Can this be done 8 pm 4/22

Modifiable Determinants of Healthcare Utilization within the African-American Population George Rust, MD, MPH; George E. Fryer Jr, MSW, PhD; Robert L. Phillips Jr, MD, MSPH; Elvan Daniels, MD; Harry Strothers, MD, MMM; and David Satcher, MD, PhD Atlanta, Georgia and Washington, DC

Background: Signikant health dipities directl affect the African-A rcan ppulaton. Mot pvious studies of dis- parities in access to and uilizan of heolthcore have focused on black-he differences rther thon fodusing on "withi-roup" analys of AMiCan Amercans.

Obctive To tease out the differential effects of modifiable risk factors (such as health isurance, usual source of care, and poverty) frbm persol Characteristics (age, gender, rural esidence) on healthcare utiio wthin the African- mercon population.

Methods: Secondory datd analyis of 3,462 records frm AfricanAmericon respondents tohef1999 Medical Expen- diu Panel Survey (MEPS) Household File, a nationll rep- resentoaive survey of the civilian, noninstitutionatized U.S. population in 1999. Relt We found significant variation in the umber of ffice visits, outpotient clinic visits, hspitl sdhdres, da hospi- tlized, and fills of prscrbed medication omong 3,462 African-Amercan rsondents Who repeet a Us. popula- fion of 36,538,639 persons. Persnad nonmodiabe charac- teriics such as age and gender we signfcntl related to healthcare utlion. Poverty and rurOl residene were also significantly correlated, bUt the stroest modifiabl dic- tors of healthcare utiztion for Afrcnmian persons in 1999 were wheth ornot individuals had health insUrance and/or a usUol source of care. Emegecy deportm t visits were the only f of care thot showed remarabl lit vari- ation based on these odifi* ridsk facts.

Conclsins: The three modifial factors of poverty, uninsur- ance, and having a pria care edical home have a dra- mai Oef on pe Core for Afican-Ae o*ts and could be indepedenty targeteO rb .

Ke words: healthcr *utian U acces U r Afcan Aekas U da

© 2004. From the Morehouse School of Medicine, Atlanta, GA (Rust, deputy director, National Center for Pnmary Care and professor of Family Medicine; Daniels, research assistant professor and associate director for Community Onented Pnmary Care, National Center for Pnmary Care; Strothers, associate professor and associate chair for education, Department of Family Medi- cine; and Satcher, director, National Center for Primary Care and professor of Community Health/Preventive Medicine) and Robert Graham Center for Policy Studies in Family Medicine & Pnmary Care, Washington, DC (Fryer, Jr; Phillips, Jr). Send correspondence and repnnt requests for J NatI Med Assoc. 2004; 96:1169-1177 to: George Rust, Deputy Director, National Center for Pri- mary Care and Professor of Family Medicine, Morehouse School of Medi- cine, 720 Westview Drive, Atlanta, GA 30310; phone: (404) 756-1236; fox: (404) 756-5767; e-mail: [email protected]

INTRODUCTION Racial and ethnic disparities in healthcare have

been documented in numerous studies,'-7 which have been summarized in national reports, such as the Institute of Medicine's Unequal Treatment: Con- fronting Racial and Ethnic Disparities in Health Care8 and various Kaiser Family Foundation reports on Race, Ethnicity, and Medical Care.9'10 Some of the most rigorously documented studies have shown racial differences in the use of cardiovascular proce- dures for patients hospitalized with coronary heart disease," '2 but clinical and demographic factors still do not adequately explain why African Americans are significantly less likely to undergo revascular- ization procedures."'3

Studies exploring disparities in cardiac proce- dures have traditionally been framed as studies of black-white differences, including racism as a cause of such disparities.'4 Far fewer studies have sought to assess the modifiable factors that drive disparities within the subgroup of African Americans. In other words, why do some African-American patients receive optimal care, while many others do not? Understanding "within-group" differences may bet- ter identify specific factors around which interven- tions could be designed to reduce health disparities in the African-American population. The Anderson and Aday model provides a conceptual framework for analyzing modifiable determinants of healthcare

JOURNAL OF THE NATIONAL MEDICAL ASSOCIATION VOL. 96, NO. 9, SEPTEMBER 2004 1169

DETERMINANTS OF HEALTHCARE UTILIZATION

utilization,'5"'6 and this framework can be further informed by Williams' sociologic perspective on personal and contextual factors that uniquely influ- ence health status among African Americans.'7"'8

Access to care through health insurance, espe- cially Medicare, has been shown to have a signifi- cant impact on improving access to care and health outcomes in high disparity groups, such as the African-American population. Zuvekas et al. con- cluded from their own analyses of MEPS data that while health insurance did not by itself explain the persistent racial & ethnic disparities, it was a signifi- cant contributing factor. 9 However, the disparities literature has tended to lump modifiable risk factors with nonmodifiable demographic characteristics of the individual in explaining variations in the use of health services. For example, studies identify being "poor, black, and uninsured" as common risk factors for lower overall healthcare utilization and costs, without differentiating the impact of modifiable risk factors (uninsurance, medical home, etc.) on mem- bers of each racial-ethnic subgroup.20 In this study, therefore, we have sought to understand the modifi- able determinants of healthcare utilization within the African-American population in order to focus on intervention points that could imp rove health outcomes for this high-disparity population. We also assessed the impact of clustered factors among the healthcare disadvantaged (poor, uninsured, with no medical home) vs. the healthcare advantaged (non- poor, insured, with a medical home).

METHODS

Data Source Data for the U.S. civilian, noninstitutionalized

population were taken from the 1999 Medical Expenditure Panel Survey (MEPS) sponsored by the Agency for Healthcare Research and Quality (AHRQ).2' MEPS is a nationally representative sur- vey of a sample of households drawn from previous National Health Interview Survey (NHIS) partici- pants. Data in 1999 were collected for 23,565 per- sons by computer-assisted personal interviews.

In this study, household component records describing the sociodemographics, health insurance coverage, and healthcare utilization of survey partic- ipants were analyzed. MEPS records are weighted for the calculation of national estimates, usually with relatively small standard errors. Race is record- ed as one of five categories in the MEPS records. African Americans were substantially oversampled in the 1999 survey. We analyzed data from all 3,462 alI-civilian, noninstitutionaIized African-American respondents to the survey, representing a U.S. African-American population of 36,538,639.

Study Variables Dependent variables related to our research ques-

tion consisted of measures of utilization in the vari- ous healthcare settings and for medications, includ- ing the following five variables:

1) Visits to physician's office (or other health pro- fessional provider's office).

2) Visits to hospital outpatient departments (hospi- tal-based clinics).

3) Inpatient hospitalizations-discharges and nights in hospital.

4) Visits to emergency departments (EDs).

5) Prescribed medications-fills.

Self-reported health status and self-reported mental health status (two variables) were also assessed as dependent variables but are not meas- ures ofhealthcare utilization.

Predictor variables used in the analysis were:

1) Age-four groups; children (under 18 years of age), adults 18-44,45-64, and the elderly (over 64 years of age).

2) Sex-female or male.

3) Residence-resident of a Metropolitan Statistical Area (MSA) county or a resident of a non-MSA county. Small samples ofAfrican Americans in some regions precluded using census region in the model.

4) Household income-reported family income divided by the federal poverty level based on family size and composition, with the resulting percentages grouped into five categories (<100% of federal poverty level, 100-124% of federal poverty level, 125-199% of federal poverty level, 200-399% of federal poverty level, and >400% of federal poverty level).

5) Health insurance-three groups: private insur- ance; public insurance only (Medicare, Medicaid and dual-eligibles); and no insurance, without regard for adequacy of coverage.

6) Usual source of care-response to the question: "Is there a particular doctor's office, clinic, health center, or other place that you go ifyou are sick or need advice about your health?

1170 JOURNAL OF THE NATIONAL MEDICAL ASSOCIATION VOL. 96, NO. 9, SEPTEMBER 2004

DETERMINANTS OF HEALTHCARE UTILIZATION

Analytical Strategy Descriptive analyses were first performed esti-

mating services received in each care setting and uti- lization of prescribed medicines. Comparisons on these measures were made across categories of the predictor variables enumerated above for African Americans (Table 1). We then evaluated potential predictor variables for inclusion in multivariate analyses. Bivariate tests indicated strong associa- tions between each of the predictor variables and measures of healthcare utilization in at least one set- ting. We also tested the associations with and with-

out children and the elderly in the analysis, and found similar patterns. Therefore, all of the listed variables were used in separate logistic and linear regressions to model healthcare use by the setting in which the service was received. Logistic regression was performed to model whether or not an individ- ual had received care at least once for each of the settings and for receipt of at least one prescribed medication fill. Adjusted odds ratios for whether or not a service was received were calculated for each predictor variable (Table 2), because of the potential for skewing of data by the large numbers of persons

Table 1. Mean Annual Healthcare Uftlization for African Americans by Demographic Characteristics and Modifiable Factors (Insurance Status and Usual Source of Care), 1999

Category/ Number Percent Office Hosp Outpt. ED Hospital Hospital Rx Fills Item Visits Clinic Visits Visits Discharges Bed-Days + Refills

Age Group Under 18 11,937,112 33.1 1.58 0.08 0.15 0.03 0.10 1.38 18-44 14,773,451 40.9 2.19 0.31 0.19 0.09 0.28 3.32 45-64 6,522,838 18.1 3.82 0.59 0.20 0.12 0.72 12.07 65+ 2,872,592 8.0 6.77 1.07 0.22 0.23 1.30 21.81 P value** <0.01 <0.01 0.18 <0.01 <0.01 <0.01

Sex Male 16,987,105 46.5 2.24 0.27 0.16 0.06 0.40 4.41 Female 19,551,534 53.5 2.96 0.41 0.20 0.12 0.48 6.81 P value** <0.01 0.13 0.03 <0.01 0.43 <0.01

MSA vs. non-MSA (Urban-rural) MSA 31,833,017 88.2 2.60 0.35 0.17 0.08 0.31 5.41 Non-MSA 4,272,976 11.8 2.96 0.31 0.25 0.15 0.93 8.13 P value* 0.25 0.78 0.02 <0.01 0.01 0.01

Family income as % fed. poverty level Poor 8,799,737 24.1 2.78 0.43 0.24 0.13 0.67 7.19 Near poor 2,328,651 6.4 1.78 0.22 0.17 0.08 0.42 5.40 Low Income 6,393,948 17.5 2.37 0.34 0.19 0.08 0.38 5.02 Middle Income 10,643,391 29.1 2.88 0.24 0.17 0.09 0.50 5.57 High Income 8,372,911 22.9 2.60 0.42 0.13 0.06 0.20 4.87 P value** 0.02 0.28 0.03 0.09 <0.01 0.10

Health Insurance Coverage Any private 21,240,269 58.1 2.61 0.33 0.15 0.07 0.31 4.93 Public only 9,999,210 27.4 3.44 0.51 0.27 0.16 0.88 9.37 Uninsured 5,299,199 14.5 1.16 0.08 0.15 0.04 0.17 1.83 P value** <0.01 <0.01 <0.01 <0.01 <0.01 <0.01

Has Usual Source of Care? Yes 28,389,137 20.0 3.08 0.41 0.19 0.10 0.44 6.89 No 7,089,761 80.0 1.08 0.13 0.15 0.04 0.13 1.55 P value** <0.01 <0.01 0.14 <0.01 <0.01 <0.01

Total African-American Population in MEPS* 36,538,639 100.0 2.63 0.34 0.18 0.09 0.45 5.69

* This represents the number of individuals in the U.S. African-American population, a number generated by applying statistical weights to raw survey data. Due to missing data on some items, totals for some categories may be less than 36,538,639; ** derived from ANOVA F values

JOURNAL OF THE NATIONAL MEDICAL ASSOCIATION VOL. 96, NO. 9, SEPTEMBER 2004 1171

DETERMINANTS OF HEALTHCARE UTILIZATION

who had zero events (zero hospital admissions or emergency visits, for example). Linear regression was done to model the actual quantity of services and medication fills received. Its beta coefficients are straightforward measures of the strength of the effect ofpredictor variables on healthcare utilization (Table 3). Analyses were done with SUDAAN to adjust variance estimates due to MEPS survey design complexity, particularly the substantial over- sampling of certain population groups.

Finally, we also assessed the clustered effect of certain predictor variables, identifying individuals at the extremes of being advantaged or disadvantaged with regard to healthcare (Table 4). Members of the healthcare-disadvantaged group have family income below the poverty level, no health insurance, and no usual source of care. The healthcare advantaged group has family income above 400% of poverty, health insurance, and a usual source of care.

RESULTS There was valid data for 3,462 African-American

respondents to the survey, representing a U.S. African-American population of 36,538,639 civil- ian, noninstitutionalized persons. Respondents had a median age of 30.0 years. Differences in self-report- ed health status and mental health status were not striking, but showed a predictable decline in health status with age, and worse health status associated with rural residence and with being publicly insured (Medicare or Medicaid). African-American individ- uals living in rural areas were also a bit more likely to be older (median age 31.0 years for non-MSA and 29.0 years for MSA). We also found that the "near poor", who may not qualify for Medicaid nor be able to afford private insurance, had worse health status than either the poor or any other income group.

The pattern of healthcare utilization for African- American persons by demographic and other char- acteristics in 1999 is depicted in Table 1. Variations are seen by age group and gender. Elderly and

Table 2. Adjusted Odds Ratios for Predictors of Healthcare Uflizafton (yes/no for each item) among African Americans, 1999 (with 95% Confidence Intervals)

Category/Item Office Visits Hosp Outpt. Clinic Visits ED Visits Hospital Discharges Rx Fills + Refills

Age Group Under 18 0.20 (0.13, 0.30) 0.16 (0.10, 0.24) 0.81 (0.55, 1.18) 0.81 (0.09, 0.30) 0.09 (0.06, 0.14) 18-44 0.23 (0.16, 0.35) 0.49 (0.32, 0.73) 1.19 (0.81, 1.76) 0.80 (0.47, 1.38) 0.17 (0.11, 0.27) 45-64 0.47 (0.31, 0.71) 0.86 (0.56, 1.30) 1.15 (0.73, 1.80) 0.96 (0.54, 1.72) 0.35 (0.22, 0.54) 65+ 1.00 1.00 1.00 1.00 1.00

Sex Male 0.65 (0.54, 0.77) 0.76 (0.60, 0.96) 0.91 (0.73, 1.14) 0.50 (0.36, 0.70) 0.70 (0.59, 0.83) Female 1.00 1.00 1.00 1.00 1.00

MSA vs. non-MSA MSA 0.92 (0.67, 1.26) 0.91 (0.59, 1.42) 0.78 (0,55, 1.09) 0.63 (0.45, 0.87) 0.79 (0.61, 1.03) Non-MSA 1.00 1.00 1.00 1.00 1.00

Income/Poverty Status Poor 0.82 (0.51, 1.34) 1.39 (0.80, 2.41) 1.63 (1.02, 2.61) 1.38 (0.75, 2.54) 0.93 (0.62, 1.38) Near Poor 0.50 (0.27, 0.93) 0.76 (0.39, 1.47) 1.10 (0.58, 2.08) 0.98 (0.50, 1.94) 0.68 (0.41, 1.12) Low Income 0.76 (0.53, 1.11) 0.92 (0.56, 1.52) 1.39 (0.85, 2.28) 1.09 (0.61, 1.96) 0.90 (0.67, 1.22) Middle Income 0.83 (0.57, 1.21) 1.00 (0.66, 1.51) 1.42 (0.97, 2.08) 1.40 (0.86, 2.26) 0.76 (0.55, 1.05) High Income 1.00 1.00 1.00 1.00 1.00

Health Insurance Any Private 3.64 (2.79, 4.75) 1.82 (1.08, 3.09) 1.28 (0.82, 1.99) 1.36 (0.72, 2.58) 2.57 (1.97, 3.35) Public Only 4.13 (2.80, 6.09) 2.07 (1.27, 3.36) 1.99 (1.23, 3.23) 3.06 (1.61, 5.81) 2.96 (2.09, 4.20) Uninsured 1.00 1.00 1.00 1.00 1.00

Usual source of care? Yes 4.47 (3.37, 5.92) 3.99 (2.02, 7.87) 1.00 (0.69, 1.44) 1.91 (1.12, 3.27) 2.73 (2.01, 3.70) No 1.00 1.00 1.00 1.00 1.00

1172 JOURNAL OF THE NATIONAL MEDICAL ASSOCIATION VOL. 96, NO. 9, SEPTEMBER 2004

DETERMINANTS OF HEALTHCARE UTILIZATION

female African-American persons used more health- care in all settings and received more medications during the year than did younger and male African Americans. With the exception of hospital outpa- tient departments, African Americans living in rural (non-MSA) areas were more frequent users of care in all settings, with rural residents spending three times as many days in the hospital as MSA resi- dents. This may reflect the greater proportion of African-American elders remaining in rural areas. After controlling for age, gender, and insurance sta- tus, rural residence was not a significant factor relat- ed to healthcare utilization of any service except hospitalization. The poorest group used services most often in every setting other than providers' offices, but the near-poor, who are less likely to qualify for Medicaid, had lower utilization.

In the 1999 MEPS sample, 14.5% of African Americans were uninsured, and 20% had no usual source of care. Health insurance status is strongly related to healthcare utilization. African Americans who were uninsured trailed both insured categories for all settings of care other than the emergency room. Those with only public health insurance were much greater consumers of healthcare, reflecting higher utilization by the elderly (Medicare/Medic- aid) and by the poor (Medicaid). For example, those with public health insurance obtained prescription medication at a rate five times higher than did the uninsured. This pattern of higher utilization for the insured vs. the uninsured (in all settings except the emergency room) held true as well for the subgroup ofnonelderly adults (i.e., when children and the eld- erly were excluded).

African Americans with a usual source of care also used outpatient visits, inpatient hospital bed- days, and prescription drugs two-to-four times more often than those without such an arrangement. Hav- ing a usual source of care did not significantly affect ED visits, which were much more strongly associat- ed with insurance status.

Results of seven separate regression procedures for utilization in each setting and prescribed medica- tions are portrayed in Table 3. It illustrates the unique contribution of each predictor characteristic and healthcare arrangement. Controlling for the effects of all other predictors, several factors were significantly associated with each utilization meas- ure, and the overall model was significant in predict- ing utilization (p<0.01). Although poverty was a sig- nificant factor for some types of utilization, age, health insurance status, and having a usual source of care were the most important predictors of utiliza- tion for African-American persons in 1999.

Table 4 depicts a comparison of two groups with very different capacities to obtain needed health-

care. Members of the disadvantaged group have family income below the poverty level, no health insurance, and no usual source of care. In the U.S. population, this represents over a half million (527,474) "healthcare disadvantaged" persons, or 1.4% of our entire sample. The advantaged group has family income above 400% of poverty, health insurance, and a usual source of care. This is actual- ly a much larger group, representing more than six million (6,169,181) African Americans (17. 1% of the study population) in 1999. These data should be interpreted with caution due to the small number of MEPS participants in each subgroup.

Utilization in all settings except the emergency room was dramatically greater for African Ameri- cans with the necessary financial means and arrangements to obtain care. There was also an inter- esting relationship between hospitalizations and ED visits. The most-disadvantaged African-American patients had only one hospital admission for every 10 ED visits, but insured, middle- and upper-income African-American patients with a usual source of care had a hospitalization for every two ED visits. Focusing on the impact of insurance status among adult patients age 18-64, the uninsured had 3.6 ED visits per hospital admission, while the insured were hospitalized once for every 1.7 ED visits.

DISCUSSION These data suggest that within each age-gender

subgroup, whether or not an individual has health insurance and a source of usual care is the most important modifiable factors driving use of needed health services within the African-American popula- tion. Poverty is also a significant and modifiable factor. Combining the three factors of health insur- ance, adequate income, and having a primary care medical home was a powerful predictor of the use of specific health services that could improve health outcomes within the African-American population.22 For example, use of doctor's office visits was four times higher among African-American patients who were "health-advantaged" (i.e., individuals with all three factors present-insured, nonpoor persons with a defined medical home) compared with the "health disadvantaged".

These disparities in healthcare utilization are important, because ultimately they drive specific racial disparities in health status and health out- comes. For example, African-American patients are significantly less likely than whites to receive influenza vaccine,23 more likely to report barriers to obtaining mammography,24 more likely to be diag- nosed with late-stage cancer, and more likely to die from cancer.25 Regardless of race or ethnicity, women participating in routine mammography

JOURNAL OF THE NATIONAL MEDICAL ASSOCIATION VOL. 96, NO. 9, SEPTEMBER 2004 1173

DETERMINANTS OF HEALTHCARE UTILIZATION

screening had earlier-stage disease by five-to- 13 percentage points."26

In previous studies, the modifiable risk factors of having health insurance and having a usual source of care strongly and independently predicted use of essential preventive services.27 Selvin et al. found that having a medical home was "the most important predictor" of cancer screening use for all racial-eth- nic groups."28 In one study, in which each patient had not only a usual source of care but also an ongoing relationship with his/her own family physician, there were no racial differences in the provision of screen- ing services, and African Americans were actually slightly more likely than whites to receive preventive counseling with regard to health behaviors!29 Unfor-

tunately, the usual source of care for African-Ameri- can individuals is less likely to be in settings that offer continuity of care with a personal physician and more likely to be in hospital outpatient depart- ments or EDs."3031

To some extent, the issues may also be quite differ- ent for each minority group. For example, in analyzing data from nationally representative surveys conducted in 1996-1997 and 1998-1999, Hargraves found that more than 80% ofthe difference between Hispanic and white, non-Hispanic respondents was due to differ- ences in measured characteristics (e.g., insurance cov- erage, income, and available safety net services), but that these factors did not sufficiently explain the black-white differences.32 For the Hispanic population,

Table 3. Multivariate Analysis-Regression Coefficients and P Values for Predictors of Healthcare Utilization among African Americans, 1999

Category/item Office Visits Hosp Outpt. Clinic Visits ED Visits Hospital Discharges Rx Fills + Refills

Beta p-value Beta p-value Beta p-value Beta p-value Beta p-value

Overall Model <0.01 <0.01 <0.01 <0.01 <0.01

Age Group <0.01 <0.01 0.03 <0.01 <0.01 Under 18 -4.93 <0.01 -0.96 <0.01 -0.03 0.35 -1.13 <0.01 -19.63 <0.01 18-44 -3.74 <0.01 -0.60 0.04 0.04 0.35 -0.71 0.04 -15.65 <0.01 45-64 -2.41 <0.01 -0.37 0.22 0.05 0.32 -0.30 0.41 -7.47 <0.01 65+ 0.00 0.00 0.00 0.00 0.00 (Reference Group)

Sex 0.25 0.51 0.14 0.48 0.04 Male -0.30 0.25 -0.06 0.51 -0.03 0.14 0.09 0.48 -0.88 0.04 Female 0.00 0.00 0.00 0.00 0.00 (Reference Group)

MSA vs. non-MSA 0.47 0.56 0.05 0.01 0.03 MSA -0.20 0.47 0.07 0.56 -0.07 0.05 -0.56 0.01 -1.91 0.03 Non-MSA 0.00 0.00 0.00 0.00 0.00 (Reference Group)

Income/Poverty 0.02 0.37 0.28 <0.01 0.01 Poor 0.36 0.40 0.08 0.78 0.07 0.08 0.35 0.03 2.88 <0.01 Near Poor -0.36 0.09 -0.14 0.52 0.02 0.72 0.23 0.20 0.82 0.40 Low Income 0.03 0.92 0.00 0.99 0.05 0.21 0.16 0.12 0.97 0.12 Middle Income 0.44 0.25 -0.14 0.23 0.04 0.12 0.31 <0.01 1.18 0.02 High Income 0.00 0.00 0.00 0.00 0.00 (Reference Group)

Health Insurance 0.01 <0.01 <0.01 0.01 <0.01 Any Private 1.03 0.02 0.23 0.07 0.02 0.71 0.18 0.04 2.32 <0.01 Public Only 1.56 <0.01 0.35 0.02 0.13 <0.01 0.51 <0.01 4.79 <0.01 Uninsured 0.00 0.00 0.00 0.00 0.00 (Reference Group)

Usual source of care? <0.01 <0.01 0.31 <0.01 <0.01 Yes 1.69 <0.01 0.24 0.41 0.03 0.31 0.28 <0.01 4.11 <0.01 No 0.00 0.00 0.00 0.00 0.00 (Reference Group)

1174 JOURNAL OF THE NATIONAL MEDICAL ASSOCIATION VOL. 96, NO. 9, SEPTEMBER 2004

DETERMINANTS OF HEALTHCARE UTILIZATION

primary language (Spanish-speaking vs. English- speaking) is also a significant predictor ofuse ofphysi- cian visits, flu vaccination, or mental health services.33

Race, poverty, and having health insurance and/or a usual source of care have also been correlated with hospitalization for ambulatory care sensitive condi- tions. National Hospital Ambulatory Medical Care Survey (NHAMCS) data also showed that follow-up arrangements for African-American patients were less likely to result in ongoing primary care."34 Previ- ous studies have also shown significant racial-ethnic differences in the use ofmedication related to specific high-disparity conditions, such as hypertension, asth- ma, diabetes, depression, schizophrenia, hyperlipi- demia, and HIV/AIDS.35A1 Inadequate treatment of any one or more of these conditions has the potential to drive disparate rates of disability and death in the African-American population. In our study, use of prescription drugs was significantly lower among the uninsured and lowest among the health-disadvan- taged. This could reflect either lower rates of care for acute and chronic conditions (fewer initial prescrip- tions) or lower prescription refill rates, or both.

The only healthcare utilization rates that did not show a significant difference between the insured and uninsured or even between the most-advantaged and most-disadvantaged African Americans was the use of the ED, the one form of access to care that is mandated by federal law. The Emergency Medical Treatment and Active Labor Act (EMTALA) makes medical emergency visits (including obstetrical patients in active labor) the only form of healthcare in America in which patients must be at least seen, evaluated, and stabilized regardless of insurance or ability to pay. By this legislation, the ED becomes the de facto safety net for low-income, uninsured patients with no primary care medical home.42 The fact that there were no differences in ED visit rates between advantaged and disadvantaged African- American persons is an interesting "within-group" finding in itself, because one would expect the disad- vantaged group to have a greater burden of disease and healthcare needs, and therefore a higher emer- gency visit rate. In fact, previous NHAMCS studies have found significant black-white differences in ED utilization rates (71% higher for African-American

Table 4. Healthcare Utflization in 1999 for Healthcare-Disadvantaged* African Americans vs. Healthcare-Advantaged* African Americans

Mean SE Mean P Value

Office-Based Provider Visits <.001 Disadvantaged* 0.37 0.17 Advantaged** 3.12 0.27

Outpatient Debt Provider Visits =0.002 Disadvantaged 0.00 0.00 Advantaged 0.46 0.15

Emergency Room Visits =0.496 Disadvantaged 0.10 0.04 Advantaged 0.13 0.02

Hospital Discharges =0.0 10 Disadvantaged 0.01 0.01 Advantaged 0.07 0.01

Nights in Hospital =0.214 Disadvantaged 0.07 0.07 Advantaged 0.17 0.04

Home Health Provider Days =0.182 Disadvantaged 0.00 0.00 Advantaged 1.12 0.84

Prescription Meds Including Refills <0.001 Disadvantaged 0.93 0.70 Advantaged 6.05 0.50

* Family income below poverty level, no health insurance, no usual source of care. ** Family income above 400% poverty level, health insurance, a usual source of care.

JOURNAL OF THE NATIONAL MEDICAL ASSOCIATION VOL. 96, NO. 9, SEPTEMBER 2004 1175

DETERMINANTS OF HEALTHCARE UTILIZATION

than for white persons). While the visit rate for elder- ly African Americans increased by 59% in just seven years (from 1992 to 1997), the visit rate for elderly whites did not change.43

However, in spite of an equal rate of reported ED visits in our study between advantaged and disadvan- taged African Americans, the ratio ofhospital admis- sions to ED visits was dramatically lower for low- income, uninsured individuals with no usual source of care (one admission for every 10 ED visits) than for insured, upper-income, individuals with a med- ical home (one admission for every two ED visits). A cross-sectional study of 29,237 admissions to 100 U.S. hospitals in 1993 and 1994 found that uninsured patients were sicker than the insured but had shorter lengths of stay and poorer health outcomes, suggest- ing that the uninsured might not be receiving neces- sary care.," The ED visit to hospital admission ratios in our data cannot be explained by suggesting that disadvantaged African-American patients are using the ED inappropriately or for nonurgent conditions, because the disadvantaged are visiting the ED at pre- cisely the same rate as are the advantaged patients.

One source of variation might be the hospitals visited by disadvantaged patients. Analysis of 1999 data from the NHAMCS suggested significant varia- tion among hospital EDs in the percent of patients admitted to the hospital."45 Another possibility is that advantaged (especially insured) patients are being admitted to the hospital with a lower threshold of clinical severity than are the disadvantaged, because there are significant financial incentives for hospitals not to admit uninsured, low-income patients. Hospi- tals may seek alternatives to hospital admission (pro- longed ED observation, or stabilization and dis- charge) especially for the uninsured, in order to avoid absorbing the cost ofan indigent hospitalization.

One potential weakness of our study is the rela- tively small proportion of individuals within the African-American population who meet all three cri- teria for being health-disadvantaged. This is an important point in itself, which is that many articles present the African-American population as being almost exclusively disadvantaged. Confounding a person's race with potentially changeable risk factors (i.e., listing "poor, black, and uninsured" in the same breath) is not only counterproductive but also inaccu- rate. However, multivariate analysis suggests that each of our three "disadvantaged" criteria contribute independently to the variation in use of healthcare services within the African-American population. While poverty is a fundamental root cause, these data also suggest that significant reductions in disparities might be achieved with focused interventions to achieve universal health insurance coverage and uni- versal access to a primary care medical home.

CONCLUSIONS The three modifiable factors of poverty, uninsur-

ance, and having a primary care medical home have a dramatic effect on patterns ofcare for African-Ameri- can patients. Targeted interventions, such as provision of health insurance, poverty reduction, and eliminat- ing barriers to having a primary care medical home, could reduce or eliminate disparities in use ofpreven- tive and primary care services (and other needed healthcare) for the African-American population.

ACKNOWLEDGEMENTS This work was supported in part by NIH/

NCMHD Grant #5 P20 MD00272-02 (EXPORT Center of Excellence on Health Disparities).

REFERENCES 1. Mayberry RM, Mili F, Ofili E. Racial and ethnic differences in access to medical care. Med Care Res Rev. 2000;57 Suppl 1 :108-45. 2. Wyatt SB, Williams DR, Calvin R, et al. Racism and cardiovascular disease in African Americans. Am J Med Sci. 2003;325:315-31. 3. Stevens GD, Shi L. Racial and ethnic disparities in the primary care experi- ences of children: a review of the literature. Med Care Res Rev. 2003;60:3-30. 4. Graham-Garcia J, Raines TL, Andrews JO, et al. Race, ethnicity, and geography: disparities in heart disease in women of color. J Transcult Nurs. 2001;12:56-67. 5. Flores G, Fuentes-Afflick E, Barbot 0, et al. The Health of Latino Children: Urgent Priorities, Unanswered Questions, and a Research Agenda. JAMA. 2002;288:82-90. 6. van Ryn M. Research on the provider contribution to race/ethnicity dis- parities in medical care. Med Care. 2002;40(1 Suppl):1140-1151. 7. Clancy CM, Stryer DB. Racial and ethnic disparities and primary care experience. Health Serv Res. 2001;36(6 Pt 1):979-86. 8. Board on Health Sciences Policy (HSP), Institute of Medicine. Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care. National Academies Press, Washington DC; 2002. 9. Ku L, Waidman. How Race/Ethnicity, Immigration Status, and Language Affect Health Insurance Coverage, Access to and Quality of Care Among the Low-lncome Population Kaiser Commission on Medicaid and the Unin- sured, Kaiser Family Foundation, 2003. Available at http://www.kff.org/ sections.cgi?section=kcmu. 10. Kaiser Commission on Medicaid and the Uninsured. Health Insurance Coverage and Access to Care Among Latinos. Kaiser Family Foundation, 2003. Available at http://www.kff.org/sections.cgi?section=kcmu. 1 1. Brown SL. Race and sex differences in the use of cardiac procedures for patients with ischemic heart disease in Maryland. J Health Care Poor Underserved. 2002;13:526-537. 12. Wenneker MB, Epstein AM. Racial inequalities in the use of procedures for patients with ischemic heart disease in Massachusetts. JAMA. 1 988;261 :253-257. 13. Funk M, Ostfeld AM, Chang VM, et al. Racial differences in the use of cardiac procedures in patients with acute myocardial infarction. Nurs Res. 2002;51 :148-1 57. 14. Wyatt SB, Williams DR, Calvin R, et al. Racism and cardiovascular dis- ease in African Americans. Am J Med Sci. 2003;325:315-331. 15. Aday LA, Andersen RM. Equity of access to medical care: a conceptu- al and empirical overview. Med Care. 1981;19:4-27. 16. Phillips KA, Morrison KR, Andersen R, et al. Understanding the context of healthcare utilization: assessing environmental and provider-related vari- ables in the behavioral model of utilization. Health Serv Res Policy. 2001;6: 183-185. 17. Williams DR. Race, socioeconomic status, and health. The added effects of racism and discrimination. Ann NYAcad Sci. 1 999;896:1 73-188. 18. Williams DR. The health of men: structured inequalities and opportuni-

1176 JOURNAL OF THE NATIONAL MEDICAL ASSOCIATION VOL. 96, NO. 9, SEPTEMBER 2004

DETERMINANTS OF HEALTH CARE UTILIZATION

ties. Am J Public Health. 2003;93:724-731. 19. Zuvekas SH, Taliaferro GS. Pathways to access: health insurance, the healthcare delivery system, and racial/ethnic disparities, 1996-1999. Health Aff. (Millwood), 2003;22:139-153. 20. Elixhauser A, Machlin SR, Zodet MW, et al. Healthcare for children and youth in the United States: 2001 annual report on access, utilization, quality, and expenditures. Ambul Pediatr. 2002;2:419-437. 21. Medical Expenditure Panel Survey. 1999 Full-Year Consolidated Data File (electronic file). Rockville, MD. Agency for Healthcare Research and Quality, 2003. 22. Dunlop DD, Manheim LM, Song J, et al. Gender and ethnic/racial dis- parities in healthcare utilization among older adults. J Gerontol B Psychol Sci Soc Sci. 2002;57:S221-33. 23. Main MG, Johanson Jr WG, Salas-Lopez D. Influenza vaccination among minority populations in the United States. Prev Med. 2002;34:235-241. 24. Rawl SM, Champion VL, Menon U, et al. The impact of age and race on mammography practices. Health Care Women tnt. 2000;21:583-597. 25. Ward E, Jemal A, Cokkinides V, et al. Cancer disparities by race/ethnici- ty and socioeconomic status. CA Cancer J Clin. 2004;54:78-93. 26. Jacobellis J, Cutter G. Mammography screening and differences in stage of disease by race/ethnicity. Am J Public Health. 2002;92:1144-1150. 27. Devoe JE, Fryer GE, Phillips R, et al. Receipt of preventive care among adults: insurance status and usual source of care. Am J Public Health. 2003;93:786-797. 28. Selvin E, Brett KM. Breast and cervical cancer screening: sociodemo- graphic predictors among white, black, and Hispanic women. Am J Public Health. 2003;93:618-623. 29. Williams RL, Flocke SA, Stange KC. Race and preventive services deliv- ery among black patients and white patients seen in primary care. Med Care. 2001;39:1260-1267. 30. Ly N, McCaig LF. National Hospital Ambulatory Medical Care Survey: 2000 outpatient department summary. Adv Data. 2002;1-27. 31. Walls CA, Rhodes KV, Kennedy JJ. The emergency department as usual source of medical care: estimates from the 1998 National Health Interview Survey. Acad Emerg Med. 2002;9:1140-1145. 32. Hargrave JL, Hadley J. The contribution of insurance coverage and community resources to reducing racial/ethnic disparities in access to care. Health Serv Res. 2003;38:809-829. 33. Fiscella K, Franks P, Doescher MP, et al. Disparities in healthcare by race, ethnicity, and language among the insured: findings from a national sam- ple. Med Care. 2002;40:52-59. 34. Oster A, Bindman AB. Emergency department visits for ambulatory care sensitive conditions: insights into preventable hospitalizations. Med Care. 2003;41 :198-207. 35. HaJar 1, Kotchen TA. Trends in prevalence, awareness, treatment, and control of hypertension in the United States, 1988-2000. JAMA. 2003. 290:199-206. 36. Lieu TA, Lozano P, Finkelstein JA, et al. Racial/ethnic variation in asthma status and management practices among children in managed medi- caid. Pediatrcs. 2002;109:857-865. 37. Schectman JM, Nadkarni MM, Voss JD. The association between dia- betes metabolic control and drug adherence in an indigent population. Diabetes Care. 2002;25:1015-21. 38. Miranda J, Duan N, Sherbourne C, et al. Improving care for minorities: can quality improvement interventions improve care and outcomes for depressed minorities? Results of a randomized, controlled trial. Health Serv Res. 2003;38:613-630. 39. Owen RR, Fischer EP, Kirchner JE, et al. Clinical practice variations in prescribing antipsychotics for patients with schizophrenia. Am J Med Qual. 2003;1 8:140-146. 40. Nelson K, Norris K, Mangione CM. Disparities in the diagnosis and phar- macologic treatment of high serum cholesterol by race and ethnicity: data from the Third National Health and Nutrition Examination Survey. Arch Intern Med, 2002;162:929-935.

41. Palacio H, Kahn JG, Richards TA, et al. Effect of race and/or ethnicity in use of antiretrovirals and prophylaxis for opportunistic infection: a review of the literature. Public Health Rep. 2002;1 17:233-251; discussion 231-232. 42. Glauser J. Rationing and the role of the emergency department as society's safety net. Acad Emerg Med. 2001;8:1101-1106. 43. McCaig LF, Burt CW. National Hospital Ambulatory Medical Care Sur- vey: 1999 emergency department summary. Adv Data. 2001;25:1-34. 44. Bradbury RC, Golec JH, Steen PM. Comparing uninsured and privately insured hospital patients: admission severity, health outcomes and resource use. Health Serv Manage Res. 2001;14:203-210. 45. McCaig LF, Burt CW. National Hospital Ambulatory Medical Care Sur- vey: 1999 emergency department summary. Adv Data. 2001;25:1-34. 1

We Welcome Your Comments The Journal of the National Medical Association welcomes your Letters to the Editor about arti- cles that appear in the JNMA or issues relevant

to minority healthcare. Address correspondence to [email protected].

Life at LiLy real people doing extraordinary things

CLINICAL RESEARCH PHYSICIANS - COMMERCIALIZATION, EARLY PHASE, AND LATE PHASE Lilly is seeking qualified physicians who would like to grow within a corporate environment and develop their regulatory knowledge and commercialization of products skills. The following positions are available:

Clinical Research Physiacan Commercialiation As a member of the US Affiliate, this individual will be responsible for the commerciatization of marketed products and the support of registration trials. This will involve working with representatives from the program and product teams, medical, marketing, and regulatory atairs.

Clinical Research Physician Early Phase Serving on the program team, this person will provide clinical expertise and leadership in support of the discovery and early phase testing of a compound. This witl involve participating in or leading a clinical action team (CAT) for each earty phase compound.

Clinical Research Physician Late Phase This person will serve on the product team to provide clinical expertise and leadership in support of the further clinical development, registration, and commercialization of a product.

Requirements for the above positions include: * An M.D. or D.O. * Board certified or eligible * Strong organizational, communication, and team relationship skills * Willingness and ability to travel * Fluency in written and spoken English * Clinical experience and experience with the clinical trial process are preferred Eli Lilly and Company is an equal opportunity employer.

wwwlilly.com/careers

Answers That Matter.

JOURNAL OF THE NATIONAL MEDICAL ASSOCIATION VOL. 96, NO. 9, SEPTEMBER 2004 1177