The United States rheumatology workforce: Supply and demand, 2005–2025




To develop and apply a model that allows prediction of current and future supply and demand for rheumatology services in the US.


A supply model was developed using the age and sex distribution of current physicians, retirement and mortality rates, the number of fellowship slots and fill rates, and practice patterns of rheumatologists. A Markov projection model was used to project needs in 5-year increments from 2005 to 2025.


The number of rheumatologists for adult patients in the US in 2005 is 4,946. Male and female rheumatologists are equally distributed up to age 44; above age 44, men predominate. The percent of women in adult rheumatology is projected to increase from 30.2% in 2005 to 43.6% in 2025. The mean number of visits per rheumatologist per year is 3,758 for male rheumatologists and 2,800 for female rheumatologists. Assuming rheumatology supply and demand are in equilibrium in 2005, the demand for rheumatologists in 2025 is projected to exceed supply by 2,576 adult and 33 pediatric rheumatologists. The primary factors in the excess demand are an aging population which will increase the number of people with rheumatic disorders, growth in the Gross Domestic Product, and flat rheumatology supply due to fixed numbers entering the workforce and to retirements. The productivity of younger rheumatologists and women, who will make up a greater percentage of the future workforce, may also have important effects on supply. Unknown effects that could influence these projections include technology advances, more efficient practice methods, changes in insurance reimbursements, and shifting lifestyles. Current data suggest that the pediatric rheumatology workforce is experiencing a substantial excess of demand versus supply.


Based on assessment of supply and demand under current scenarios, the demand for rheumatologists is expected to exceed supply in the coming decades. Strategies for the profession to adapt to this changing health care landscape include increasing the number of fellows each year, utilizing physician assistants and nurse practitioners in greater numbers, and improving practice efficiency.

The American College of Rheumatology (ACR) created the Workforce Study Advisory Group and retained The Lewin Group to conduct a new workforce study of US rheumatologists in 2005, to project demand for rheumatology services, in order to guide policy regarding rheumatology manpower for the next 2 decades. The Lewin Group is an applied research and consulting group specializing in health policy issues. While workforce projections are not an exact science, they attempt to model future needs using information and assumptions from the present.

The Graduate Medical Education National Advisory Committee (GMENAC) was founded in 1976 to estimate the need for physicians by specialty and used estimates based on the judgments of experts, a method that has come to be known as the GMENAC method. Centralized planning of physician supply became more explicit with the establishment of the Council of Graduate Medical Education (COGME) in 1986. The emphasis shifted from concern that there were too few physicians to concern that there were too many specialists. Centralized planning of residency positions became explicit in the Clinton administration's 1993 proposal for health care reform, the Health Security Act. Such planning of residency positions required determining whether a given specialty was a shortage or a surplus specialty, and it sparked renewed interest in models of the physician workforce. In 1994 Weiner (1) projected that there would be a substantial excess supply of specialists by the year 2000, primarily as a result of the assumed continued growth of staff-model health maintenance organizations (HMOs) and the prediction that this would restrain utilization of specialists.

Recent physician workforce analyses by COGME and by Cooper (2) projected a shortage of ∼85,000 physicians, mostly specialists, by 2020. This shift in projections occurred in part because of the failure of HMOs to limit specialty physicians and the assumption that continued economic growth would increase demand for specialty care. Newer methodologies for workforce planning are multimodal and apply a mixture of approaches to measure and analyze available data. By diversifying the measurement tools, economists hope to improve the accuracy of workforce projections.

A 1996 workforce study (submitted to the ACR, not published) predicted a surplus of rheumatologists based on assumed increases in staff-model HMOs and management of rheumatic diseases by primary care physicians, neither of which materialized. The ACR has projected a decline in rheumatologists beginning in 2016, based on the age distribution of its membership and the number of entering fellows. The consensus of practicing rheumatologists is that a shortage exists now, and that it is likely to worsen. This workforce study assumes that supply and demand are in balance in 2005 and projects to the future; however, the model will allow adjustment of baseline supply and demand figures if data are available.

The rheumatology workforce will no doubt be challenged by the increased incidence of musculoskeletal diseases in the aging US population, and expanding therapeutic options that will increase demand for our services. Hootman and Helmick updated projected age- and sex-specific prevalence rates of arthritis, doctor-diagnosed arthritis, and arthritis-attributable activity limitations in US adults ages 18 years and older for the years 2005–2030 (3). The latest case definitions were estimated from the 2003 National Health Interview Survey of 31,000 adults. Using future population projections from the US Census Bureau, prevalence rates were projected to increase from 47.8 million in 2005 to nearly 67 million by 2030. The changing sex distribution and work intensity of younger specialists may affect productivity per provider. Demand for rheumatologists' time could be reduced by lower insurance coverage, increased direct patient costs, restrictions on expensive therapeutic options, and more efficient approaches to providing care. Predicting how these factors will evolve and influence one another is a major challenge to forecasters, economists, and policy makers.

The purpose of this study was to identify the significant factors affecting supply and demand for rheumatology services, to quantify these factors, and to develop a computer-based model that will project supply and demand between 2005 and 2025. The Workforce Study Advisory Group was tasked with addressing the complexities and uncertainties that have made past workforce projections inaccurate. We report the main results and discuss the implications for our specialty.


The study was organized into tasks which included the formation of an advisory panel by the ACR Committee on Training and Workforce, an analysis of public databases and review of the literature by The Lewin Group, the development of a survey instrument that was sent to rheumatologists, the development of a computer-based model of supply and demand, and presentation of the findings to the ACR. The advisory panel (Appendix A) was chosen to represent the diverse constituencies in the ACR, including physicians in practice, academicians, pediatric rheumatologists, and allied health professionals. The Lewin Group was interviewed by the members of Committee on Training and Workforce and was contracted by the ACR to perform the study. The advisory panel met with staff of The Lewin Group and regularly discussed critical issues pertaining to development of the survey and computer model.

The American Medical Association (AMA) master file was the primary source of data on the current number of rheumatologists. The numbers are consistent with (but not identical to) estimates from the American Board of Internal Medicine. However, the number of individuals classified as academic rheumatologists appeared to be understated in the AMA file. The ACR membership file was used to supplement the AMA file. Because of the small numbers of pediatric rheumatologists, the ACR membership file was used to supplement the AMA file and an additional 39 pediatric rheumatologists were added. A rheumatologist was defined as a self-reported board-certified and/or fellowship-trained rheumatologist or pediatric rheumatologist. Health care utilization data on Medicare and non-Medicare populations were obtained from Medicare Part B and the National Ambulatory Medical Care Survey. Population data were derived from the US Census Bureau population projections and labor force participation rates for professionals. Data on per capita income were from the Bureau of Economic Analysis.

To obtain more detailed and current information about rheumatologists, a survey instrument was designed, addressing educational and professional qualifications, work effort and productivity, rheumatologic conditions treated, distribution of time and effort, employment setting characteristics, measures of excess demand, job satisfaction, retirement plans, and sources and level of income. Rheumatologists were randomly selected using a “skip factor” with a random starting value, automatically generated by SAS. A total of 1,683 rheumatologists were asked to complete the survey, and 627 responded (37% return rate) (Table 1). In addition to the randomly selected sample, all pediatric rheumatologists and rural rheumatologists (defined as not residing in a metropolitan statistical area) were sampled to assure that the sample size of these groups would be sufficient for statistical estimates (90% confidence interval ± 5–8%).

Table 1. Sample size and response rate for the survey instrument*
Rheumatologist categorySample sizeNo. of respondentsResponse rate, %
  • *

    Forty-six pediatric rheumatologists had boards in both adult and pediatric rheumatology. It is estimated that the majority (83% [38 of 46]) see primarily patients over age 18; therefore, for model projections regarding pediatric rheumatologists, the number used was 218 (256 − 38). Thus, total US rheumatology workforce in 2005 = 4,946 adult rheumatologists and 218 pediatric rheumatologists.

  • Original sample size and number of surveys sent was 1,735. Fifty-two were returned as undeliverable, yielding a total sample size of 1,683.

  • Main practice area outside a metropolitan statistical area as defined by the US Census Bureau.

<40 years of age1845328.8

A computer-based Markov projection model was created to project supply and demand and permit sensitivity analyses of factors affecting the future workforce. The demand factors included prevalence rates of selected diseases, population size and demographic mix, insurance coverage by type of insurance for population, and per capita household income. Factors affecting supply included the number and demographic distribution of the current workforce, hours of work, retirement and mortality rates, and the number and fill rates of fellowship positions.

Supply-side assumptions in the model included the following: as of 2005, the supply and demand for rheumatology services are in equilibrium, the number of fellow positions will remain static at the 2004–2005 level for adult positions (n = 378) and pediatric positions (n = 56), international medical school graduates (IMGs) will account for 35% and 29% of the adult and pediatric fellows, respectively (and 80% of IMGs will practice in the US), the proportion of women in fellowship training will remain at the 2004–2005 level (49% and 69% of adult and pediatric fellows, respectively), male and female rheumatologists will continue to provide the same average number of visits as in the year 2005, and rheumatologists will remain in the workforce after age 65. Demand-side assumptions included the following: the US population will grow at rates projected by the US Census Bureau, per capita income will grow ∼1% annually, and the rate of uninsured will remain the same.


The 2005 supply of rheumatologists is 4,946 and is shown in Table 1 (AMA file supplemented with ACR membership list). The age and sex distributions of adult and pediatric rheumatologists are shown in Figure 1. The number of male and female adult rheumatologists is approximately equal up to age 44 years, after which men predominate. The trend is for an increasing percentage of women in the adult and pediatric rheumatology workforce (Figure 2). Although 49% of pediatric rheumatologists are women, 67% of pediatric rheumatologists under age 40 are women, while the majority over age 55 are men. The mean number of patient visits provided annually by rheumatologists in clinical practice is shown in Figure 3. The average male rheumatologist provides 34% more visits than the average female rheumatologist (mean number of visits per rheumatologist per year 3,758 for male rheumatologists and 2,800 for female rheumatologists). Male rheumatologists under age 40 provide fewer visits than those in the age 40–49 and age 50–59 groups. Complete survey results can be seen on the ACR Web site,

Figure 1.

Numbers of adult (left) and pediatric (right) rheumatologists in the US, by age and sex.

Figure 2.

Projected percentage of women among pediatric rheumatologists (left) and adult rheumatologists (right), 2005–2025. The percentage of women in the pediatric rheumatology workforce is projected to rise from 51.6% in 2005 to 62.7% in 2025, and that of women in the adult rheumatology workforce from 30.2% in 2005 to 43.6% in 2025.

Figure 3.

Mean ± SD number of patient visits provided by female and male rheumatologists, by age. Men provide 35% more patient visits than women. The survey sample did not contain enough women over age 60 for valid estimates of patient visit numbers; these women were combined with the 50–59 age group.

Table 2 shows the model's predictions for adult rheumatologists. The number of rheumatologists needed to meet the excess demand is predicted to be 1,029 by 2015 and 2,576 by 2025. The stochastic model predicts a slight decline in the number of 2005-equivalent rheumatologists (the term adjusts for the predicted decrease in productivity of the workforce) between 2010 and 2015. Forty percent of the observed increase in demand is related to growth in the US population, holding the age distribution of the US population constant. Eleven percent of the increase is related to the increased proportion of the population over age 64, and the remaining 49% is related to per capita income growth. The demand for health care services, like most goods and services, has been shown to increase with real income growth. Evidence for this increased demand, particularly for medical specialty services, is reviewed by Cooper (2). If one assumes that 2005 demand exceeds supply by 10% (∼500 rheumatologists), the excess demand in 2025 increases to 3,296 adult rheumatologists.

Table 2. Projected supply and demand for adult and pediatric rheumatologists
  • *

    Rheumatologists under age 40 and female rheumatologists are predicted to see fewer patients than the older male workforce of 2005. Since their numbers will increase as a percent of the overall workforce, more rheumatologists will be needed to see the same number of patients in the future. The term “equivalent rheumatologists” refers to this predicted change in productivity.

Adult rheumatologists     
 Supply of 2005-equivalent rheumatologists*4,9465,0194,9404,8064,643
 No. of rheumatologists4,9465,1985,2585,1785,008
Pediatric rheumatologists     
 Supply of 2005-equivalent rheumatologists*218238247252254
 No. of rheumatologists218244258266271

The model's predictions for pediatric rheumatologists are also presented in Table 2. The model indicates a need for 33 pediatric rheumatologists to meet excess demand by 2025. Approximately 60% of the excess demand is related to per capita income growth, and 40% to increases in the size of the population under age 18.


The supply/demand model predicts a substantial excess in demand for adult rheumatology services relative to workforce supply between 2005 and 2025. The most important contributors to this shortfall are an increasing prevalence of musculoskeletal diseases due to growth and aging of the US population and the modeled 1% increase in per capita income. This increased demand occurs with no predicted increase in the number of practicing rheumatologists and, based on survey results, a lower average workload of younger rheumatologists compared with more senior practitioners. There is also an increasing percentage of women in the workforce who, based on survey data, see fewer patients (a greater percentage of women work part-time).

The survey conducted as part of this workforce study appears to show current excess demand. The average wait time for new patients is 38 days, and 30% of practices are currently hiring, and 50% planning to hire, a rheumatologist, nurse practitioner (NP), or physician assistant (PA) over the next 5 years.

The predicted excess demand for pediatric rheumatology services does not widen to the same extent as for adults, in part because of the younger age of the workforce and later predicted times of retirement; however, the gap between supply and demand would be influenced by current shortage estimates. For pediatric rheumatology there is reason to believe that there is a significant current excess demand. Data from the state of Washington have shown that 62% of adult rheumatologists care for children as young as 5.3 years of age (4). A California survey showed that adult rheumatologists see an average of 3.1 children per week, or 43% of all children seen for pediatric rheumatic diseases (5). In 2005 ∼10 states and 18 metropolitan statistical areas with >700,000 people were without a pediatric rheumatologist. These data suggest a current imbalance. If one assumes that 2005 demand exceeds supply by 25% or 50%, the excess demand in 2025 increases from 33 pediatric rheumatologists in the baseline case to 104 or 176, respectively.

The baseline case assumes a 1% increase in per capita income. If instead there were no increase between 2005 and 2025, the predicted deficit of adult rheumatologists would decrease from 2,576 to 1,100 by 2025, still a substantial shortfall. A change to 3% simulated income growth results in a 40% increase in demand for adult rheumatology services in 2025 (7,219 rheumatologists with a 1% growth in the Gross Domestic Product [GDP]; 10,081 with a 3% growth in the GDP) and a 37% increase in demand for pediatric rheumatology services (287 pediatric rheumatologists with a 1% growth in the GDP; 394 with a 3% growth in the GDP). GDP growth between 1967 and 1986 was between 2.5% and 2.9%; the assumption of 3% would be an upper boundary on possible income growth. A decline in the uninsured population from 15% to 8% would increase demand by only 4.8%.

Past projections of physician workforce have often been unreliable because the assumptions used to build the models were inaccurate. Our projections of future supply are based on current number and age distributions of rheumatologists, estimates from recent historical data on mortality and retirement rates of physicians by age and sex, and the number and fill rates of fellowship positions. They rely on survey data for estimates of workload capacity of rheumatologists by age and sex, but these data are not the primary source for supply or demand estimates. Our demand estimates are based on statistical estimates of the relationship between demand for the services of rheumatologists and population, the age distribution of the population, the insurance coverage of the population, and the per capita real income of the population. The single largest influence on demand is population, which can be projected over periods of 15–20 years with a high degree of precision. GMENAC based its demand projections almost entirely on expert judgment, using the Delphi method.

Efforts were made to improve the current model by collecting data from a large cohort of rheumatologists, working with acknowledged experts in health system analysis, and carefully identifying key contributors to the physician supply/demand relationship. The creation of a computer-based model of supply and demand will allow the ACR to adjust projections on an ongoing basis when any of our key assumptions change.

If this model is correct in predicting a growing shortage of rheumatology services, one would reasonably want to develop a strategic plan to address the looming shortage. Interventions that could reduce excess demand might include expanding the number of fellowship positions, redesigning practice patterns, and expanding use of NPs and PAs.

On the supply side, the most obvious response would be to expand training programs to fill the gap. The model predicts that a 30% increase in fellowship positions in each 5-year period beginning in 2005 would be needed in order to meet the predicted shortfall. The additional 188 first-year fellowship positions would more than double the current number. Funding this expansion is problematic given current federal budgetary caps on GME positions and the uncertain status of Medicare funding. It is also unclear whether new positions could be filled with qualified residents. The 2005 match had ∼350 applicants for 162 positions beginning in 2007, but there are no data on the qualifications of the unmatched applicant pool. Furthermore, it is not clear how academic resources, faculty and clinic facilities, and patient populations could be expanded and funded to support increased training. Nor does training of more rheumatologists solve the geographic maldistribution; currently, 10 states have no pediatric rheumatologist.

Problems with expanding current federal funding through the National Institutes of Health make attracting and retaining faculty to academic training programs difficult. Salary lines for clinician educators and program directors are limited. The academic component of this workforce study reveals concerns about the health of the academic enterprise. First, academic salaries are ∼30% lower than private practice compensation, according to the Medical Group Management Association Compensation and Productivity Survey. Only 28% of academic rheumatologists are tenured, while another 16% are eligible for tenure. Two-thirds of academic faculty members require at least 5 years after fellowship to achieve independent investigator status and half require at least 7 years (survey results). The increased competition for R01 funding could further delay the maturation of independent investigator status. These findings may be a harbinger of a significant shortage of academic rheumatologists.

The task force also considered what other circumstances and proactive interventions might reduce the projected supply/demand imbalance. On the demand side, improvement would occur if rheumatology scope of service or the number of referrals were to decrease. Alternatives include changing from traditional to managed new patient scheduling (6) and reaching more rational agreements with referring physicians as to how problems would be managed across diseases and levels of severity. Reducing disease activity through more effective treatment would decrease followup visits, as would adopting more advanced appointment scheduling practices for established patients (7, 8). If access to rheumatologists becomes severely and persistently reduced, patients may seek treatment elsewhere and other providers may compete successfully for rheumatologists' traditional patient base.

Given these realities, redesigning rheumatology practice appears to be the most realistic and responsible option for addressing the projected excess of demand over supply. In 2001, the Institute of Medicine published Crossing the Quality Chasm: A New Health Care System for the 21st Century, which documents the waste in and underperformance of the US health system in chronic disease care and outlines the potential for improved outcomes and costs through practice redesign (9). The supply/demand problems projected in that report are heavily dependent on the largely traditional approaches described by the surveyed rheumatologists. More effective and efficient practice processes have great potential to improve rheumatology access, patient throughput, service, disease outcomes, and costs, independent of increasing provider numbers.

Methods for health care process improvement have been borrowed from other US industries and implemented in some rheumatology practices already (10). Key strategies include standardizing and streamlining visits and data management, involving other health care professionals in team-based practice, managing new referrals through preappointment management (6), improving access through scheduling process changes (8), improving disease control by measuring disease activity and dependably accelerating treatment (7, 11, 12), reducing dependence on physician visits for providing necessary care such as for monitoring medications and filling prescriptions, using nurses to coordinate care, collect data, and educate patients, using patient-generated data (13–16), and exploiting task management software (17). If established rheumatologists and new trainees learn and implement process redesign, as recommended by the Institute of Medicine, Institute for Healthcare Improvement, American Council for Graduate Medical Education, and others, the imbalance between supply and demand predicted in this report can be reduced with fewer additional rheumatologists.

An interdisciplinary consensus on which patients each specialty should most appropriately manage could result in more appropriate referrals, closer coordination of care among providers, reduced duplication and waste, and shorter wait times. Primary care physicians may be expected to care for patients with acute low back pain, early osteoarthritis, and somatoform pain disorders, and provide osteoporosis prevention. Rheumatologists could spend more time managing autoimmune diseases and inflammatory arthritis and prescribing disease-modifying drugs at an early point in the disease, when the impact on long-term outcomes can be the greatest. Alternatively, the future could see breakthroughs in treatment of osteoarthritis that would require considerable expertise and increase the need for rheumatologist management, and the early use of biologic agents might result in less need for long-term management of inflammatory disorders by rheumatologists.

Expanding the roles of NPs and PAs in rheumatology practices is a strategy that is being increasingly used in a number of settings to enhance services. The role of the rheumatologist remains central in these practices, but with transfer of selected tasks the efficiency of the practices is generally improved. Because of the proficiency achieved by repetition, some of the common and less complex elements of care may be provided more effectively than by physicians (18). As noted above, ∼25% of all rheumatologists work with an NP or PA, and another one-quarter of those surveyed reported they intend to employ one within the next 5 years. However, rheumatology practices will have to compete for NPs and PAs with other specialties that offer greater compensation (surgery and cardiology, for example). An organizational approach to specialty training may help recruitment. The ACR is currently developing a Web-based rheumatology curriculum for NPs and PAs in order to reduce the in-office training time required for clinical competence.

If demand for rheumatology services exceeds supply as predicted, then long wait times for appointments, patient dissatisfaction, and a decline in quality of care would result. For rheumatologists, full schedules would be guaranteed, allowing selection of patients based on economic factors, and maintenance of income. Increasing demand for services could lead to more stress, longer work hours, and a change in career satisfaction. The computer-based model developed for this workforce study will allow the ACR to adjust supply and demand estimates in the future as changes in health care occur.

If the goal of the ACR is to deliver high-quality care in a timely manner to patients who need special services, then the major available options now are increasing fellow positions, adding NPs and PAs, and practice redesign. If more rheumatologists are produced and the design of practice models remains unchanged, there will be an inevitable further rise in health care costs, which are already creating a threat to the viability of our economy.

Practice redesign should be an increasing priority for the ACR. Improving delivery of care needs to be supported as actively as expanding knowledge. Academic practices must become laboratories for developing and teaching redesign and implementing new, more effective processes. Value- rather than volume-based compensation must be forcefully advocated with payers in order to strongly encourage improvement in and support for the work and resources required for redesign. In return, rheumatologists should take responsibility for controlling costs by reducing patient visits and providing better management of chronic diseases. Rheumatologists must lead in changing the current zero-sum competition in the US health system to a shared pursuit of highest value for patients (19).


Dr. Deal had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Deal, Hooker, Harrington, Birnbaum, Hogan, Bouchery, Klein-Gitelman, Barr, and Advisory Group members.

Acquisition of data. Hogan, Bouchery.

Analysis and interpretation of data. Deal, Hooker, Harrington, Birnbaum, Hogan, Bouchery, Klein-Gitelman, Barr.

Manuscript preparation. Deal, Hooker, Harrington, Birnbaum, Hogan, Bouchery, Klein-Gitelman, Barr.

Statistical analysis. Hogan, Bouchery.



Members of the Workforce Study Advisory Group were as follows: Chad L. Deal, MD (Chair, Workforce Study Advisory Group), Walter Barr, MD (Chair, Committee on Training and Workforce), Neal Birnbaum, MD (President, ACR [2006–2007]), Dennis Boulware, MD, Paul Caldron, MD, MPH, Timothy Harrington, MD, Marc Hochberg, MD, MPH, Roderick Hooker, PhD, PA, Marisa Klein-Gitelman, MD, MPH, Julianne Orlowski, DO, Stephen Paget, MD, Christy Park, MD, Audrey Uknis, MD, Patience White, MD, Ellen Bouchery (The Lewin Group), Paul Hogan (The Lewin Group).