Assessing the productivity of PAs and NPs : JAAPA

Secondary Logo

Journal Logo

AAPA Members can view Full text articles for FREE. Not a Member? Join today!
Original Research

Assessing the productivity of PAs and NPs

Stefos, Theodore PhD; Moran, Eileen A. MPA; Poe, Stacy A. PhD; Hooker, Roderick S. PhD, MBA, PA

Author Information
JAAPA 35(11):p 44-50, November 2022. | DOI: 10.1097/01.JAA.0000885152.52758.48
  • Free


We set out to evaluate the clinical practice of physician associates/assistants (PAs) and NPs in the Department of Veterans Affairs (VA) Veterans Health Administration (VHA). Understanding where efficiency can reduce costs is of interest to healthcare facility managers. As one part of VA improvement, we identified factors that may affect clinical productivity. The work builds on a growing interest in the broadening roles of PAs and NPs in American medicine.1 This VA study examined the division of medical labor in the VHA system. Such undertaking occurs in an era characterized by high US healthcare spending and inadequate health outcomes. Our aim was to explore productivity in the vertically integrated healthcare system of the VHA and in the broadening the roles of PAs and NPs.2,3 This interest centers on the economic theory that productivity gains could be achieved by altering the labor inputs used in the healthcare sector.4

The economics of PA and NP employment in the VHA is an element of growing human resource interest. Moran and colleagues showed that of the about 6,000 PAs and NPs in the VA, the annual average productivity of PAs was 8% higher than NPs.5 This difference remained even after stratifying for a wide spectrum of variables such as geography, sex, supervisory arrangements, type of VA medical center, years of experience, and scope of practice. Since the 2016 study, VA medical staffing has seen a dramatic increase in PAs and NPs, from 6,000 to 8,000 by 2020, and a greater number used in specialty areas.6 The growing presence of PAs and NPs in medicine is consistent with their roles in the broader US landscape of medical specialization.2,7 Concurrently, VA data and refinements of the data have grown in quality and quantity, and analysis strategies have improved.

We set out to understand the persistent differences in productivity levels between PAs and NPs in the context of clinical specialty. Three aims were set forth:

  • Compare PA and NP productivity levels overall, and then parse the findings by specialty area.
  • Identify important factors affecting PA and NP productivity.
  • Estimate adjusted productivity levels for PAs and NPs that will control for differences generated by essential variables.


This was an observational study of 8,001 PAs and NPs in 134 VA medical centers (including community-based outpatient clinics as satellites of VA medical centers) during fiscal year 2016 (October 1, 2015, to September 30, 2016). The unit of observation was the individual healthcare provider. Basic PA and NP productivity comparisons were made by clinical specialty area. Two techniques were used:

  • General linear models, a statistical strategy that in this case examined the effect of categorical factors on a continuous response variable, assessed the overall statistical productivity differences between PAs and NPs and the PA/NP productivity differences in and between specialties.
  • A clinician-level, multivariate generalized linear model was used to simultaneously control for various factors such as medical or surgical specialty, practice location, and the complexity of the practice institution that might affect the estimation of relative clinician productivity. A generalized linear model is a flexible generalization of linear regression that allows the response variable to have an error distribution other than a normal distribution. In this model, productivity, the dependent variable, was measured by the annual relative value units (RVUs) produced by each PA or NP in the sample. The policy variable of interest is an indicator of the PA's or NP's discipline: The discipline indicator, NP, takes the value of 1 if the clinician was an NP and the value zero if the clinician was a PA. The binomial discipline variable measures PA and NP productivity differences; the model controlled for other vital factors that may influence productivity. We controlled the varying clinical practice concentration of PAs and NPs with indicator variables for each clinical specialty. We used internal medicine as the reference group for practice specialties. We used indicator variables for the geographic and administrative location of the clinician's practice using VA Integrated Network Indicators (VISN). The VISN clusters medical centers into regional administrative networks (Figure 1).
VISNsSource: US Department of Veterans Affairs.

We also included a set of indicator variables for the clinical complexity of the clinician's practice hospital (medical center group, or MCG, Table 1). Finally, a set of interaction variables was included, which were the product of the discipline indicator and the clinical specialty indicator. The interactions measured the relative productivity differences between PAs and NPs in each clinical specialty. Covariates were specified a priori and were entered in the model simultaneously after examination for collinearity using collinearity diagnostics developed by Belsley, Kuh, and Welsch.8 All analyses were conducted with the SAS analysis system.

TABLE 1. - The five MCGs in order of complexity
  • Level 1—Highest complexity. Facilities with high-volume, high-risk patients, most complex clinical programs, and large research and teaching programs.

  • Level 2—High complexity. Facilities with medium-high-volume, high-risk patients, many complex clinical programs, and medium-large research and teaching programs.

  • Level 3—Mid- to high complexity. Facilities with medium-high-volume, medium-risk patients, some complex clinical programs, and medium-sized research and teaching programs.

  • Level 4—Medium complexity. Facilities with medium-volume, low-risk patients, few complex clinical programs, and small or no research and teaching programs.

  • Level 5—Facilities with low-volume, low-risk patients, few, or no complex clinical programs, and small or no research and teaching programs.


All VHA clinicians were assigned a clinical specialty based on their encounter workload across clinical areas or clinic designations called stop codes. Stop codes were first mapped to specialty areas (including primary care and mental health), and then a single practicing specialty was assigned to the clinician by assessing the specialty representing the majority of the clinician's clinical encounter workload. In general, the practicing specialty category stop code was mapped as follows: internal medicine and primary care stop codes were mapped to primary care, mental health codes were mapped to mental health, and all other codes that constitute a majority of workload were mapped to a specialty care area (medical, surgical, or ancillary) and then to clinical specialty (for example, cardiology, orthopedic surgery, physical medicine/rehabilitation, etc.). Clinicians who did not have a greater than 50% encounter workload in a single clinical specialty were assigned a clinical specialty of multipractice.

Productivity was estimated by dividing work RVUs by direct clinical full-time employee (FTE) for each provider. Clinical FTE was estimated by dividing the annual clinical time for the provider by 2,080 hours.

Discipline (PA/NP clinical role), clinical specialty, VISN, and the clinical/management complexity of the parent hospital (MCG) served as covariates in the analysis. VISNs are 21 geographical clusters of the 149 VA medical centers (Figure 1). MCGs vary in complexity from tertiary hospitals associated with academic health centers to small hospitals without postgraduate trainees (Table 1).

Because the PA and NP productivity data are highly skewed, we provided a more normal data distribution and removed implausible productivity values by eliminating from the analysis any clinician with less than 100 RVUs/year or more than 7,600 RVUs/year. Suggested outliers included brief employment periods of individuals or a small number of employees who accumulated many overtime hours. We also removed potential outliers using the Cook distance (Cook D) measure calculated from ordinary least squares estimation using SAS.


Eliminating outliers left information for 7,336 PAs and NPs with average productivity of 1,793 RVUs and a median productivity level of 1,644 RVUs. PAs generated average productivity of 1,851 RVUs (SD = 1,034); NPs averaged 1,769 RVUs (SD = 1,040), 82 RVUs fewer than PAs. The difference was statistically significantly different (F = 9.62, P = .0019).

Clinical specialty

The 2,125 PAs and 5,211 NPs were located across 21 VISNs. Fifty-six percent worked in primary care and mental health; the remaining 44% worked in a medical or surgical specialty area. PAs and NPs generally had higher productivity in mental health and primary care than in specialty care and multipractice settings (Figure 2). Productivity levels for all practice specialties were statistically significantly different (F = 195.56, P < .0001).

Average annual productivity (RVUs) by specialty area

Subdividing the overall data by specialty area and discipline (Figure 3) revealed minor or statistically insignificant differences in NP and PA productivity in primary care (F = .14, P = .7077) and mental health (F = 1.72, P = .1898). Differences in PA and NP productivity in specialty care (278 RVUs, F = 61.08, P < .0001) and multipractice (273 RVUs, F = 12.01, P = .0006) were somewhat larger and statistically significant. Examining the productivity of NPs and PAs across specialty areas (Figure 3), we saw the highest productivity for NPs practicing in mental health, followed by primary care, multipractice, and specialty care. The levels of NP productivity across all specialties were statistically significantly different (F = 204.93, P < .0001). Similarly, PAs practicing in mental health saw the highest PA productivity. However, PAs practicing in primary care relative to PAs practicing in multipractice generated somewhat similar productivity (P = .0884). PAs practicing in specialty care also generated a similar workload to PAs in multipractice (P = .0991).

Average annual productivity (RVUs) by specialty area

Medical and surgical specialties

We further subdivided the specialty care area into specialty medicine and specialty surgery (Figure 4). PAs and NPs working in specialty surgery generated a statistically significant 102 RVUs more than PAs and NPs in the specialty medicine environment (P = .0067). NPs in specialty surgery, on average, generated 101 RVUs per year, more than their counterparts in specialty medicine. In comparison, PAs in specialty medicine generated an average 95 RVUs more per year than their counterparts in specialty surgery. Statistically significant differences also were found between PAs and NPs in specialty surgery and specialty medicine. PAs in specialty surgery, on average, generated a statistically significant 159 RVUs per year more than NPs in specialty surgery (F = 7.87, P = .0051). On average, PAs in specialty medicine generated a statistically significant 355 RVUs per year more than NPs in specialty medicine (F = 51.96, P < .0001).

Average annual productivity (RVUs) for specialty care

Looking in the medical and surgical specialty areas indicated a wide variation in workload produced across practicing specialties (Figures 5 and 6). Geriatrics had the lowest medical area of PA and NP productivity (1,170 and 1,078 RVUs, respectively), followed by four subspecialties: neurology, infectious disease, nephrology, and hematology-oncology: (1,449 to 1,078 RVUs, respectively). Dermatology PAs and NPs exhibited the highest average productivity (3,798 and 3,321 RVUs, respectively).

Average productivity (RVUs) by medical clinical specialty∗statistically significant differences in PA and NP specialty means at the .05 level
Average productivity (RVUs) in surgical clinical specialties∗statistically significant differences in NP and PA specialty means at the .05 level


To measure the statistical relationship between productivity, discipline, location, and hospital complexity while interacting with the clinician's discipline (PA or NP) and clinical specialty, we employed a Poisson distributed, identity-linked generalized linear model. This is a flexible regression method that allows for the nonnormal distribution of dependent variables. Results are shown in Table 2. The results indicated that a PA practicing primarily in general internal medicine, located in a VISN 1 (Northeast United States), Level 5 facility (the least clinically complex), would be expected to have an average productivity of 1,699 RVUs per year. Controlling for specialty, the complexity of the clinician's hospital, and location of practice (VISN), NPs generated an average of 79 RVUs per FTE per year, more than PAs.

TABLE 2. - Analysis of generalized linear model estimates
Parameter Estimate Pr > ChiSq Parameter Estimate Pr > ChiSq
Intercept 1,699 < .0001 Cardiology∗ NP 193 < .0001
NP 79 < .0001 Critical care/pulmonary∗ NP 365 < .0001
Cardiology -309 < .0001 Dermatology∗ NP -910 < .0001
Critical care/pulmonary -444 < .0001 Endocrinology∗ NP -229 < .0001
Dermatology 2,056 < .0001 Gastroenterology∗ NP -314 < .0001
Endocrinology -8 .6191 Geriatric medicine∗ NP -43 < .0001
Gastroenterology -10 .0574 Hematology-oncology∗ NP 22 .0024
Geriatric medicine -817 < .0001 Infectious disease∗ NP -136 < .0001
Hematology-oncology -549 < .0001 Mental health∗ NP 213 < .0001
Infectious disease -501 < .0001 Nephrology∗ NP -209 < .0001
Mental health 191 < .0001 Neurological surgery∗ NP 92 < .0001
Nephrology -388 < .0001 Neurology∗ NP 61 < .0001
Neurological surgery -1,092 < .0001 Obstetrics-gynecology∗ NP 114 .0036
Neurology -770 < .0001 Ophthalmology∗ NP -369 < .0001
Obstetrics-gynecology -678 < .0001 Orthopedic surgery∗ NP -413 < .0001
Ophthalmology -349 < .0001 Otolaryngology∗ NP -378 < .0001
Orthopedic surgery 20 < .0001 Pain medicine∗ NP -68 < .0001
Otolaryngology 323 < .0001 Physical medicine-rehabilitation∗ NP -1,077 < .0001
Pain medicine -565 < .0001 Plastic surgery∗ NP 67 .0076
Physical medicine-rehabilitation 455 < .0001 Rheumatology∗ NP -230 < .0001
Plastic surgery -330 < .0001 Surgery -216 < .0001
Rheumatology -393 < .0001 Thoracic surgery∗ NP 172 < .0001
Surgery -460 < .0001 Multipractice∗ NP -310 < .0001
Thoracic surgery -1,259 < .0001 Urology∗ NP -128 < .0001
Multipractice -46 < .0001 Vascular surgery∗ NP -31 .0002
Urology 212 < .0001
Vascular surgery -736 < .0001
Level 1 4.3613 .0164 VISN10 371.4174 < .0001
Level 2 -45.6762 < .0001 VISN11 139.2381 < .0001
Level 3 205.081 < .0001 VISN12 216.5606 < .0001
Level 4 59.9614 < .0001 VISN15 192.5066 < .0001
VISN2 147.4828 < .0001 VISN16 192.838 < .0001
VISN3 -70.5359 < .0001 VISN17 266.3279 < .0001
VISN4 418.1537 < .0001 VISN18 96.6335 < .0001
VISN5 154.1577 < .0001 VISN19 16.1866 < .0001
VISN6 90.363 < .0001 VISN20 294.6922 < .0001
VISN7 194.9996 < .0001 VISN21 -237.889 < .0001
VISN8 237.1167 < .0001 VISN22 271.3516 < .0001
VISN9 153.7859 < .0001 VISN23 371.0811 < .0001

Relative productivity varied across the 25 practicing specialties (Table 2). PAs and NPs practicing in thoracic surgery produced 1,259 fewer RVUs per year than those in general internal medicine. PAs and NPs practicing in dermatology produced 2,056 RVUs per year more than those in general internal medicine. RVU production in specialties also appeared to significantly vary by discipline, as represented by the interaction of discipline status and clinical subspecialty. For example, NPs in cardiology can be expected on average to produce 193 RVUs per year, more than PAs in cardiology. In several surgical subspecialties such as ophthalmology, orthopedic surgery, otolaryngology, and general surgery, PAs generated more RVUs per year than their NP counterparts (369, 413, 378, and 216, respectively).

The clinical complexity (level) of the hospital/MCG had a significant effect on clinician productivity. PAs and NPs in Level 2 facilities generated on average 46 RVUs per year less than those in a typical Level 5 institution. PAs and NPs in Level 3 institutions generated on average 205 RVUs per year more than those in Level 5 hospitals. The geographic region also appeared to have a significant effect on productivity. Other conditions remaining the same, we estimated a range from 418 RVUs/year more than a VISN 1 (Northeast region) location to 238 RVUs/year in all other regions (other than a VISN 1 Northeast region).


Within the VA data set, it appears that clinical specialty, medical center clinical complexity level, and the geographic location of the practice were significant predictors of PA and NP productivity. Unadjusted for these factors, PAs averaged 82 annual RVUs more than NPs, a difference of 4%. In the specialty and multipractice areas, PAs averaged a statistically significant 270 to 280 more RVUs per year than NPs. Further parsing the domains of specialty areas, PAs in specialty medicine and specialty surgery averaged a statistically significant 159 to 355 more RVUs per year than NPs. However, in mental health and primary care, NPs averaged a statistically insignificant 16 to 134 RVUs per year more than PAs.

These productivity differences belie the tendency for clinicians in different practice specialties to inherently see patients of varying RVU complexity, the ability of more clinical complexity institutions to treat a larger number of patients with proportionally higher RVU value, and the effect of practice location on RVU production. Once these factors were simultaneously controlled in a generalized linear model, we saw a more detailed picture of productivity differences across disciplines, practice specialties, the VISN complexity, and the location of the practice.

Overall, after controlling for clinical specialty, institutional complexity, and location, NPs averaged 79 adjusted RVUs per year more than PAs. However, although this was statistically significant for all practical purposes, it was an insignificant 4% clinical difference in overall PA and NP productivity. The productivity variation across practice specialties was considerable and generally of much larger magnitude, ranging from thoracic surgery, where PAs and NPs generated an adjusted productivity level of 1,259 RVUs less than the level generated by those in internal medicine, to dermatology, where PAs and NPs averaged 2,056 RVUs per year more than those in internal medicine. The effect of discipline (PA or NP) in a clinical specialty also appeared to be significant. We estimated average PA and NP productivity difference within specialties ranging from 31 RVUs/year (vascular surgery) to 1,077 RVUs/year (physical medicine and rehabilitation).

The effect of institution clinical complexity appeared to be of a smaller magnitude than the effects of clinical specialty, ranging from 45 RVUs/year less than a Level 3 hospital to 205 RVUs/year more than a Level 3 institution. Geographic location also appeared to have had a statistically significant effect on clinician productivity, ranging from 71 RVUs/year (VISN3) to 418 RVUs/year (VISN4).

Reasons for annual productivity differences are not readily apparent from this overview of the large VHA system. For example, new employees may not be as productive as those on board for more than a few years because onboarding takes time.9 Another reason for small productivity differences is that some employees may be temporary or suspend their clinical service role for various reasons.

Since the millennium, the VHA has undergone many changes, expansions, and modernization.10 As pointed out in this study, employment for PAs and NPs grew from 6,000 to 8,000 in 2 years. Along with 25,000 physicians, PAs and NPs are spread across 1,221 VA outpatient care sites and 144 hospitals. The growth of VHA service facilities requires a large cadre of healthcare professionals. Not unexpectedly, the recruitment of PAs and NPs is ongoing, and their productivity is of interest to human resource managers.

Healthcare delivery in the VA is of particular interest to health services researchers because it is vertically integrated, comprehensive, does not impose a cost to the veteran member, and the locations are widely distributed across the nation in 50 states, the District of Columbia, and five territories.10,11 As a result, size, improved data collection strategies, and outcomes of care and productivity have become central themes for several VA studies. How it delivers this care and the growing dependency on PAs and NPs to deliver that care continues to be a subject of interest to observers of healthcare delivery systems. For example, Ahmed and colleagues and Faza and colleagues have shown not only is the outcome of care in diabetes and cardiovascular disease indistinguishable between PAs and NPs, it is equal or somewhat better than physicians.12,13 Morgan and colleagues found that clinician type might be at least somewhat associated with patient-, clinic-, and state-level factors representing medical and social complexity.14 These results indicate that VA patients assigned to physicians may be somewhat more medically complex than those assigned to NPs or PAs.15 This medical and social complexity builds on the work by Jackson and colleagues analyzing VA visits at the patient level, at the same time analyzing PAs and NPs separately.16 Outcomes of chronic disease care based on 5-year longitudinal systems in the VA reveal that the three primary types of medical providers are producing significant and comparable care.17 Finally, the complexity of clinician interpersonal care in the VA and its effect on some chronic conditions is only now being investigated.18


The data and models in this study attempted to describe the how of PA and NP productivity differences but did not explore why these differences exist. The models set aside fixed or random effects at the facility or even clinician level of experience. We expect future studies will employ fixed or random effects to help control for unobservable factors. This is based on the probability that PAs and NPs in facilities are likely to have correlated RVU production measures. Clearly further data are needed to control for other factors such as how PAs and NPs may be used differently in different specialties, clinician age, number of years in practice, geographic location, and sex. Information on the degree of teaching at a facility, the riskiness of a clinician's patient panel, and market characteristics will enhance future models. Longitudinal tracking is needed to see if these stratified cross-sectional findings remain. Finally, a full production function model would estimate the degree of complementarity or substitutability of work RVUs across many clinicians in a hospital setting, including physicians, RNs, psychologists, social workers, and others.


PAs and NPs in the VHA show little difference in overall productivity levels even when deconstructed by many variables. After controlling for clinical specialty, discipline, clinical complexity, and location, the expected difference between PAs and NPs across all specialties was less than 82 RVUs/year, about 4%. Based on this analysis, any management desire to create separate productivity targets for PAs and NPs would not be supported using the findings presented here. However, given the estimated broad variation in productivity levels across specialty areas, these data and analyses may provide some useful information to help create clinician productivity targets by clinical specialty area, especially if designating broad groups of specialties.


1. Pittman P, Park J, Bass E, Luo QE. Understanding why nurse practitioner (NP) and physician assistant (PA) productivity varies across community health centers (CHCs): a comparative qualitative analysis. Med Care Res Rev. 2021;78(1_suppl):18S–29S.
2. Ray KN, Martsolf GR, Mehrotra A, Barnett ML. Trends in visits to specialist physicians involving nurse practitioners and physician assistants, 2001 to 2013. JAMA Intern Med. 2017;177(8):1213–1217.
3. van Tuyl L, Vrijhoef B, Laurant M, et al. Broadening the scope of task shifting in the organisation of healthcare. Int J Care Coordin. 2021;24(3-4):91–95.
4. Adams EK, Markowitz S. Improving efficiency in the health-care system: removing anticompetitive barriers for advanced practice registered nurses and physician assistants. Accessed July 28, 2022.
5. Moran EA, Basa E, Gao J, et al. PA and NP productivity in the Veterans Health Administration. JAAPA. 2016;29(7):1–6.
6. US Department of Veterans Affairs. Veterans Health Administration. Our providers. Accessed August 22, 2022.
7. Dall TM, Gallo PD, Chakrabarti R, et al. An aging population and growing disease burden will require a large and specialized health care workforce by 2025. Health Aff (Millwood). 2013;32(11):2013–2020.
8. Belsley DA, Kuh E, Welsch RE. Regression Diagnostics: Identifying Influential Data and Sources of Collinearity. New York, NY: John Wiley & Sons; 1980.
9. Morgan P, Sanchez M, Anglin L, et al. Emerging practices in onboarding programs for PAs and NPs. JAAPA. 2020;33(3):40–46.
10. Kizer KW. Veterans and the Affordable Care Act. JAMA. 2012;307(8):789–790.
11. Rose L, Tran LD, Asch SM, Vashi A. Assessment of changes in US Veterans Health Administration care delivery methods during the COVID-19 pandemic. JAMA Netw Open. 2021;4(10):e2129139.
12. Ahmed ST, Mahtta D, Rehman H, et al. Association between frequency of primary care provider visits and evidence-based statin prescribing and statin adherence: findings from the Veterans Affairs system. Am Heart J. 2020;221:9–18.
13. Faza NN, Akeroyd JM, Ramsey DJ, et al. Effectiveness of NPs and PAs in managing diabetes and cardiovascular disease. JAAPA. 2018;31(7):39–45.
14. Morgan P, Everett CM, Smith VA, et al. Factors associated with having a physician, nurse practitioner, or physician assistant as primary care provider for veterans with diabetes mellitus. Inquiry. 2017;54:46958017712762.
15. Morgan PA, Smith VA, Berkowitz TS, et al. Impact of physicians, nurse practitioners, and physician assistants on utilization and costs for complex patients. Health Aff (Millwood). 2019;38(6):1028–1036.
16. Jackson GL, Lee SD, Edelman D, et al. Employment of mid-level providers in primary care and control of diabetes. Prim Care Diabetes. 2011;5(1):25–31.
17. Yang Y, Long Q, Jackson SL, et al. Nurse practitioners, physician assistants, and physicians are comparable in managing the first five years of diabetes. Am J Med. 2018;131(3):276–283.
18. Everett CM, Morgan P, Smith VA, et al. Interpersonal continuity of primary care of veterans with diabetes: a cohort study using electronic health record data. BMC Fam Pract. 2018;19(1):132.

Department of Veterans Affairs; economics; efficiency; PA; NP; medical specialty

Copyright © 2022 American Academy of Physician Associates