OBJECTIVE. The objective of this study was to describe geographic proximity, quantify variation in supply, and estimate pediatric population increments that are needed to support providers across pediatric subspecialties.
METHODS. Data from the American Board of Pediatrics and the Claritas Pop-Facts Database were used to calculate subspecialty-specific straight-line distances between each zip code and the nearest board-certified subspecialist. These data sources also were used to estimate the percentage of hospital referral regions with providers and calculate physician-to-population ratios for each of 16 pediatric medical subspecialties. Coefficients of variation for the ratios were used to assess intraspecialty variation in supply across markets. Estimates of the pediatric population that is needed to support an initial or additional physician in a market were generated using subspecialty-specific ordered logit analyses.
RESULTS. The population-weighted average distance to a subspecialist ranged from 15 miles for neonatology to 78 miles for pediatric sports medicine. For most pediatric subspecialties, more than two thirds of children live within 40 miles of a certified physician. For 7 of 16 of pediatric subspecialties, fewer than one half of hospital referral regions have a provider. Coefficients of variation vary across subspecialties and are lowest for neonatology at 76% and greatest for pediatric sports medicine at 287%. Pediatric population thresholds likewise vary with a tendency toward lower thresholds for procedural specialties, such as cardiology and critical care medicine.
CONCLUSIONS. The practice locations of pediatric subspecialists parallel the geographic distribution of children in the United States, yet many hospital referral regions lack pediatric subspecialists and coefficients of variation vary widely across subspecialties. These findings suggest that either the supply of pediatric subspecialists is inadequate, pediatric subspecialists are distributed inequitably, or the market for pediatric subspecialists is larger than the hospital referral regions. Furthermore, population thresholds for many cognitive pediatric subspecialties are high; the extent to which high thresholds reflect low disease prevalence versus other factors, such as inadequate reimbursement, is not established.
In a 2000 study of state Title V directors, one third of respondents described the availability of pediatric subspecialists in their state as not very or not at all adequate due, in part, to an uneven distribution of physicians within the state.1 Although many studies have investigated geographic proximity to physician services, none has detailed children’s geographic proximity to pediatric subspecialty care. Pediatric medical subspecialists merit separate consideration for a number of reasons. Given the low prevalence of many pediatric conditions, demand for subspecialty care is low relative to the adult population; consequently, pediatric subspecialties are among the smallest medical specialties in the United States. Furthermore, most pediatric subspecialists practice in academic settings, where professional demands include patient care, research, and medical education.2 As a result, pediatric subspecialists may choose their practice location on factors other than patient demand.
Although many factors, such as insurance status and provider reimbursement, influence access to care, geographic proximity is 1 important component. With the exception of neonatology,3–6 no one has assessed the geographic distribution of pediatric medical subspecialists in the United States or investigated how the supply of these physicians relates to the distribution of the pediatric (ie, under-18), population. Moreover, there have been no efforts to determine the pediatric population size that is needed to support pediatric medical subspecialty practice. Given the dearth of published data on the availability of pediatric subspecialty care in the United States and how supply differs throughout the nation, estimates of these ratios will provide useful data to policy makers and medical educators.
This study uses diplomate data from the American Board of Pediatrics (ABP) and zip code–level data from Claritas,7 a proprietary database, to generate specialty-specific estimates for each of the following measures: (1) the population-weighted average distance to a provider, (2) the percentage of children who live within relatively short driving distance of a provider, (3) the number of hospital referral regions (HRRs) that have providers, (4) provider-to-pediatric population ratios, and (5) pediatric population thresholds that are needed to support pediatric subspecialists within HRRs. Study data also are used to produce maps that depict the practice locations of pediatric medical subspecialists in the United States.
Using 2003 diplomate data from the ABP, 16 pediatric medical subspecialties were studied. Other pediatric subspecialties exist, but their certification data are maintained by other boards. The 16 subspecialties that were included in this analysis vary in terms of their scope of practice (cognitive versus procedural), the number of years since certification became available, and the number of certified physicians. The 2003 ABP diplomate file includes data for 14780 board-certified pediatric subspecialists and excludes physicians who are known to be deceased or retired; physicians with current certification in a subspecialty (97%) were included and classified by subspecialty. Physicians with 2 or more current certifications were included in analyses for each of the subspecialties for which they had a certification because our data did not allow us to discern the discipline in which the physician was currently active.
Although similar data are available in the American Medical Association (AMA) Physician Masterfile, discrepancies between the 2002 Physician Masterfile and the 2003 ABP diplomate file suggested that the latter was preferable; these discrepancies have been described elsewhere.8 The most significant problems with the 2002 AMA Physician Masterfile were undercounts of board-certified pediatric subspecialists relative to the ABP data, which is the gold standard for board-certified pediatric subspecialists, and limited data on newly certified specialties such as developmental behavioral pediatrics and neurodevelopmental disabilities. Only the analyses using ABP diplomate data are reported here; sensitivity analyses that were performed using the 2002 Physician Masterfile are available on request.
Practice zip codes were used to produce maps for each pediatric medical subspecialty. Maps were generated in MapInfo 7.0 (MapInfo Corporation, Troy, NJ) using a circle to denote the presence of 1 or more pediatric subspecialists in each zip code.
Distance to Care
Straight-line distance from residence to the nearest provider is a commonly used measure of travel impedance9 and accounts for nearly all of the variation in actual travel time.10 For each pediatric specialty, the straight-line distances between each zip code in the United States and the nearest specialist were calculated using latitude and longitude data for the zip code centroids. Zip codes in which providers were located were assigned a distance estimate of 0. Zip code–level estimates of the pediatric population were obtained from Claritas and merged to distance estimates. For generation of national estimates of the average distance to a subspecialist, the distances between each zip code and the nearest pediatric subspecialist were averaged across all zip codes for each subspecialty, using the under-18 population as a weight.
To date, no individual or group has established the desired distance between patients and providers of pediatric subspecialty care. In a 1980 report, the Graduate Medical Education National Advisory Committee recommended 95% of the population have a maximum travel time of 30 minutes for pediatric care but provided no specific recommendation about travel times to pediatric subspecialty services.11 In the absence of any universally accepted threshold, zip codes were classified as being within 10 miles, 11 to 20 miles, 21 to 40 miles, 41 to 80 miles, or 80 miles or more of a provider. The thresholds were used assuming that 20 miles corresponds to a drive time of 30 minutes.12 Specialty-specific percentages of the overall under-18 population living within these distances of a provider then were calculated.
Defining Market Areas
With the exception of neonatology,3 past studies have not generated market areas for pediatric subspecialty physicians. A reasonable conceptualization of market area for these subspecialties is an area in which inflows and outflows of patients are minimized.13 HRRs are mutually exclusive market areas on the basis of Medicare data on referrals for specialized care.14 They have not been validated for use in children but may serve as a reasonable proxy for pediatric subspecialty market areas. For all market-level analyses, data were aggregated to the HRR (n = 306).
Supply Relative to Pediatric Population
Physician-to-population ratios allow straightforward, gross comparisons of supply across services areas.9 Specialty-specific estimates of percentage of HRRs with a provider, the average ratio of pediatric subspecialists to pediatric population, and the coefficients of variation (COVs) of the ratios were estimated. The COV equals (SD/mean) × 100 and is a measure of dispersion. A large SD of the physician-to-population ratio, relative to the mean, would suggest that physician-to-population ratios vary considerably across markets. COVs can be compared across subspecialties to assess relative dispersion.
Population Thresholds Needed to Support Pediatric Subspecialists
The predominance of academic practice among pediatric subspecialties makes it unlikely that an individual provider would make a practice location decision solely on the basis of patient demand; it is reasonable to assume, however, that academic medical centers consider the level of patient demand in their staffing decisions. Using counts of physicians and population data aggregated at the HRRs, subspecialty-specific ordered logit regressions were performed. In these models, the dependent variable of interest, the number of subspecialty physicians in an HRR, was treated as a function of the pediatric population. Ordered logit coefficients are not interpreted easily. The coefficients that were generated from these regressions were used to estimate the pediatric population size that is needed to support an initial subspecialist. These results also were used to estimate the population increments (ie, the number of additional children) needed to support each additional subspecialty physician. Neonatology was excluded from the order logit analysis because it included a wide range of values that were inconsistent with the preferred form of a dependent variable in ordered logit. In the interest of space, the coefficients from the 16 ordered logit models are omitted here but are available on request.
Subspecialties were classified as procedural or cognitive. The terms “procedural” and “cognitive” often are used to divide physician specialties, yet an accepted definition of this distinction is not readily available. For this analysis, procedural medical subspecialties were those in which the physician personally spends much of his or her practice in performing or overseeing invasive procedures such as endoscopy or dialysis as well as those in intensivist subspecialties; all other subspecialties were considered cognitive. The categorization of specialties in this analysis was reviewed by 2 physician researchers. The institutional review board at the University of North Carolina at Chapel Hill approved these analyses.
Nationwide, the dispersion of providers varied considerably across pediatric subspecialties (www.shepscenter.unc.edu/mayer_pedsubspec_maps042706.pdf). Likewise, the average population-weighted distance to care for board-certified pediatric subspecialists varied widely from 15 miles for neonatology to 78 miles for pediatric sports medicine (Table 1). The average population-weighted distance to the nearest board-certified pediatric subspecialist correlated with the number of physicians who were certified in that specialty (ρ = −0.68) and with the year that subspecialty certification became available (ρ = 0.64). For the majority of pediatric subspecialties, approximately half of the pediatric population lived within 10 miles of a provider and more than two thirds of the pediatric population lived within 40 miles of a board-certified provider (Table 2). Exceptions included adolescent medicine, developmental behavioral pediatrics, neurodevelopmental disabilities, pediatric rheumatology, and pediatric sports medicine.
Seven of the 16 pediatric medical subspecialties studied were present in fewer than one half of HRRs (Table 3). Neonatologists were the most widely available, with 1 or more practicing in 88% of HRRs; only relatively remote hospital regions, such as Tyler, Texas, lacked certified neonatologists. Pediatric sports medicine was available in <20% of HRRs As expected, given the differences in incidence rates for diseases that are managed by these specialties, average physician-to-population ratios varied considerably across specialties. Of greater interest were the differences in the COVs for the ratios across specialties, which suggest greater variation in provider ratios across areas within certain specialties than in other specialties. The COV was lowest for neonatology, at 76%, and greatest for pediatric sports medicine, at 287%.
On average across HRRs, there was approximately 1 neonatologist per 25000 children but >1 million children per pediatric sports medicine physician. For most subspecialties, there are between 100000 and 200000 children per provider, on average, across HRRs. The average number of children per provider in each HRR was particularly high for developmental behavioral pediatrics (337222), neurodevelopmental pediatrics (570784), and pediatric rheumatology (628410).
Estimates from the ordered logistic regression analyses suggested that the pediatric population that is needed to support an initial provider in an HRR varied considerably across pediatric subspecialties (Fig 1). Pediatric cardiology had the lowest predicted HRR pediatric population threshold needed to support an initial provider at 89089 children. In contrast, ordered logit models predicted that 563471 children were needed to support the first pediatric sports medicine physician in an HRR.
For most procedural and intensivist pediatric subspecialties, models estimated that <150000 children were needed to support an initial provider in an HRR (Fig 2). The pediatric population increment that is needed to support the second provider was uniformly lower than the pediatric population increment that is needed to support the initial provider across all procedural and intensivist subspecialties. For example, the regression estimated that ∼129794 children were needed to support the first pediatric critical care physician but that only an additional 26767 were needed to support the second physician.
Compared with procedural subspecialties, greater pediatric populations were needed to support physicians in many cognitive pediatric subspecialties, such as pediatric rheumatology, neurodevelopmental disabilities, and developmental and behavioral pediatrics (Fig 3). These subspecialties require large population increments to support the initial as well as additional physicians in a market. In contrast, pediatric infectious diseases, pediatric hematology-oncology, and pediatric endocrinology needed smaller pediatric populations to support providers. For these 3 specialties, the number of additional children who are needed to support the second provider ranges from ∼53959 for pediatric hematology oncology to 95429 for pediatric infectious diseases.
Although there are many determinants of access, geographic proximity is 1 important component. Mean population-weighted distance to care varies across pediatric subspecialties but generally is lower in older pediatric subspecialties and those with more physicians. There is some evidence that increased numbers of physicians may lead to increased dispersion among pediatric subspecialties. Although past studies of other physician specialties have disagreed about the extent to which increases in physician supply enhance dispersion,12,15 none has studied pediatric subspecialties. Because many pediatric subspecialties are smaller and younger, in terms of availability of certification, increases in supply may have different implications for dispersion in these disciplines. The extent to which increases in the number of certified pediatric subspecialists lead to more equitable distribution depends, nonetheless, on a number of other factors, such as practice locations chosen, location of training programs, financial viability of private practice, ability of pediatric subspecialists to generate sufficient clinical revenue in academic medical centers, and substitution of other professional activities, such as research and teaching, for patient care activities. Future studies of emerging specialties, such as developmental and behavioral pediatrics, may provide insight into how a pediatric subspecialty diffuses over time as the number of providers increases. Studies of which communities gain 1 of these subspecialists for the first time may be especially insightful.
Approximately half of the children in the United States live within 10 miles of most pediatric subspecialists. For 11 of the 16 pediatric medical subspecialties, two thirds of the pediatric population lives within 40 miles of a board-certified provider; therefore, the geographic distribution of pediatric subspecialists in the United States closely parallels the distribution of the under-18 population. Despite this, nearly 1 in 3 children must travel 40 miles or more to receive care from a pediatrician who is certified in adolescent medicine, developmental behavioral pediatrics, neurodevelopmental disabilities, pediatric pulmonology, pediatric emergency medicine, pediatric nephrology, pediatric rheumatology, and pediatric sports medicine. The extent to which such distances are barriers to subspecialty care is not known, but they may be problematic for low-income families with limited access to transportation and/or sick leave benefits.
Geographic proximity, however, does not equal access. For many pediatric subspecialties, a surprising number of HRRs lack providers. Because HRRs have not been validated for pediatric care, it is not clear whether the large number of HRRs without pediatric subspecialists reflects poor supply or the inappropriateness of using HRRs as markets for pediatric subspecialty care. Across all HRRs, mean provider-to-pediatric population ratios range from 0.09 per 100000 for pediatric sports medicine to 4.10 per 100000 for neonatology. Only neonatology, cardiology, and hematology-oncology have <100000 children per provider on average across HRRs. Although estimates of the number of children per provider raise concerns, it is not possible to know whether the estimates are excessive without combined prevalence rates for the conditions that are treated by each pediatric subspecialty and specialty-specific data on average practice size. With better specialty-specific estimates of combined prevalence and practice size, benchmarking could be used to assess the pediatric subspecialty workforce and make workforce recommendations.16
The extent to which the supply of pediatric subspecialists is adequate, even in areas that have physicians, partially depends on disease prevalence rates, availability of substitute providers, provider involvement in patient care, and intensity of physician involvement in individual cases. Consequently, variation in physician-to-population ratios across specialties is expected and not necessarily a cause for concern. COVs suggest, however, that there is considerable variation in provider-to-pediatric population ratios across HRRs within pediatric subspecialties. These findings reflect the results of Goodman et al3,4 that neonatologists are not distributed evenly in the United States. These variations in relative supply across markets may reflect inequitable distribution; differences in demand levels across markets that may or may not reflect differences in need; and/or differences in the level of competition from other physicians, such as internist subspecialists. More research is needed to assess the relationship between pediatric subspecialist supply and outcomes for relevant pediatric conditions and to ascertain the adequacy of the pediatric subspecialty workforce relative to patient need.
Estimates of the population thresholds that are needed to support initial or additional pediatric subspecialists in HRRs vary considerably across specialties. The prevalence of conditions that are managed by a given specialty naturally will influence the size of the pediatric population that is needed to support a provider, but other factors may particularly influence population thresholds among cognitive specialties. Reimbursement for pediatric evaluation and management services is inadequate for a number of reasons17; consequently, difficulties in generating clinical revenue for cognitive pediatric subspecialties may require larger pediatric populations to ensure practice viability, leading to large population thresholds for specialties such as pediatric rheumatology. In addition, the lack of parity in coverage for mental health care, as well as the relative recency of board certification, may contribute to the higher population thresholds that are observed for developmental-behavioral and neurodevelopmental pediatrics. In contrast, pediatric infectious diseases, which is an important consult service for revenue-generating hospital cost centers such as neonatal intensive care, has a much lower threshold. Finally, the availability of similarly trained internal medicine physicians in fields such as adolescent medicine and sports medicine may lead to location in a limited number of large population centers.
Population increments that are needed to support a second provider are almost uniformly lower than thresholds for the initial physicians; these findings are similar to those of Brasure et al.18 This consistent decline in the population increments suggests that the fixed costs of starting a pediatric subspecialty practice, such as overhead expenses and capital investment, may lead to a higher threshold for the initial provider but that once a practice is in place, fewer additional patients are needed to support an additional provider. Other benefits of having an additional provider, such as shared call and professional support, also may contribute to the lower threshold needed to attract an additional provider.
These analyses have several limitations. It has been noted previously that distance to care measures overweight providers who are located at the periphery of a boundary and fail to account for border crossing.9 A recent study found that use of physician street address and census block location did not change qualitatively study results, especially for less diffuse physician specialties.12 Because the present analyses are done at the zip code level and pediatric subspecialists tend to be in a limited number of geographic areas, concerns related to the imprecision of zip code likely are less relevant than in studies of more diffusely located providers, such as primary care physicians. In addition, these analyses could not account for the availability of pediatric subspecialty care through traveling pediatric subspecialty clinics or telemedicine, which certainly reduce geographic barriers to care.
Our ABP-based analysis also uses “head counts” of providers and could not distinguish between physicians who are active in direct patient care and those who are not. In addition, this study uses data on board-certified pediatric subspecialists only. These estimates, therefore, fail to account for physicians who are trained in these subspecialties but not yet certified as well as those who do not have training and are practicing in these specialties. The AMA-based analyses, which limited the investigation to providers who are active in direct patient care and included noncertified providers, had comparable findings for 14 of the 16 subspecialties studied. Findings for allergy immunology and adolescent medicine are not comparable between the ABP and AMA data sources because the latter unavoidably includes some internists who have specialized in these fields.
These analyses also are limited in that they reflect the current supply of pediatric subspecialists and may not reflect truly the population size that could support a pediatric subspecialist. It may be that a pediatric population of <500000 could support a pediatric sports medicine physician but that, for other reasons, providers simply have not moved into smaller markets. Furthermore, the availability of a pediatric subspecialist may actually increase demand for their services for mild conditions. Parents may be more likely, for example, to bring a child with chronic constipation to a gastroenterologist if 1 practices nearby. In addition, constraints in the number of training slots could decrease supply artificially and limit the diffusion of providers, yielding higher population thresholds.
These analyses use HRR as market areas. One study suggested that HRR-based markets may not be valid for pediatric care.14 The lack of a national source of data on pediatric care akin to Medicare data continues to be an impediment to the creation of pediatric-specific market areas. Expansion of the Kids’ Inpatient Database to include all states needs to be a national health care data priority. A national database of outpatient pediatric care that oversamples children with special health care needs also is sorely needed. Without improved data sources, researchers are unable to depict the patterns of pediatric subspecialty care receipt accurately.
Finally, these analyses likely underestimate the availability of physicians in fields that have internal medicine counterparts that deal with a similar population, such as adolescent medicine and allergy immunology. Estimates that are based on AMA data for these specialties find shorter distances to care, smaller physician-to-population ratios, and smaller population thresholds than those that are based on ABP data.
Despite the limitations, these findings are the first national estimates of the distances that children travel, on average, to receive pediatric subspecialty care. Although the majority of the pediatric population lives within a 1-hour drive to a pediatric subspecialty provider, care is less widely available for certain subspecialties and in certain regions. Additional research is needed on recruitment difficulties and patient wait times for initial and follow-up appointments, which may elucidate supply constraints in these subspecialties. Furthermore, calculation of specialty-specific combined prevalence rates and average practice sizes will allow benchmarking of the pediatric subspecialty workforce. Finally, detailed investigations into how the health care system meets the needs of children with chronic conditions and serious acute conditions that typically are treated by pediatric subspecialists, particularly in regions that are virtually devoid of these providers, may uncover innovative approaches to care that compensate for limited subspecialty access.
This research was supported through grant 1-K02-HS013309-01A1 from the Agency for Healthcare Research and Quality. The funding agency was not involved in the performance of the study.
Ann Howard was instrumental in performing distance analyses, and Katie Gaul produced the national maps. Special thanks to the Works in Progress Lunch at the Sheps Center for Health Services Research; the Sprint Manuscript Team at the Department of Health Policy and Administration at the University of North Carolina at Chapel Hill; Donald Pathman, MD; and anonymous reviewers for comments on previous drafts.
- Accepted August 29, 2006.
- Address correspondence to Michelle L. Mayer, PhD, MPH, Department of Health Policy and Administration, UNC Chapel Hill School of Public Health, Cecil G. Sheps Center for Health Services Research, CB 7590, Chapel Hill NC 27599-7590. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
Dr Mayer 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.
- ↵Limb S, McManus M, Fox H. Pediatric Provider Capacity for Children With Special Health Care Needs: Results From a National Survey of State Title V Directors. Washington, DC: Maternal Child Health Policy Research Center; 2001
- ↵Stoddard JJ, Cull WL, Jewett EA, Brotherton SE, Mulvey HJ, Alden ER. Providing pediatric subspecialty care: a workforce analysis. AAP Committee on Pediatric Workforce Subcommittee on Subspecialty Workforce. Pediatrics.2000;106 :1325– 1333
- ↵Goodman DC, Fisher ES, Little GA, Stukel TA, Chang CH. Are neonatal intensive care resources located according to need? Regional variation in neonatologists, beds, and low birth weight newborns. Pediatrics.2001;108 :426– 431
- ↵Thompson LA, Goodman DC, Little GA. Is more neonatal intensive care always better? Insights from a cross-national comparison of reproductive care. Pediatrics.2002;109 :1036– 1043
- ↵Pop-Facts Database for ZIP Codes [computer program]. San Diego, CA: Claritas; 2003
- ↵Mayer ML, Preisser JS. The changing composition of the pediatric medical subspecialty workforce. Pediatrics.2005;116 :833– 840
- ↵Phibbs CS, Luft HS. Correlation of travel time on roads versus straight line distance. Med Care Res Rev.1995;52 :532– 542
- ↵US Department of Health and Human Services. GMENAC Summary Report. Washington, DC: US Department of Health and Human Services, Health Resources Administration; 1980
- ↵Rosenthal MB, Zaslavsky A, Newhouse JP. The geographic distribution of physicians revisited. Health Serv Res.2005;40(pt 1) :1931– 1952
- ↵Morrisey MA, Sloan FA, Valvona J. Defining geographic markets for hospital care. Law Contemp Probl. 1988;51(2) :165– 194
- ↵Council on Graduate Medical Education. Physician Distribution and Health Care Challenges In rural and Inner City Areas, 10th report. Rockville, MD: Council on Graduate Medical Education; 1998
- ↵Jewett EA, Anderson MR, Gilchrist GS. The pediatric subspecialty workforce: public policy and forces for change. Pediatrics.2005;116 :1192– 1202
- ↵Brasure M, Stearns SC, Norton EC, Ricketts T 3rd. Competitive behavior in local physician markets. Med Care Res Rev.1999;56 :395– 414
- Copyright © 2006 by the American Academy of Pediatrics