Where Do Graduating Pediatric Residents Seek Practice Positions?

Where Do Graduating Pediatric Residents Seek Practice Positions? William L. Cull, PhD; Chiang-hua Chang, MS; David C. Goodman, MD, MS Objective.—To profile the characteristics of areas that graduating pediatric residents target in their job searches and to explore whether residents applying to primary care markets with higher pediatrician supplies experience job-search difficulty. Study Design.—A national random sample of 500 graduating categorical pediatric residents was surveyed. The communities that the pediatric residents targeted for jobs were linked to local-area characteristics by using Dartmouth Primary Care Service Areas (PCSAs), which are discrete markets that represent patient travel patterns for primary care services. PCSA population and providers were characterized by using data from the 2000 US Census and American Medical Association/American Osteopathic Association Physician Masterfiles. Results.—A total of 308 graduating residents (62%) completed the survey. Of the respondents, 136 (44%) applied for general-practice positions. The characteristics of the PCSAs that residents applied to differed from the PCSAs without an application. Residents’ first-choice areas had higher ratios of general pediatricians, had higher median household incomes, and were more likely to be urban areas. Residents applying to higher-supply areas were significantly more likely to report moderate or considerable difficulty in their job searches than were residents applying to lower-supply areas. Residents applying to medium- and higher-supply areas sent out more total applications for positions and received lower starting salaries than did residents applying to lower-supply areas. Conclusions.—Residents continue to prefer high-supply areas in their job searches, despite experiencing greater search difficulty in these areas. Current targeted incentives and market forces are unlikely to redress geographic variation in pediatrician supply. KEY WORDS:

general pediatrics; maldistribution; pediatric workforce; resident education

Ambulatory Pediatrics 2005;5:228 234

T

he number of children under 18 years of age in the United States is estimated to be over 78 million, with 38 457 general pediatricians in practice to care for them.1 This is equivalent to 49 pediatricians for every 100 000 children, which represents a 27% increase in the per-child supply of general pediatricians over the past decade.2 The number of pediatricians will likely continue to increase, resulting in a one-third rise in the ratio of pediatricians to children by 2020.1,3 Despite growth in the number of pediatricians relative to children in the United States, researchers disagree concerning whether there will be sufficient pediatricians to meet future health care requirements.1–12 Several factors may affect the desired balance of pediatricians and physicians for the future, including increased interest in part-time practice arrangements13 and increased visits to pediatricians as family physicians care for aging baby boomers.14,15 However, even when these and many other factors were tested by sensitivity analyses, some analysts predict an increase in the number of adjusted pediatricians per children.1

National physician-supply projections, however, do not take into account the regional variation in supply that exists throughout the United States. For example, the sizable increase in the number of pediatricians per children from 1982 to 1992 led to only a modest improvement in the geographic distribution of pediatricians.16 Areas of very high supply remain across the country, but many other areas are without a pediatrician.17 Graduating residents’ job-search experiences may serve as a labor-market indicator to examine the requirement for pediatricians, nationally and locally. Since 1996, the American Academy of Pediatrics (AAP) has surveyed a sample of graduating pediatric residents annually to examine their training and job-search experiences. Recent trends have shown some tightening in residents’ jobsearch success nationally.18 Meanwhile, residents’ debt levels continue to rise, theoretically increasing the potential effectiveness of loan-forgiveness programs that target the geographic distribution of physicians.19,20 The goal of the current study was to examine the relationship of local pediatrician supply and graduating residents’ job-search behaviors and experiences. We specifically aimed to describe the areas that residents target in their job searches, to test the correlation of residents’ job-search success with local-area pediatrician-supply levels, and to examine resident interest in debt-relief programs to encourage physician diffusion.

From the Division of Health Services Research, American Academy of Pediatrics, Elk Grove Village, Ill (Dr Cull); and the Department of Pediatrics and the Center for the Evaluative Clinical Sciences, Dartmouth Medical School, Hanover, NH (Ms Chang and Dr Goodman). Address correspondence to William L. Cull, PhD, Division of Health Services Research, American Academy of Pediatrics, 141 Northwest Point Blvd, Elk Grove Village, IL 60007 (e-mail: [email protected]). Received for publication September 8, 2004; accepted January 18, 2005. AMBULATORY PEDIATRICS Copyright q 2005 by Ambulatory Pediatric Association

METHODS Data from 2 sources, the AAP Graduating Resident Survey and the Dartmouth Primary Care Service Area project, were combined to conduct this study. A national

228

Volume 5, Number 4 July–August 2005

AMBULATORY PEDIATRICS

random sample of 500 third-year categorical pediatrics residents received the 2003 AAP Graduating Resident Survey as they concluded training or immediately afterward (May–August). The residents surveyed were randomly selected from a database of all Accreditation Council for Graduate Medical Education–accredited US pediatric residency programs that the AAP uses to offer the Pediatrics Review and Education Program free to residents. Residents in combined programs, such as internal medicine pediatrics residents, were excluded from this survey. The survey was mailed up to 4 times to nonrespondents. No incentives for participation were offered to residents. The survey asked residents about their residency experiences, career intentions, and job-search experiences. Demographic information about residents’ race and ethnicity, gender, medical school, marital status, family characteristics, and educational debt were also collected. The AAP Institutional Review Board approved this survey. A total of 308 (62%) residents from 173 US pediatric residency programs (86% of programs) responded to the survey. The database contained several demographic variables that allowed comparisons of responders and nonresponders. No significant difference was apparent between the responder and the nonresponder groups for gender (69% vs 66% women, P 5 .16), but responders were slightly younger than nonresponders (mean years: 31.3 vs 32.1, P 5 .03). To profile the areas that residents applied to, zip code information from residents’ first- and second-most desired positions was linked with the characteristics of the Primary Care Service Area (PCSA) within which each application zip code was located. Created by the Dartmouth Center for the Evaluative Clinical Sciences, PCSAs are contiguous areas arising from a patient-origin study of 1999 Medicare Part B and Outpatient claims data. PCSAs are geographic markets of primary care derived from patients’ travel patterns for primary care services. Because PCSAs represent actual travel, PCSA boundaries can extend across geopolitical borders such as county and state lines. Validity analyses have demonstrated PCSA utility for pediatric workforce studies.21 A more detailed description of the methods used to create these areas is available elsewhere.21 Descriptions of the local areas were then created by using population characteristics from the 2000 US Census and physician-supply information from the American Medical Association/American Osteopathic Association Masterfile.22 Pediatrician supply was based on postgraduate medical-education physicians with the majority of their professional time in clinical medicine. The PCSA information included variables such as the number of children and their racial breakdown, the median household income, the percentage of children in poverty, and the supply of general pediatricians in the area. Job-search analyses were limited to the 2 most desired applications of the residents actively seeking general-practice positions. Fifteen residents listed the same zip code for both their first- and second-choice positions. The first

Resident Job Search

229

set of analyses separated all PCSAs based on whether any residents sent their first-choice applications there. The characteristics of these areas, including pediatrician supply, were then contrasted with the areas where residents did not apply. We then limited the analysis to only urban areas to see if residents’ targeted urban areas differed from the urban areas they avoided. A PCSA was defined as an urban area if more than 50% of its population lived in urbanized areas or urban clusters.23 This definition includes both urban inner-city areas and urban non–innercity areas. For all remaining analyses, residents seeking generalpractice positions were categorized into 3 subgroups based on the general pediatrician-to-child ratio within the area of residents’ most desired jobs. The pool of job seekers was categorized into equal thirds (44 in each group) based on the cut points ,54, 54.1–97.1, and .97.1 pediatricians per 100 000 children. We compared the 3 groups of lower, medium, and higher supply for various job-search and job-outcome variables. Residents were also asked a series of hypothetical questions about practicing in rural areas for varying levels of debt relief that were compared among the groups. Several statistical tests were used to analyze the data. Chi-square analyses were used for categorical outcome variables. For continuous outcome variables, comparisons were made by using independent-groups Student’s t tests and 1-way analyses of variance. Nonparametric tests, including the Kruskal-Wallis and Mann-Whitney tests, were also used to examine variables with high skewness values or to compare groups that were very discrepant in size. Follow-up comparisons of the supply groups were conducted only when all-group tests revealed differences among the groups.24 Some values were missing for the response or the predictor variables, producing slight variations in the number of cases used for analysis. A P value # .05 was considered statistically significant for all analyses except for the PCSA-level analyses with thousands of cases, where P # .001 was used. RESULTS A total of 136 graduating pediatric residents (44%) applied for general-practice positions. The remainder pursued subspecialty fellowships (29%), chief residencies (11%), other jobs (12%), or time off (3%). These analyses focus on the experiences of those residents seeking a general-practice position. To profile residents’ job searches, a comparison was first made between the PCSAs for which residents did and did not apply for jobs (Table 1). The areas that residents targeted in their job searches were very different from the rest of the country. Of residents’ first-choice areas, 91% were urban compared with 45% for the areas without firstchoice applications. The supply of pediatricians per 100 000 children in the first-choice areas was triple that for other areas (97.8 vs 30.0). These first-choice areas also tended to have other characteristics typically found in urban areas, including a higher percentage of Hispanic children (15% vs 9%) and black children (16% vs 9%), higher

230

Cull et al

AMBULATORY PEDIATRICS

Table 1. Characteristics of Areas Where Residents Sought Jobs First-Choice Application Area Characteristics† Number of PCSAs Urban area, % Metropolitan area, % General pediatricians per 100 000 children, mean Median household income, mean Percentage of children in area who are black, mean Percentage of children in area who are Hispanic, mean Percentage of children in area in poverty, mean Presence of rural health center, % Presence of community health center, %

PCSAs‡ With an Application

PCSAs Without an Application

119 91 90

6423 45* 47*

97.8 47 737

30.0* 38 344*

Second-Choice Application PCSAs With an Application

PCSAs Without an Application

96 88 90

6446 45* 47*

96.9 49 413

30.2* 38 353*

16

9*

14

10*

15

9*

15

9*

15 18 36

17 44* 20*

15 13 38

17* 44* 20*

†Statistical testing performed by using x2 and Mann-Whitney tests. ‡PCSAs indicates Primary Care Service Areas. *P , .001.

median population incomes ($47 737 vs $38 344), and a greater percentage of community health centers (36% vs 20%). The large differences between the PCSAs with and without applications were also apparent for residents’ second-choice positions. The correlation between the supply of pediatricians within the areas that residents targeted with their first and second choices was significant (r 5 .35, P , .001). The urban-rural distinction does not, however, fully explain the differences in the areas that residents targeted. When analyses were limited to urban areas, residents still preferred the higher-supply areas (Figure 1). Limiting the analysis to urban areas attenuated the effect slightly, but the per-child supply of general pediatricians in the targeted areas was still nearly double that of the urban areas that the residents did not target (105 vs 55). Other differences that remained when contrasting urban areas with and without applications included higher median population incomes ($49 523 vs $43 498) and a higher percentage of black children (17% vs 11%). The remaining analyses compared demographics and job-search experiences for the 3 groups of residents on

Figure 1. Supply of general pediatricians for areas with and without a resident first-choice application.

the basis of the supply of general pediatricians per children in the area of each resident’s first choice. Residents applying to lower-supply areas tended to come from smaller residency programs (Table 2). The higher-supply group had a higher percentage of female residents, but that difference was not statistically significant. Job searches went more smoothly for residents who applied to the lower-supply areas (Table 3). Residents in the medium- and higher-supply groups averaged over 20 applications sent compared with only 5.5 for the lower-supply group. When directly asked about their job-search difficulty, markedly more residents from the medium- and higher-supply groups (43% and 48%) reported moderate or considerable difficulty in their job searches compared Table 2. Resident Characteristics by Job-Search Groups Pediatricians Per 100 000 Children Resident Characteristics†

Lower Supply (,54)

Medium Supply (54.1–97.1)

N 44 44 Woman, % 68 73 International medical graduate, % 11 14 Under represented minority, % 9 5 Married or living with partner, % 71 77 Have children, % 35 30 Residents per class, 17.0 mean 13.2*,** Educational debt (including spouse), mean 98 851 91 716

Higher Supply (.97.1) 44 80 19 12 72 40 17.7

99 138

†Statistical testing performed by using x2 tests and 1-way analyses of variance. *P , .05 for comparison between lower- and medium-supply groups. **P , .05 for comparison between lower- and higher-supply groups.

AMBULATORY PEDIATRICS

Resident Job Search

Table 3. Job-Search Experiences by Job-Search Groups

Table 4. Job Outcomes by Job-Search Groups

Pediatricians Per 100 000 Children Job-Search Measures† N Applications sent, mean Included part-time positions in job search, % Reported that spouse or partner’s career plan or family situation limited selection of positions, % Reported very few or no general pediatric jobs within 50 miles of training, % Experienced moderate or considerable difficult in job search, % Would choose pediatrics again if redid residency, % Had accepted a general practice position when surveyed, %

Lower Supply (,54)

Medium Supply (54.1–97.1)

Higher Supply (.97.1)

44 5.5*,**

44 27.9

44 22.2

19**

36

50

63

63

28

35

231

70

21

Pediatricians Per 100 000 Children Job-Search Measures†

Lower Supply (,54)

N 37 Starting salary, mean $114 486*,** Expected work hours in practice, mean 40.8*,** Position is part time, % 9** Job is located in same state as residency, % 56 Job is located in same city or area as residency, % 19*,**

Medium Supply (54.1–97.1) 30

Higher Supply (.97.1) 31

$102 759*** $86 960

36.1

34.2

20

28

60

73

41

60

†Statistical testing performed by using x tests and 1-way analyses of variance. *P , .05 for comparison between lower- and medium-supply groups. **P , .05 for comparison between lower- and higher-supply groups. ***P , .05 for comparison between medium- and higher-supply groups. 2

26**

95

43

48

100

93

70

71

84

†Statistical testing performed by using x and Kruskal-Wallis tests. *P , .05 for comparison between lower- and medium-supply groups. **P , .05 for comparison between lower- and higher-supply groups. 2

with residents in the lower-supply group (26%). More residents sought part-time positions in the medium- and higher-supply groups. However, interest in part-time positions does not appear to have confounded the relationship of supply and moderate or considerable search difficulty, for the relationship remained after removing parttime job seekers from the analysis (lower: 21%, medium: 33%, higher: 48%, P 5 .044). The bottom of Table 3 shows a nonsignificant trend where residents in the lower-supply group were more likely to have successfully obtained general-practice positions when surveyed. For those who did accept general-practice positions, the characteristics of those positions were also compared (Table 4). In all groups, the majority of residents accepted positions in the same state as their residency training. Residents in the medium- and higher-supply groups were significantly more likely to accept positions in the same city or area as their residency and accept part-time positions, as is reflected in the total number of practice hours that residents expected to work. The starting salaries for residents were more than $25 000 higher on average (32% relative difference) for the positions accepted by residents in the lower-supply group compared with the higher-supply group. The higher salaries in the lower-supply area were not a simple function of less parttime work. When part-time positions were excluded from the analysis, the mean expected work hours increased to 39.9 hours, but the salary difference between the highsupply ($95 722) and low-supply ($118 859) groups re-

mained over $23 000 (24% relative difference). These starting salaries may reflect even greater purchasing power difference because the median household incomes for the lower-supply areas were also lower (lower: $40 331, medium: $50 614, higher: $49 628). Residents in all groups were asked whether certain factors had discouraged their interest in practicing in rural areas or small towns. No significant differences in the pattern of responses were found across the supply groups. The most commonly reported discouraging factors in the groups were social life (53%), call coverage (48%), adequacy of professional back-up (47%), quality of schools (33%), and practice viability (22%). Figure 2 shows residents’ interests in working in a small town or rural area for 3 years in exchange for varying levels of hypothetical debt relief. Steep curves were apparent for all groups, showing that the vast majority of

Figure 2. Resident interest in rural areas as a function of hypothetical debt relief.

232

Cull et al

AMBULATORY PEDIATRICS

residents would consider working in these areas for substantial debt relief. As expected, residents from the lowersupply group had lower thresholds for reporting interest in these areas. The 40% debt-relief level showed the largest differences among the groups (P 5 .045). The crossover of the medium- and higher-supply groups in the upper ranges of debt relief may reflect the fact that the higher-supply group had greater levels of debt than did the medium-supply group (see Table 2). For the residents who reported any debt, the average amount of debt across all groups was $123 644 and the median debt amount was $110 000. When 40% debt relief is applied to the mean debt amount, for example, it computes to $49 458. DISCUSSION The job-search patterns of the newest group of pediatricians indicate little further improvement in the distribution of pediatricians in the United States. Currently, 48% of the PCSAs in the United States do not have a general pediatrician.17 Our results, however, show that residents continue to apply to areas with already high supplies of pediatricians. These findings suggest 2 important questions. First, what can be done to promote the geographic diffusion of pediatricians throughout the country? Second, are there areas of the country where the pediatrician-concentration levels are high enough to produce job-market pressures and resident job-search difficulty? Geographic Maldistribution Part of the maldistribution problem stems from the employee-owner conundrum. Pediatricians less frequently own their own practices,25,26 and new pediatricians are especially likely to be employed in group practices. Where residents locate, therefore, is tied to the current locations of pediatric practices and multispecialty groups, who are the most likely employers of newly trained pediatricians. To break this cycle, pediatric residents would need sources of start-up capital and better office-management training and career support. Even with these strategies, pediatricians would assume the financial risk of the practice and would need to find other clinicians to share call. Limits on practice viability in small towns, however, cannot be overlooked. Communities may have too few children to support a pediatric practice. Further research should examine the number of health care markets that cannot support a pediatrician. Perhaps because of greater health care utilization from the aging baby-boom generation, family physicians may be more amenable in the future to cooperative arrangements with a pediatrician to ensure adequate care for children in their area. Even with viable practice areas and promising practice arrangements, the important question remains of whether residents have interest in working in smaller towns and rural areas. Our results are somewhat mixed on this issue. On the one hand, most residents report that their spouse or partner’s career plans or their family situation limited their selection of positions, along with many other specific factors discouraging their consideration of these areas. On the other hand, smaller residency programs were more

likely to train residents who applied to the lower- or medium-supply areas. This may reflect the geographic areas where these programs are located, of the residents who enrolled in the smaller programs, or of aspects of training that may differ from larger programs. Regardless of their job-search group, a very high percentage of residents carrying debt reported that they would consider working in a small town or rural area only in exchange for substantial debt relief. For complete debt relief, 80% of residents indicated that they would consider these areas. Our results suggest that the maldistribution of pediatricians involves more than a simple distinction between urban and rural areas. When limiting the analysis of applications to residents sent to urban areas only, we found residents still preferred the higher-supply PCSAs over the lower-supply PCSAs. It is interesting that the percentage of children living in poverty did not differ from the desired and undesired urban areas. Two factors linked to the most desired areas were parttime opportunities and the location of residency programs. For part-time opportunities, residents may first desire parttime positions and then choose higher-supply areas in order to increase the chances of arranging such positions. Alternatively, residents may choose the higher-supply areas and then accept part-time positions because of competition for full-time positions in the area. For residency program locations, residents and their families establish many links to their communities during training and thus may prefer to stay there. This pattern, however, underscores the importance of the original policy decisions for locating residency programs and training clinics. More work is needed to better understand these and the other factors that make certain urban areas desirable and to examine whether these areas can absorb more pediatricians. Overall, these results suggest that the improvement of pediatrician geographic distribution will require efforts beyond market forces and current incentive programs like the National Health Service Corps.27 Factors that other specialties have examined include selective recruitment of medical students from rural areas, greater exposure to underserved areas during residency training, and loan forgiveness.19,20,28 Local-Area Pediatrician Concentration The results of the current study show a clear relationship between the per-child supply of pediatricians in a local area and the difficulty of residents’ job searches. Residents applying to the highest-supply areas sent out the most applications, reported the most difficulty in their job searches, and received the lowest starting salaries despite a higher household income in those areas. Residents’ strong interest in the high-supply areas adds to the gradual trend of declining job-search success for general-practice positions.18 These findings do not mean that residents cannot find practice positions. Even in the highest-supply areas, resiliency in residents’ attitudes and in the areas’ ability to absorb new pediatricians was apparent. In the higher-supply group, 93% of residents indicated that they would

AMBULATORY PEDIATRICS

choose pediatrics again if they redid their residency, and over 70% of the higher-supply group succeeded in obtaining general-practice positions at the time that they completed their surveys. Many other residents may also end up working in these areas as hospitalists or as pediatric subspecialists after fellowship training. There is no current definition of a physician-oversupply area. Although residents’ job-search experiences may be part of such a definition, the current results may fall just short. Half of residents applying to areas with more than 100 general pediatricians per 100 000 children did not report job-search difficulty, and most residents seeking general-practice positions in these areas did find them. Still, the profiles for these areas suggest real limits on the number of children available to practices. For example, the ratio of 100 general pediatricians to 100 000 children leaves 1000 children per pediatrician in the population, whereas staffing ratios commonly exceed 1500 children per pediatrician.29 These profiles and the greater market pressure seen for these areas relative to other areas suggest a need to continue to monitor residents’ job-search experiences at a local level. More research should focus on whether higher-supply levels translate into better overall health and well-being for children in the area. On the more positive side, our results show that residents with more interest in low- or medium-supply areas had much smoother job searches. These residents sent out only a handful of applications on average and received the highest-paying jobs. Also, the percentage of residents entering fellowships remained high, which may mean good news for the various pediatric subspecialties in need of additional fellows.30 This study has several limitations. First, respondents to the survey were slightly younger than nonrespondents, and the results also may not represent the experiences of nontraditional residents. Second, insofar as fewer than half of residents sought general-practice positions, the number of residents in each group was fewer than 50, which limits the generalizability of our findings. The finding of so many relationships despite this limitation in power reflects the robustness of the various effects. Third, because of the timing of the survey, some residents may have found jobs soon after completing their surveys, which may have underrepresented residents’ search success overall. Finally, our results do not reflect the jobsearch experiences of pediatricians who relocated. In conclusion, residents were more likely to seek jobs in urban areas with higher pediatrician supply and household income. Residents applying for positions in lowersupply regions experienced easier job searches. Many residents expressed interest in working in a small town or rural area for a limited time but only if substantial debt relief were offered. To improve the geographic distribution of pediatricians, targeted incentives beyond current programs and market forces will be needed. ACKNOWLEDGMENTS This research was partially funded by the Bureau of Primary Health Care and the Bureau of Health Professions, Health Resources

Resident Job Search

233

and Services Administration, US Department of Health and Human Services.

REFERENCES 1. Shipman SA, Lurie JD, Goodman DC. The general pediatrician: projecting future workforce supply and requirements. Pediatrics. 2004;113:435–442. 2. American Academy of Pediatrics, Committee on Pediatric Workforce. AAP Technical Report, The Pediatrician Workforce: Current status and future prospects. Pediatrics. In press. 3. Freed GL, Nahra TA, Wheeler JRC. Predicting the pediatric workforce: use of trend analysis. J Pediatr. 2003;143:570–575. 4. Chesney RW. How many pediatricians? Pitfalls in analysis and possible approaches. J Pediatr. 2003;143:550–552. 5. Cooper RA, Getzen TE, McKee HJ, Laud P. Economic and demographic trends signal an impending physician shortage. Health Aff (Millwood). 2002;21:140–154. 6. Grumbach K. The ramifications of specialty-dominated medicine. Health Aff (Millwood). 2002;21:155–157. 7. Mullan F. Some thoughts on the white-follows-green law. Health Aff (Millwood). 2002;21:158–159. 8. Weiner JP. A shortage of physicians or a surplus of assumptions? Health Aff (Millwood). 2002;21:160–162. 9. Mundinger MO. Through a different looking glass. Health Aff (Millwood). 2002;21:163–164. 10. Reinhardt UE. Analyzing cause and effect in the U.S. physician workforce. Health Aff (Millwood). 2002;21:165–166. 11. Snyderman R, Sheldon GF, Bischoff TA. Gauging supply and demand: the challenging quest to predict the future physician workforce. Health Aff (Millwood). 2002;21:167–168. 12. Barer M. New opportunities for old mistakes. Health Aff (Millwood). 2002;21:169–171. 13. Cull WL, Mulvey HJ, O’Connor KG, et al. Pediatricians working part time: past, present, and future. Pediatrics. 2002;109: 1015–1020. 14. Freed GL, Nahra TA, Wheeler JRC. Which physicians are providing health care to America’s children: trends and changes during the past 20 years. Arch Pediatr Adolesc Med. 2004;158: 22–26. 15. Forrest CB. Counting child health care professionals: will the United States ever have a coherent workforce policy for children’s health care? Arch Pediatr Adolesc Med. 2003;158:13–14. 16. 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. 17. American Academy of Pediatrics. Mapping healthcare delivery for America’s children. Available at: http://www.aap.org/ mapping/. Accessed December 29, 2004. 18. Cull WL, Yudkowsky BK, Shipman SA, Pan RJ. Pediatric training and job market trends: results from the AAP Third Resident Survey 1997–2002. Pediatrics. 2003;112:787–792. 19. Pathman DE, Taylor DH, Konrad TR, et al. State scholarship, loan forgiveness, and related programs: the unheralded safety net. JAMA. 2000;284:2084–2092. 20. Pathman DE, Konrad TR, King TS, et al. Outcomes of states’ scholarship, loan repayment, and related programs for physicians. Med Care. 2004;42:560–568. 21. Goodman DC, Mick SS, Bott D, et al. Primary Care Service Areas: a new tool for the evaluation of primary care services. Health Serv Res. 2003;38:287–309. 22. Pasko T, Smart DR. Physician Characteristics and Distribution in the US: 2003–2004 Edition. Chicago, Ill: American Medical Association; 2003. 23. US Bureau of the Census. Urban cluster definition. Available at: http://www.census.gov/geo/www/ua/uap2k.html. Accessed December 29, 2004. 24. Keppel G, Zedeck S. Data Analysis for Research Designs: Analysis of Variance and Multiple Regression/Correlation Approaches. New York, NY: WH Freeman & Co; 1989:145–182.

234

Cull et al

25. Brotherton SE, Tang SF, O’Connor KG. Trends in practice characteristics: analyses of 19 periodic surveys (1987–1992) of fellows of the American Academy of Pediatrics. Pediatrics. 1997; 100:8–18. 26. Kletke PR, Emmons DW, Gillis KD. Current trends in physicians’ practice arrangements: from owners to employees. JAMA. 1996;276:555–560. 27. Politzer RM, Trible LQ, Robinson TD, et al. The National Health Service Corps for the 21st century. J Ambul Care Manage. 2000;23:70–85.

AMBULATORY PEDIATRICS 28. Rabinowitz H, Diamond J, Markham F, Paynter N. Critical factors for designing programs to increase the supply and retention of rural primary care physicians. JAMA. 2001;286: 1041–1048. 29. Bocian AB, Wasserman RC, Slora EJ, et al. Size and age-sex distribution of pediatric practice: a study from Pediatric Research in Office Settings. Arch Pediatr Adolesc Med. 1999;153:8. 30. American Academy of Pediatrics, Committee on Pediatric Workforce. Pediatrician workforce statement. Pediatrics. In press.