On both the macro and individual levels, theories of social resources share a common belief that such resources enhance health. On a societal level, theorists from Durkheim to Putnam have argued that social integration and connection to the larger society promotes health enhancing behaviors through social regulation and attachment, as well as providing the tools needed for achieving health, through social capital and shared social resources 13. On an interpersonal level, social connectedness is also seen as health-promoting for most individuals, through both social role obligations and social support 14. The relationship, however, between interpersonal social bonds and societal level roles is less clear.

When considering the role of social resources on health, it is possible to view societies as composed of nested social networks, working at various levels in a complementary way. On the other hand, it can be argued that affiliations within social groups come by definition only with exclusion of other groups, and that building of social capital for one group comes at the expense of the power of another 15. From such a perspective, group identities and loyalties increase symbolic and material conflict between groups, leading not to social integration but disintegration.

Within societies, one interpretation of discrimination is as the result of struggles of groups to achieve occurring at the expense of other groups. For social researchers, as in the discussion of concepts such as stress, discrimination has emerged as multifaceted – it can be defined as the intent of the perpetrator, the interpretation by the recipient, or as the effect, or potential effect, of events and actions. Intent relies on the report of the perpetrator, and interpretation on the report by its recipient, while effect can be observed and measured by third parties.

As researchers, depending on our scientific philosophy, we must ask a related question as to whether our epistemology of discrimination is essentially based in a positivistic or more subjective, interpretivistic meaning and measurement 16. Empirical studies of discrimination and health reflect this tension between externally defined, objective injustices and constructed or perceived discrimination.

The first is the measurement of incidents or processes that are defined as inherently discriminatory practices, or create de facto situations of discrimination, ranging from individual actions to discriminatory laws. These measures do not rely on the perceptions of the disadvantaged individual. Krieger 17 has labeled this type of discrimination as "indirect" because it measures result rather than intent or interpretation. The credibility and utility of such a consensus-based objectivistic approach is vulnerable to changing legal or cultural definitions of discrimination 18.

Perceived or reported discrimination, however, requires that individuals experience a situation in which they perceive themselves to be at a disadvantage compared to others, attribute that disadvantage to discrimination, and choose to disclose it to the questioner. Defining discrimination as an essentially subjective phenomenon means that if respondents report no discrimination, if they 1) do not feel they have received less, 2) attribute the difference to reasons other than their group membership, or 3) chose not to disclose their perceptions, then we must accept their authority in the interpretations of events. For example, in our previous use of questionnaire items on perceived discrimination in schooling, some older African Americans answered "No, I never experienced racial discrimination, because Blacks and Whites went to different schools," while other of their contemporaries answered affirmatively, reflecting the more common view that racially segregated schools were fundamentally discriminatory. Although a more positivist perspective would view these differing responses as problematic, a researcher whose goal was to understand "perceived discrimination" would focus on exploring these differences.

There is an international literature on ethnic and racial discrimination, focusing largely on discrimination toward either indigenous or immigrant non-white ethnic groups by economically or socially dominant white ethnic groups 19202122232425. This literature shows wide variation in the discriminatory experiences reported, with variation depending on methodology and measures used, the characteristics of the discriminatory acts asked about (timing, type, setting, etc) and the sociodemographic characteristics of the respondents being questioned. This variation is also reflected in U.S. literature exploring racial discrimination among non-white groups. Historically, the majority of U.S. studies focus on the African-American experience; thus this literature provides the greatest evidence of the complexity of this issue.

Despite the prevalence of discriminatory practices in US society, national surveys 26 show that, for example, only 49% of Blacks report lifetime occurrence of major events of discrimination, while 71% report day-to-day discrimination as occurring "often" or "sometimes." Younger persons, and those with higher educational status are consistently more likely to report discrimination 26. Adams and Dressler 27 found greater racism reported by African-Americans who had greater perceived personal influence, concluding "persons who see themselves as able to make changes are also more likely to perceive conditions that need changing." Others suggest that, for ethnic minorities, achieving greater social status clarifies discrimination as race- and not class-based 28. These patterns are, on first glance, counter-intuitive, because we would anticipate that, if using an objective measure of negative experiences, those worse off would be most, rather than least, likely to report discrimination. Given that all African-Americans are subject to adverse conditions, it appears that those with greater personal resources are more likely to recognize, attribute, and willingly disclose discrimination. This paradox makes it difficult to separate, especially in cross-sectional measurement, the conditions accompanying or causing discrimination, the conditions facilitating its recognition, attribution and disclosure, and the conditions discrimination in turn truly produces. Measurement and framing effects further complicate comparisons across groups 2930.

When considering older, African-American women living in poverty, who have lived for many years as members of not one but multiple groups subject to discrimination in the US – African-Americans, low income persons, and women – can we speculate on how labeling oneself a recipient of discrimination affects one's well-being? A conflict model would predict that this would build group consciousness, and move one from being a single victim to being a member of a larger struggle 31. However, a social structural model would argue that recognizing one's distance from the majority society is not essentially an empowering experience, and quite the reverse, may serve to increase hopelessness and anomie.

Anomie has been conceptualized as a characteristic of societies 32 as well as individuals 33, and is a loss of orientation or norms, accompanied by, at the individual level, feelings of emptiness, meaninglessness and apathy. Merton [33, p.230] described the "social typography of anomie" as "the structural places in American society... where the disjunction between the cultural values enjoining people to aim for certain goals and the patterned possibilities for living up to these values is at a maximum."

The relevance of this theory for understanding health-related behaviors such as preventive screening in disadvantaged populations is striking. Older low-income African-American women are often described to be at "quadruple jeopardy" of negative health or social outcomes due to four types of disadvantage – race, age, gender and social class disparities 3435. Yet from a sociological perspective these women are seldom described at being at risk for anomie, and disengagement from the social structures of their communities. Indeed, they are often stereotyped as singularly positive social resources within their weakened families, institutions, and communities 36. How can we interpret this seeming hardiness against the negative effects of disadvantage, especially as it may relate to engagement in health-related preventive behaviors?

Clark 37 uses a stress-coping model to argue that the psychological effects of discrimination can only occur when individuals recognize and respond to discrimination as a threat. Thus appraisal is a critical element in producing a range of responses, from maladaptive to adaptive. Some research suggests that the negative health consequences are greatest when discrimination is perceived and active steps are taken to combat it with limited resources, a psychological phenomenon labeled John Henryism 38. Other researchers have found that not challenging situations of perceived discrimination was related to negative health outcomes ranging from increased blood pressure 39 to reduced access to health procedures 40. It has been argued that both denial and overreaction to discrimination can be harmful; Krieger 17 offers evidence that both those who cannot identify any such experiences, and those who identify many, are more at risk for hypertension. However, Broman 41 found no relationship between racism and hypertension, and Jackson et al.'s 42 longitudinal study of reported racism found both positive and negative health effects. Overall, there is stronger evidence of a connection between discrimination and adverse mental health outcomes 434445 such as depression 46 as well maladaptive behaviors including smoking 47, alcohol use 48, and violence 49.

In addition to the effects of discrimination in all aspects of life, discrimination specifically within the health care setting has warranted special focus as a proximal and powerful influence on health-related behaviors. Disadvantage in the quality of medical care received covers both process and outcome, and both are seen as distinct, yet interrelated.

A substantial body of evidence demonstrates that disadvantage exists in the receipt and quality of medical services, and that this directly influences the health of many groups in society 750. Equally well studied is disadvantage in the medical care

process

When studying racial and gender effects in medical care it is often argued that racial concordance between individual provider and patient can improve patient experiences for ethnic minority patients 5859. There is at least cross-sectional evidence that patients fearing discrimination are more likely to prefer same race providers 5960, and that among patients preferring this, racial concordance leads to greater patient satisfaction 61. However, there is also evidence that minority patients have fewer choices in medical care, and that minority providers have fewer choices of practice settings. Furthermore, even when there is racial concordance, if many minority physicians practice in, and many minority patients receive care from, lower resource medical environments, can cultural understanding, despite its importance, replace the material resources needed for high quality healthcare?

Although there are ample reasons for addressing historical inequalities of access in medical training and employment, the argument that a goal of patient-provider racial, cultural or gender concordance is, in and of itself, a solution to inequalities in health should be made cautiously. To do otherwise is to ignore the multiple pathways leading to these inequalities.

There is only a sparse literature, with varying measures, which directly examines the relationship between perceived racism, either globally or specifically within healthcare, and uptake of preventive services, especially specific to cancer screening. Structured reviews of the literature on disparities in colorectal 62, cervical 63, and breast 64 cancer found no studies examining the role of racism in relation to screening. Using 2001 national survey data, Blanchard 65 found mixed results, in that respondents believing they had been treated unfairly because of race were more likely to have optimal cancer screening, equally likely to report an exam within the past year, but were less likely to have optimal chronic disease screening, to follow doctors advice, and were more likely to delay care. Trivedi 66 found in the California Health Interview Survey that perceived discrimination in receipt of recent health care attributed to any reason (age, race, language, disability, insurance status, weight, income, gender or medical beliefs) was significantly predictive of lower rates of flu shots, hemoglobin A1c and cholesterol testing, and foot exam, but not prostate specific antigen (PSA) testing, or aspirin use.

Data used in these analyses come from a multi-year National Cancer Institute-funded study of breast cancer screening among African-American women in Baltimore, Maryland, a large US city. Methods and related findings have been previously published 2676869 and will be briefly described here. With the original purpose of evaluating the impact of a no-cost screening intervention within communities at risk for poor screening, we recruited all screening program participants age 45 and older residing in the 10 contiguous zipcodes of East Baltimore which served as the target catchment area of this program. This area, comprising roughly 40% of the City, contains both working class and extremely low income areas. We also recruited an age (± 5 years) and neighborhood-matched sample of participant-nominated friends and neighbors not attending the program. The 90-minute, in-home audiotaped interview was conducted by African-American female interviewers. During 1997 and 1998, we interviewed 576 women between the ages of 45 and 93, representing response rates of 83% and 86% from the clinic and nominated control sampling frames respectively. All participants provided written informed consent, and received $25 for participation. The study was approved by the Johns Hopkins Medical Institutions institutional review board.

The original case-control design was chosen to evaluate the impact of the screening program 2. In addition, we geocoded all respondents by residential address; our comparison of respondents, using U.S. Census blockgroup-level data70, to the sociodemographic characteristics of women in their neighborhoods supports analysis of the total group as a representative population of low and moderate income urban African-American women in East Baltimore, for questions not specifically related to the no-cost program 676869.

In order to examine cancer and health in the context of older African-American women's lives, we developed our survey instrument using both open-ended questions eliciting each woman's views in her own words, as well as structured measures, chosen from our own or others' work in African-American and women's health.

In these analyses, we use nine psychosocial covariates, including three sociodemographic measures: age, years of formal education, and self-reported household income, and two measures of physical and mental health status: each woman's rating of her health, and her responses on an abbreviated version of the CES-D 71 to measure depressive symptoms during the past week (Cronbach's Alpha = .82) 72. In these analyses, we also incorporate four types of social connectedness: whether or not respondents currently worked, were homeowners, attended weekly religious activities, or were active in community events.

For these analyses, we used several measures to explore multiple aspects of the respondents' perspectives and experiences with power, both on a societal level and also within the health care system. We used two items which measured perspectives, conceptualized as shaped by but distinct from a woman's own experiences. A five item version of Scrole's 73 scale of anomie measured generalized hopelessness (Cronbach's Alpha = .74). We used thirteen items from Green's 74 Perceptions of Racism Scale to capture views on inequities facing African-American women in various areas of American society, including general racism (2 items), medical care (6), courts and government (2), job-seeking (1), education (1), and social class (1), measuring four levels of respondent agreement or disagreement with statements such as "Judges are harder on African-Americans than whites." (Cronbach's Alpha = .79).

Krieger et al.'s well-validated measures 75 were used to capture personal experience with, and response to, racial discrimination. Respondents were asked, when faced with unfair treatment, whether they generally "accepted it as a fact of life" or "tried to do something about it." Similarly, they were asked if they generally "talked to other people" about such experiences or "kept it to yourself." They were then asked if they had ever experienced "discrimination, been prevented from doing something, been hassled or made to feel inferior because of your race or color" in each of six types of settings (school, job hiring, work, housing, medical care, police/courts). We created a single dichotomous item indicating any experience of perceived racism, and two dichotomous possible types of reactions: talking to others, and trying to do something. To distinguish between measures, we label Green's Perceptions of Racism Scale as "Societal Racism" and responses to Krieger's measure of experiences of perceived discrimination "due to race or color" as "Reported (or perceived) Racism".

We used two questions to create a four-category measure of whether or not the patient currently received medical care from a provider with whose race she felt comfortable. In a likert scale, we asked patients whether they strongly agreed, somewhat agreed, somewhat disagreed, or strongly disagreed with the statement, "I would be more comfortable seeing a doctor who was African-American than a doctor of another race." Elsewhere, we asked respondents whether their current primary provider was African-American. Women were grouped according to whether they had a primary provider who was African-American or not, and whether they agreed that they would be more comfortable with an African-American provider.

To specifically measure fear of deception in medical care, we asked the following: "Some people are afraid of being treated at big research hospitals like Johns Hopkins, because they are afraid they might be part of a research experiment without knowing it. Would you be concerned about that?"

The focus of this investigation is attitudes and beliefs about the secondary prevention of cancer, rather than actual behaviors. In these data, consistent with existing literature, we have found that a woman's actual receipt of screening is influenced by many facilitators and barriers in addition to attitudes, including access to care, costs, and physician recommendation 26768. For these reasons, in this analysis, we chose to examine screening-related motivation, an important psychological component of health behavior in its own right, rather than the respondents' actual patterns of screening.

We operationalized our outcome variable as an index (appendix), summing respondents' answers to eleven questionnaire items regarding breast cancer and screening. We theorize that women with high scores on this index had an understanding of breast cancer and mammography compatible with cancer control strategies promulgated by the medical community, as well as willingness to use the majority culture medical system as a partner in managing their health. This index had a Cronbach's alpha of .71, indicating moderate reliability consistent with its use in this type of exploratory analysis 72.

Consistent with the strong literature demonstrating the link between prevention attitudes and behaviors, we found that these attitudes were indeed predictive of mammography behaviors. In testing the construct validity of this measure, we found it to be significantly and positively correlated with both time since last mammography and intention to receive future mammography.

We were interested first in understanding the prevalence of the experiences and perspectives of interest in our study population, and also how these experiences and perspectives varied in different subgroups of our population. We conducted a bivariate analysis to examine relationships between our nine psychosocial characteristics of interest, and our measures of attitudes, experiences and screening index scores. In Tables 1 and 2, we report means and t-tests for continuous measures, and Chi Square statistic for categorical measures. In Table 3, we report the pairwise associations between attitudes, experiences, and screening motivation index scores, using Pearson correlation coefficients.

In Table 4, we use multivariate linear regression to examine our outcome of interest, a positive attitude towards mammography, in relation to psychosocial characteristics, perspectives and experiences. We present two multivariate models: a full model, including all independent predictors, and a final most parsimonious model, including only those variables significant at the p < .05 level, using backwards elimination. For the multivariate analysis, we standardized our continuous measures of age, years of formal schooling, CES-D score, anomie score, and societal racism score, by centering at the population mean, and dividing by the standard deviation. (Such arithmetic operations do not change relationships for main effects, but allow for the interpretation of interaction terms at values relevant in the population, such as the mean, rather than extreme values 76). To examine modifying effects, after building the most parsimonious model of main effects, we tested whether model fit was improved by adding, one at a time, relevant two-way interaction terms of psychosocial characteristics, attitudes and experiences. We tested whether the effects of anomie, reported racism, societal racism, talking to others when experiencing discrimination, and physician race preference varied significantly by age, education level, or depressive symptoms.

We used mediational analysis 7778 to explore further the mediating effects of worldviews and interpretations on the relationship between reported racism and screening motivation, as theorized in our model in Figure 1. To explore the relationships on the left side of the model, between reported racism and worldview, we first used simple linear regression to estimate the relationship between reported racism and screening motivation, as well as reported racism and each of two potential mediating variables: societal racism and the respondent's reported typical response when experiencing racism (talking to another versus keeping it to herself). Next, we modeled two independent variable linear regression equations, predicting screening motivation from both reported racism and these two potential mediators.

To explore relationships on the right side of the model, between world views and more proximal attitudes about medical care, we first used simple linear regression to estimate the relationship between anomie and screening motivation, as well as anomie and preference for a Black medical provider. Next, we modeled a two independent variable linear regression equation, predicting screening motivation from both anomie and preference for a Black provider.

The results of these analyses are displayed in Figure 2. SPSS statistical software 79 was used for all analyses.

Table 1 results illustrate both the social diversity of this population of older low income women, and also the relationship of these diverse attributes to life perspectives and experiences. Study participants ranged in age from 45 to 93, with 62% younger than 65. Years of formal schooling ranged from only 3 to over 20, with 56% having fewer than twelve years. (In Maryland and across the southern US prior to integration of schools, few African-American high schools offered a twelfth grade.) Forty-eight percent resided in households with incomes of less than $10,000 per year (compared to only 9.5% of all US households in 1999 80).

In terms of social spheres, all worked at some time in their lives, with 41% employed at the time of the study. Fifty-six percent lived in a home owned by themselves or a family member. Almost two-thirds attended religious services or activities once a week or more, and 49% had been involved in community affairs in the past year.

Anomie scores were fairly high overall in this group (mean = 15.5, index range 5–20), indicating a strong level of underlying hopelessness. However, as Durkheimian theory would predict, significantly lower levels of anomie were found among women with comparatively greater social, psychological, and physical resources. Lower rates of anomie were reported by those with more years of education and having more income. In addition, better physical and mental health were also related to lower rates of anomie. All four measures of social connectedness – homeownership, employment, religious and community involvement – were also significantly associated with lower rates of social alienation.

Similarly, this population in aggregate sees a significant level of societal racism against African-American women (mean = 35.1, range 13–52). However, in contrast to anomie, there was little variation by social characteristics. Women not working outside the home were less likely to view U.S. society as discriminatory against African-American women, perhaps identifying a group somewhat protected from public interactions, with less opportunity to observe societal racism.

Regarding their own experiences, 55% of respondents report perceived racial discrimination. Younger women were significantly more likely than older women to report perceived racism, as were more educated women. Reported racism was positively associated with higher levels of community involvement.

In reaction to unfair experiences, the great majority of women (86%) reported that they talk to others, rather than keeping it to themselves, while fewer (65%) reported trying to do something, rather than accepting it as a fact of life. These two strategies were more prevalent among groups with greater personal and social resources – younger women, women with more years of schooling, those with better health or higher incomes, more frequent church attendance, and community involvement.

Despite their low resources, respondents mirrored national patterns for older women in that they were, for the most part, consumers of medical care: 91% reported having a regular source of care, and 78% had some form of health insurance (data not shown). In Table 2, we see that experiences related to health care varied by psychosocial characteristics.

First we examined racial concordance with current medical provider, as well as comfort level with African-American versus other race physicians. Overall, 32% of respondents agreed with the statement that they would be more comfortable with an African-American doctor. Explanatory audiotaped comments included both rejection of race preference – "A good doctor is a good doctor" – as well as cultural preferences taking precedence over race – "He does not have to be African-American, just so long as he is some kind of American." (In comparison, 53% of respondents agreed that they would feel more comfortable seeing a woman physician than a man.)

However, only 24% of respondents reported having a primary care provider who was African-American. (The remaining 76% represent 67% whose primary care providers were not African-American and 9% who reported not having one usual source of primary care). Having a black provider was more common among women who expressed greater comfort with same-race providers (34%) than among those who said they did not agree with the statement (19%), although in these cross-sectional data, we cannot assess whether comfort level preceded, and possibly influenced provider choice, or vice versa.

These patterns of comfort and actual provider race varied by respondent age, work status, income, and CES-D symptoms. Younger, better educated, higher income, employed, or less depressed women were less likely to express provider race preference than older, less educated, non-working, poorer, or more depressed women, who were especially likely to not have a black provider, but wish for one.

The data reveal evidence of mistrust of at least some of the health care institutions within their communities. Fifty-nine percent of the respondents would be concerned about receiving care from research institutions, for fear of being deceived about research involvement. The only women with significantly greater fear were the less educated. However, it is fair to say that this fear was common, as there is no subgroup category in which the majority of respondents did not express this concern.

Finally, in Table 2, we examined the average score on the motivation for screening index among subgroups of respondents (mean score = 31.2, standard deviation = 5.5). As predicted, groups with higher motivation to be screened on a regular basis for breast cancer included younger, better educated, and wealthier women, as well as those in better physical and mental health. Additionally, working women, homeowners, and those who were involved in their communities were also more motivated to be screened. Religious participation was not associated with screening motivation in the bivariate analysis, perhaps due to greater religious involvement among older women.

In Table 3, results indicate that these experiences and perspectives did not represent a single phenomenon, and were differentially held by subgroups within the survey population, as Tables 1 and 2 suggested. Racial awareness appears to have taken several forms in this population. Perceived powerlessness, as measured by anomie, was weakly associated with preferring an African-American physician (r = .16, p < .001), and fearing research-related victimization at large hospitals (r = .13, p < .001). However, anomie was not significantly related to either societal racism (r = .05, p = .23), or to reported perceived racism (r = -.04, p = .38). As might be anticipated, it was weakly negatively related to talking about (r = -.13, p < .001) or taking action against discrimination (r = -.17, p < .001).

Views on societal racism against African-American women were modestly correlated with reports of one's own experiences of racism (r = .25, p < .001), and to a lesser degree with talking about racism (r = .09, p < .02), wanting an African-American provider (r = .09, p = .04), and fearing research (r = .08, p = .06). Interestingly, reporting perceived racism was not related to provider preference or research fears; rather it was the general coping strategies a woman says she typically takes that predicts her views on medical care. Those who talk to others or take action when experiencing racism were less likely to express preference for African-American providers (r = -.12, p < .01).

The last column in table 3 describes the correlations between the eight measures of perspectives and experiences and scores on the screening motivation index. On a bivariate level, anomie and greater comfort with an African-American provider have moderately negative correlation with screening motivation (r = -.39, p < .001); more modest, but still statistically significant negative correlations are seen between fear of research (r = -.16, p < .001) and currently having an African-American provider (r= -.10, p < .02). Positive correlations with screening motivation are seen with societal racism (r = .12, p = .004), reported perceived racism (r = .12, p < .006), talking about (r = .30, p < .001) and doing something about racism (r = .24, p < .001).