|Home | About | Journals | Submit | Contact Us | Français|
More than a quarter of human immunodeficiency virus (HIV)-infected persons are undiagnosed and therefore unaware of their HIV+ status. African Americans are disproportionately infected. Although perceived racism influences their attitudes toward HIV prevention, how racism influences their behaviors is unknown. We sought to determine whether perceiving everyday racism and racial segregation influence African American HIV testing behavior.
This was a clinic-based, multi-level study. Eligibility was limited to African Americans (N=373) seeking sexually transmitted disease diagnoses or screening. We collected survey data, block group characteristics, and lab-confirmed HIV testing behavior. We estimated associations using logistic regression with generalized estimating equations (GEE).
More than 90% of the sample perceived racism, which was associated with higher odds of HIV testing (OR=1.64, 95% CI=1.07, 2.52) controlling for residential segregation and other covariates. Neither patient satisfaction nor mechanisms for coping with stress explained the association.
African Americans are not merely victims of racism but also exercise agency within their social contexts. Those who perceive everyday racism may draw upon health promoting assets relative to their behaviors. How segregation influences testing warrants further research.
Despite decreases in mortality due to acquired immune deficiency syndrome (AIDS), the prevalence of human immunodeficiency virus (HIV) infection in the United States remains high overall and diagnoses among African Americans continue to increase.1 Although African Americans represent less than 13% of the US population, they account for 42% of prevalent HIV infections and 54% of annual diagnoses.1 An estimated one-fourth of all HIV-infected US residents have not been diagnosed.2 That HIV positive blacks delay care-seeking more, progress to AIDS faster, and die from AIDS sooner than whites underscores the need to improve HIV screening in this population.3, 4
Sexually transmitted disease (STD) clinics are an important setting for reaching persons at elevated risk of sexual transmission, the primary mode by which HIV transmission occurs.5 These clinics provide testing regardless of an individual's ability to pay. The prevalence of undiagnosed HIV infection is higher in STD clinic populations than in the general population and persons engaging in STD risk behaviors are by definition at risk for HIV transmission. Furthermore, although every exposure to HIV does not result in seropositivity, epidemiologic synergy between HIV and classic STDs such as gonorrhea renders STD-infected persons more susceptible to HIV-infection upon exposure to the virus.6
Population-based surveys suggest that blacks obtain HIV testing at higher rates than do other racial/ethnic groups7; however, self-reports may overestimate actual testing behavior. In one nationally representative study, 25% of blacks reporting prior HIV tests had assumed they were tested during some clinical visit where they had neither requested nor consented to a test.8 Among STD clinic patients, blacks may actually be less likely to test.9
For African Americans, negative attitudes toward HIV prevention are linked to racism.10 Racism has been defined as “an organized system, rooted in an ideology of inferiority that categorizes, ranks, and differentially allocates societal resources to human population groups”.11 (p.76) Racism is a multilevel construct fundamentally influenced by macro-level factors such as residential segregation.12-14 Research suggests that perceiving or experiencing racial discrimination contributes to hypertension15-17, pre-term birth18, mental health outcomes19, 20, and unhealthy coping behaviors (e.g., cigarette smoking and alcohol use)21. Exactly how individuals respond to perceived racism also is important. In the CARDIA study, for instance, persons who perceived and challenged racism on the job had lower systolic blood pressures (i.e., better outcomes) than those whom investigators described as internalizing it.16 For some African Americans, however, overachieving in response to racism may adversely impact health, a phenomenon described as “John Henryism”.22
Racism can be thought of as an element in the social environment; perceived racism is the extent to which individuals are aware of that element. Perceived everyday racism reflects individuals' assessments of potentially negative routine interactions (e.g., being followed while shopping in a store) as resulting from racism rather than as due to other causes (e.g. coincidence).23 In some contexts perceiving racism is detrimental while in others it is self-protective.24
Most racism-related HIV prevention research examines extreme forms of racism rather than everyday racism. These studies indicate that awareness of the US Public Health Services' study of untreated syphilis in African American men and beliefs that “the government is…using AIDS as a way of killing off minority groups”25 are prevalent and associated with negative attitudes about HIV prevention.26-29 Few studies have examined perceived racism's association with HIV preventive behaviors and these primarily assess self-reports of behavior. A national phone survey28 of African Americans (N=500) and a Houston-based survey30 of a multi-racial sample (N=1494) found negative associations between conspiracy beliefs and self-reported condom use for African American men.28, 30 One study31 found perceived everyday racism positively associated with condom use. These studies did not account for residential context.
Perceptions about racism, however, are influenced by interracial interactions. More integrated blacks perceive more racism.32 Outside the workplace, residential areas are the most likely arenas for interactions; often, however, these are racially segregated. The systematic residential isolation of blacks from whites through de facto segregation is a fundamental cause of disparities, differentially influencing access to health care, socioeconomic status, and quality of services.13, 33, 34 Segregation historically has impacted communities in the US South. The most widely assessed dimension of segregation is unevenness. Calculated via the dissimilarity index, D, it indicates an area's relative proportions of minority and majority populations.35
Studies using census geographies such as block groups permit monitoring of area socioeconomic and demographic trends across time and place; census designations are well-defined units of analysis and the socioeconomic data are systematically collected.36 Block groups, which average 1,500 residents, are the smallest geographic units for which the US Census Bureau provides sample data.37 Although many studies operationalize neighborhoods as tracts or zip code regions, block groups may be more appropriate units when studying smaller cities or regions where neighborhood boundaries change rapidly.
The purpose of this study was to examine perceived everyday racism's association with routine HIV testing among at-risk African Americans while accounting for racialized residential contexts.
Data were collected from March to June 2003 in a public STD clinic located in a North Carolina city with high HIV and STD prevalences. African Americans make up 20% of the county's and 28% of the city's population.37 The city's residential segregation (Dissimilarity Index=0.54) exceeds that of 85% of US cities with comparable population size.38 The African American population resides primarily within two contiguous zip code regions where STD and HIV prevalences were highest. Most persons who obtain care at the clinic reside in these neighborhoods. As established in extensive formative research39, access (e.g., transportation) to the clinic was not a barrier to testing in this population. The clinic is the primary source of HIV tests in the county, providing between 3,000 and 4,000 HIV tests annually.40 Blacks account for more than 60% of the clinic's patients.
This was a multilevel, cross-sectional study estimating individual-level associations relative to the behavioral outcome, visit-specific uptake of routine HIV testing, while accounting for population-averaged residential (i.e., “neighborhood”) characteristics and other factors. The conceptual framework guiding the study integrated Critical Race Theory41 concepts and Andersen's Access to Care Model42. As standard practice at the clinic, STD patients seeking non-follow-up care were automatically offered an HIV antibody test during their visit. The University of North Carolina's School of Public Health Institutional Review Board approved all aspects of the study.
Eligible clinic participants were consecutively enrolled during routine clinic hours. Only outpatients seeking diagnosis or screening for STD and/or HIV infection were eligible for participation; persons seeking follow-up care, information or other services were not eligible. Inclusion criteria were self-reported race as black or African American, age 18 years or older and presenting for diagnosis of or screening for possible STD infection. Of those eligible and invited (N=474), 61 (41 men, 20 women) declined participation, resulting in an 87% response rate; 413 patients consented to and enrolled in the study. Responses from 38 participants (9%) were excluded from the analyses because data on their visit-specific HIV testing behavior, the outcome of interest, were missing. Two additional observations were excluded because of incomplete questionnaires. The final sample size for individual-level analyses was 373, 56% (n=210) women. Block group analyses further excluded 61 observations because participant addresses were unreported (e.g., due to homelessness) or impossible to validate using a geographic information system (GIS). The final sample size for block group analyses was 312 individuals living within 117 block groups.
Participants completed the 101-item questionnaire while seated in the clinic lobby. To enable privacy and to accommodate possible low levels of literacy, participants listened to an audio-taped version of the questionnaire using headsets and marked responses on a corresponding paper form. The audio-taped recording was completed by an African American research assistant native to the region. Each participant's visit-specific HIV testing behavior was ascertained from the clinic's daily log of diagnostic tests and recorded in a manner blinded to questionnaire data collection. All names were then removed from questionnaires.
The outcome was visit-specific uptake of HIV testing via blood draw with enzyme-linked immunosorbent assay (ELISA), as recorded in the clinic lab's daily log of administered tests. Test uptake was coded yes=1 or no=0. Each time a study participant was seen, lab technicians summarized patient information (name, chart number, and whether HIV testing was performed) in the daily log relevant to HIV antibody test acceptance.
Perceived racism and residential segregation were the key explanatory variables. The questionnaire assessed perceived racism43, 44, coping mechanisms for stress, individuals' demographic characteristics, HIV prevention-related constructs (e.g., perceived risk, HIV knowledge), and clinical encounter factors (e.g., previous clinic use). In the pilot study, internal consistency was high (Cronbach's α≥0.70) for the perceived racism43, 44, patient satisfaction45, HIV knowledge46 and perceived HIV susceptibility47 scales. Reliability and validity of the perceived racism scale had been established previously for in-person 43 and telephone administration44. We used the 10-item sub-scale of the perceived racism scale to assess respondents' perceptions about the extent to which blacks generally encounter several types of white-on-black perceived racist experiences on the job, in social settings and in public settings. For instance, “In general, when blacks shop, they are followed or watched by white security guards or white clerks.” Response options ranged across a four-point Likert-type scale from strongly disagree to strongly agree.
We operationalized residential areas as the Census 2000 block groups in which study respondents resided on the basis of addresses sample members reported in the questionnaires. We obtained data on block group racial composition from the census summary file one and assessed segregation using the dissimilarity index (D). This index reflects the proportion of blacks who must move in order to achieve racially equal population distributions.35 Block group segregation was derived from each block group's constituent blocks' racial composition and compared across the block groups comprising the study area. Index values may range from 0-1, with one indicating complete segregation.
Covariates included patient satisfaction, a potential confounder. Based upon focus groups, we derived a 5-item, ordinal scale from Marshall and Hays' 18-item questionnaire45. Each response was scored on a five-point Likert-type scale (strongly agree to strongly disagree). Stress coping mechanisms were assessed via responses to the statement “How often do you cope with stress by…” Responses were categorized as passive (e.g., sleeping), healthy (e.g., exercising), or negative (e.g., drinking). Standard HIV prevention-related constructs such as perceived HIV risk were assessed using measures borrowed from the CDC's Behavioral Risk Factor Surveillance System (BRFSS)47, 48.
Additional variables reflecting neighborhood deprivation or inequity were derived from census summary files one and three. These included relative income (ratio of block group median income to median income for the region), concentrated poverty (40% or greater poverty within a block group)49, percentage of black residents (proportion of block group population that was black alone or in combination with some other race relative to the block group total population), percentage of vacant households, percentage unemployed, mean educational attainment, and percentage of female-headed households.
We used non-automatic, backward elimination to specify multivariable logistic regression models50 with generalized estimating equations (GEE) to account for variance clustering that occurs when one level of data (i.e., individual) is nested within another (i.e., block group). GEE is preferred to other multilevel approaches (e.g., mixed models) when, as in this study, the group level units are not a random sample of some universe of block groups.51 We specified an exchangeable correlation structure (i.e., any two responses within a cluster have the same correlation) and used robust variance estimators to derive 95% confidence intervals around each estimate. In exploratory analyses we examined variable distributions, collinearity, interaction, and potential statistical confounding. We also compared estimates obtained using the dissimilarity index to those using percentage black. Analyses were conducted using Stata Version 8.52
Table 1 displays sample demographic characteristics. Perceived risk of HIV infection was low across age groups and gender categories. Forty-six percent of sample members were seeking care because they had symptoms of an STD; 10% had been referred by another provider.
Perceived racism scores ranged from the absolute minimum, 10, indicating no perceived racism, to the maximum, 40, indicating the highest level of perceived racism assessed on the scale. The observed mean scores revealed, on average, agreement with nearly all of the items on the scale. The range, mean and median scores did not vary by gender.
Fifty-five percent of participants obtained HIV tests during their clinical visits. Proportionally fewer men than women (χ2=6.25, DF=1, p=0.01) tested. Relative to younger participants (i.e., age<30 years), greater proportions of participants aged 45 years old or older tested (54% versus 68%), but this difference was not statistically significant. HIV knowledge and perceived HIV risk were unrelated to testing behavior.
Figure 1 displays summary data on residential characteristics. Spatially, residences clustered around the clinic and loosely followed major roads (data not shown). Between one and 14 participants resided in each block group. Block group dissimilarity index scores ranged from D=0.13 to 0.94; however, they were heavily centered around the median, 0.40. Testers resided in more integrated areas and non-testers resided in more segregated areas (data not shown).
The final statistical model is shown in Table 2. The full model included perceived racism, individual demographic characteristics, patient satisfaction, symptoms, ever previously obtaining an HIV test, perceived HIV risk, coping mechanisms, HIV knowledge and block group characteristics. Perceived racism was associated with higher odds of obtaining routine HIV testing during the clinic visit in crude analyses; this association persisted in the final model (OR=1.64, 1.07, 2.52). Each 10-unit increase (e.g. from “medium” to “high”) on the 30-unit Perceived Racism Scale corresponded with approximately 60% higher odds of testing. The point estimate for the dissimilarity index (OR=0.32, 0.05, 2.18) in the final model suggested higher levels of segregation were associated with not testing; however, the 95% CI included the null value. In models replacing the dissimilarity index with percentage black, the percentage of blacks in block groups was unassociated with HIV testing (OR=0.99, 0.99, 1.00).
Nearly all participants in this study perceived everyday racism; the more racism perceived the higher the odds of being tested for HIV during an STD clinic visit. Neither patient satisfaction nor stress coping mechanisms explained this association. Previous research linking perceived racism to attitudes has suggested it may negatively influence behaviors.53 Our findings did not support that hypothesis. One reason may be that while this study examined everyday racism, most previous research has focused on extreme forms of racism (e.g., HIV conspiracy beliefs). In racially stratified societies individuals are constantly exposed to everyday racism43; so, it is conceivable that there may be contexts in which perceiving it is health protective.
Although the positive association between perceived racism and HIV testing may seem counterintuitive, support for it exists in previous research. Others have found that perceiving racism is not inherently detrimental and that African Americans who notice it in their social environments and challenge it may have healthier outcomes than do those who deny its existence or blame themselves or other minorities for observed disparities.16 The findings also concur with those of the one other study found that examined perceived everyday racism and HIV preventive behaviors.31 Investigators using the 38-item Racism and Life Experiences Scale (SP Harrell, PhD, unpublished data, 1997) observed that perceived racism was prevalent and positively associated with condom-related HIV preventive behaviors among African American women.
We controlled for coping mechanisms because how individuals deal with racism they perceive has implications for their well-being.23 Resiliency, which we did not directly measure, may explain positive associations between perceived racism and health protective behaviors such as HIV testing. According to the resiliency hypothesis, blacks who perceive racism and develop ways to function day-to-day in spite of it have higher resiliency levels and cope better with racism-related stressors. An alternative explanation is that social support, also unmeasured in this study, could have confounded the association between perceived racism and testing if persons with greater perceived support felt more comfortable both reporting perceived racism and obtaining HIV tests.
Fewer residents of more segregated block groups got tested. This finding is consistent both with the hypothesis that more segregated blacks perceive racism less54 and with the observed association in this study between lower perceived racism and not getting tested. While segregation often signals blacks' poor access to care, extensive formative research preceding this study indicated access to the STD clinic was not a barrier.39 Rather, segregation was included because of its conceptual relevance to perceived racism. The small geographic area studied, and its predominately black composition, however, biased any associations around segregation toward the null. The findings contrast with a Los Angeles study in which, regardless of race/ethnicity, persons living in “predominately black” areas were more likely to obtain HIV tests.55 That study operationalized neighborhoods using ZIP codes rather than the census designations recommended for monitoring population health.56 It also used percentage black as a proxy measure of segregation, rather than a segregation index as used in this study. Percentage black may measure unspecified socioeconomic or racial/ethnic factors. Then, too, this research was conducted in the US South where distinctive racial relations persist and racial groups are less heterogeneous. The complex relationships between segregation, network characteristics such as network-specific HIV prevalences, and behaviors require further research.57-59
That perceived HIV risk was low corroborates other research60, 61. Denial and fear of becoming HIV infected may explain this common finding.62 Primary prevention to promote accurate risk assessments will remain important in the increasingly screening-oriented prevention climate.63
Greater conceptual and empirical clarification of ‘racism’ constructs is also needed. This study assessed two related concepts: perceived racism, an individual level factor, and residential segregation, a macro level factor. Most research emphasizes interpersonal discrimination rather than structural factors, which may underlie persistent disparities.64
The study had several limitations. Thirty eight observations were excluded because the participants' testing behaviors were not recorded in the clinic's log. This only occurred if individuals reported a false name to the study and/or clinic, suggesting distrust and/or stigma in this population. Sensitivity analyses using single unit contrasts on the 30-unit perceived racism scale indicated that these missing cases did not substantially distort the findings. Although slightly lower mean perceived racism scores biased the results slightly away from the null, similar point estimates and substantially overlapping confidence intervals were obtained whether these 38 observations were excluded (OR=1.04; 95% CI=1.01, 1.09), recoded to “no test” (OR=1.05; 95% CI=1.02,1.09), or recoded to “test” (OR=1.03; 95% CI=1.00, 1.06). The study did not control for risk behaviors such as injection drug use, which could influence testing behavior. Although the dissimilarity index is the most widely-used measure of residential segregation, it is insensitive to extremely high or low minority populations, may underestimate overall levels of segregation and does not address geographic considerations captured by other measures (e.g., local area spatial autocorrelations).35, 65 The geographic region studied here was small and limited in its racial heterogeneity. Block group estimates, therefore, were biased toward the null. Lastly, data were sparse (n≤5) in most block groups. One strength of GEE is that it accommodates sparse or missing data66; nonetheless, we confirmed the main association by conducting logistic regression analyses that excluded block group level variables and found basically the same results (OR=1.64; 95% CI: 1.06, 2.55).
Use of actual, not self-reported, HIV testing reassures us of the validity of this measure of stigmatized behavior. Similarly, our operationalization of residential areas as block groups may better approximate neighborhood patterns in small cities, where neighborhood boundaries may be small or may shift quickly. The public health literature on racism, which primarily targets mental health, chronic disease and birth-related outcomes, is advanced by this research on an infectious disease-related outcome. Although social context was not empirically confirmed as a risk factor for testing or failure to test, the field is advanced conceptually by the inclusion of context in the study's design.67
African Americans are not merely victims of racism but also exercise agency within and regarding their social contexts. Those who perceive everyday racism may draw upon health promoting assets relative to their behaviors. We recommend that practitioners work closely with this population to identify transferable, health protective skills to promote preventive behaviors among other blacks and that future research help to clarify the settings, outcomes and subpopulations in which perceived racism is detrimental versus constructive.
Fullilove suggests that racism is the “elephant in the room” when providers and educators deliver counseling, testing and education in this population.68 By showing that perceived racism was not a barrier to HIV prevention, our findings like Fullilove's argument challenge assumptions that awareness of racism necessarily inhibits HIV prevention among blacks. Intervention research is needed to explore whether discussing racism during counseling, testing or education can reduce distrust and improve communication between African Americans and prevention professionals. For residents of more segregated areas who may be less likely to obtain clinic-based HIV testing, outreach may prove to be effective for reaching this population.
Policies attentive to assets-based interventions, cultural competence and equity are needed to guide the development and implementation of any prevention strategies (i.e., targeted outreach) that aim to integrate racism-related concerns and prevention efforts.69
This research was supported by the National Institute of Allergy and Infectious Diseases (Fellowship #: 1 F31 AI058914-01), the Agency for Healthcare Research and Quality (Grant #'s: T32-HS00032, T32-HS00031) and the W. K. Kellogg Foundation Kellogg Health Scholars Program (Grant # P0117943) administered via the Center for the Advancement of Health. The following units at the University of North Carolina also provided support: Office of the Vice-Chancellor for Research and Development, Department of Social Medicine, Cecil G. Sheps Center for Health Services Research, Center for AIDS Research, North Carolina Rural Health Research Program, and Center for the Study of the American South. The authors acknowledge Anissa Vines, Lori Carter-Edwards and Dionne Godette whose feedback strengthened the manuscript.
Contributors: Chandra Ford conceived of the study, led in conducting the analyses and led in writing the manuscript. Mark Daniel assisted with study design, analyses and manuscript revisions. Jo Anne Earp assisted with study design, interpretation of findings and manuscript revisions. Jay Kaufman assisted with analyses, interpretations of findings and manuscript revisions. Carol Golin and William Miller assisted with study design and manuscript revisions.
Human Participant Protection: The study was approved by the Institutional Review Board of the University of North Carolina's School of Public Health.