Is Patient HMO Insurance or Physician HMO Participation Related to Racial Disparities in Primary Care?

The American Journal of Managed Care, June 2005, Volume 11, Issue 6

Objective: To examine the relationship between racial disparitiesin common primary care procedures and patient HMO membershipand physician level of HMO participation.

Design: Cross-sectional analysis.

Methods: Data were obtained from a nationally representativesample of primary care office visits documented in the NationalAmbulatory Medical Care Survey for 1985, 1989-1992, and 1997-2000. Patient HMO membership was assessed based on reports byprimary care physicians (defined as family physicians/general practitioners,internists, or obstetrician-gynecologists). Physician HMOparticipation was assessed based on the proportion of the physician'spatients who were in an HMO. Patient characteristics (age,sex, race, insurance, diagnoses) and office procedures or interventionswere determined by examining the physician report. Patientswere adults aged 19 years or older.

Results: In adjusted analyses, African Americans, comparedwith whites, had lower odds of receiving a Pap test (adjusted oddsratio [AOR] = 0.76; 95% confidence interval [CI] = 0.65, .90), arectal exam (AOR = 0.67; 95% CI = 0.54, 0.84), smoking cessationadvice (AOR = 0.72; 95% CI = 0.58, 0.91), and mental healthadvice (AOR = 0.46; 95% CI = 0.29, 0.72), but had higher odds ofreceiving advice on diet and weight, and a follow-up appointment.Notably, there were no significant interactions between eitherpatient HMO membership or physician level of HMO participation,patient race, and receipt of primary care services.

Conclusion: Neither patient HMO membership nor physicianlevel of HMO participation is substantially associated with racialdisparities in primary care.

(Am J Manag Care. 2005;11:397-402)

The impact of HMOs on racial and ethnic disparitiesin healthcare remains uncertain. On onehand, HMOs focus on the quality of healthcarefor a defined population and thus might be expected toreduce disparities in healthcare. On the other hand,HMO cost-containment strategies and administrativebarriers to care might disproportionately affect racialand ethnic minorities. Empirical data are mixed; somestudies show attenuation of disparities in some areas,1-5while others show worse or no effects.6-10 However,these studies all have examined effects of HMO participationat the individual patient level, with limitedaccounting for the managed care arrangements affectingthose patients. Significantly, none of the studies examinedwhether the level of HMO participation by individualphysicians affects healthcare disparities.

To some extent, HMOs may affect physicians' practicepatterns. Any HMO-induced changes in practicestyle may generalize to patients not in HMOs.11,12 Theimpact on physician practice style may depend on theproportion of patients seen who are enrolled in HMOs.

In this study, we used national data for primary careoffice visits to determine whether patient HMO membershipor physician level of HMO participation affectedvisit-level racial disparities. (Most reductions in disparitiesshould be observed at the visit level.) These dataprimarily reflect procedures recommended or conductedby the physician at the time of the visit.Consequently, racial variation in these procedures (notexplained by differences in patient case mix) primarilyreflect differences in physician decision making, asopposed to patient adherence. Given the growth andevolution of HMOs over the last 20 years,13 we examinedwhether relationships between the level of physicianHMO participation and racial disparities in carechanged over time.



The data for this study were derived from theNational Ambulatory Medical Care Surveys (NAMCSs)for 1985-2001. The survey was not conducted from1986 to 1988, and different versions were used in 1993and 1994. For this survey, a nationally representativesample of office-based physicians complete a short surveyon approximately 20 patient encounters. Immediatelyafter the encounter, the physicians complete apatient record for every encounter, regardless of billingsource, which includes age, race, insurance status, up to3 diagnoses assigned, medications prescribed, whetheror not the patient had been seen before for the presentingproblem or other problems, services provided, advicegiven, disposition, and visit duration. Not all questionsare asked each year. Information also is collected onspecialty, whether the physician's practice is located ina standard metropolitan statistical area (ie, whether thepractice is urban or rural), and geographic region.Complete details about the surveys are available at:

The subset of encounters selected for this analysiswere visits by African American or white patients 19years of age and older to primary care physicians(internists, family physicians/general practitioners,and obstetrician-gynecologists). Visits that were theresult of a referral were excluded (given our focus onprimary care). Elements were selected for analysisbased on a review of the current literature on racialdisparities and those activities documented frequentlyenough in the surveys to generate stable estimates.These included advice giving (exercise, diet, cholesterol,smoking cessation, breast health, and mentalhealth), prevention procedures (Pap test, mammogramin women over age 40 years, breast examination,pelvic examination, rectal examination, bloodpressure check, vision examination, cholesterol test,and prostate-specific antigen testing in men over age50 years), visit duration, and follow-up plans (specificfollow-up and referral to another physician).

Independent Variables

The primary independent variables were patient HMOmembership and physician level of HMO participation.Patient HMO membership was defined based on check-offby the physician that the patient was in an HMO. In 1985,an HMO was defined in NAMCS as "charges includedunder a health maintenance organization plan or otherpre-payment plan. Include IPAs, PPOs, etc." By 1992,HMOs were defined in NAMCS as "health maintenanceorganizations (HMOs), independent practice organizations(IPAs), and all other prepaid health care plans." By1997, the definition was "HMO is defined as a health caredelivery system that offers comprehensive health servicesprovided by an established panel or network of providersto a voluntarily enrolled population for a prepaid fixedfee and whose members are required to utilize serviceswithin the panel of contracted providers."

Physician level of HMO participation wasdefined based on the proportion of patientsseen by the physician during the samplingperiod who were in an HMO. There were 2 categories:<50%, and &#8805;50%. We also examinedthis variable as a continuous measure.


The following measures were used: age(years), sex, race (African American or white),insurance (private, Medicaid, Medicare), rurality(living in a metropolitan statistical area ornot), region of the country (Northeast,Midwest, South, or West), year of visit, casemix (based on the number of medications prescribedand the ambulatory care groups[ACGs] described below), physician specialty(family physician/general practitioner, internist,or obstetrician-gynecologist), and percentageof Medicaid patients seen.

Case-Mix Adjustment

International Classification of


Case-mix adjustment was based on the ACGsystem14 using the ambulatory diagnosticgroups (ADGs) of the ACG system. The ADGscomprise 32 diagnostic and preventive clustersto which each diagnostic and preventive code can be uniquelyassigned. Each patient is assigned a series of dummy(0,1) values for each ADG, depending on the up to 3diagnoses assigned by the physician for the visit. Use ofthe ADGs for case-mix adjustment has previously beenfound to be valid in this dataset.15

Statistical Analyses

The National Ambulatory Medical Care Surveys use acomplex survey design, involving the clustering of visitswithin each physician's practice and the use of weightsdesigned to yield population estimates of encounters.The data were analyzed with STATA (version 8.2,StataCorp, College Station, Tex) to adjust for the samplingdesign and weights. The weights on the public-usedata adjust each physician-patient encounter accordingto its sampling probability and the probability of physiciannonresponse to yield unbiased national estimatesof annual total visits.

The relationship between patient race (AfricanAmerican vs white) and each visit characteristic wasexamined using logistic regression models, except forvisit duration, which was examined using linear regression.We adjusted for potential confounding by physicianspecialty, whether or not the patient was seenbefore, patient sociodemographics (age, sex, insurancestatus, rurality, region, year of visit), and case mix(based on the number of medications prescribed andthe ADGs coded).

The key term of interest was the interaction betweenrace and patient membership in an HMO or physicianlevel of HMO participation. This term was included inall analyses. The proportion of HMO patients seen bythe physician was examined both as a continuous variableand a dichotomized one (<50%, and &#8805;50%). Toreduce the risk of confounding by practice setting, wealso included a term for the proportion of patients seenwith Medicaid insurance. We examined 2 sets of stratified subsamples of the data: first, we repeated the analysesstratified by patient HMO status, and we againrepeated the analyses stratified by physician level ofHMO participation. Finally, to explore temporal trends,we conducted analyses including interaction termsbetween race and year, and betweenrace, proportion of HMO patients,and year.


The sample included 62 348 visits,10.7% by African Americans, to2112 primary care providers (Table1). Patients with HMO insuranceaccounted for 23% of the visits;physicians with practices consistingof >50% HMO patients saw 16.4% ofall patients. Compared with whitepatients, African Americans wereyounger; more likely to reside inmetropolitan statistical areas, to befemale, have Medicaid coverage, andreside in the South; and were prescribedmore medications. AfricanAmericans were less likely, however,to have Medicare or to have beenseen previously.

Compared with whites, AfricanAmericans were less likely to havePap smear screening, mammographyordered, rectal or breast exams,mental health advice, breast self-examadvice, and tobacco counseling(Table 2). African Americans,however, were more likely toreceive blood pressure testing anddiet/weight counseling, and be givena scheduled follow-up appointment(Table 2). In adjusted analyses,African Americans had lower oddsof receiving a Pap test, a rectalexam, mental health advice, andsmoking cessation advice, buthad higher odds of receivingdiet/weight counseling and a follow-up appointment.

We found no significant interactionsbetween race and eitherpatient HMO membership or physicianlevel of HMO participation.There also was no significant interactionwith time, Medicaid insurance, or percentage ofMedicaid patients in the physician's practice. The stratifiedresults by patient HMO status are shown in Table3 and by physician HMO participation in Table 4. It canbe seen that there is little evidence that any trends inpatient HMO membership or physicianlevel of HMO participation areassociated with systematic effectson racial disparities.


Previous findings have beenmixed regarding the impact ofHMOs on racial and ethnic disparitiesin healthcare.1-3,6-9 However, toour knowledge, this is the firstanalysis of disparities at the visitlevel that examined the relationshipbetween physician level of HMOparticipation and disparities in primarycare procedures. Prior studiesexamined the effect of patient HMOmembership, but no previous studiesexamined the impact of greaterphysician HMO participation.

Given the limitations of NAMCS,our findings may underestimateracial disparities in primary careand thus our ability to detect significantinteractions between race andphysician level of HMO participation.Furthermore, despite aggregation ofthe NAMCS data across years, ourpower to detect modest interactionswas limited. We cannot excludethe possibility of small effects.

We also cannot exclude the possibilitythat HMO membershipeither hinders or promotes access toprimary care. We conducted visit-levelanalyses, not patient-or population-level analyses, so we couldonly assess disparities that occurredduring the visit.

Self-reported data by physiciansmay lead to underestimation of racial disparity in theircare. A comparison of NAMCS data with direct observationof office visits showed high specificity (range of 90%to 99%) for office procedures and counseling, but moderateto low sensitivity (range of 0.12 to 0.84).16 Thereliability of physician coding of patient HMO status inNAMCS has not been formally assessed. However, thesedata should be readily available from patient insuranceinformation typically included on patient encounterforms. Any measurement error would likely bias resultstowards the null.

Other limitations include the use of a short data-collectionform; under-representation of AfricanAmericans due to restriction of the sample to office visitsas opposed to clinic visits; unmeasured confounders,particularly physician characteristics; absence of datafor selected years; slight changes in the definition of relevantvariables, including HMOs; and the potential forinadequate comorbidity adjustment.

In summary, our findings suggest that neither patientHMO membership nor physician level of HMO participationsignificantly affect disparities in receipt of primary care visits. Changes in HMO membership aloneare unlikely to affect disparities in receipt of primarycare for better or worse. Other, more focused interventionswill likely be required to address disparities.

From the Departments of Family Medicine and Community & Preventive Medicine,University of Rochester School of Medicine & Dentistry, Rochester, NY (KF); and theDepartment of Family and Community Medicine, Center for Health Services Research inPrimary Care, University of California, Davis (PF).

This study was supported by a grant (R01 HS10910-02) from the Agency for HealthcareResearch and Quality.

Address correspondence to: Kevin Fiscella, MD, MPH, 1381 South Ave, Rochester, NY14620. E-mail:

J Fam Pract.

1. Clancy CM, Franks P. Utilization of specialty and primary care: the impact ofHMO insurance and patient-related factors. 1997;45:500-508.

Admin Policy Mental Health.

2. Crawford K, Fisher WH, McDermeit M. Racial/ethnic disparities in admissionsto public and private psychiatric inpatient settings: the effect of managed care.1998;26:101-109.

Med Care.

3. Haas JS, Phillips KA, Sonneborn D, McCulloch CE, Liang SY. Effect of managedcare insurance on the use of preventive care for specific ethnic groups in theUnited States. 2002;40:743-751.

Health Serv Res.

4. Bindman AB, Chattopadhyay A, Osmond DH, Huen W, Bacchetti P. Theimpact of Medicaid managed care on hospitalizations for ambulatory care sensitiveconditions. 2005;40:19-38.

J Gen Intern Med.

5. DeLaet DE, Shea S, Carrasquillo O. Receipt of preventive services among privatelyinsured minorities in managed care versus fee-for-service insurance plans.2002;17:451-457.

Health Aff.

6. Davis K, Collins KS, Schoen C, Morris C. Choice matters: enrollees' views oftheir health plans. 1995;14(2):99-112.


7. Schneider EC, Cleary PD, Zaslavsky AM, Epstein AM. Racial disparity ininfluenza vaccination: does managed care narrow the gap between AfricanAmericans and whites? 2001;286:1455-1460.

J Health Care Finance.

8. O'Connell L, Brown SL. Do nonprofit HMOs eliminate racial disparities in cardiaccare? 2003;30:84-94.

Arch Pediatr Adolesc Med.

9. Stevens GD, Shi L. Effect of managed care on children's relationships with theirprimary care physicians: differences by race. 2002;156:369-377.

Med Care.

10. Stepanikova I, Cook KS. Insurance policies and perceived quality of primarycare among privately insured patients: do features of managed care widen theracial, ethnic, and language-based gaps? 2004;42:966-974.d

J Health Econ.

11. Glied S, Zivin JG. How do doctors behave when some (but not all) of theirpatients are in managed care? 2002;21:337-353.

Health Serv Res.

12. Hadley J, Mitchell JM, Sulmasy DP, Bloche MG. Perceived financial incentives,HMO market penetration, and physicians' practice styles and satisfaction.1999;34(1 pt 2):307-321.

N Engl J Med.

13. Dudley RA, Luft HS. Managed care in transition. 2001;344:1087-1092.



14. Weiner JP, Starfield BH, Steinwachs DM, Mumford LM. Development andapplication of a population-oriented measure of ambulatory care case-mix. 1991;29:452-472.

J Fam Pract.

15. Blumenthal D, Causino N, Chang YC, et al. The duration of ambulatory visitsto physicians. 1999;48:264-271.

Med Care.

16. Gilchrist VJ, Stange KC, Flocke SA, McCord G, Bourguet CC. A comparison ofthe National Ambulatory Medical Care Survey (NAMCS) measurement approachwith direct observation of outpatient visits. 2004;42:276-280.