rural–urban differences in primary care physicians' practice patterns, characteristics, and...

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

Rural–Urban Differences in Primary CarePhysicians’ Practice Patterns, Characteristics,and IncomesWilliam B. Weeks, MD, MBA;1,2,3,4 and Amy E. Wallace, MD, MPH1,3

ABSTRACT: Context: Low salaries and difficult workconditions are perceived as a major barrier to therecruitment of primary care physicians to rural settings.Purpose: To examine rural–urban differences inphysician work effort, physician characteristics, andpractice characteristics, and to determine whether, afteradjusting for any observed differences, rural primary carephysicians’ incomes were lower than those of urbanprimary care physicians. Methods: Using survey datafrom actively practicing office-based general practitioners(1,157), family physicians (1,378), general internists(2,811), or pediatricians (1,752) who responded to theAmerican Medical Association’s annual survey ofphysicians between 1992 and 2002, we used linearregression modeling to determine the association betweenpracticing in a rural (nonmetropolitan) or urban(standard metropolitan statistical area) setting andphysicians’ annual incomes after controlling for specialty,work effort, provider characteristics, and practicecharacteristics. Findings: Rural primary care physicians’unadjusted annual incomes were similar to their urbancounterparts, but they tended to work longer hours,complete more patient visits, and have a much greaterproportion of Medicaid patients. After adjusting for workeffort, physician characteristics, and practicecharacteristics, primary care physicians who practiced inrural settings made $9,585 (5%) less than their urbancounterparts (95% confidence intervals: −$14,569,−$4,602, P < .001). In particular, rural practicinggeneral internists and pediatricians experienced lowerincomes than did their urban counterparts. Conclusions:Addressing rural physicians’ lower incomes, longer workhours, and greater dependence on Medicaidreimbursement may improve the ability to ensure that anadequate supply of primary care physicians practice inrural settings.

Access to a full spectrum of healthcare services for rural populations isoften limited by the ability to successfullyrecruit and retain physicians to workin rural settings.1-6 Although federally

funded7,8 and nonfederal9,10 programs have beendesigned to improve recruitment of physicians to ruralareas, physicians continue to be reluctant to locate theirpractices in rural settings.2,11,12 In particular,administrators who manage community health centersthat are located in rural settings report having muchmore difficulty recruiting and retaining physicians thanthose who manage urban community health centers, achallenge that is reflected in substantially highervacancy rates for primary care physicians such asfamily physicians and pediatricians.1

There are several explanations for the difficultiesrecruiting and retaining physicians to rural practicesettings, including limited spousal job opportunities,cultural and professional isolation, and lack of controlover work hours.1,5,13 However, inadequate totalprofessional reimbursement is frequently cited as amajor barrier to recruitment of physicians to work inrural settings.1,3-5,7,12,14-16

1VA Outcomes Group Research Enhancement Award Program,White River Junction VA Medical Center, White River Junction, Vt.2Department of Community and Family Medicine, DartmouthMedical School, Hanover, NH.3Department of Psychiatry, Dartmouth Medical School, Hanover,NH.4The Dartmouth Institute for Health Policy and Clinical Practice,Lebanon, NH.

This work was supported in part by VA Health Services Researchand Development Grant REA 03-098. The views expressed in thisarticle do not necessarily represent the views of the Departmentof Veterans Affairs or of the United States government. Dr. Weekshad full access to all of the data in the study and takesresponsibility for the integrity of the data and the accuracy of thedata analysis. For further information, contact: William B. Weeks,MD, MBA, VA Medical Center (11Q), White River Junction, VT 05009;e-mail [email protected].

C© 2008 National Rural Health Association 161 Spring 2008


While 1 recent report that used data from theAmerican Medical Association for 1986-1996 to examinetrends in physician incomes for general practitionersand family physicians who practiced in rural settingsfound that rural physicians’ incomes were growing at aslower rate than urban physicians’ incomes while theirpatient care hours per week were increasing faster, thatstudy did not incorporate differences in work effort orphysician characteristics into its analysis.17 A recentstudy challenged the assertion that physicians whopractice in rural settings are poorly compensated whencompared to their urban counterparts: while ruralprimary care physicians had lower unadjusted incomesthan their urban counterparts, after adjustment forlower costs of living and practicing in rural settings, thestudy found that rural primary care physicians’incomes were 10%-13% higher than those of their urbancounterparts.18 However, that study used only a singleyear’s data from the Center for Studying Health SystemChange’s Community Tracking Study PhysicianSurvey19 and did not adjust for a number of importantvariables including race, the inverted U lifetimeearnings pattern that is typical for physicians, and theUS Census region practice location that we havepreviously shown to be associated with the annualincomes of general internists20 and family physicians.21

Accurately assessing the differences in workcharacteristics of physicians who live in rural ascompared to urban settings is important for 2 reasons.First, lack of availability of physicians who work inrural settings may restrict access to health care for therural population.5,12,22 Rural families tend to have lowerincomes than do urban families23,24 and are less likely tohave health insurance that is employment-based,25,26

factors that may further restrict access to care.27,28 Thislack of access to health care could contribute to thelower health-related quality of life seen in the general29

and veteran28 rural populations.Second, policymakers and recruiters have

historically used a combination of incentives such asbonuses and educational debt relief to recruitphysicians to rural settings.9,10,14,30 There is someevidence that current efforts are inadequate. First,foreign medical school graduates, who may have fewerpractice options than US graduates, reportedly aremore likely to work in rural settings.31 In addition, theongoing difficulty of retaining physicians in ruralsettings32 suggests that current incentive plans areinadequate, although additional, nonfinancial factorsare undoubtedly important as well.1,4

Therefore, we used data collected through theAmerican Medical Association’s SocioeconomicMonitoring System from 1992 to 200133 to examinerural–urban differences in primary care physicians’

characteristics, practice patterns, and incomes with aneye toward understanding interventions that mightimprove recruitment and retention of rural physicians.

MethodStudy Design and Data. We used survey data to

conduct a cross-sectional retrospective analysis ofphysicians’ annual incomes. Between 1992 and 2001,the American Medical Association (AMA) conductedregular telephone surveys of physicians that collected abroad variety of individual physician level data,including weeks and hours of practice, number ofpatient visits seen, provider characteristics, practicecharacteristics, and physician incomes.33 The surveywas designed to provide representative information onthe population of actively practicing nonfederalphysicians who spend the greatest proportion of theirtime in patient care activities; weights for eachrespondent were calculated to correct for potential biascreated by unit nonresponse, survey eligibility, and toensure that physician responders reflected the nationaldistribution of physicians.33 This study was approvedby Dartmouth Medical School’s Committee for theProtection of Human Subjects, Hanover, NH (CPHS #17707).

Survey Methods. Each year, the AMA conducted atelephone-administered survey on a random sample ofeligible physicians from the AMA Masterfile. Thefollowing physicians were excluded: doctors ofosteopathy, foreign medical graduates with temporarylicensure, inactive physicians, physicians who weresampled during the last 5 years, physicians who are onthe “do not contact” list, physicians not practicing inthe United States, and physicians who have no license.In addition, after initial screening, federally employedphysicians and physicians who spent less than 20 hourseach week in patient care activities were excluded.

The following field procedures were developed tominimize nonresponse bias: 2 weeks prior to datacollection, advance letters were sent describing theprocess and the survey; many specialty organizationsprovided endorsement letters; and summaries of thetype of questions to be asked were provided in advanceof the survey. In addition, a minimum of 4 callbacks torespondents were made before abandoning interviewefforts, letters encouraging participation were sent tophysicians who initially refused participation, andrefusal conversion attempts were made by selectinterviewers.33 In general, response rates were verygood and ranged 55%-77% over the years examined.

Survey Weights. Survey weights were derived byfirst dividing the AMA Physician Masterfile population

The Journal of Rural Health 162 Vol. 24, No. 2


and survey respondents into 200 cells defined byspecialty, years since the respondent received an MD,AMA membership status, and board certificationstatus. Unit response rates were constructed as the ratioof the number of physicians in the population to thenumber of respondents in each cell. Second, aneligibility correction was employed, as only nonfederalpatient care physicians—excluding residents–wereeligible. The eligibility weight was determined bydividing the subset of the population for whicheligibility was known into 40 cells (according to yearsin practice, AMA membership status, gender, andboard certification) and calculating the proportion ofphysicians in each cell who were eligible. The overallweight applied for a given respondent was the productof the unit response weight and the eligibility weight.33

Sample. Although the survey had been conductedfor much longer, this analysis was limited to datacollected between 1992 and 2001 for 2 reasons. First,during the study period, the survey methods allowedus to categorize physicians into well-defined specialtygroups in a way that allowed for the disaggregation ofresponses from different medical specialists. Previously,general internists and internal medicine subspecialtieswere aggregated, as were general practitioners andfamily medicine physicians. Second, these were themost recent data available for analysis, and thereforelikely to be the most relevant to the currently practicingphysician workforce.

To ensure that all the physicians analyzed werecomparable, that variables critical to the analysis wereavailable for each subject, and that extreme outliers didnot drive results, a sequential process of elimination ofsurvey respondents was used. First, we includedphysicians identified as practicing in an “office basedpractice” and who were in 1 of 4 primary carespecialties—general practice (1,820), family medicine(2,168), general internal medicine (4,479), or pediatrics(2,693)—in the study. This eliminated the minority ofphysicians who worked primarily doing research, asmedical educators, as administrators, or in hospitalsettings. We included only respondents who providedinformation on annual incomes and other key variables(annual hours worked, number of years practicingmedicine, ownership interest, and practice location).Finally, because we were concerned that outliers mightboth skew results and represent data entry errors, welimited our analysis to physicians who had annualincomes and numbers of patient visits that werebetween the 1st and 99th percentile of those metrics fortheir specialty. This process left 1,157 generalpractitioners, 1,378 family physicians, 2,811 general

internists, and 1,752 pediatricians for analysis(weighted to 1,205 general practitioners, 1,301 familyphysicians, 3,293 general internists, and 1,627pediatricians). The number of respondents by specialtyand percentage that were rural are shown in Figure 1.The proportion of rural respondents for each specialtywas not markedly different when comparing the finalsample for analysis to all respondents.

Variables Proposed to Influence Physicians’Incomes. From the AMA data set, we extracted 3 typesof independent variables likely to influence thedependent variable—net annual income:

1. Physician workload as measured by annual hours workedand number of patient visits completed. Both hoursworked34-36 and number of patient visitscompleted20,21 have been used to account fordifferences in primary care physicians’ incomes.

2. Provider characteristics. When making gendercomparisons of physician incomes, age hascommonly been used as an adjustment factor.34,35

Over the working lifetime, incomes follow an“inverted U” pattern37 that typically peaks near age55 for primary care physicians,38,39 or after 20-25years of practicing medicine. To dispel a concernthat graduation from a foreign medical school mightinfluence the age at which a physician enteredmedical school, and therefore bias results, weincorporated the number of years that respondentshad been practicing medicine into the analysisinstead of physician age. In addition, becausepractice arrangements, such as having an ownershipinterest in the practice, are associated withdifferences in annual income among physicians,40

we included whether the physician was anemployee, as opposed to a full or partial owner ofthe practice, in the analysis. Because boardcertification is associated with higher physicianincomes,41 we included board certification status asan independent variable in the analysis. Finally,because race,20,21 gender,20,21,35 and internationalmedical school graduate status31 have beenassociated with work effort, practice characteristics,and incomes of primary care physicians, weincluded those factors in our analysis.

3. Practice characteristics. Physician incomes varyaccording to US Census region of practice location33;therefore, information on the US Census region, inwhich the practice was located, was collected.

Rural–Urban Designation. The SocioeconomicMonitoring Survey assigned standard metropolitanstatistical area codes to individual respondents, basedon the county in which they practiced. The United

Weeks and Wallace 163 Spring 2008


Figure 1. Exclusions and Method of Identifyingthe Sample Studied. The Numbers inParentheses Provide the Number ofRespondents and the Proportion WhoPractice in Rural Settings.

States Office of Management and Budget definesmetropolitan statistical areas according to publishedstandards that are applied to Census Bureau data.42 Inthis study, physicians in nonmetropolitan areas wereclassified as rural and those in metropolitan areas asurban.

Calculated and Dummy Variables. We used theconsumer price index43 to adjust reported net annualincome to constant 2004 dollars—so-calledinflation-adjusted annual incomes. For instance, toinflate income reported for 1995 to 2004 dollars, wemultiplied the reported income in 1995 by theconsumer price index in 2004 (188.9) and then dividedthat figure by the consumer price index in 1995 (152.4).Following a common practice,20,21,35,36,38,39,44-46 wemultiplied the reported number of weeks worked in thelast year by the total number of hours worked in thelast week to calculate the annual number of hoursworked. To account for the “inverted U” relationshipbetween number of years practicing medicine andannual incomes, we constructed dummy variables thatreflected the categorization of years practicing medicineinto 5-year increments, from 0 to 5 years practicingthrough 40-plus years practicing. Finally, to account forthe possibility of overall trends over the time periodstudied, we included a variable representing the yearthat the survey was conducted in theanalysis.

Analysis. We used the Student’s t-test forcontinuous variables and the chi-square test fordichotomous variables when comparing results forrural physicians to results urban for physicians, foreach specialty. To determine the association betweenliving in a rural setting and physicians’ incomes, afteradjustment for specialty, work hours, and practice andprovider characteristics, we used a linear regressionmodel that simultaneously entered the independentvariables detailed above and used the consumer priceindex-adjusted annual income as the dependentvariable. We used dummy variables to representdifferent binary conditions such as being in a particularyears-in-practice category, being board certified, orliving in a particular US Census region. We developed 2linear regression models that incorporated type ofspecialty to calculate an overall rural to urbancomparison of adjusted incomes for primary carephysicians: one used inflation-adjusted incomes andthe second used natural log-transformedinflation-adjusted incomes. The regression coefficientsof the first model represent dollar differences and theregression coefficients of the second representpercentage differences after correcting for all othervariables in the model. We used SPSS (Version 13.0,Base and Advanced ModelsTM add-on module using



the General linear models procedure, Chicago, Ill) andthe survey weights described above for all analyses.

ResultsAfter adjusting only for inflation in the sample

examined, we found that only rural generalpractitioners had statistically significantly differentannual incomes than their urban counterparts (Table 1).Rural general practitioners’ inflation-adjusted incomeswere $8,958 (6%) higher than were urban generalpractitioners’ incomes (P = .036). Only amongpediatricians were rural primary care physicians’inflation-adjusted incomes lower than urban primarycare physicians’ incomes, but not statistically significantso. Rural primary care physicians completedsubstantially more patient visits (22% more for generalpractitioners, 14% for family physicians, 19% forgeneral internists, and 13% for pediatricians, all P <

.001) than did their urban counterparts. Except forpediatricians, rural primary care physicians alsoworked more annual hours (8% more for generalpractitioners, 7% for family physicians, and 4% forgeneral internists, all P < .01) than their urbancounterparts.

In the sample examined, when compared to theirurban counterparts, rural primary care physicians weremore likely to be an owner of the practice (statisticallysignificant for all specialties except general internalmedicine), less likely to be female (statisticallysignificant for all specialties except pediatrics), andequally likely to be board certified. Rural generalpractitioners and family physicians were more likely tobe white and less likely to be international medicalschool graduates than were their urban counterparts.Rural primary care physicians were more likely to livein the Southern Census region and less likely to live inthe Northeast Census region. For every primary carespecialty examined, physicians who practiced in ruralsettings had about 1.5 times as many patients onMedicaid as did their urban counterparts (P < .001 forall).

As expected, in our regression models we found astatistically significant relationship between additionalwork effort and higher annual incomes within thesample (Table 2). Similarly, the inverted U-shapedincome-earning curve was captured in the model, withphysician incomes peaking at 20–24.9 years of practice,and declining thereafter. Board certification wasassociated with higher annual incomes, while femalegender was associated with lower annual incomes.While white race was not associated with incomedifferences, international medical school graduatesshowed a trend toward having higher incomes than US

medical school graduates. As groups, after adjusting forother variables, general practitioners, familyphysicians, and pediatricians had lower annualincomes than general internists. For the primary carespecialties examined, practices located in the Northeasthad lower annual incomes than those in other Censusregions. A greater proportion of Medicaid enrollees wasassociated with lower annual incomes, but Medicareprovider status was not a statistically significantcontributor to the model. Our results suggest that, afteradjustment for inflation and the other variables in themodel, there was a downward trend of about 0.3% peryear in overall incomes for primary care physiciansduring the time period examined. After adjusting forwork effort, physician characteristics, and practicecharacteristics, primary care physicians who practicedin rural settings made $9,585 less than their urbancounterparts (95% confidence intervals: −$14,569,−$4,602, P < .001). Using natural log-transformed data,we found that primary care physicians who worked inrural settings made 5.0% lower annual incomes than dotheir urban counterparts (95% confidence intervals:−7.9%, −2.1%, P = .001).

The results of the regression models performed atthe specialty level are shown in Figure 2. Within thesample, the adjusted annual incomes for generalistswho practiced in rural settings (3.7% lower, 95% CI:−9.6%, 2.3%) and family physicians who practiced inrural settings (1.6% lower, 95% CI: −6.8%, 3.6%) werenot statistically different than the incomes of those whopracticed in urban settings. However, the adjustedannual incomes for general internists who practiced inrural settings (5.4% lower, 95% CI: −10.3%, −0.4%) andpediatricians who practiced in rural settings (8.8%lower, 95% CI: −16%, −1.6%) were significantly lowerthan that of their urban counterparts.

DiscussionThis study examined provider and practice

characteristics that were likely to be associated withprimary care physicians’ annual incomes and revealeddifferences attributable to whether they practiced in arural or urban setting. We found that rural primary carephysicians have similar annual incomes to their urbancounterparts, but they tend to work longer hours,complete more patient visits, and care for a muchgreater proportion of Medicaid enrollees. Afteradjusting net annual incomes for observed differences,we found that living in a rural setting wasindependently associated with modestly lower netannual incomes among primary care physicians,specifically so for general internists and pediatricians.



Table 1. Sample Characteristics of Primary Care Physicians by Rural–Urban Status

Rural Urban P Value

General PracticeInflation-adjusted annual income (2004 dollars) $153,775 $144,817 .036Physician work effort

Total annual visits 7,402 6,049 <.001Total annual hours worked 2,898 2,672 <.001

Provider characteristicsYears in medical practice 19.8 18.1 .02Physician is a nonowner 23.0% 36.0% <.001Physician is board certified 60.0% 60.7% .81Physician is female 8.1% 17.8% <.001Physician is white 90.1% 80.3% <.001Physician is an international medical school graduate 10.1% 19.2% <.001

Practice characteristicsCensus region of practice

Northeast Census region 9.8% 14.9% .014North Central Census region 34.1% 23.1% <.001Southern Census region 37.4% 31.7% .05Western Census region 18.5% 30.4% <.001

Service populationProportion providing Medicare 97.2% 94.5% .03Proportion of patients on Medicaid 19.7% 13.1% <.001

Family medicineInflation-adjusted annual income (2004 dollars) $169,811 $161,783 .07Physician work effort

Total annual visits 6,664 5,854 <.001Total annual hours worked 2,825 2,641 <.001

Provider characteristicsYears in medical practice 17.0 15.9 .053Physician is a nonowner 32.3% 49.7% <.001Physician is board certified 83.6% 83.6% .99Physician is female 13.7% 19.0% .022Physician is white 94.4% 84.0% <.001Physician is an international medical school graduate 6.3% 15.6% <.001


Northeast Census region 11.0% 17.5% .003North Central Census region 34.7% 24.1% <.001Southern Census region 38.8% 31.0% .007Western Census region 15.6% 27.4% <.001


General internal medicineInflation-adjusted annual income (2004 dollars) $190,868 $186,393 .38Physician work effort

Total annual visits 6,092 5,140 <.001Total annual hours worked 3,019 2,897 .009

Provider characteristicsYears in medical practice 14.0 15.1 .007Physician is a nonowner 35.9% 40.6% .08Physician is board certified 76.9% 76.0% .67Physician is female 11.9% 16.8% .016Physician is white 71.0% 76.3% .028Physician is an international medical school graduate 27.7% 26.5% .62


Northeast Census region 16.3% 27.9% <.001North Central Census region 26.0% 20.5% .013



Table 1. Continued

Rural Urban P Value

Southern Census region 44.4% 30.6% <.001Western Census region 13.2% 21.1% <.001


PediatricsInflation-adjusted annual income (2004 dollars) $157,406 $165,809 .19Physician work effort

Total annual visits 6,722 5,934 .001Total annual hours worked 2,695 2,711 .82

Provider characteristicsYears in medical practice 16.3 15.0 .10Physician is a nonowner 33.3% 45.1% .005Physician is board certified 83.0% 85.4% .44Physician is female 30.1% 35.0% .22Physician is white 75.5% 75.6% .98Physician is an international medical school graduate 18.4% 23.7% .14


Northeast Census region 9.8% 24.7% <.001North Central Census region 20.4% 19.5% .78Southern Census region 49.0% 33.2% <.001Western Census region 20.4% 22.7% .51


The reduced income that we found is remarkablyconsistent with findings based on the Center forStudying Health System Change’s CommunityTracking Study (CTS) Physician Survey that in 2001rural physicians had annual incomes that were 4%-9%lower than their urban counterparts.18 Using a differentsource of data, our analysis—which includedadjustment for race, gender, and international medicalschool graduate status and inclusion of Census region,number of patient visits, and the typical lifetimeearnings pattern—confirmed those results. Therefore, itseems likely that rural primary care physicians havemodestly lower incomes after adjusting for work hoursand physician characteristics.

This analysis has several limitations. First, thestudy was limited by the survey methodology used bythe American Medical Association. This establishedsurvey of physicians demonstrated substantialyear-to-year variation in number of respondents duringthe time period examined. However, the ability tocombine 10 years of data strengthened the study andoffered a much more robust data set than would havebeen the case had fewer years of data been available.Second, although we used the most recent data

available, the data that we analyzed are becomingdated. It is possible that more recent data, should theybecome available, would show different results.

Third, the study was inherently limited byavailable data. Although it would have been interestingto explore alternative explanations for the incomedifferences that were found, such as the standard ofliving in rural as compared to urban settings, theproportion of charity care provided by rural ascompared to urban dwelling primary care physicians,or respondents’ levels of satisfaction with theirpractices, the data required to answer these questionswere not available. Our regression models accountedfor only about 23% of the variance in physicianincomes; clearly, additional factors that were notincorporated into the analysis are likely to influenceexpected physician incomes and might mitigate thedifferences found here.

Finally, our findings are of an associative, notcausative, nature. Additional study is required todetermine causal pathways that might be associatedwith the lower incomes that rural primary carephysicians received. A variety of potential explanationsfor our findings may exist—for instance, rural



Table 2. Coefficients and 95% Confidence Intervals for the Regression Model

Model Using Inflation- Model Using NaturalAdjusted Incomes Log of Inflation-Adjusted Incomes

95% Confidence 95% ConfidenceIntervals Intervals

P Lower Upper P Lower UpperCoefficient CoefficientValue Bound Bound Value Bound Bound

Work effortAdditional 100 patient visits $854 <0.001 $773 $935 0.54% <.001 0.50% 0.59%Additional 100 work hours $406 0.001 $164 $649 0.34% <.001 0.20% 0.47%

Physician characteristicsPracticing less than 5 years −$25,155 <0.001 −$33,159 −$17,150 −20% <.001 −25% −16%Practicing 5-9.9 years Referent ReferentPracticing 10-14.9 years $12,338 <0.001 $6,686 $17,990 6% <.001 3% 9%Practicing 15-19.9 years $12,801 <0.001 $6,992 $18,610 6% .001 2% 9%Practicing 20-24.9 years $23,336 <0.001 $16,457 $30,215 12% <.001 8% 16%Practicing 25-29.9 years $2,555 0.512 −$5,074 $10,183 2% .468 −3% 6%Practicing 30-34.9 years $3,037 0.529 −$6,424 $12,497 −1% .822 −6% 5%Practicing 35-39.9 years −$23,845 <0.001 −$33,520 −$14,170 −19% <.001 −24% −13%Practicing 40 or more years −$16,274 0.042 −$31,975 −$574 −12% .006 −21% −3%Nonowner of practice −$16,554 <0.001 −$20,589 −$12,520 −5% <.001 −7% −2%Board certified $23,379 <0.001 $18,796 $27,963 17% <.001 15% 20%Female −$30,561 <0.001 −$35,385 −$25,738 −20% <.001 −23% −17%White −$701 0.795 −$5,986 $4,585 −0.3% .824 −3% 3%International medical school $4,821 0.071 −$416 $10,058 3% .076 −0.3% 6%

graduateSpecialty

General practice physician −$39,921 <0.001 −$45,613 −$34,229 −23% <.001 −26% −20%Family medicine physician −$27,383 <0.001 −$33,016 −$21,751 −14% <.001 −17% −11%General internist Referent ReferentPediatrician −$21,308 <0.001 −$29,387 −$13,229 −10% <.001 −15% −6%

Practice characteristicsLocated in Northeast Census region −$10,226 <0.001 −$15,314 −$5,138 −5% <.001 −8% −2%Located in North Central Census −$4,199 0.093 −$9,103 $705 −0.2% .876 −3% 3%

regionLocated in Southern Census region Referent ReferentLocated in Western Census region −$4,475 0.077 −$9,441 $491 −1.5% .312 −4.3% 1.4%Additional% of Medicaid enrollees −$116 0.027 −$218 −$13 −0.08% .004 −0.14% −0.03%

in practiceMedicare provider −$267 0.947 −$8,136 $7,603 0.4% .851 −4% 5%

Survey year −$856 0.028 −$1,621 −$92 −0.3% .182 −0.7% 0.1%Practices in a rural setting −$9,585 <0.001 −$14,569 −$4,602 −5.0% .001 −7.8% −2.1%

The model accounted for 22.7% of the variance in inflation-adjusted physicians’ incomes and 26.2% of the variance in naturallog-transformed inflation-adjusted physicians’ incomes.

physicians may have different practice arrangements,subspecialty practice types, and procedure rates. Morestudy is required to fully understand the differencesthat we found.

Despite these limitations, the results of this studysuggest that primary care physicians who work in ruralsettings work longer hours, make less money, and aremore likely to serve patients covered by Medicaid(which provides relatively low reimbursement). In

concert with the reality that tuition costs have asubstantial influence on the returns on educationalinvestment,38,39 the lack of evidence that ruralphysicians have lower debt, and the acknowledgementthat primary care physicians have lower returns ontheir educational investment than procedure-basedphysicians,38,44,45 our findings suggest that evenmodestly lower expected incomes will markedlydecrease measures of return on educational investment



Figure 2. Adjusted Incomes for Rural andSuburban Physicians as a Proportionof That for Urban Physicians, bySpecialty and in Aggregate.

80% 90% 100% 110%

All combined

Pediatricians

General internists

Family physicians

General physicians

Adjusted income as a percentage of that for urban physicians

for physicians who practice in rural settings. However,even if real incomes of rural physicians—adjusted forpurchasing power parity—are higher than those oftheir urban counterparts,18 the fact is that practices inrural areas continue to have difficulty recruitingphysicians. This suggests that to ensure an adequatesupply of primary care physicians practicing in ruralsettings, rural practice needs to be made moreappealing. Our findings suggest that this might beaccomplished by increasing incomes, reducing workhours, or some combination of the two.

References1. Rosenblatt RA, Audrilla CH, Curtin T, Hart LG. Shortages of

medical personnel at community health centers: implications forplanned expansion. JAMA. 2006;295(9):1042-1049.

2. A National Agenda for Rural Minority Health Series:Recruitment and training of health professionals. Kansas City,Mo: National Rural Health Association; 2001. Available at:http://www.nrharural.org/advocacy/sub/issuepapers/ipaper18.html. Accessed May 26, 2006.

3. Pathman DE, Konrad TR, King TS, Spaulding C, Taylor DH.Medical training debt and service commitments: the ruralconsequences. J Rural Health. 2000;16(3):264-272.

4. 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.

5. Recruitment and Retention of a Quality Health Workforce inRural Areas: Defining the Issues and the Principles ofRecruitment and Retention—May 2005. Kansas City, Mo:National Rural Health Association; 2005. Available at:

http://www.nrharural.org/advocacy/sub/issuepapers/RcrtmntRtntion_1.pdf. Accessed May 26, 2006.

6. Anonymous, Rural primary care. American College ofPhysicians. Ann Intern Med. 1995;122(5):380-390.

7. Chan L, Hart LG, Ricketts T, Beaver S. An analysis of Medicare’sIncentive Payment program for physicians in health professionalshortage areas. J Rural Health. 2004;20(2):109-117.

8. Rosenblatt RA, Saunders G, Shreffler J, Pirani MJ, Larson EH,Hart LG. Beyond retention: National Health Service Corpsparticipation and subsequent practice locations of a cohort ofrural family physicians. J Am Board Fam Pract 1996;9(1):23-30.

9. Rabinowitz HK, Diamond JJ, Markham FW, Hazelwood CE. Aprogram to increase the number of family physicians in rural andunderserved areas: impact after 22 years. JAMA.1999;281(3):255-260.

10. Pathman DE, Steiner BD, Jones BD, Konrad TR. Preparing andretaining rural physicians through medical education. Acad Med.1999;74(7):810-820.

11. Rabinowitz H, Diamond J, Markham F, Paynter N. Critical factorsfor designing programs to increase the supply and retention ofrural primary care physicians. JAMA. 2001;286:1041-1048.

12. Health Care Workforce Distribution and Shortage Issues in RuralAmerica. Kansas City, Mo: National Rural Health Association;2003. Available at: http://www.nrharural.org/advocacy/sub/policybriefs/WorkforceBrief.pdf. Accessed May 26, 2006.

13. Rabinowitz HK, Paynter NP. MSJAMA. The rural vs. urbanpractice decision. JAMA. 2002;287(1):113.

14. Weil TP. Attracting qualified physicians to underserved areas.Part 2. Pay physicians more to practice in underserved areas.Physician Exec. 1999;25(1):53-63.

15. Woolf MA, Uchill VL, Jacoby I. Demographic factors associatedwith physician staffing in rural areas: the experience of theNational Health Service Corps. Med Care. 1981:444-451.

16. Stensland J, Brasure M, Moscovice I. Why do rural primary carephysicians sell their practices? J Rural Health. 2002;18(1):93-108.

17. Wright GE. The economics of rural practice, Pages 275-288 In:Geyman JP, Norris TE, Hart LG, eds. Textbook of Rural Medicine.New York: McGraw-Hill Companies; 2001.

18. Reschovsky JD, Staiti AB. Physician incomes in rural and urbanAmerica. Issue Brief Cent Stud Health Syst Change. 2005;92:1-4.

19. Community Tracking Study Physician Survey. Available at:http://hschange.org/index.cgi?data=04. Accessed May 26, 2006.

20. Weeks WB, Wallace AE. Race and gender differences in generalinternists’ annual incomes. J Gen Intern Med. 2006;21(11):1167-1171.

21. Weeks WB, Wallace AE. The influence of race and gender onfamily physicians’ annual incomes. J Am Board Fam Pract2006;19(6):548-556.

22. Hart LG, Salsberg E, Phillips DM, Lishner DM. Rural health careproviders in the United States. J Rural Health. 2002;18:S211-S232.

23. Wadsworth ME, Montgomery SM, Bartley MJ. The persistingeffect of unemployment on health and social well-being in menearly in working life. Soc Sci Med. 1999;48(10):1491-1499.

24. Beck RW, Jijon CR, Edwards JB. The relationships among gender,perceived financial barriers to care, and health status in a ruralpopulation. J Rural Health. 1996;12(3):188-189.

25. Frenzen PD. Health insurance coverage in U.S. urban and ruralareas. J Rural Health. 1993;9(3):204-214.

26. Larson SL, Hill SC. Rural–urban differences in employment-related health insurance. J Rural Health. 2005;21(1):21-30.

27. Weeks WB, Mahar PJ, Wright SM. Utilization of VA and Medicareservices by Medicare-eligible veterans in Northern New England:the impact of additional access points in a rural setting. J HealthCare Manage. 2005;50(2):95-106.



28. West AN, Weeks WB. Physical and mental health and access tocare among nonmetropolitan VA patients younger than 65 years.J Rural Health. 2006;22(1):9-16.

29. Eberhardt M, Ingram D, Makuc D. Urban and Rural HealthChartbook. Health, United States, 2001. Hyattsville, Md: NationalCenter fo Health Statistics; 2001.

30. Kukulka G, Christianson J, Moscovice I, DeVries R.Community-oriented primary care. Implementation of a nationalrural demonstration. Arch Fam Med. 1994;3:495-501.

31. Mick SS, Lee SY. Are there need-based geographical differencesbetween international medical graduates and U.S. medicalgraduates in rural U.S. counties? J Rural Health. 1999;15(1):26-43.

32. Sempowski IP. Effectiveness of financial incentives in exchangefor rural and underserviced area return-of-service commitments:systematic review of the literature. Can J Rural Med.2004;9(2):82-88.

33. Wassenaar J, Thran S. Physician Socioeconomic Statistics,2000–2002. Chicago, Ill: American Medical Association Center forHealth Policy Research; 2001.

34. Baker L. Differences in earnings between male and femalephysicians. N Engl J Med 1996;334(15):960-964.

35. Wallace AE, Weeks WB. Differences in income between male andfemale primary care physicians. J Am Med Womens Assoc.2002;57(4):180-184.

36. Weeks WB, Wallace AE. Time and money: a retrospectiveevaluation of the inputs, outputs, efficiency, and incomes ofphysicians. Arch Intern Med. 2003; 163(8):944-948.

37. Polachek SW, Siebert WS. The Economics of Earnings. Cambridge:Cambridge University Press; 1993.

38. Weeks WB, Wallace AE, Wallace MM, Welch HG. A comparisonof educational costs and incomes of physicians and otherprofessionals. N Engl J Med. 1994;330:1280-1286.

39. Weeks WB, Wallace AE. The more things change: revisiting acomparison of educational costs and incomes of physicians andother professionals. Acad Med. 2002;77(4):312-319.

40. Ness R, Ukoli F, Hunt S, et al. Salary equity among male andfemale internists in Pennsylvania. Ann Intern Med.2000;133:104-110.

41. Becker ER, Culler S, Ohsfeldt R. Impact of board certification onphysician practice characteristics. J Med Educ. 1985;60(1):9-15.

42. Metropolitan and Micropolitan Statistical Areas. U.S. CensusBureau. Available at: http://www.census.gov/population/www/estimates/metroarea.html. Accessed January 15, 2008.

43. Consumer price index. Available at: http://www.bls.gov/cpi/.Accessed May 22, 2006.

44. Weeks WB, Wallace AE. Financial returns on specialty trainingfor surgeons. Surgery. 2002;132(5):795-802.

45. Weeks WB, Wallace AE. Long-term financial implications ofspecialty training for physicians. Am J Med. 2002;113(5):393-399.

46. Weeks WB, Wallace AE. Return on educational investment ingeriatrics training. J Am Geriatr Soc. 2004;52(11):1940-1945.


rural–urban differences in primary care physicians' practice patterns, characteristics, and...

Documents