High School Census Tract Information Predicts Practice in Rural and Minority Communities

High School Census Tract Information PredictsPractice in Rural and Minority CommunitiesSusan Hughes, MS;1 John Zweifler, MD, MPH;1 Sean Schafer, MD;2,3 Mark A. Smith, MD;4

Sukhdeep Athwal, BS;5 and H. John Blossom, MD1,6

ABSTRACT: Purpose: Identify census-derivedcharacteristics of residency graduates’ high schoolcommunities that predict practice in rural, medicallyunderserved, and high minority-population settings.Methods: Cohort study of 214 graduates of the Universityof California, San Francisco-Fresno Family PracticeResidency Program (UCSF-Fresno) from its establishmentin 1970 through 2000. Rural-urban commuting area code;education, racial, and ethnic distribution; median income;population; and federal designation as a medicallyunderserved area were collected for census tracts of eachgraduate’s (1) high school address and (2) practicelocation. Findings: Twenty-one percent of graduatespractice in rural areas, 28% practice in areas with highproportions of minority population (high minority areas),and 35% practice in federally designated medicallyunderserved areas. Graduation from high school in a ruralcensus tract was associated with rural practice (P , .01).Of those practicing in a rural site, 32% graduated froma rural high school, as compared with 11% of nonruralpractitioners. Graduation from high school in a censustract with a higher proportion of minorities was associatedwith practice in a proportionally high minority community(P¼ .01). For those practicing in a high-minority setting,the median minority percentage of the high school censustract was 31%, compared with 16% for people notpracticing in a high minority area. No characteristics ofthe high school census tract were predictive of practice ina medically underserved area. Conclusion: Census datafrom the residency graduate’s high school predicted ruralpractice and practice in a proportionally high minoritycommunity, but not in a federally designated medicallyunderserved area.

Although overall numbers seem adequate,many communities in the US needmore physicians.1 These communitiesare disproportionately rural,socioeconomically disadvantaged, and/

or have proportionally larger minority populations.1,2

Consequently, much research has been dedicated to the

study of determinants of rural and underserved medicalpractices. Gender, race and ethnicity of the physician,specialty, personal and societal considerations, andincome expectations all have been linked to practice siteselection.3-9

Using self-report instruments and observationaldesigns, investigators have shown that physiciansraised in rural areas more commonly enter ruralpractice.9,10 There are fewer studies looking atpredictors of practice in areas with high proportions ofminority population (high minority) and underservedareas.4,5 Studies like these suggest that physicians havea greater tendency to enter practice in rural and/orunderserved communities when their own backgroundsare socioeconomically and demographically similar.

The present study was designed to explore thehypothesis that practice in rural, underserved, andhigh minority communities can be predicted by US

1Family and Community Medicine, University of California,

San Francisco Fresno Medical Education Program, Fresno, Calif.2Office of Disease Prevention and Epidemiology,

Oregon Department of Human Services, Portland, Ore.3Office of Workforce and Career Development,

Centers for Disease Control and Prevention, Atlanta, Ga.4La Clinica de La Raza, Vallejo, Calif.5College of Osteopathic Medicine, Touro University, Vallejo, Calif.6California Area Health Education Center, Fresno, Calif.

This manuscript was partially supported by funding provided by the

University of California, San Francisco-Fresno Latino Center for

Medical Education and Research Mini-grant Program. Susan

Hughes was supported by Health Resources and Services

Administration Border Health Education and Training Center grant

No. 5 D39 HP 00023-09. H. John Blossom was supported by Area

Health Education Center grant No. 2 U77 HP 03015-59. For further

information, contact: Susan Hughes, MS; UCSF Fresno, Family and

Community Medicine, 155 N Fresno St., Fresno, CA 93701;

e-mail [email protected].

. . . . . Workforce Issues . . . . .

The Journal of Rural Health 228 Vol. 21, No. 3

Census–derived demographic and socioeconomicfeatures of the physician’s community of origin.

MethodsSample Population. There have been 214 graduates

of the University of California, San Francisco-FresnoFamily Practice Residency Program (UCSF-Fresno) fromits inception in 1970 through 2000. UCSF-Fresnograduates were enumerated from the Coastal ResearchDatabase (CRD), which is based on annual surveys ofCalifornia family practice graduates,11 and comparedagainst residency archives. One graduate not present inthe CRD was appended to the CRD list. Residencygraduates who graduated from high school in a foreigncountry or were currently in practice with the militarywere excluded from analyses. We sent a letter tograduates explaining our study and asking them toupdate their current practice address. This research wasapproved by the University of California, San FranciscoCommittee on Human Research (CHR number:H7572-17775-01).

Independent Variables. Each graduate’s age,gender, ethnicity, medical school attended, andgraduating high school name were taken from the CRD.When high school information was missing, wecontacted the graduate. We used the US Department ofEducation Web site12 to find the street address of eachgraduate’s high school. When street addresses were notavailable, we called the schools.

We identified the census tracts where the graduates’high schools were located using the Web site of theFederal Financial Institutions Examinations Council.13

Education distribution, racial distribution, ethnicdistribution, median income, and population wereobtained for all high school census tracts using the 1990US Census summary tape file 3A.14 We also obtaineda rural-urban commuting area code (RUCA), land area,population, and county metropolitan status for all1990 census tracts from the Economic Research Serviceof the US Department of Agriculture.15 We thencalculated population density from land area andpopulation information. RUCA is based on citypopulation and percentage of people who commuteto another city for work. RUCA values range from1 to 10: 1 to 3 define commuting patterns for urbanareas, 4 to 6 pertain to large towns (population 10,000to 50,000), 7 to 9 relate to small towns (population2,500 to 10,000) and 10 is for rural areas (less than2,500 population). The US Department of Health andHuman Services, Health Resources and ServicesAdministration (HRSA) designates census tracts asMedically Underserved Areas or Populations (MUAs).16

Dependent Variables. Clinical addresses obtainedfrom the CRD were checked against residencyarchives. In cases of a discrepancy, we referred toInternet yellow pages and the Administrators inMedicine, Association of State Medical BoardExecutive Directors Docfinder Web site.17 We usedthe same methods described above for the graduates’high school census tracts to identify the graduates’current practice census tracts and recorded the samedemographic variables and medically underservedstatus for each. We decided to use the last knownpractice address for the graduates who did notrespond or whose letters were returned, becausestudies of physician movement suggest that despitepractice site changes by individual physicians, theoverall number of physicians in any specified regionremains stable.18

Analysis. Data were entered into Microsoft Excelspreadsheets and analyzed using SAS (SAS InstituteInc., Release 8.2, Cary, NC) statistical analysis software.We calculated simple proportions and medians, whereappropriate, for demographic information derived fromCRD records of graduates. We divided all high schooland current practice census tracts into 2 levels on 3different domains: (1) Rural area; (2) MUA; and (3) Highminority area. Rural area was determined from theRUCA by considering all numbers greater than or equalto 4 as rural. MUA is a federal census tract leveldesignation. We defined as high minority census tractswith greater than 40% minority populations. We thencompared the census characteristics of the graduates’high school location with the current practice locationusing v2 tests for categorical variables (eg, high schoolin/not in MUA vs practice in/not in MUA). The SAS 1-way nonparametric analysis of variance median testwas used to analyze continuous variables (eg, medianfamily income of high school census tract vs practicein/not in MUA). Three multivariate logisticregressions were used to adjust for individualcharacteristics in each of the 3 domains previouslyspecified. Independent variables employed in eachmodel were: age, gender, minority race, Hispanicethnicity, rural training track, high school in rural area,high school in MUA, and high school census tractpercentage minority population.

ResultsAt UCSF-Fresno, 214 family practice residents

graduated between 1970 and 2000. Of the graduates,65% were men; 60% were non-Hispanic Caucasian,20% were Hispanic, 14% were Asian, and 6% wereother races. Twenty percent of the graduates


Hughes, Zweifler, Schafer, Smith, Athwal and Blossom 229 Summer 2005

participated in a rural continuity of care trainingexperience. Of the 214 graduates, 27 graduated from aninternational high school and 5 had unidentifiable UShigh school addresses and were excluded from thestudy. Of the remaining 182 US graduates, 1 was servingin the military, 1 was deceased, and 3 had unidentifiablepractice addresses. We chose to include the deceasedgraduate’s last practice address, because the death wasfrom natural causes unrelated to practice location. Alltold, 178 cases were left for analysis.

Of the graduates, 21% practice in rural areas, 35%practice in federally designated MUAs, and 28%practice in high minority areas. Eighty-two percent ofgraduates practice in California, where 12% of all censustracts are rural, 21% of all census tracts are designatedMUA, and 31% of the population is minority. Table 1shows the high school census tract characteristics for the3 different domains of practice location: rural, medicallyunderserved, and high minority. Rural high schoollocation was significantly associated with practice inrural areas (P , .01). For graduates practicing in a rurallocation, 32% (12 of 38) graduated from a rural highschool, compared with 11% (16 of 140) of those ina nonrural practice. None of the high school census tractvariables were associated with practice in MUAs.Practice in a high minority area was associated withlocation of high school in a higher minority setting(P¼ .01). The median minority percentage of thepopulation was 31% for those practicing in a highminority location, compared with 16% of those notpracticing in a high minority location.

Table 2 reports the findings from each logisticregression. Graduation from a rural high school (OR ¼

5.7; 95% CI, 2.0 to 16.4) was significantly associated withpractice in a rural area. Rural training during residencywas the only individual characteristic associated withrural practice (OR ¼ 2.7; 95% CI, 1.2 to 6.4). Nocharacteristics of our graduates (individual orotherwise) were associated with eventual practice inmedically underserved areas. Graduation from a highschool in a higher minority area was significantlyassociated with practice in areas with more than 40%minorities (OR ¼ 6.5; 95% CI, 1.1 to 37.6), but noindividual characteristics were significant.

DiscussionUsing US Census data based on high school

attended as a measure of characteristics of communityof origin, we have found that even after adjusting foran individual’s age, race, and gender, UCSF-Fresnofamily practice graduates raised in rural settings morecommonly enter rural practice, and those fromcommunities with more minorities practice in highminority communities. These findings complementprevious studies based on self-report of ruralupbringing9,10 and race or ethnicity of theindividual physician.4,5 The consistency of thesefindings with earlier research that used alternativemethods also provides some evidence of convergentvalidity for our use of census tract data from thehigh school attended.

We were somewhat surprised not to finda relationship between ethnic makeup of thecommunity, population density, or any othercensus-derived characteristic of high school attended

Table 1. Graduates’ High School Characteristics by Practice Location

High School Census Tract Characteristic

HighSchool inRural Area

High Schoolin MedicallyUnderservedArea (MUA)

MedianFamilyIncome

MedianMinority

Population

MedianHispanic

Population

Median AdultPopulation WithLess Than HighSchool Diploma

All graduates (n ¼ 178) 16% (28) 16% (29) $38,994 18% 11% 16%

Rural practice (n ¼ 38) 32% (12)* 13% (5) $38,203 12% 8% 15%Nonrural practice (n ¼140) 11% (16) 17% (24) $39,102 19% 12% 16%

High minority practice (n ¼ 49) 16% (8) 18% (9) $38,569 31% 21% 18%Not high minority practice

(n ¼129)16% (20) 16% (20) $39,141 16% 10% 15%

MUA practice (n ¼ 62) 15% (9) 13% (8) $40,691 18% 11% 14%Not MUA practice (n¼116) 16% (19) 18% (21) $38,660 18% 12% 17%

* Bolded values significant at a¼ 0.05 level.



and medically underserved practice as defined bythe federal government. Besides the possibility thatthere is no underlying relationship between thedemographics of high school community andunderserved practice, there are other explanations forour negative findings. First, sample size may havelimited our ability to detect associations. Second, highschool graduated may not be the best surrogatemarker for early environmental influences. Third, ouruse of census data from one point in time mayweaken the association between community oforigin and practice location. Finally, census tractsthemselves may not capture the true demographicconstitution of the communities where our graduateswere raised.

There is also the question of whether MUAs area true representation of medically needy communities.We settled on the federal MUA designation as the mostconvenient, appropriate, and accessible source ofinformation on underserved status. However, localentities must apply for the MUA designation. It is likelythat some impoverished or underserved communitiesmay not be designated MUA and that other less

disadvantaged communities, recognizing an economicor political advantage, would have sought and havereceived this designation. Each of these circumstancescould shift the association between high school attendedand practice in an MUA in either direction.

Despite the relative convenience of census data,our study had limitations. Census analysis is limitedto the information collected by the census. Thus, ourstudy does not contribute to our understanding ofwhy physicians from rural or high minoritycommunities are more likely to return to practice insimilar communities. We used information from onlyone point in time, the 1990 census. Data from the2000 census were not available electronically at thetime of our analysis. Census data from the 1960s,1970s, and 1980s (when many of the program’sgraduates were in high school) were not availableelectronically, making it a practical impossibility forus to correct for secular demographic trends and thevariability of associations between high school andpractice location. Similarly, the use of 1990 censusinformation misses demographic changes that mayhave occurred during the ensuing 10 years. Using onlyone point in time may weaken associations betweenpractice location and community of origin, biasingtoward the null hypothesis.

The definition of rurality was elusive. Thefederal government also uses a definition of‘‘nonmetropolitan area.’’ However, nonmetropolitanareas are defined by county borders. Sprawlingcounties may include urban communities, as well assparsely populated regions. A good example isFresno County, California, which is a majoragricultural county, yet is defined as a metropolitanarea because it includes the eighth largest city inCalifornia (47th largest city in the US).19 Populationdensity is another measure associated with ruralitythat can be calculated. Other investigators have usedUS Department of Agriculture–defined commutingcodes to better define rural locations. We eventuallysettled on commuting codes for simplicity, butexplored all 3 methods of defining rurality andfound associations with high school of graduation andpractice location for each (data not shown).

Clear evidence of characteristics that predict practicein rural and/or underserved settings would bea powerful tool for health policy makers and admissionscommittees at medical schools and residency programswho have the goal of supplying physicians tounderserved communities. Our study supportsassertions that rural physicians come from ruralbackgrounds and train in rurally oriented programs.The study also supports the claim that physicians fromminority areas go on to practice in high minority areas.

Table 2. Adjusted Odds Ratio for Rural, HighMinority, and Medically Underserved(MUA) Practice

Variable

Model

RuralPractice*

HighMinorityPracticey

MUAPractice�

High school in ruralcensus tract 5.7 (2.0-16.4)§ 1.4 (0.5-4.0) 1.0 (0.4-2.7)

Minority % populationof high schoolcensus tract 1.3 (0.2-10.5) 6.5 (1.1-37.6) 0.8 (0.2-4.5)

High school in MUAcensus tract 0.3 (0.1-1.3) 0.7 (0.2-2.0) 0.7 (0.2-2.0)

Graduate in ruraltraining track 2.7 (1.2-6.4) 1.0 (0.4-2.3) 1.1 (0.5-2.5)

Age of graduate 1.1 (1.0-1.2) 1.0 (0.9-1.1) 1.0 (0.9-1.1)Gender of graduate 0.7 (0.3-1.7) 1.5 (0.7-3.2) 1.4 (0.7-2.8)Minority graduate 0.8 (0.2-2.9) 1.3 (0.5-3.6) 1.2 (0.5-3.2)Hispanic graduate 1.0 (0.3-2.7) 1.1 (0.4-2.7) 1.8 (0.8-4.2)

* Rural Practice defined as a Rural-Urban Commuting AreaCode �4.y High Minority Practice defined as a census tract with .40%

minority population.� Medically Underserved Area (MUA) Practice is a federal census

tract designation.§ Bolded values are significant at a ¼ 0.05 level.


Hughes, Zweifler, Schafer, Smith, Athwal and Blossom 231 Summer 2005

However, it is important to emphasize that we did notuncover any significant associations betweencommunity of origin as defined by high school ofgraduation and eventual practice in federallydesignated MUAs. Thus, selecting medical studentsfrom rural and high minority communities does notnecessarily address the needs of the underservedpopulations in these communities. Although noassociation between the socioeconomic variables of thecommunity of origin and eventual MUA practice wasfound, it is possible that the socioeconomiccharacteristics of the individual or their family mayplay a role. Further study aimed at identifyingcharacteristics of individuals who practice inunderserved areas is still needed.

ConclusionsThis study used census tracts of physician practices

and high schools attended as demographic proxies.Previous studies demonstrating the influence of ruralupbringing on rural practice were supported. Theassociation between high school education in higherminority areas leading to practice in higher minorityareas also complements earlier findings. However,location of high school graduated did not predictpractice within a federally designated MUA. Medicalschools need stronger predictors than high school oforigin if they hope to train students who will eventuallypractice in underserved settings. Similarly, healthworkforce planners may need to continue to rely onincentive programs that attract physicians tounderserved areas regardless of where they grew up.

References1. Council on Graduate Medical Education. Physician Distribution and

Health Care Challenges in Rural and Inner-City Areas: Tenth Report.Washington, DC: US Department of Health and Human Services;February 1998.

2. Coffman JM, Young JQ, Vranizan K, Blick N, Grumbach K.California Needs Better Medicine. San Francisco: University ofCalifornia, San Francisco Center for the Health Professions; 1997.

3. Spenny ML, Ellsbury KE. Perceptions of practice among ruralfamily physicians—is there a gender difference? J Am Board FamPract. 2000;13(3):183-187.

4. Komaromy M, Grumbach K, Drake M, et al. The role of black andHispanic physicians in providing health care for underservedpopulations. N Engl J Med. 1996;334(20):1305-1310.

5. Cantor JC, Miles EL, Baker LC, Barker DC. Physician service to theunderserved: implications for affirmative action in medicaleducation. Inquiry. 1996;33(2):167-180.

6. Schmittling G, Black RR, Stern TL, Clinton C. Practice locations offamily practice residency graduates. J Med Educ.1981;56(9; Pt 1):709-716.

7. Kiker BF, Zeh M. Relative income expectations, expectedmalpractice premium costs, and other determinants of physicianspecialty choice. J Health Soc Behav. 1998;39(2):152-167.

8. Escarce JJ, Polsky D, Wozniak GD, Pauly MV, Kletke PR. Healthmaintenance organization penetration and the practice locationchoices of new physicians: a study of large metropolitan areas inthe United States. Med Care. 1998;36(11):1555-1566.

9. Rabinowitz HK, Diamond JJ, Veloski JJ, Gayle JA. The impact ofmultiple predictors on generalist physicians’ care of underservedpopulations. Am J Public Health. 2000;90(8):1225-1228.

10. Rabinowitz HK, Diamond JJ, Markham FW, Paynter NP. Criticalfactors for designing programs to increase the supply andretention of rural primary care physicians. JAMA.2001;286(9):1041-1048.

11. Coastal Research Group. Coastal Research Group Web site.Available at: http://www.coastalresearch.org. AccessedOctober 31, 2001.

12. Find a College, K-12 School, or Library. US Department ofEducation Web site. Available at http://www.ed.gov. AccessedAugust 2000-April 2001.

13. FFIEC Geocoding System. Federal Financial InstitutionsExamination Council Web site. Available at http://www.ffiec.gov/geocode/default.htm. Accessed August 2000-April 2001.

14. 1990 Census Lookup. US Census Bureau Web site. Available athttp://www.venus.census.gov/cdrom/lookup. Accessed August2000-April 2001.

15. Rural-urban Commuting Area Code. Economic Research Serviceof the US Department of Agriculture Web site. Available at http://www.ers.usda.gov/emphases/rural/data/ruca/. AccessedOctober 2000.

16. Data and statistics: medically underserved areas/medicallyunderserved populations. US Department of Health and HumanServices, Health Resources and Services Administration Web site.Available at http://www.hrsa.gov/data.htm. AccessedNovember 2000.

17. AIM docfinder. Administrators in Medicine, Association ofState Medical Board Executive Directors Web site. Available athttp://www.docboard.org/. Accessed August 2000-April 2001.

18. Ricketts TC, Tropman SE, Slifkin RT, Konrad TR. Migration ofobstetrician-gynecologists into and out of rural areas. Med Care.1996;34(5):428-438.

19. Population of the 100 largest urban places: 1990. US CensusBureau Web site. Available at http://www.census.gov/population/documentation/twps0027/tab22.txt. AccessedMay 2001.



High School Census Tract Information Predicts Practice in Rural and Minority Communities

Documents

Transcript of High School Census Tract Information Predicts Practice in Rural and Minority Communities