- Email: [email protected]
Geographic Analysis of the Radiation Oncology Workforce
Int. J. Radiation Oncology Biol. Phys., Vol. 82, No. 5, pp. 1723–1729, 2012 Copyright Ó 2012 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/$ - see front matter
doi:10.1016/j.ijrobp.2011.01.070
CLINICAL INVESTIGATION
The Profession
GEOGRAPHIC ANALYSIS OF THE RADIATION ONCOLOGY WORKFORCE SANJAY ANEJA, B.S.,*y BENJAMIN D. SMITH, M.D.,z CARY P. GROSS, M.D.,yx LYNN D. WILSON, M.D., M.P.H.,* BRUCE G. HAFFTY, M.D.,{ KENNETH ROBERTS, M.D.,* y AND JAMES B. YU, M.D.* *Department of Therapeutic Radiology, Yale University School of Medicine, New Haven, CT; yCancer Outcomes, Policy, and Effectiveness Research Center at Yale, New Haven, CT; zUniversity of Texas M. D. Anderson Cancer Center, Houston, TX; x Department of General Internal Medicine, Yale University School of Medicine, New Haven, CT; and {Cancer Institute of New Jersey, New Brunswick, NJ Purpose: To evaluate trends in the geographic distribution of the radiation oncology (RO) workforce. Methods and Materials: We used the 1995 and 2007 versions of the Area Resource File to map the ratio of RO to the population aged 65 years or older (ROR) within different health service areas (HSA) within the United States. We used regression analysis to find associations between population variables and 2007 ROR. We calculated Gini coefficients for ROR to assess the evenness of RO distribution and compared that with primary care physicians and total physicians. Results: There was a 24% increase in the RO workforce from 1995 to 2007. The overall growth in the RO workforce was less than that of primary care or the overall physician workforce. The mean ROR among HSAs increased by more than one radiation oncologist per 100,000 people aged 65 years or older, from 5.08 per 100,000 to 6.16 per 100,000. However, there remained consistent geographic variability concerning RO distribution, specifically affecting the non-metropolitan HSAs. Regression analysis found higher ROR in HSAs that possessed higher education (p = 0.001), higher income (p < 0.001), lower unemployment rates (p < 0.001), and higher minority population (p = 0.022). Gini coefficients showed RO distribution less even than for both primary care physicians and total physicians (0.326 compared with 0.196 and 0.292, respectively). Conclusions: Despite a modest growth in the RO workforce, there exists persistent geographic maldistribution of radiation oncologists allocated along socioeconomic and racial lines. To solve problems surrounding the RO workforce, issues concerning both gross numbers and geographic distribution must be addressed. Ó 2012 Elsevier Inc. Workforce, Geography, Maldistribution, Disparity, Gini.
INTRODUCTION
diotherapy will increase 10 times faster than the supply of radiation oncologists (4). The radiation oncology workforce problem has likely exacerbated since ASTRO’s first report for at least three reasons. First, technological advancements in the field of radiation oncology have markedly increased the overall demand for radiotherapy to treat cancer, despite recent declines in cancer incidence (4, 5). Second, a supply gap has been created by modest increases in the number of training programs in the face of larger increases in clinical demands (4, 6). Third, technological advancements in the field that not only increased the quantity of treatment but also physician planning time have potentially decreased the overall efficiency of the radiation oncologist workforce (4, 7). Previous analyses of the radiation oncology workforce have focused solely on radiation oncologist numbers and projecting the need for the future. Few studies, however, have addressed the geographic distribution of radiation oncologists. Geographic access to care has known associations
The physician workforce crisis has remained a focal point of recent healthcare legislation (1). With the expansion of healthcare coverage nationally, the Center for Workforce Studies projects the gap in the physician workforce to increase by 50% by the year 2050. Specifically, the shortage of specialists is expected to quadruple (2). Because of this known association between access to care and healthcare quality, the Patient Protection and Affordable Care Act attempted to address the physician workforce issue through the establishment of the National Healthcare Workforce Commission, whose purpose is to monitor the supply and distribution of physicians. The workforce issue is not new to the oncology or radiotherapy communities. In 2002 the American Society for Radiation Oncology (ASTRO) Workforce Committee was the first to note shortages in the future of radiation oncology (3). More recently, however, it has been projected that the demand for ra-
Reprint requests to: James B. Yu, M.D., Yale School of Medicine, Smilow LL 511B, 333 Cedar Street, New Haven, CT 06520. Tel: (203) 785-5703; Fax: (203) 785-4622; E-mail: james. [email protected]
Conflicts of interest: none. Received Nov 28, 2010, and in revised form Jan 21, 2011. Accepted for publication Jan 29, 2011. 1723
I. J. Radiation Oncology d Biology d Physics
1724
with cancer-related outcomes. Increased urologist density has been shown to be associated with decreased prostate and kidney cancer mortality (8). Though some disciplines may adapt to a lack of geographic distribution via the use of telemedicine or physician extenders, access to radiation oncology services is particularly dependent on geographic distribution for several reasons. First, a clinical course of radiotherapy requires multiple daily trips for treatment, for up to 8 or 9 weeks in some cases. Second, unlike other specialties, gaps in care caused by lack of radiation oncologists cannot be as readily filled by primary care physicians and physician extenders because of the highly technical nature of radiation oncology. Third, radiotherapy requires significant amounts of large, immobile, and highly technical equipment that cannot feasibly be transported (9). Descriptive analysis of the geographic distribution of the radiation oncology workforce is a necessary first step to help inform policymakers and clinicians regarding ways to best provide radiotherapy and multidisciplinary cancer care. To investigate these issues, we examined the geographic distribution of radiation oncologists in 1995 and 2007. Specifically, we studied changes in the ratio of radiation oncologists to population aged 65 years or older (ROR) among health service areas (HSAs) within the United States. To determine how the radiation oncology distribution compares with that of the primary care and total physician workforces, we compared the evenness of ROR distribution with equivalent ratios of primary care physicians (PCPs) and total physicians (MDs). We also examined the population characteristics that are associated with the current geographic distribution of the radiation oncologists. METHODS AND MATERIALS Data sources A multistep approach was used to construct the dataset for this study. Using the 1995 and 2007 editions of the Area Resource File (ARF), we obtained demographic, population, and physician distribution data (10). Published by the Health Resources and Services Administration of the U.S. Department of Health and Human Services, the ARF is a collection of data from more than 50 sources, including the American Medical Association (AMA), American Hospitalization Association, U.S. Census, and National Center for U.S. Health Statistics. The ARF aggregates information concerning the healthcare professionals, healthcare facilities, and population for each county in the United States. Specifically, the ARF includes the number of specialists within each county according to data from the AMA Physician Masterfile. The geographic units of analysis were the 949 HSAs within the United States as defined by the National Center for Health Statistics. A health service area is defined as a single county or group of contiguous counties that remain self-contained with respect to hospital care. Health service areas were chosen as the unit of analysis because they best represent geographic access to healthcare within a region. County-level data from the ARF was aggregated to HSAs using simple summation for physician and population variables and population-weighted sums for the descriptive variables.
Volume 82, Number 5, 2012
Ratio of radiation oncologists to population aged 65 years or older The physician-to-population ratio remains a frequently used measure of physician distribution within a region (8, 11). The primary outcome of our study was the ROR. The elderly were chosen as the population of interest because they represent a group that has the highest prevalence of cancer and represent a demographic that utilizes a significant portion of healthcare services in the United States (4, 12). In addition, elderly patients may be less mobile than younger patients and may be more hesitant to travel long distances for their care, meaning geographic proximity is of utmost importance. The ROR of each HSA was calculated as the number of radiation oncologists per 100,000 people aged 65 years or older. The RORs for each HSA were generated for the years 1995 and 2007. To examine the relationship between radiation oncology and other physician specialties, equivalent ratios were calculated for the PCP (PCPR) and total physician (MDR) workforces. Primary care physicians were defined as physicians in general practice, family practice, or general internal medicine.
Descriptive analysis To evaluate changes in the distribution of the workforce, mean RORs were calculated for 1995 and 2007 and compared with equivalent changes in PCPR and MDR. To assess the current distribution of radiation oncologists, each HSA was ranked according to ROR and compared with equivalent rankings of PCPRs and MDRs. In an effort to visually compare the distribution of radiation oncologists between different HSAs and trends in their distribution, ROR values were mapped to corresponding HSAs using the geographical information system ArcGIS, version 9.2 (Environmental Systems Research Institute, Redlands, CA).
Lorenz curves and Gini coefficients Traditionally used to assess distribution of wealth, Gini coefficients and Lorenz curves have recently been used to evaluate physician distribution (11, 13, 14). The Lorenz curve was constructed by graphing the cumulative percentage of radiation oncologists as ranked by RORs vs. the cumulative percentage of the population age 65 years or older. A 45 line representing a perfectly even distribution was drawn from the origin to the maximum point of the Lorenz curve. The Gini coefficient was calculated by dividing the area between the Lorenz curve and the 45 line by the total area under the 45 line. The Gini coefficient ranges in value from 0 (complete equity) to 1 (complete inequity) and serves as a quantitative way to describe relative ‘‘evenness’’ of physician distribution. Because Gini coefficients possess a linear relationship, changes in value can be used to evaluate changes in the evenness of physician distributions over time. Lorenz curves and Gini coefficients were generated to examine the radiation oncologist distribution among all HSAs between 1995 and 2007. Subgroup analysis was performed to compare the evenness of radiation oncologist distribution, specifically in less-populated HSAs, using population quartiles. To assess the uniqueness of the radiation oncologist distribution compared with all other physicians and PCPs, Lorenz curves and Gini coefficients were also calculated for all physicians and PCPs between 1995 and 2007. The Lorenz curve and Gini coefficient calculations were conducted using Stata, version 9.2 (StataCorp, College Station, TX).
Regression analysis To evaluate population characteristics potentially associated with the current geographic distribution of radiation oncologists,
Radiation workforce in the United States d S. ANEJA et al.
regression analysis was used. After assessing distribution, mean, and variance of HSA-level ROR, zero-inflated negative binomial regression was chosen. The dependent variable was ROR for the year 2007. The independent variables included county-level population race, income, education, and unemployment rate that were aggregated to HSA level using weighted sums. Population race was defined as percentage white population within each HSA from the 2000 Census. Population income was defined as the median household income according to the 2007 Census update. Population education was defined as percentage population aged 25 years or older with at least a high school education according to the 2000 Census. Finally, unemployment rates for each HSA were from the 2007 Bureau of Labor Statistics estimates. Independent multivariate regression models were built using backwardsstepwise selection with a univariate p < 0.15 for inclusion into the model. A Vuong test was used to assess the appropriateness of the model. Statistical significance was determined at p < 0.05. Statistical analysis was conducted using Stata 9.2.
RESULTS Changes in radiation oncology workforce In the 12-year period of our study, the radiation oncology workforce grew approximately 24%, from 3515 radiation oncologists in 1995 to 4378 in 2007. Over the same time period, the PCP workforce grew approximately 31%, from 213,619 physicians to 278,728. Similarly, the overall physician workforce grew 29%, from 617,362 physicians to 794,184. The mean ROR increased by slightly more than one radiation oncologist per 100,000 people (Table 1). In contrast, mean PCPR and MDR increased by 85 physicians per 100,000 people and 145 physicians per 100,000 people, respectively. Increases in the mean ROR, although modest, suggest a growth in the radiation oncology workforce that outpaced the growth of the elderly population. The distribution of radiation oncologists was significantly more skewed than for PCPs and MDs. In the year 2007, approximately 44% of HSAs within the United States lacked a radiation oncologist. Comparatively, approximately 0.74% of HSAs lacked a PCP, and 0.63% lacked a physician of any kind. Moreover, approximately 75% of HSAs had two or fewer radiation oncologists per 100,000 people aged 65 years or older. Additionally, there existed consistent geographic maldistribution of radiation oncologists from 1995 to 2007 (Fig. 1). Health service areas within the Northeast, California, and Florida exhibited high RORs in both 1995 and 2007, whereas rural HSAs within the Midwest generally
1725
exhibited lower RORs in 1995 and 2007 relative to the rest of the country (Fig. 2). Equitable distribution of the radiation oncologist workforce Gini coefficient calculations confirmed the maldistribution of the radiation oncology workforce. Radiation oncologists were less evenly distributed than both PCPs and MDs in both 1995 and 2007 (Table 2), However, the Gini coefficients of all three groups exhibited a percentage decrease, indicating an improvement in distribution toward equity. Radiation oncologists exhibited the largest change (10.93%), followed by PCPs (4.85%) and finally MDs (2.34%). Analysis of Gini coefficients within population quartiles highlighted the uneven distribution of radiation oncologists in non-metropolitan areas. Specifically, we found that, unlike PCPs and MDs, who possessed relatively similar Gini coefficients among different population quartiles, the Gini coefficient of radiation oncologists in the lowest population quartile was an inequitable 0.951, compared with 0.251 in the highest population quartile (Table 3). The difference between Gini coefficients among population quartiles suggests that the geographic maldistribution of radiation oncologists stems primarily from inequity in less-populated non-metropolitan HSAs. Factors associated with radiation oncologist distribution Regression analysis found an association between radiation oncologist distribution and population characteristics. Univariate zero inflated negative binomial regression showed that increased ROR in 2007 was associated with HSAs with lower unemployment rates (p < 0.001), higher household incomes (p < 0.001), and higher education rates (p = 0.001) (Table 4). Surprisingly, decreased ROR in 2007 was associated with HSAs that had higher percent white population (p = 0.022). Multivariate analysis confirmed the association between increased ROR and lower unemployment rates (p < 0.001), higher household incomes (p < 0.001), and increased minority population (p = 0.010). Higher education rates proved to be insignificant in the multivariate model (p = 0.461). Because the association between minority population and ROR was unexpected, a separate model using percent African American population was built to confirm the association with minority groups. The confirmatory model yielded similar results (p < 0.001). DISCUSSION
Table 1. Changes in mean physician to population aged 65 years or older ratio, 1995–2007 Parameter Radiation oncologists per 100,000 PCP per 100,000 MD per 100,000
1995
2007
% Change
5.08
6.16
21.26
521.67 1056.44
606.50 1201.35
16.26 13.72
Abbreviations: PCP = primary care physician; MD = physician (doctor of medicine).
From 1995 through 2007, although the radiation oncology workforce increased slightly, it remained geographically maldistributed. Radiation oncologists remained concentrated in primarily highly populated metropolitan HSAs of the country, leaving large segments of the United States lacking access to radiotherapy. Additionally, geographic access to radiotherapy is associated with HSAs that have higher socioeconomic characteristics and higher minority makeup. Recent evidence has suggested that the demand
1726
I. J. Radiation Oncology d Biology d Physics
Volume 82, Number 5, 2012
Fig. 1. Distribution of radiation oncologists among health service areas: 1995 and 2007.
for radiotherapy will undoubtedly outpace the growth in the workforce, potentially exacerbating current geographic disparities (4). Our study suggests that the physician workforce problem should not merely focus on physician numbers but also the geographic distribution of physicians.
The growth in the radiation oncology workforce was modest compared with the overall growth in the physician workforce. The radiation oncology workforce grew 5% less than the overall physician workforce and 9% less than the primary care workforce. In a 12-year period, the mean ROR
Fig. 2. Ratios of radiation oncologists to population aged 65 years or older among health service areas: 1995 and 2007.
Radiation workforce in the United States d S. ANEJA et al.
Table 2. Trends of Gini coefficients among different physicians, 1995–2007 Physician type
1995
2007
Change
% Change
Radiation oncologist PCP MD
0.366 0.206 0.299
0.326 0.196 0.292
0.040 0.010 0.007
10.93 4.85 2.34
Abbreviation as in Table 1.
increased by slightly more than one radiation oncologist per 100,000 people aged 65 years or older, compared with larger increases in the PCPR and MDR. The modest growth in the radiation oncology workforce is most likely due to increasing residency positions for radiation oncology (15–17), Despite growth in the workforce, there was persistent geographic variation in the distribution of radiation oncologists in 1995 and 2007. Mapping of the RORs for 1995 and 2007 showed large geographic segments of the elderly population having little or no access to radiotherapy services, specifically in non-metropolitan areas. In 2007, an alarming 44% of HSAs in the United States lacked radiation oncologists. This translated to 3,137,580 people aged 65 years or older without access to radiotherapy in their HSA. Comparatively, Gini analysis showed that radiation oncologists were less evenly distributed than both PCPs and the overall physician workforce. Our finding that PCPs were the most evenly distributed replicates findings from previous studies and suggests that recent policy initiatives to increase geographic access to primary care have been successful (11, 18). More interestingly, the main driver in the uneven geographic distribution of radiation oncologists was an inequitable distribution in less-populated non-metropolitan HSAs. The lack of access to radiation oncologists in less-populated non-metropolitan areas is perhaps due to the large capital investment required to obtain the equipment and resources necessary to establish a radiation oncology practice. These impediments are increased in rural areas, where patient population levels are potentially less. Conversely, large pockets of radiation oncologists were centered in metropolitan areas with large academic centers. This finding corroborates previous findings from the ASTRO workforce committee that groups of 10 or more radiation oncologists are more likely to be in academic centers (3). The clustering of radiation oncologists in academic centers is perhaps due to recent increases in academic radiation oncology due to an influx of physician scientists in radiation oncology residency programs, or due to the attractiveness of technological resources available at academic centers (19). Table 3. Gini coefficients among population quartiles, 2007 Physician type Radiation oncologist PCP MD
Quartile 1
Quartile 2
Quartile 3
Quartile 4
0.951
0.675
0.402
0.251
0.268 0.333
0.233 0.258
0.232 0.265
0.169 0.236
Abbreviation as in Table 1.
1727
Our analysis found that access to radiation oncology services was allocated along socioeconomic and racial characteristics and provides information for policymakers regarding potential factors that affect radiation oncologist distribution, and could be used for future legislation aimed at physician recruitment to underserved areas. Although Hayanga et al. (20) found access to radiation oncologists to be associated with decreased minority population on the county level, our analysis found a positive association between minority population and ROR within HSAs. The reason for this difference is partly that our negative binomial regression model more accurately predicted radiation oncologist density given the distribution compared with the linear regression used in their analysis. Additionally, our geographic units of analysis were HSAs, which better predict patterns of healthcare usage compared with counties. Additionally, unlike previous analysis, our ROR outcome variable accounts for the variation in population density across the United States and is more accurate than gross physician totals. Moreover, unlike previous analyses, our analysis excluded counties located in U.S. territories, such as Guam, Puerto Rico, and the U.S. Virgin Islands, that generally have disproportionate minority populations and skewed access to specialists. Our findings suggest that, although racial disparities in cancer outcomes have been well documented, minority populations generally reside within HSAs in close proximity to radiation oncologists (21). Our findings highlight the problem of physician recruitment to rural areas. Recently, rural physician recruitment efforts have faced obstacles because of the technological gap that exists in rural settings. This problem is evermore important in a highly technical field like radiation oncology and has translated to patient dissatisfaction as physician retention rates in rural areas have decreased (19). Traditional policymakers have theorized that market forces and rural incentives would decrease geographic disparities as the workforce increases (22). However, we found little or no decrease in geographic variability of radiation oncologists to accompany a growth in the workforce. The significant non-physician personnel needed to operate radiotherapy facilities could also be a factor in the geographic maldistribution of radiation oncologists. The workforce shortages of radiation oncologists exacerbated with corresponding workforce shortages in non-physician personnel, such as dosimetrists and medical physicists (23), perhaps making it difficult provide cost-effective quality radiotherapy in rural areas in the United States. Recently, studies have suggested that the solution to geographic maldistribution is not increased recruitment to rural areas, but rather technical innovations that increase the efficiency of healthcare for rural populations (24–27). An example of a treatment innovation, specifically in radiotherapy, is hypofractionated treatment, as performed for breast radiotherapy in Canada (the so-called ‘‘Canadian fractionation’’), where geographic distance between radiation oncology centers is even more severe than in the United States.
I. J. Radiation Oncology d Biology d Physics
1728
Volume 82, Number 5, 2012
Table 4. Zero inflated negative binomial regression analysis using 2007 ROR as dependent variable Univariate analysis
Multivariate analysis
Independent variable
IRR
p
95% CI
IRR
p
% White population Unemployment rate % High school education Median household income
0.995859 0.889519 1.000001 1.000016
0.022 <0.001 0.001 <0.001
0.992335–0.999395 0.858483–0.921678 1.000000–1.000002 1.000011–1.00002
0.995351 0.906783 1.000005 1.000010
0.010 <0.001 0.461 <0.001
95% CI 0.991843 0.872964 0.999999 1.000005
0.978397 0.891379 1.000015 1.000015
Abbreviations: ROR = ratio of radiation oncologists to population aged 65 years or older; IRR = incidence rate ratio; CI = confidence interval.
Our analysis has several limitations. First, the physician location data from the AMA Masterfile does not capture physicians who have multiple practices in different regions, and therefore it is possible that we have overestimated the geographic clustering of radiation oncologists. However, although we believe that radiation oncologists are more likely to have multiple physician practice locations compared with primary care providers, we suspect that if radiation oncologists have multiple practices that they would be within the same HSA. Unfortunately, this cannot be confirmed by our data. Second, the Masterfile has been shown to underestimate physician shortages in rural areas (28). Third, our study does not capture other barriers to care, such as lack of health insurance and whether radiation oncology centers accept Medicaid payment for their services. Finally, our analysis cannot provide an optimal ROR because we did not relate geographic distribution to clinical outcomes. Future analysis
should focus on finding the radiation oncologist distribution that will optimize cancer outcomes. CONCLUSION Our study is the first geographic analysis of the radiation oncology workforce. Despite an overall growth in the radiation oncology workforce from 1995 through 2007, there remains persistent geographic maldistribution of radiation oncologists, with radiation oncologists concentrated in urban and metropolitan regions. Additionally, the distribution of radiation oncologists was related to the socioeconomic and racial characteristics of a region. To effectively solve the workforce issues facing radiation oncology and provide adequate care, we must focus not only on gross numbers of radiation oncologists, but also geographic distribution.
REFERENCES 1. American Association of Medical Colleges Center for Workforce Studies. Health care reform and the health workforce. 2010. Available at: https://www.aamc.org/download/142078/data/health_ workforce_provisions_and_health_care_reform.pdf. Last accessed April 4, 2011. 2. Wang S. Med schools: Health-care overhaul to accelerate doctor shortage. Wall Street Journal 2010. Available at: http://blogs.wsj.com/health/2010/09/30/med-schools-healthcare-overhaul-to-accelerate-doctor-shortage/ Last accessed April 4, 2011. 3. ASTRO Workforce Committee. 2002 Radiation Oncology Workforce Study: American Society for Therapeutic Radiology and Oncology. Int J Radiat Oncol Biol Phys 2003;56:309–318. 4. Smith BD, Haffty BG, Wilson LD, et al. The future of radiation oncology in the United States from 2010 to 2020: Will supply keep pace with demand? J Clin Oncol 2010;28:5160–5165. 5. Bogdanich W. The radiation boom: As technology surges, radiation safeguards lag. New York Times 2010. Available at: http://www.nytimes.com/2010/01/27/us/27radiation.html?_r=1. Last accessed April 4, 2011. 6. Lewis RS, Sunshine JH. Radiation oncologists in the United States. Int J Radiat Oncol Biol Phys 2007;69:518–527. 7. Elkin EB, Bach PB. Cancer’s next frontier: Addressing high and increasing costs. JAMA 2010;303:1086–1087. 8. Odisho AY, Cooperberg MR, Fradet V, et al. Urologist density and county-level urologic cancer mortality. J Clin Oncol 2010; 28:2499–2504. 9. Amory CF, Levine SR. Is telemedicine the answer for delivering stroke care to rural areas? Nat Clin Pract Neurol 2007;3:192–193.
10. U.S. Department of Health and Human Services, Health Resources and Services Administration. Area resource file (ARF): National county-level health resource information database. Available at: http://www.arfsys.com. Last accessed June 14, 2010. 11. Chang RK, Halfon N. Geographic distribution of pediatricians in the United States: An analysis of the fifty states and Washington, DC. Pediatrics 1997;100:172–179. 12. Smith BD, Smith GL, Hurria A, et al. Future of cancer incidence in the United States: Burdens upon an aging, changing nation. J Clin Oncol 2009;27:2758–2765. 13. Matsumoto M, Inoue K, Farmer J, et al. Geographic distribution of primary care physicians in Japan and Britain. Health Place 2010;16:164–166. 14. Matsumoto M, Inoue K, Bowman R, et al. Self-employment, specialty choice, and geographical distribution of physicians in Japan: A comparison with the United States. Health Policy 2010;96:239–244. 15. Wilson LD, Yu JB, Haffty BG. National Residency Matching Program (NRMP) results for radiation oncology: 2007 update. Int J Radiat Oncol Biol Phys 2007;69:326–327. 16. Wilson LD, Haffty BG. National Residency Matching Program results for radiation oncology: 2006 update. Int J Radiat Oncol Biol Phys 2006;66:321–322. 17. Wilson LD, Haffty BG. National Residency Matching Program results for radiation oncology, 2005 update. Int J Radiat Oncol Biol Phys 2005;63:3–4. 18. Rittenhouse DR, Fryer GE Jr., Phillips RL Jr., et al. Impact of Title VII training programs on community health center
Radiation workforce in the United States d S. ANEJA et al.
19.
20.
21. 22.
staffing and National Health Service Corps participation. Ann Fam Med 2008;6:397–405. Fears D. Renewed effort to lure doctors to rural areas faces obstacles. Washington Post 2010. Available at: http://www. washingtonpost.com/wp-dyn/content/article/2010/08/08/AR20 10080802832.html?nav=emailpage. Last accessed April 4, 2011. Hayanga AJ, Waljee AK, Kaiser HE, et al. Racial clustering and access to colorectal surgeons, gastroenterologists, and radiation oncologists by African Americans and Asian Americans in the United States: A county-level data analysis. Arch Surg 2009;144:532–535. Gross CP, Smith BD, Wolf E, et al. Racial disparities in cancer therapy: Did the gap narrow between 1992 and 2002? Cancer 2008;112:900–908. Newhouse JP. Geographic access to physician services. Annu Rev Public Health 1990;11:207–230.
1729
23. Mills MD, Thornewill J, Esterhay RJ. Future trends in the supply and demand for radiation oncology physicists. J Appl Clin Med Phys 2010;11:3005. 24. Goodman DC, Grumbach K. Does having more physicians lead to better health system performance? JAMA 2008;299: 335–337. 25. Goodman DC, Fisher ES. Physician workforce crisis? Wrong diagnosis, wrong prescription. N Engl J Med 2008;358:1658–1661. 26. Goodman DC, Fisher E, Stukel TA, et al. The distance to community medical care and the likelihood of hospitalization: Is closer always better? Am J Public Health 1997;87:1144–1150. 27. Goodman DC. Do we need more physicians? Health Aff (Millwood) 2004. Suppl Web Exclusives:W4-67–;W4–69. 28. Shea JA, Kletke PR, Wozniak GD, et al. Self-reported physician specialties and the primary care content of medical practice: A study of the AMA physician masterfile. American Medical Association. Med Care 1999;37:333–338.
doi:10.1016/j.ijrobp.2011.01.070
CLINICAL INVESTIGATION
The Profession
GEOGRAPHIC ANALYSIS OF THE RADIATION ONCOLOGY WORKFORCE SANJAY ANEJA, B.S.,*y BENJAMIN D. SMITH, M.D.,z CARY P. GROSS, M.D.,yx LYNN D. WILSON, M.D., M.P.H.,* BRUCE G. HAFFTY, M.D.,{ KENNETH ROBERTS, M.D.,* y AND JAMES B. YU, M.D.* *Department of Therapeutic Radiology, Yale University School of Medicine, New Haven, CT; yCancer Outcomes, Policy, and Effectiveness Research Center at Yale, New Haven, CT; zUniversity of Texas M. D. Anderson Cancer Center, Houston, TX; x Department of General Internal Medicine, Yale University School of Medicine, New Haven, CT; and {Cancer Institute of New Jersey, New Brunswick, NJ Purpose: To evaluate trends in the geographic distribution of the radiation oncology (RO) workforce. Methods and Materials: We used the 1995 and 2007 versions of the Area Resource File to map the ratio of RO to the population aged 65 years or older (ROR) within different health service areas (HSA) within the United States. We used regression analysis to find associations between population variables and 2007 ROR. We calculated Gini coefficients for ROR to assess the evenness of RO distribution and compared that with primary care physicians and total physicians. Results: There was a 24% increase in the RO workforce from 1995 to 2007. The overall growth in the RO workforce was less than that of primary care or the overall physician workforce. The mean ROR among HSAs increased by more than one radiation oncologist per 100,000 people aged 65 years or older, from 5.08 per 100,000 to 6.16 per 100,000. However, there remained consistent geographic variability concerning RO distribution, specifically affecting the non-metropolitan HSAs. Regression analysis found higher ROR in HSAs that possessed higher education (p = 0.001), higher income (p < 0.001), lower unemployment rates (p < 0.001), and higher minority population (p = 0.022). Gini coefficients showed RO distribution less even than for both primary care physicians and total physicians (0.326 compared with 0.196 and 0.292, respectively). Conclusions: Despite a modest growth in the RO workforce, there exists persistent geographic maldistribution of radiation oncologists allocated along socioeconomic and racial lines. To solve problems surrounding the RO workforce, issues concerning both gross numbers and geographic distribution must be addressed. Ó 2012 Elsevier Inc. Workforce, Geography, Maldistribution, Disparity, Gini.
INTRODUCTION
diotherapy will increase 10 times faster than the supply of radiation oncologists (4). The radiation oncology workforce problem has likely exacerbated since ASTRO’s first report for at least three reasons. First, technological advancements in the field of radiation oncology have markedly increased the overall demand for radiotherapy to treat cancer, despite recent declines in cancer incidence (4, 5). Second, a supply gap has been created by modest increases in the number of training programs in the face of larger increases in clinical demands (4, 6). Third, technological advancements in the field that not only increased the quantity of treatment but also physician planning time have potentially decreased the overall efficiency of the radiation oncologist workforce (4, 7). Previous analyses of the radiation oncology workforce have focused solely on radiation oncologist numbers and projecting the need for the future. Few studies, however, have addressed the geographic distribution of radiation oncologists. Geographic access to care has known associations
The physician workforce crisis has remained a focal point of recent healthcare legislation (1). With the expansion of healthcare coverage nationally, the Center for Workforce Studies projects the gap in the physician workforce to increase by 50% by the year 2050. Specifically, the shortage of specialists is expected to quadruple (2). Because of this known association between access to care and healthcare quality, the Patient Protection and Affordable Care Act attempted to address the physician workforce issue through the establishment of the National Healthcare Workforce Commission, whose purpose is to monitor the supply and distribution of physicians. The workforce issue is not new to the oncology or radiotherapy communities. In 2002 the American Society for Radiation Oncology (ASTRO) Workforce Committee was the first to note shortages in the future of radiation oncology (3). More recently, however, it has been projected that the demand for ra-
Reprint requests to: James B. Yu, M.D., Yale School of Medicine, Smilow LL 511B, 333 Cedar Street, New Haven, CT 06520. Tel: (203) 785-5703; Fax: (203) 785-4622; E-mail: james. [email protected]
Conflicts of interest: none. Received Nov 28, 2010, and in revised form Jan 21, 2011. Accepted for publication Jan 29, 2011. 1723
I. J. Radiation Oncology d Biology d Physics
1724
with cancer-related outcomes. Increased urologist density has been shown to be associated with decreased prostate and kidney cancer mortality (8). Though some disciplines may adapt to a lack of geographic distribution via the use of telemedicine or physician extenders, access to radiation oncology services is particularly dependent on geographic distribution for several reasons. First, a clinical course of radiotherapy requires multiple daily trips for treatment, for up to 8 or 9 weeks in some cases. Second, unlike other specialties, gaps in care caused by lack of radiation oncologists cannot be as readily filled by primary care physicians and physician extenders because of the highly technical nature of radiation oncology. Third, radiotherapy requires significant amounts of large, immobile, and highly technical equipment that cannot feasibly be transported (9). Descriptive analysis of the geographic distribution of the radiation oncology workforce is a necessary first step to help inform policymakers and clinicians regarding ways to best provide radiotherapy and multidisciplinary cancer care. To investigate these issues, we examined the geographic distribution of radiation oncologists in 1995 and 2007. Specifically, we studied changes in the ratio of radiation oncologists to population aged 65 years or older (ROR) among health service areas (HSAs) within the United States. To determine how the radiation oncology distribution compares with that of the primary care and total physician workforces, we compared the evenness of ROR distribution with equivalent ratios of primary care physicians (PCPs) and total physicians (MDs). We also examined the population characteristics that are associated with the current geographic distribution of the radiation oncologists. METHODS AND MATERIALS Data sources A multistep approach was used to construct the dataset for this study. Using the 1995 and 2007 editions of the Area Resource File (ARF), we obtained demographic, population, and physician distribution data (10). Published by the Health Resources and Services Administration of the U.S. Department of Health and Human Services, the ARF is a collection of data from more than 50 sources, including the American Medical Association (AMA), American Hospitalization Association, U.S. Census, and National Center for U.S. Health Statistics. The ARF aggregates information concerning the healthcare professionals, healthcare facilities, and population for each county in the United States. Specifically, the ARF includes the number of specialists within each county according to data from the AMA Physician Masterfile. The geographic units of analysis were the 949 HSAs within the United States as defined by the National Center for Health Statistics. A health service area is defined as a single county or group of contiguous counties that remain self-contained with respect to hospital care. Health service areas were chosen as the unit of analysis because they best represent geographic access to healthcare within a region. County-level data from the ARF was aggregated to HSAs using simple summation for physician and population variables and population-weighted sums for the descriptive variables.
Volume 82, Number 5, 2012
Ratio of radiation oncologists to population aged 65 years or older The physician-to-population ratio remains a frequently used measure of physician distribution within a region (8, 11). The primary outcome of our study was the ROR. The elderly were chosen as the population of interest because they represent a group that has the highest prevalence of cancer and represent a demographic that utilizes a significant portion of healthcare services in the United States (4, 12). In addition, elderly patients may be less mobile than younger patients and may be more hesitant to travel long distances for their care, meaning geographic proximity is of utmost importance. The ROR of each HSA was calculated as the number of radiation oncologists per 100,000 people aged 65 years or older. The RORs for each HSA were generated for the years 1995 and 2007. To examine the relationship between radiation oncology and other physician specialties, equivalent ratios were calculated for the PCP (PCPR) and total physician (MDR) workforces. Primary care physicians were defined as physicians in general practice, family practice, or general internal medicine.
Descriptive analysis To evaluate changes in the distribution of the workforce, mean RORs were calculated for 1995 and 2007 and compared with equivalent changes in PCPR and MDR. To assess the current distribution of radiation oncologists, each HSA was ranked according to ROR and compared with equivalent rankings of PCPRs and MDRs. In an effort to visually compare the distribution of radiation oncologists between different HSAs and trends in their distribution, ROR values were mapped to corresponding HSAs using the geographical information system ArcGIS, version 9.2 (Environmental Systems Research Institute, Redlands, CA).
Lorenz curves and Gini coefficients Traditionally used to assess distribution of wealth, Gini coefficients and Lorenz curves have recently been used to evaluate physician distribution (11, 13, 14). The Lorenz curve was constructed by graphing the cumulative percentage of radiation oncologists as ranked by RORs vs. the cumulative percentage of the population age 65 years or older. A 45 line representing a perfectly even distribution was drawn from the origin to the maximum point of the Lorenz curve. The Gini coefficient was calculated by dividing the area between the Lorenz curve and the 45 line by the total area under the 45 line. The Gini coefficient ranges in value from 0 (complete equity) to 1 (complete inequity) and serves as a quantitative way to describe relative ‘‘evenness’’ of physician distribution. Because Gini coefficients possess a linear relationship, changes in value can be used to evaluate changes in the evenness of physician distributions over time. Lorenz curves and Gini coefficients were generated to examine the radiation oncologist distribution among all HSAs between 1995 and 2007. Subgroup analysis was performed to compare the evenness of radiation oncologist distribution, specifically in less-populated HSAs, using population quartiles. To assess the uniqueness of the radiation oncologist distribution compared with all other physicians and PCPs, Lorenz curves and Gini coefficients were also calculated for all physicians and PCPs between 1995 and 2007. The Lorenz curve and Gini coefficient calculations were conducted using Stata, version 9.2 (StataCorp, College Station, TX).
Regression analysis To evaluate population characteristics potentially associated with the current geographic distribution of radiation oncologists,
Radiation workforce in the United States d S. ANEJA et al.
regression analysis was used. After assessing distribution, mean, and variance of HSA-level ROR, zero-inflated negative binomial regression was chosen. The dependent variable was ROR for the year 2007. The independent variables included county-level population race, income, education, and unemployment rate that were aggregated to HSA level using weighted sums. Population race was defined as percentage white population within each HSA from the 2000 Census. Population income was defined as the median household income according to the 2007 Census update. Population education was defined as percentage population aged 25 years or older with at least a high school education according to the 2000 Census. Finally, unemployment rates for each HSA were from the 2007 Bureau of Labor Statistics estimates. Independent multivariate regression models were built using backwardsstepwise selection with a univariate p < 0.15 for inclusion into the model. A Vuong test was used to assess the appropriateness of the model. Statistical significance was determined at p < 0.05. Statistical analysis was conducted using Stata 9.2.
RESULTS Changes in radiation oncology workforce In the 12-year period of our study, the radiation oncology workforce grew approximately 24%, from 3515 radiation oncologists in 1995 to 4378 in 2007. Over the same time period, the PCP workforce grew approximately 31%, from 213,619 physicians to 278,728. Similarly, the overall physician workforce grew 29%, from 617,362 physicians to 794,184. The mean ROR increased by slightly more than one radiation oncologist per 100,000 people (Table 1). In contrast, mean PCPR and MDR increased by 85 physicians per 100,000 people and 145 physicians per 100,000 people, respectively. Increases in the mean ROR, although modest, suggest a growth in the radiation oncology workforce that outpaced the growth of the elderly population. The distribution of radiation oncologists was significantly more skewed than for PCPs and MDs. In the year 2007, approximately 44% of HSAs within the United States lacked a radiation oncologist. Comparatively, approximately 0.74% of HSAs lacked a PCP, and 0.63% lacked a physician of any kind. Moreover, approximately 75% of HSAs had two or fewer radiation oncologists per 100,000 people aged 65 years or older. Additionally, there existed consistent geographic maldistribution of radiation oncologists from 1995 to 2007 (Fig. 1). Health service areas within the Northeast, California, and Florida exhibited high RORs in both 1995 and 2007, whereas rural HSAs within the Midwest generally
1725
exhibited lower RORs in 1995 and 2007 relative to the rest of the country (Fig. 2). Equitable distribution of the radiation oncologist workforce Gini coefficient calculations confirmed the maldistribution of the radiation oncology workforce. Radiation oncologists were less evenly distributed than both PCPs and MDs in both 1995 and 2007 (Table 2), However, the Gini coefficients of all three groups exhibited a percentage decrease, indicating an improvement in distribution toward equity. Radiation oncologists exhibited the largest change (10.93%), followed by PCPs (4.85%) and finally MDs (2.34%). Analysis of Gini coefficients within population quartiles highlighted the uneven distribution of radiation oncologists in non-metropolitan areas. Specifically, we found that, unlike PCPs and MDs, who possessed relatively similar Gini coefficients among different population quartiles, the Gini coefficient of radiation oncologists in the lowest population quartile was an inequitable 0.951, compared with 0.251 in the highest population quartile (Table 3). The difference between Gini coefficients among population quartiles suggests that the geographic maldistribution of radiation oncologists stems primarily from inequity in less-populated non-metropolitan HSAs. Factors associated with radiation oncologist distribution Regression analysis found an association between radiation oncologist distribution and population characteristics. Univariate zero inflated negative binomial regression showed that increased ROR in 2007 was associated with HSAs with lower unemployment rates (p < 0.001), higher household incomes (p < 0.001), and higher education rates (p = 0.001) (Table 4). Surprisingly, decreased ROR in 2007 was associated with HSAs that had higher percent white population (p = 0.022). Multivariate analysis confirmed the association between increased ROR and lower unemployment rates (p < 0.001), higher household incomes (p < 0.001), and increased minority population (p = 0.010). Higher education rates proved to be insignificant in the multivariate model (p = 0.461). Because the association between minority population and ROR was unexpected, a separate model using percent African American population was built to confirm the association with minority groups. The confirmatory model yielded similar results (p < 0.001). DISCUSSION
Table 1. Changes in mean physician to population aged 65 years or older ratio, 1995–2007 Parameter Radiation oncologists per 100,000 PCP per 100,000 MD per 100,000
1995
2007
% Change
5.08
6.16
21.26
521.67 1056.44
606.50 1201.35
16.26 13.72
Abbreviations: PCP = primary care physician; MD = physician (doctor of medicine).
From 1995 through 2007, although the radiation oncology workforce increased slightly, it remained geographically maldistributed. Radiation oncologists remained concentrated in primarily highly populated metropolitan HSAs of the country, leaving large segments of the United States lacking access to radiotherapy. Additionally, geographic access to radiotherapy is associated with HSAs that have higher socioeconomic characteristics and higher minority makeup. Recent evidence has suggested that the demand
1726
I. J. Radiation Oncology d Biology d Physics
Volume 82, Number 5, 2012
Fig. 1. Distribution of radiation oncologists among health service areas: 1995 and 2007.
for radiotherapy will undoubtedly outpace the growth in the workforce, potentially exacerbating current geographic disparities (4). Our study suggests that the physician workforce problem should not merely focus on physician numbers but also the geographic distribution of physicians.
The growth in the radiation oncology workforce was modest compared with the overall growth in the physician workforce. The radiation oncology workforce grew 5% less than the overall physician workforce and 9% less than the primary care workforce. In a 12-year period, the mean ROR
Fig. 2. Ratios of radiation oncologists to population aged 65 years or older among health service areas: 1995 and 2007.
Radiation workforce in the United States d S. ANEJA et al.
Table 2. Trends of Gini coefficients among different physicians, 1995–2007 Physician type
1995
2007
Change
% Change
Radiation oncologist PCP MD
0.366 0.206 0.299
0.326 0.196 0.292
0.040 0.010 0.007
10.93 4.85 2.34
Abbreviation as in Table 1.
increased by slightly more than one radiation oncologist per 100,000 people aged 65 years or older, compared with larger increases in the PCPR and MDR. The modest growth in the radiation oncology workforce is most likely due to increasing residency positions for radiation oncology (15–17), Despite growth in the workforce, there was persistent geographic variation in the distribution of radiation oncologists in 1995 and 2007. Mapping of the RORs for 1995 and 2007 showed large geographic segments of the elderly population having little or no access to radiotherapy services, specifically in non-metropolitan areas. In 2007, an alarming 44% of HSAs in the United States lacked radiation oncologists. This translated to 3,137,580 people aged 65 years or older without access to radiotherapy in their HSA. Comparatively, Gini analysis showed that radiation oncologists were less evenly distributed than both PCPs and the overall physician workforce. Our finding that PCPs were the most evenly distributed replicates findings from previous studies and suggests that recent policy initiatives to increase geographic access to primary care have been successful (11, 18). More interestingly, the main driver in the uneven geographic distribution of radiation oncologists was an inequitable distribution in less-populated non-metropolitan HSAs. The lack of access to radiation oncologists in less-populated non-metropolitan areas is perhaps due to the large capital investment required to obtain the equipment and resources necessary to establish a radiation oncology practice. These impediments are increased in rural areas, where patient population levels are potentially less. Conversely, large pockets of radiation oncologists were centered in metropolitan areas with large academic centers. This finding corroborates previous findings from the ASTRO workforce committee that groups of 10 or more radiation oncologists are more likely to be in academic centers (3). The clustering of radiation oncologists in academic centers is perhaps due to recent increases in academic radiation oncology due to an influx of physician scientists in radiation oncology residency programs, or due to the attractiveness of technological resources available at academic centers (19). Table 3. Gini coefficients among population quartiles, 2007 Physician type Radiation oncologist PCP MD
Quartile 1
Quartile 2
Quartile 3
Quartile 4
0.951
0.675
0.402
0.251
0.268 0.333
0.233 0.258
0.232 0.265
0.169 0.236
Abbreviation as in Table 1.
1727
Our analysis found that access to radiation oncology services was allocated along socioeconomic and racial characteristics and provides information for policymakers regarding potential factors that affect radiation oncologist distribution, and could be used for future legislation aimed at physician recruitment to underserved areas. Although Hayanga et al. (20) found access to radiation oncologists to be associated with decreased minority population on the county level, our analysis found a positive association between minority population and ROR within HSAs. The reason for this difference is partly that our negative binomial regression model more accurately predicted radiation oncologist density given the distribution compared with the linear regression used in their analysis. Additionally, our geographic units of analysis were HSAs, which better predict patterns of healthcare usage compared with counties. Additionally, unlike previous analysis, our ROR outcome variable accounts for the variation in population density across the United States and is more accurate than gross physician totals. Moreover, unlike previous analyses, our analysis excluded counties located in U.S. territories, such as Guam, Puerto Rico, and the U.S. Virgin Islands, that generally have disproportionate minority populations and skewed access to specialists. Our findings suggest that, although racial disparities in cancer outcomes have been well documented, minority populations generally reside within HSAs in close proximity to radiation oncologists (21). Our findings highlight the problem of physician recruitment to rural areas. Recently, rural physician recruitment efforts have faced obstacles because of the technological gap that exists in rural settings. This problem is evermore important in a highly technical field like radiation oncology and has translated to patient dissatisfaction as physician retention rates in rural areas have decreased (19). Traditional policymakers have theorized that market forces and rural incentives would decrease geographic disparities as the workforce increases (22). However, we found little or no decrease in geographic variability of radiation oncologists to accompany a growth in the workforce. The significant non-physician personnel needed to operate radiotherapy facilities could also be a factor in the geographic maldistribution of radiation oncologists. The workforce shortages of radiation oncologists exacerbated with corresponding workforce shortages in non-physician personnel, such as dosimetrists and medical physicists (23), perhaps making it difficult provide cost-effective quality radiotherapy in rural areas in the United States. Recently, studies have suggested that the solution to geographic maldistribution is not increased recruitment to rural areas, but rather technical innovations that increase the efficiency of healthcare for rural populations (24–27). An example of a treatment innovation, specifically in radiotherapy, is hypofractionated treatment, as performed for breast radiotherapy in Canada (the so-called ‘‘Canadian fractionation’’), where geographic distance between radiation oncology centers is even more severe than in the United States.
I. J. Radiation Oncology d Biology d Physics
1728
Volume 82, Number 5, 2012
Table 4. Zero inflated negative binomial regression analysis using 2007 ROR as dependent variable Univariate analysis
Multivariate analysis
Independent variable
IRR
p
95% CI
IRR
p
% White population Unemployment rate % High school education Median household income
0.995859 0.889519 1.000001 1.000016
0.022 <0.001 0.001 <0.001
0.992335–0.999395 0.858483–0.921678 1.000000–1.000002 1.000011–1.00002
0.995351 0.906783 1.000005 1.000010
0.010 <0.001 0.461 <0.001
95% CI 0.991843 0.872964 0.999999 1.000005
0.978397 0.891379 1.000015 1.000015
Abbreviations: ROR = ratio of radiation oncologists to population aged 65 years or older; IRR = incidence rate ratio; CI = confidence interval.
Our analysis has several limitations. First, the physician location data from the AMA Masterfile does not capture physicians who have multiple practices in different regions, and therefore it is possible that we have overestimated the geographic clustering of radiation oncologists. However, although we believe that radiation oncologists are more likely to have multiple physician practice locations compared with primary care providers, we suspect that if radiation oncologists have multiple practices that they would be within the same HSA. Unfortunately, this cannot be confirmed by our data. Second, the Masterfile has been shown to underestimate physician shortages in rural areas (28). Third, our study does not capture other barriers to care, such as lack of health insurance and whether radiation oncology centers accept Medicaid payment for their services. Finally, our analysis cannot provide an optimal ROR because we did not relate geographic distribution to clinical outcomes. Future analysis
should focus on finding the radiation oncologist distribution that will optimize cancer outcomes. CONCLUSION Our study is the first geographic analysis of the radiation oncology workforce. Despite an overall growth in the radiation oncology workforce from 1995 through 2007, there remains persistent geographic maldistribution of radiation oncologists, with radiation oncologists concentrated in urban and metropolitan regions. Additionally, the distribution of radiation oncologists was related to the socioeconomic and racial characteristics of a region. To effectively solve the workforce issues facing radiation oncology and provide adequate care, we must focus not only on gross numbers of radiation oncologists, but also geographic distribution.
REFERENCES 1. American Association of Medical Colleges Center for Workforce Studies. Health care reform and the health workforce. 2010. Available at: https://www.aamc.org/download/142078/data/health_ workforce_provisions_and_health_care_reform.pdf. Last accessed April 4, 2011. 2. Wang S. Med schools: Health-care overhaul to accelerate doctor shortage. Wall Street Journal 2010. Available at: http://blogs.wsj.com/health/2010/09/30/med-schools-healthcare-overhaul-to-accelerate-doctor-shortage/ Last accessed April 4, 2011. 3. ASTRO Workforce Committee. 2002 Radiation Oncology Workforce Study: American Society for Therapeutic Radiology and Oncology. Int J Radiat Oncol Biol Phys 2003;56:309–318. 4. Smith BD, Haffty BG, Wilson LD, et al. The future of radiation oncology in the United States from 2010 to 2020: Will supply keep pace with demand? J Clin Oncol 2010;28:5160–5165. 5. Bogdanich W. The radiation boom: As technology surges, radiation safeguards lag. New York Times 2010. Available at: http://www.nytimes.com/2010/01/27/us/27radiation.html?_r=1. Last accessed April 4, 2011. 6. Lewis RS, Sunshine JH. Radiation oncologists in the United States. Int J Radiat Oncol Biol Phys 2007;69:518–527. 7. Elkin EB, Bach PB. Cancer’s next frontier: Addressing high and increasing costs. JAMA 2010;303:1086–1087. 8. Odisho AY, Cooperberg MR, Fradet V, et al. Urologist density and county-level urologic cancer mortality. J Clin Oncol 2010; 28:2499–2504. 9. Amory CF, Levine SR. Is telemedicine the answer for delivering stroke care to rural areas? Nat Clin Pract Neurol 2007;3:192–193.
10. U.S. Department of Health and Human Services, Health Resources and Services Administration. Area resource file (ARF): National county-level health resource information database. Available at: http://www.arfsys.com. Last accessed June 14, 2010. 11. Chang RK, Halfon N. Geographic distribution of pediatricians in the United States: An analysis of the fifty states and Washington, DC. Pediatrics 1997;100:172–179. 12. Smith BD, Smith GL, Hurria A, et al. Future of cancer incidence in the United States: Burdens upon an aging, changing nation. J Clin Oncol 2009;27:2758–2765. 13. Matsumoto M, Inoue K, Farmer J, et al. Geographic distribution of primary care physicians in Japan and Britain. Health Place 2010;16:164–166. 14. Matsumoto M, Inoue K, Bowman R, et al. Self-employment, specialty choice, and geographical distribution of physicians in Japan: A comparison with the United States. Health Policy 2010;96:239–244. 15. Wilson LD, Yu JB, Haffty BG. National Residency Matching Program (NRMP) results for radiation oncology: 2007 update. Int J Radiat Oncol Biol Phys 2007;69:326–327. 16. Wilson LD, Haffty BG. National Residency Matching Program results for radiation oncology: 2006 update. Int J Radiat Oncol Biol Phys 2006;66:321–322. 17. Wilson LD, Haffty BG. National Residency Matching Program results for radiation oncology, 2005 update. Int J Radiat Oncol Biol Phys 2005;63:3–4. 18. Rittenhouse DR, Fryer GE Jr., Phillips RL Jr., et al. Impact of Title VII training programs on community health center
Radiation workforce in the United States d S. ANEJA et al.
19.
20.
21. 22.
staffing and National Health Service Corps participation. Ann Fam Med 2008;6:397–405. Fears D. Renewed effort to lure doctors to rural areas faces obstacles. Washington Post 2010. Available at: http://www. washingtonpost.com/wp-dyn/content/article/2010/08/08/AR20 10080802832.html?nav=emailpage. Last accessed April 4, 2011. Hayanga AJ, Waljee AK, Kaiser HE, et al. Racial clustering and access to colorectal surgeons, gastroenterologists, and radiation oncologists by African Americans and Asian Americans in the United States: A county-level data analysis. Arch Surg 2009;144:532–535. Gross CP, Smith BD, Wolf E, et al. Racial disparities in cancer therapy: Did the gap narrow between 1992 and 2002? Cancer 2008;112:900–908. Newhouse JP. Geographic access to physician services. Annu Rev Public Health 1990;11:207–230.
1729
23. Mills MD, Thornewill J, Esterhay RJ. Future trends in the supply and demand for radiation oncology physicists. J Appl Clin Med Phys 2010;11:3005. 24. Goodman DC, Grumbach K. Does having more physicians lead to better health system performance? JAMA 2008;299: 335–337. 25. Goodman DC, Fisher ES. Physician workforce crisis? Wrong diagnosis, wrong prescription. N Engl J Med 2008;358:1658–1661. 26. Goodman DC, Fisher E, Stukel TA, et al. The distance to community medical care and the likelihood of hospitalization: Is closer always better? Am J Public Health 1997;87:1144–1150. 27. Goodman DC. Do we need more physicians? Health Aff (Millwood) 2004. Suppl Web Exclusives:W4-67–;W4–69. 28. Shea JA, Kletke PR, Wozniak GD, et al. Self-reported physician specialties and the primary care content of medical practice: A study of the AMA physician masterfile. American Medical Association. Med Care 1999;37:333–338.