To analyze the distribution of rheumatology practices in the US and factors associated with that distribution, in order to better understand the supply of the rheumatology workforce.
To analyze the distribution of rheumatology practices in the US and factors associated with that distribution, in order to better understand the supply of the rheumatology workforce.
Using the American College of Rheumatology membership database, all practicing adult rheumatologist office addresses were mapped with ArcView software. The number of rheumatologists per Core Based Statistical Area (CBSA) was calculated. To investigate whether sociodemographic factors correlated with clustering of rheumatologists, covariates from the 2010 US Census for each CBSA, including age, sex, race/ethnicity, and median household income, were modeled.
Many CBSAs, predominantly smaller micropolitan areas, did not have a practicing rheumatologist. For some of these smaller micropolitan areas (with populations of at least 40,000), the closest practicing rheumatologist was more than 200 miles away. However, we also identified several more-populous areas (populations of 200,000 or more) without a practicing rheumatologist. Greater numbers of rheumatologists were more likely to practice in areas with higher population densities and higher median incomes. More rheumatologists were also found in CBSAs in which there were rheumatology training programs.
These findings demonstrate that many smaller regions of the country have no or few practicing adult rheumatologists. Patients with chronic rheumatic conditions in these areas likely have limited access to rheumatology care. Policy changes could address potential regional rheumatology workforce shortages, but limitations of the current data would need to be addressed prior to implementation of such changes.
The American College of Rheumatology is an independent, professional, medical and scientific society which does not guarantee, warrant, or endorse any commercial product or service.
In 2005–2006, the American College of Rheumatology (ACR) conducted an extensive workforce study administered by the Lewin Group (and referred to as the Lewin report) (). At that time it was estimated that there were 4,908 practicing adult rheumatologists in the 50 US states, the District of Columbia, and Puerto Rico (1.67 adult rheumatologists per 100,000 persons). The report further stated that demand was in balance with the supply of adult rheumatologists in 2005. However, the aging demographics of the US population and projected lack of growth in the number of rheumatologists over the next 10–20 years prompted the investigators to predict that demand would outstrip supply, projecting that this would result in shortages of adult rheumatologists of ∼400 (in 2010) to >2,500 (in 2025). The ACR responded with an increase in funding for the training of new rheumatologists, by targeting programs with unfilled Accreditation Council for Graduate Medical Education slots (ref. and Portfolio Review: Core Programs November 2012. Rheumatology Research Foundation. Advancing Treatment, Finding Cures).
However, these national estimates do not take into account regional variation in the distribution of rheumatologists. Clustering of rheumatologists in some regions can leave other areas of the country with too few adult rheumatologists. Regional shortages of physicians have been identified for other provider groups, including primary care physicians (), general surgeons (), and geriatricians (). The Department of Health and Human Services defines these as Medically Underserved Areas () and Health Profession Shortage Areas (). In these regions, researchers have demonstrated poor quality of care, as documented by lower use of statins and restricted access to specialists (for cardiovascular disease), than in nonshortage areas. () Federal resources are committed to Medically Underserved Areas through Federally Qualified Health Centers. However, specialists have not directly participated in such programs, and these centers' providers report problems obtaining access to specialty consultation ().
To better understand how the regional supply of rheumatologists might create locally underserved markets, the ACR Committee on Rheumatology Training and Workforce Issues examined the distribution of rheumatologists and the factors associated with that distribution across the 48 contiguous United States and the District of Columbia. The results of that study are reported herein.
From the ACR 2010 member database, all nontrainee practicing adult rheumatologist office addresses were identified and were geocoded using ArcView software (ArcGIS). The member database includes self-reported business address and practice type (adult versus pediatric, patient care) ascertained at the time individuals become members. With each annual renewal, these data are updated.
For purposes of regional analysis, Core Based Statistical Areas (CBSAs) (more commonly known as metropolitan and micropolitan areas) were selected. These areas are defined by regions with a high degree of social and economic integration (determined by commute to work) around a central urban core (). Metropolitan areas are defined by populations of ≥50,000 around a central urban core, whereas micropolitan areas have a population of ≥10,000 (but <50,000). In rare exceptions, a micropolitan area can have >50,000 people if it is centered around a less-dense urban cluster (as opposed to the more common dense urban core). As an example, Seaford CBSA (Sussex County, DE), the largest designated micropolitan area in the US, has a population of 197,145, while the population of the city of Seaford is only 6,928 (). For purposes of the present analysis, all other regions are defined as rural.
ArcView maps boundaries of CBSAs using data from the 2010 US Census. The number of rheumatologists within a CBSA can then be tallied to provide a count of rheumatologists per CBSA and number of rheumatologists per 100,000 persons per CBSA (density)—the primary outcome measure of the present study.
To investigate whether sociodemographic factors correlated with clustering of rheumatologists, we abstracted covariates from the 2010 US Census Summary File 2 () for each CBSA, using the US Census FactFinder tool. The covariates included age (as either median age or proportion of persons >65 years old), sex, and race/ethnicity. We abstracted median household income for each CBSA from the 2010 US Census American Community Survey ().
To investigate the association between rheumatology training programs and rheumatologist practice location, the location of each rheumatology program was mapped and the counts of available training slots, as well as the number of rheumatology fellows enrolled during 2010, were counted per CBSA. Data on training programs were obtained from the ACR training program database.
An important implied objective of this analysis was to identify areas that are potentially underserved with regard to rheumatology care. The assumptions behind this analysis are crude and limited by the data (see below). As a simple definition, we first used the Lewin report–defined threshold of 1.67 rheumatologists per 100,000 persons to identify potentially underserved areas. We then refined this definition to include travel distances to the nearest rheumatologist. Finally, we plotted the CBSAs without practicing rheumatologists against travel distance to the nearest rheumatologist. ArcView network analyst was used to calculate driving distance from the center of the CBSA to the nearest practicing rheumatologist.
Multivariate analyses were conducted to evaluate correlations between CBSA descriptors (described above) and the total number of rheumatologists per CBSA, controlling for the size of the CBSA population. Negative binomial regressions were used to model the number of rheumatologists per CBSA, controlling for population per CBSA and other covariates as described above. This modeling was performed for all regions (full sample) and then separately for metropolitan areas only. As data on median household income are available only for metropolitan and larger micropolitan areas, this variable was excluded from the regression model that included the full sample from the 48 states. All analyses were performed using SAS version 9.2.
In the 48 contiguous states and the District of Columbia, in 2010, there were 3,920 practicing adult rheumatologists in the ACR database. The majority of member rheumatologists (3,512 [90%]) practiced in metropolitan areas. There were 134 members practicing in micropolitan areas (3%) and 274 members in rural regions (7%). Notably, a greater proportion of rheumatologists were practicing in metropolitan areas (versus micropolitan areas) than would be expected based on population distribution alone (P < 0.001). While only 31 metropolitan areas (9%) did not have a practicing member rheumatologist, the majority of micropolitan areas (84%) did not have a rheumatologist (Table 1).
|Metropolitan areas (n = 360)||Micropolitan areas (n = 573)||Rural areas|
|Definition, population||Urban area ≥50,000||Urban area ≥10,000 and <50,000, or urban cluster||Non-metropolitan/non-micropolitan|
|Total population, millions (% of study population)||256 (83)||31 (10)||22 (7)|
|No. of practicing adult rheumatologists (% of total 3,920)||3,512 (90)||134 (3)||274 (7)|
|Rheumatologists per 100,000 population, no. (%) of CBSAs||NA|
|CBSAs with no rheumatologist||31 (9)||479 (84)|
|CBSAs with <1 rheumatologist||127 (35)||10 (2)|
|CBSAs with 1–2 rheumatologists||145 (40)||46 (8)|
|CBSAs with ≥2 rheumatologists||57 (16)||38 (6)|
|Travel distance to nearest rheumatologist among CBSAs with no rheumatologist, no. of CBSAs||NA|
Among CBSAs that did not have a rheumatologist, the distance from the center of the CBSA to nearest practicing rheumatologist varied widely. While only a few metropolitan areas (1%) had travel distances of >75 miles, there were almost 100 micropolitan areas (16%) with travel distances of >75 miles (Table 1).
To further evaluate areas of potential unmet need, we then plotted distance to nearest practicing rheumatologist against population size of the CBSA. The plot (Figure 1) illustrates that there were some large CBSAs without a practicing rheumatologist, all with populations of >200,000, and some with travel distance to the nearest rheumatologist as great as 94 miles. For some smaller micropolitan regions, the travel distance to the nearest rheumatologist was >200 miles.
The geographic distribution of rheumatologists clearly demonstrated that there are many regions of the country without a practicing rheumatologist (Figure 2). Figure 3 illustrates the aggregated number of rheumatologists by CBSA, to describe the number of rheumatologists per 100,000 persons. Based on simply using the number of 1.67 rheumatologists per 100,000 persons as a definition of insufficient rheumatology supply (from the Lewin report) (), a majority of CBSAs (793 [85%]) would be defined as potentially underserved. However, it is more sensible to restrict that definition of insufficient supply to CBSAs where there is no practicing rheumatologist within a 50-mile travel distance; accordingly, 224 CBSAs (24%) and 18.9 million persons (7% of total population) would be affected. Restriction to a 75-mile or 100-mile travel distance would reduce the number of CBSAs and population count to 97 CBSAs (10%) with 9.3 million people (3%) (with a 75-mile cut point) or 51 CBSAs (5%) with 2.5 million people (1%) (with a 100-mile cut point).
To examine factors associated with rheumatologist distribution, we conducted multivariate analysis of correlates with the number of practicing rheumatologists per CBSA, for the entire population (933 CBSAs) and subsequently for only the 360 metropolitan areas (representing 83% of the entire population) (Table 2). In the full sample, greater number of rheumatologists per CBSA was associated with larger populations and greater population density, higher proportion of the population being younger, female, Caucasian, and Asian, and presence of an active adult rheumatology fellowship training program. Among the 360 metropolitan areas, for which data on median household income per CBSA were available, greater numbers of rheumatologists were found in CBSAs with higher income; however, race and population density were no longer significant.
|Full sample (n = 933 CBSAs)||Metropolitan sample (n = 360 CBSAs)|
|Total population count||+||+|
|Population density (persons per square mile)||+||NS|
|Proportion of population age ≥65 years||–||NS|
|Proportion of population female||+||+|
|Proportion of population by race|
|Proportion of population of Hispanic ethnicity||NS||NS|
|Median household income||NA||+|
|Active rheumatology fellowship training program||+||+|
The magnitude of these effects is depicted in Table 3, which shows the crude characteristics of CBSAs stratified by density of rheumatologists for the 360 metropolitan areas. For example, the results show a linear trend of increasing median household income with greater density of rheumatologists per CBSA. Not surprisingly, the number of rheumatology training programs was also correlated with greater density of rheumatologists. Analysis of other demographic variables, however, was more complex. Though statistically significant, the magnitude of the effect of male/female distribution was small. The magnitude of the effect of differences in median age was small and nonlinear across the rheumatologist density strata. While there were meaningful differences related to race/ethnicity variables, the associations were nonlinear and difficult to fully explain given the limitations of the administrative database. Detailed statistical results from the multivariate analysis are provided in Supplementary Table 1 (available on the Arthritis & Rheumatism web site at http://onlinelibrary.wiley.com/doi/10.1002/art.38167/abstract).
|No rheumatologists (n = 31 CBSAs)||<1 rheumatologist (n = 127 CBSAs)||1–2 rheumatologists (n = 145 CBSAs)||≥2 rheumatologists (n = 57 CBSAs)|
|No. of rheumatologists||0||576||1,985||951|
|Total population/mean population per CBSA||4,666,997/150,548||75,706,241/596,112||138,502,845/955,192||37,689,975/661,228|
|Total square miles/mean square miles per CBSA||43,325/1,398||354,106/2,788||386,204/2,663||110,804/1,944|
|Population density, per square mile||108||214||359||340|
|Population age ≥65 years, %||14||12||13||13|
|Weighted average median age, years||36.7||35.3||37.2||37.6|
|Hispanic ethnicity, %||13||23||18||10|
|Median annual household income, dollarsa||42,172||44,784||46,238||47,982|
|Rheumatology training programs and fellows|
|No. of programs||0||8||61||36|
|No. of slots||0||40||237||116|
|No. of fellows||0||10||98||53|
Prior reports on rheumatology workforce did not address regional variation in supply (or demand) in any detail (). Even so, those reports describe a growing shortage of rheumatologists based on an expected increased demand in the face of flat supply of new rheumatologists. Shortages of rheumatologists could lead to poorer patient outcomes. In circumstances where resources are limited, reduced utilization or utilization of lower-quality resources is often reported. In regions with fewer dual x-ray absorptiometry scanners, screening for osteoporosis is decreased (). In rural areas, patients are more likely to undergo elective joint replacement in low-volume hospitals (), where poorer patient outcomes are reported.
Our findings highlight that regional shortages in the workforce already exist. While this is no surprise to affected local patients, practitioners, or policy makers, the present study identifies potential target communities (based on population size and travel distance to the nearest rheumatologist) that might benefit most from addition of a local rheumatologist.
As noted above, both the federal government and other specialty organizations have identified this problem (with regard to generalists). The federal government identifies regions with shortages and poor health outcomes and provides funding through Federally Qualified Health Centers.
There are several potential interventions that could be implemented to increase the supply of rheumatologists in potentially underserved communities. Simply providing up-to-date information about the local supply of rheumatologists could attract more rheumatologists to underserved regions through migration, expansion (opening of a second practice site), or attraction of new rheumatologists (graduating fellows).
Given the association between rheumatologist supply and rheumatology training programs and trainees, increasing the supply of trainees in regions of unmet need would also likely improve local supply. Among the 793 CBSAs with <1.67 rheumatologists per 100,000 persons, there were 34 training programs with a total of 144 unfilled slots. Using the more restrictive definition of underserved area based on minimum 50-mile travel distance, there were 2 programs with 7 unfilled slots. Committing additional funds to training programs in underserved areas that otherwise lack adequate financial resources could be a viable policy option to increase the supply of local rheumatologists.
To address the general practitioner shortage in Medically Underserved Areas, through the Affordable Care Act, the federal government allocated $168 million for training primary care physicians () and $290 million through loan forgiveness programs () to support new physician practices. However, rheumatologists (and other specialists) are not eligible for this program.
Other novel solutions to address rural care have been implemented, which include examples such as traveling clinics, e-mail, or videoconsultation as either direct care rheumatologist–to-patient interview or peer-to-peer (P2P) consultation. ([17, 18]) Project ECHO (Extension for Community Healthcare Outcomes) at the University of New Mexico School of Medicine is one of the better-described telemedicine systems, which seeks to deliver specialty support for common but complex problems (such as hepatitis C or rheumatoid arthritis management) (). Reports cite greater physician satisfaction with P2P videoconferencing over P2P e-mail or visiting clinics ().
Deal and colleagues, summarizing the Lewin report (), and Birnbaum, responding to the Lewin report on behalf of the ACR (), suggested that midlevel practitioners could help fill the projected shortages in the rheumatology workforce. Several randomized controlled trials (in other areas of care) have demonstrated that midlevel practitioners provide efficient and effective delivery of care ().
The important and significant limitations of the present research project need to be considered. The most critical limitation of these analyses is the fact that practicing rheumatologists who are not members of the ACR are not included in the data set. There are several potential sources to supplement the data, including state licensure files and the American Medical Association (AMA) Masterfile. Unfortunately, however, there is no single data source that provides a complete list of practicing rheumatologists. The Lewin study showed that the AMA Masterfile and the ACR member database each included individuals not listed in the other. The ACR member database contains 90% of the rheumatologists identified in the AMA Masterfile. There are several benefits to the ACR member database for these analyses as it shows the type of rheumatology practice (adult versus pediatric), whether the individual is clinically active in providing patient care, and the office address.
It is possible that nonmember rheumatologists are more likely to practice in remote regions. A simple Internet search may identify rheumatologists practicing in areas not represented in the ACR member database. Missing data on a single rheumatologist from a CBSA would markedly alter the analyses that focused on CBSAs without a rheumatologist (e.g., Figure 1 and portions of Table 1). For this reason it is critical that these specific analyses should be considered preliminary and exploratory. With a more complete database, the methodology could be applied to target areas with greatest need. However, as the remainders of the analyses were based on continuous count data (e.g., Tables 2 and 3), those analyses are less susceptible to single missing observations. Prior to implementing policy decisions, it would be essential that the ACR obtain a more comprehensive database of practicing rheumatologists.
Furthermore, the present analyses treat all rheumatologists as equal full-time units, without consideration for part-time practice or mixed internal medicine/rheumatology practice. Nor does the analysis account for multiple location sites. (However, multiple practice locations are unlikely to affect these analyses, as separate offices would likely reside within the same CBSA.) The ACR is collecting these data with ongoing new clinical practice surveys.
These analyses also assume that demand for rheumatology care is uniform across populations. Though demand factors were not directly modeled, in the multivariate analyses we did adjust for age, sex, and race/ethnicity, all of which are correlated with regional variation in rheumatic conditions (and therefore demand for rheumatic care) ().
Competing sources of musculoskeletal care, such as internists, family practitioners, geriatricians, or midlevel providers, were also not taken into account in the present study. A more sophisticated supply model, in which these providers are considered as part of the workforce for musculoskeletal care (particularly for noninflammatory conditions), would provide more comprehensive information. Information about a rheumatologist's service area, which would be dependent upon population density and patient willingness to travel to see a rheumatologist (which would likely vary by diagnosis), is also needed.
Patient access to rheumatologists in their area also was not taken into account. In areas with few rheumatologists, those who are there are likely to be overworked, possibly not accepting new patients or having long wait times for scheduling of appointments. Furthermore, in busy practices, physicians might be more selective about seeing patients with no or poor health insurance coverage, further limiting access to care and exacerbating physician shortages for poorer patients.
The analyses of factors associated with number of rheumatologists per CBSA have important limitations. In addition to the usual limits of administrative data, income estimates were not available for 423 of the 573 smaller CBSA micropolitan areas, and this variable was excluded from the analysis of the full sample. While the magnitude and direction of the estimates were similar between the full model and the model addressing only the metropolitan areas, the results of tests of significance varied, suggesting that the analysis was sensitive to either the smaller sample size in the metropolitan area–only analysis or the fact that metropolitan and micropolitan areas differed by number of rheumatologists and other covariate factors (e.g., race/ethnicity). As noted above, the relationship between covariates and number of rheumatologists per CBSA is complex, and only partially addressed by this analysis.
In conclusion, in evaluating the rheumatology workforce, addressing workforce distribution and conducting small-area analyses are critical for identifying regions with underserved populations. However, better information is needed about factors that affect access to rheumatologists in underserved areas, including information about physician practices (such as part-time work, proportion of rheumatology care, number of office locations, wait times, local insurance mix, and distances patients will travel to see a consulting physician). As described above, the present analyses should be considered preliminary and exploratory. More comprehensive practice databases and additional research are needed prior to implementation of policies or reallocation of current resources. However, the present report identifies ongoing regional shortages in the rheumatology workforce and highlights potential target communities that might benefit most from addition of a local rheumatologist.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. FitzGerald 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 anaylsis.
Study conception and design. FitzGerald, Benford, Brown, Chakravarty, Abelson.
Acquisition of data. FitzGerald, Benford, Abelson.
Analysis and interpretation of data. FitzGerald, Benford, Battistone, Cannella, Elashoff, Gelber, Lozada, Punaro, Slusher, Abelson.
We are grateful to Christine Stamatos, ANP-C for providing feedback on content of the manuscript during its preparation.