Coincidentally, the US annual PA attrition rate averaged 2.9% for both methods (retirement and death).

Figure 2
Figure 2
Image Tools

Inventory model From the base year supply of 84,064 PAs in 2013, the model predicts that the PA workforce will increase to 127,773 by January 2026, with an upper bound of 132,434 and a lower bound of 121,328 (Figure 3). The upper bound and population-based lower bound are the sensitivity analyses used in this model, based on high program growth in contrast to modest population growth.

Figure 3
Figure 3
Image Tools
Back to Top | Article Outline

DISCUSSION

This modeling of American PAs identifies a supply trend beginning in 2013 and projected to 2026. The model is stocked with 84,064 licensed PAs in the United States in 2013 and ends with a conservative estimate of 125,847 by 2026. The growth is based on programs in development that have requested an accreditation visit by ARC-PA and uses a projection of graduates based on historical graduation rates of accredited programs. Age of graduates is determined through matriculates and attrition (based on profiles of retirees and those intended to retire). This US PA growth, from a historical standpoint, was the result of a surge in new programs resulting in an increase in the annual graduating cohort.3 The annual 4.7% trend in new entrants to the workforce is offset by a 2.9% annual attrition rate.

For PAs as a whole, a large group of early career women is replacing a smaller but older group of men; 80% of all PAs are under the age of 55 years.7 This relatively youthful profile coupled with high job satisfaction may be a significant factor in the stability of the PA workforce.13 The mean career span is 29 years with a range of 20 to 38 years.7,8 More research is needed to determine whether more women in the PA workforce will have any effect on average career longevity.

Policy implications of this analysis are significant. The main implication is that some analysts have underestimated the PA workforce and such improved modeling could help with the anticipated medical supply gap.14–16 Although allopathic and osteopathic medical school growth is under way, and a domestic graduation cohort of 25,000 by 2020 is anticipated, these numbers still represent a declining ratio of physicians per capita.17 PA growth, as this model projects, will be between 9,000 and 10,000 graduates annually by 2026, resulting in a large cadre of active clinicians.

This is not the first supply model of PAs. A 2010 model used AAPA survey data and assumed that all accredited PA programs were producing graduates.2 The estimate of 127,800 by 2025 is within the margin of error of the 125,847 by 2026 predicted by this model. Such closeness in calculation suggests validation and perhaps usefulness of both models. The NCCPA calculation of 85,320 certified and employed is remarkably close to this estimate of 84,064 licensed-–both as of end of year 2012. Because modeling uses precise mathematics and imprecise data, periodic reviews of modeling calculations are needed for confidence as to which technique provides the greatest usefulness.

Although this is a supply model, the demand for PA services is expected to continue in the short run, according to the federal Health Resources and Services Administration.18 For more granular examples of demand, PAs and NPs in community health centers (CHCs) are the providers of record for 30% to 50% of all visits and are staffed at twice the rate compared with private practice offices.19 In the Veterans Health Administration (VHA), PAs and NPs attend to 30% of all primary care visits and the complexity of their patients is only slightly less than those of physicians in the same setting.20 Both the VHA and CHCs are in expansion phases to meet a growing population and have specifically targeted PAs and NPs as providers of importance.21,22 PA and NP expansion in hospital-based clinics, both urban and rural, doubled between 2000 and 2010.23 Finally, evidence is mounting that PAs and NPs are influencing their market appeal through improved scope-of-practice laws and regulations.24

The potential barriers to PA increase could be the competition for clinical training sites among medical, osteopathic, and NP educational programs. Relaxing immigration policy for overseas-trained doctors (both Americans educated abroad and international medical graduates) could help offset demand and constrain jobs. The growth of NPs in the clinician labor pool could crowd out opportunities for PAs as well. Finally, demand can abate in almost any service industry, and this labor pool is unlikely to be an exception.

Back to Top | Article Outline

LIMITATIONS

All forecasts have inherent limitations, and the projection model presented here is no exception. This is an analysis using a dataset new to health professionals that relies on licensing records and differs from self-report sources. Although P360 is considered valid and reliable for insurance billing purposes and pharmacy benefit projections, it has been slow to appear in the literature. The upside is that it uses different methods of data accumulation from traditional sources such as the AAPA and NCCPA, which permit corroboration studies and comparative analyses. Multiple data sources help validate the accuracy of the P360 information.

The refinement of health professions workforce modeling is under way and many players are engaged in finding the right method. However, all models suffer in not understanding the supply and demand effect from technology, a growing and aging population, declining birth rate, economic perturbations, sustainability of chronic diseases, and increasing efficiency in service delivery. A 2013 Dutch study on predictive modeling using a 5- and 10-year backtested strategy of physicians in a country the size of Maryland illustrated that 10-year projections are less reliable than those for shorter periods.25 Such reassessment of prediction models from time to time is needed to verify economic assumptions, test new models, and assess the reliability of equations. Furthermore, evaluating the accuracy of projections is needed for policies that support certain sectors of health profession growth.

The opportunity cost of education is growing and constraints in graduate medical education funding looms large, which could affect supply as well. New models of care include patient-centered medical homes, accountable care organizations, direct care (concierge) medicine, and team-based care. These and other externalities produce effects on supply that are difficult to predict.

Back to Top | Article Outline

CONCLUSION

A half-century ago, America began experimenting with a new type of health professional to augment the role of the busy physician. The role of PAs has become so successful that they are involved in all discussions of health policy and healthcare reform. A transformative labor movement, PAs may have accounted for 10% of available providers in 2013 and their accessibility for service delivery is growing. The profession also is increasingly youthful with a likely career span of 2 to 3 decades. What is offered here is an upgraded PA supply model using a new, different, and novel source of data but arriving at a conclusion not too dissimilar of other models. Such triangulation of effort offers confidence that newer models can arrive at similar conclusions and sets the stage for other validation work. Finally, PA supply is a social measurement, as is any population prediction. This supply should be assessed accurately and frequently with a maximum of efficiency and a minimum standard of error. To undertake more contemporary PA modeling, research requires refined longitudinal cohort studies.

Back to Top | Article Outline

REFERENCES

1. Cooper RA. Health care workforce for the twenty-first century: the impact of nonphysician clinicians. Annu Rev Med. 2001;52:51–61.

2. Hooker RS, Cawley JF, Everett CM. Predictive modeling the physician assistant supply: 2010-2025. Public Health Rep. 2011;126(5):708–716.

3. Cawley JF, Jones PE. Institutional sponsorship, student debt, and specialty choice in physician assistant education. J Physician Assist Educ. 2013;24(4):4–8.

4. Schafft GE, Cawley JF. The Physician Assistant in a Changing Health Care Environment. Rockville, MD: Aspen Publishers; 1987.

5. Hooker RS, Williams JH, Papneja J, et al. Dietetics supply and demand: 2010–2020. J Acad Nutr Diet. 2012;112(3 suppl):S75–S91.

6. Glicken AD, Miller AA. Physician assistants: from pipeline to practice. Acad Med. 2013;88(12):1883–1889.

7. Coombs JM, Hooker RS, Brunisholz KD. What do we know about retired physician assistants? A preliminary study. JAAPA. 2013;26(3):44–48.

8. Coombs J, Hooker RS, Brunisholz K. Physician assistants and their intent to retire. Am J Manag Care. 2013;19(7):e256–e262.

9. American Academy of Physician Assistants. AAPA Physician Assistant Census Report. Alexandria, VA: American Academy of Physician Assistants; 2011.

10. Danielsen RD, Lathrop J, Arbet S. The certified physician assistant in the United States: a 2011 snapshot. JAAPA. 2012;25(4):58.

11. Sherbrooke CC. Optimal Inventory Modeling of Systems: Multi-echelon Techniques. Springer; 2004.

12. Arias E. United States life tables, 2008. Centers for Disease Control and Prevention. National Vital Statistics Reports. September 24, 2012. http://www.cdc.gov/nchs/data/nvsr/nvsr61/nvsr61_03.pdf. Accessed December 16, 2013.

13. LaBarbera DM. Gender differences in the vocational satisfaction of physician assistants. JAAPA. 2010;23(10):33–34,36–39.

14. Petterson SM, Liaw WR, Phillips RL Jr, et al. Projecting US primary care physician workforce needs: 2010–2025. Ann Fam Med. 2012;10(6):503–509.

15. Colwill JM, Cultice JM, Kruse RL. Will generalist physician supply meet demands of an increasing and aging population. Health Aff (Millwood). 2008;27(3):w232–w241.

16. Sargen M, Hooker RS, Cooper RA. Gaps in the supply of physicians, advanced practice nurses, and physician assistants. J Am Coll Surg. 2011;212(6):991–999.

17. Grover A, Niecko-Najjum LM. Building a health care workforce for the future: more physicians, professional reforms, and technological advances. Health Aff (Millwood). 2013;32(11):1922–1927.

18. Health Services Resources Administration, National Center for Health Workforce Analysis. Projecting the Supply and Demand for Primary Care Practitioners Through 2020; 2013.

19. Hing E, Hooker RS, Ashman JJ. Primary health care in community health centers and comparison with office-based practice. J Community Health. 2011;36(3):406–413.

20. Morgan PA, Abbott DH, McNeil RB, Fisher DA. Characteristics of primary care office visits to nurse practitioners, physician assistants and physicians in United States Veterans Health Administration facilities, 2005 to 2010: a retrospective cross-sectional analysis. Hum Resour Health. 2012;10(1):42.

21. Woodmansee DJ, Hooker RS. Physician assistants working in the Department of Veterans Affairs. JAAPA. 2010;23(11):41–44.

22. Hing E, Uddin S. Visits to primary care delivery sites: United States, 2008. NCHS Data Brief. 2010;(47):1–8.

23. Hooker RS, Benitez JA, Coplan BH, Dehn RW. Ambulatory and chronic disease care by physician assistants and nurse practitioners. J Ambul Care Manage. 2013;36(4):293–301.

24. Stange K. How does provider supply and regulation influence health care markets? Evidence from nurse practitioners and physician assistants. J Health Econ. 2014;33:1–27.

25. Van Greuningen M, Batenburg RS, Van der Velden LF. The accuracy of general practitioner workforce projections. Hum Resour Health. 2013;11(1):31.

Keywords:

physician assistant; supply; health workforce; predictive modeling; primary care; career trends

© 2014 American Academy of Physician Assistants.

Login

Article Tools

Images

Share

Search for Similar Articles
You may search for similar articles that contain these same keywords or you may modify the keyword list to augment your search.