Based on our findings, we next created instruments to measure the four components of stigma (enacted, vicarious, felt normative, and internalized) in order to test hypothesized associations in our theoretical model (Nyblade & MacQuarrie, 2006
). We anticipated that psychological distress (e.g. depression) would be associated with higher levels of enacted, felt normative, and internalized stigma. However, we predicted that the associations between depression and intrapersonal forms of stigma (felt normative, internalized) would be mediated by the use of disclosure-avoidance strategies that limit individuals' access to social support (Hays et al., 1993
; Simbayi et al., 2007
). Finally, we anticipated that both enacted stigma and the newest framework component, vicarious stigma, would influence perceptions about the prevalence of stigmatizing attitudes in the community (felt normative stigma).
Participants were recruited as part of a larger study of adherence to antiretroviral therapy (ART). All participants were 18 years of age or older, HIV-seropositive, and had taken ART medications for at least one month. Most respondents were referred via the outpatient general medicine department of a large, urban, private hospital in southern India. We intentionally sought to over-represent women (who comprised only 10% of HIV patients at the hospital) to enhance our likelihood of detecting gender differences. A screener in the medical records department flagged the charts of potentially eligible individuals, who were then invited by the clinic charge nurse to meet with a project interviewer while waiting for their appointment. The remaining patients were informed of the study by their physician or by a service provider at local non-governmental agencies (NGO) that served HIV-infected individuals. All referred individuals met privately with a study interviewer, who described the research purpose and obtained informed consent.
Participants completed an interviewer-administered, baseline survey on the day of their enrollment into the study. It lasted one hour and assessed a variety of topics, including demographics and health history, HIV stigma, and psychological well-being. Surveys were completed in Kannada, Telugu, Tamil, and English. The instruments were developed in English and then translated into the three Indian languages. All translations were independently back-translated into English to ensure semantic equivalence.
An initial pool of items for the HIV stigma instruments was drawn from anecdotes provided by participants in Study 1, findings from prior work in India (Bharat, Aggleton, & Tyrer, 2001
), and, when appropriate, existing measures (e.g., Berger, Ferrans, & Lashley, 2001
; Herek, 2006
). In order to reduce respondent fatigue and interview time in future research, we intended to use an initial pool of items to identify a smaller set of items to be used in the final version of each scale. The complete sets of items used in all stigma instruments can be found in the appendices (available from the authors on request).
An initial pool of 21 items was assembled to assess if participants had experienced specific discriminatory acts because of their HIV infection. Responses were no (0) or yes (1). Because we expected the incidence of enacted stigma to be fairly low and did not predict that different experiences of enacted stigma would necessarily be correlated, we did not conceptualize our measure of enacted stigma as a scale per se. Rather, it was meant to be an index, capturing the kinds of discriminatory events that people experienced. As such, we initially inspected the outcomes on this measure by looking at the distribution. The 21 enacted stigma items differed widely in terms of the frequency with which they were endorsed, ranging from 0.4% to 26.2% of the sample. We selected 10 items that included both severe and mild enactments and that varied in terms of the frequency of endorsement.
Total scores on the brief and longer versions of the instrument were highly correlated, r (229) = .92, p < .01. Furthermore, the results of all analyses were identical, regardless of which version was used. These findings suggest that the shorter version is representative of the enacted stigma experiences measured by the longer version. The items for this brief version are displayed in . The appendices display the original pool of items and the frequency with which each item was endorsed by both male and female participants.
Items Selected for Use in the Stigma Assessments
An initial pool of 20 items captured the frequency with which participants had heard stories about people being mistreated because of HIV infection. For each item, the participant responded on a 4-point scale that ranged from 0 (never) to 3 (frequently).
Felt Normative Stigma
An initial pool of 15 items assessed participants' perceptions of the prevalence of HIV stigmatizing attitudes. They reported their expectations of how many people in their community would engage in a discriminatory behavior or would endorse a stigmatizing belief. Responses were given on a 4-point scale ranging from 0 (no one) to 3 (most people).
Using parallel versions of the 15 felt normative stigma items, participants reported the extent to which they believed that, as HIV-infected people, they should be treated in a discriminatory manner or be a target of stigmatizing beliefs. Responses were given on a 4-point scale running from 0 (not at all) to 3 (a great deal).
Psychometric reduction of Stigma measures
For vicarious, felt normative, and internalized stigma, we entered items from each scale into an exploratory principal axis factor analysis with an oblique (promax) rotation. Initially, we permitted an analysis to return all factors with eigenvalues above 1.0. We then inspected the scree plot and factor loadings to determine if the full solution was interpretable or if it was advisable to limit the number of factors. As appropriate, we ran subsequent analyses specifying the number of factors to be extracted; 2-factor solutions emerged, with the first factor accounting for most of the explained variance. However, for all three sets of items, the two factors were highly correlated (vicarious: r = .63; felt normative: r = .78; internalized: r = .43), suggesting that, in each case, they were facets of the same stigma construct and could be appropriately combined into a single unidimensional scale. Therefore, for each scale, we selected five of the highest loading items from each of the original factors to create 10-item measures. To ensure equivalence between felt normative and internalized stigma, we selected parallel items. The final 10-item versions of the scales all demonstrated acceptable reliability (Vicarious Stigma Scale: α = .88; Felt Normative Stigma Scale: α = .94; Internalized Stigma Scale: α = .83). provides the final set of items in each scale. The appendices display the original multi-factor solutions and the final single factor solutions.
Other Assessments Disclosure Avoidance
Study 1 findings confirmed our expectation that felt normative stigma motivates many PLWH to limit their disclosure of HIV. Thus, we developed a 14-item measure assessing the use of strategies to avoid revealing that one has HIV. Examples included hiding medications, describing one's illness as tuberculosis, and seeking care away from one's local village. For each item, participants indicated the frequency with which they employed a disclosure-avoidance technique using a 4-point scale ranging from 0 (never) to 3 (often). Scores were derived by averaging responses to the questions (Cronbach's α= .82). The scale's items are provided in the appendices.
Beck Depression Inventory
To assess psychological distress, which we predicted would result from stigma and use of disclosure-avoidance techniques, we included a variant of the Beck Depression Inventory (Version I) (BDI) that had been validated previously in Southern India (Chandra, Gandhi, Satishchandra, Kamat, Desai, Ravi et al., 2006
). The scale was available in Kannada. We translated it into Tamil and Telugu, and then translated all Indian language versions into English to ensure semantic equivalence. The Indian version of the BDI uses the same items as found in the United States. However, some wording is modified to reflect local cultural norms or to make statements understandable in Indian languages. The scale had acceptable reliability in our sample (Cronbach's α= .90).
Visibility of HIV Status
We opted to include an additional measure that captured the degree to which a person's HIV status is known to others. This concept is not represented formally in our model () because disclosure avoidance is thought to be potentially harmful regardless of its success at helping a person pass as a member of the nonstigmatized majority (e.g, Scambler, 1989
). Nonetheless, in spite of disclosure avoidance, a substantial number of participants in Study 1 reported that their HIV status had been disclosed to others without permission.
In our measure, participants were asked to identify (by name or pseudonym) the people to whom they turned for emotional support and assistance, and to indicate how many of these people knew of their serostatus, regardless of the source. We computed the ratio of the number of people who knew of the person's HIV status to the total number of people she or he listed as sources of support. Scores could range from 0 (none of support network knows of infection) to 1 (100% of support network knows of infection).
Participant demographic and health history characteristics
Individuals were asked to describe their gender, age, marital status, household composition, employment status, monthly income, years of education, religion, and length of time since HIV diagnosis.
Composite scores for all stigma measures were correlated with psychological and behavioral outcomes. To examine the potential for mediation among our measures, we regressed a presumed mediator onto the respective predictor variable, and the outcome variable onto a predictor variable both when the mediator was included in and excluded from the model (Baron & Kenny, 1986
). Analyses were conducted using SPSS 13.0
The sample consisted of 229 respondents. A majority were male, married, and Hindu (). Most had obtained 10 or fewer years of education and reported earnings of 3000 or fewer rupees per month, indicative of low socioeconomic status. Participants reported living with a variety of relatives, including spouses (n = 160; 70%), children (n = 161; 70%), parents (n = 81, 35%), siblings (n = 54; 23%), and other extended family (n = 45; 20%). A majority could read Kannada (n = 161; 70%), whereas smaller percentages could read English (n = 110; 48%), Hindi (n = 56; 25%), Telegu (n = 52; 23%), and Tamil (n = 28; 12%). Participants had been diagnosed with HIV for an average of 3.85 years (range: 0.13-17.22 years). Although we do not have information on individuals who refused participation, the age and religious profile of our sample was similar to the profile of all patients in the HIV clinic at the hospital where the study was conducted. The sample's representation of women was higher (31% female vs. 10% female among all HIV patients at the hospital), a reflection of our intentional efforts to oversample females.
Demographic profile of the cohort participants
Mean Levels of Stigma
Reports of enacted stigma on the 10 item scale were low (M = 0.73, SD = 1.41). Seventy-one percent of participants (n = 163) did not report any instance of discrimination as a result of their HIV status. Vicarious, felt normative, and internalized stigma scores were computed by summing the responses and dividing by the number of items. Scores could range between 0 and 3, with higher scores indicating greater levels of stigma. Participants reported more felt normative stigma (M = 1.19, SD = 1.00) than internalized stigma (M = 0.60, SD = 0.64; Z = 7.63, p < .001) or vicarious stigma (M = 0.82, SD = 0.72; Z = 5.75, p < .001).
There were a limited number of significant correlations between the stigma instruments and demographic characteristics. Women had experienced more vicarious stigma than men: Women: M = 0.99, SD = 0.84; Men: M = 0.75, SD = 0.65; F(1, 227) = 5.33, p < .05, η2 = .02. However men had higher scores than women on internalized stigma: Men: M = 0.65, SD = 0.65; Women: M = 0.47, SD = 0.57; F(1, 227) = 4.30, p < .05, η2 = .02. There were no gender differences for enacted or felt normative stigma. Having been diagnosed with HIV for longer periods of time was correlated with higher scores for vicarious stigma, r (229) = .25, p < .01, but lower scores for internalized stigma, r (229) = −.13, p < .05. And increasing age was negatively associated with vicarious stigma, r (229) = −.20, p < .01.
Association of HIV Stigma with Disclosure, Disclosure Avoidance, and Depression
displays correlations among the constructs in our model (), i.e., the four stigma instruments and the measures of avoidant coping (M = 0.82, SD = 0.59) and depression (M = 12.69, SD = 11.42). Greater levels of depression were associated with felt normative, internalized, and enacted stigma, but not with vicarious stigma. Furthermore, the use of disclosure avoidance techniques was associated positively with all forms of stigma, as well as with depression. This pattern of associations did not differ by gender, nor were there gender differences in the overall reported levels of disclosure-avoidance and depression.
Correlations Among Stigma Instruments, HIV Serostatus Disclosure Avoidance, and Depression
Mediation Among Stigma, Disclosure Avoidance, and Depression
The conceptual model depicted in predicted that several of the observed correlations (specifically, the associations between vicarious stigma and disclosure avoidance, enacted stigma and disclosure avoidance, felt normative stigma and depression, and internalized stigma and depression) would be partially or fully mediated by associations with other variables. To test for the possibility of mediation, we looked to see if the size of the beta-weight for a predictor variable was reduced significantly by including a mediator in a regression model (Baron & Kenny, 1986
). diagrams four successful tests of mediation. In each depiction, the beta weights represented along the bottom (horizontal) arrow are critical. The first number is the beta weight of the predictor without the mediator in the model. The second number (in parentheses) is the reduced weight for the predictor with the mediator included in the model. The Sobel test identifies if this reduction in the beta weight size is significant.
Mediation Analyses Examining Associations Among Stigma, Disclosure Avoidance, and Depression
For the correlation between vicarious stigma and disclosure avoidance, we predicted that the association would be mediated by felt normative stigma. As seen in , Cell A, this mediated model was fully supported. The association between vicarious stigma and disclosure avoidance was reduced to near 0 when felt normative stigma was included in the regression. Similarly, as seen in , Cell B, felt normative stigma fully mediated the association between enacted stigma and disclosure avoidance, consistent with our predictions in . We next examined whether disclosure avoidance mediated the associations between felt normative stigma and depression (, Cell C) and between internalized stigma and depression (, Cell D). The regression weight for felt normative stigma dropped to non-significant levels when disclosure avoidance was included in the model, suggestive of full mediation as predicted by . Similarly, the correlation between internalized stigma and depression was reduced significantly with the inclusion of disclosure avoidance in a regression model. However, internalized stigma still remained a significant unique predictor of depression. This pattern of results suggests that disclosure avoidance only partially mediates the association between internalized stigma and depression, which is consistent with .
HIV Serostatus Visibility
Although not formally diagrammed in our model, we included a measure of HIV serostatus visibility (M = 0.44, SD = 0.40; median number of individuals who had knowledge of a participant's serostatus = 4) in order to examine the degree to which disclosure avoidance strategies were successful. Interestingly, there were notable gender differences in the average levels of HIV status visibility (men: M = 0.38, SD = .39; women: M = 0.59, SD = 0.38; F(1, 223) = 13.65, p < .001, η2 = .06) and in how visibility correlated with disclosure avoidance and with stigma. For men, higher serostatus visibility was significantly associated with lower levels of disclosure avoidance, r (158) = −.22, p < .01, but was uncorrelated with reports of enacted, r (158) = .07, p = .33; vicarious r (158) = .05, p = .50; or felt normative stigma, r (158) = .05, p =.46. For women, however, the pattern of results was almost exactly opposite. Greater serostatus visibility was unrelated to disclosure avoidance, r (67) = −.01, p = .95, but was significantly correlated with enacted, r (67) = .24, p = .05 and vicarious stigma, r (67) = .35, p < .01. Furthermore, although the association between serostatus visibility and felt normative stigma was not significant among women, (r (67) = .16, p = .19), the size of correlation was substantially greater than that observed among men and would likely be significant with a larger sample of women. This pattern of results suggests that nondisclosure efforts may be more successful for men than women in hiding HIV infection.
We developed four new instruments to assess different facets of HIV stigma—enacted, vicarious, felt normative, and internalized. All demonstrated good psychometric properties and proved useful in understanding experiences and beliefs in the Indian context. Participants reported very few enactments of stigma. Rather, they were more likely to perceive high levels of HIV stigma in the community (felt normative stigma), and to personally hold mixed views about the validity of this stigma (internalized stigma). Because stigma enactments were low, felt normative stigma was driven most strongly by stories that participant had heard of other PLWH experiencing discrimination. The addition of vicarious stigma to our framework was critical. It helped explain the observed pattern of results by documenting a channel for the communication of HIV information. Participants did not need to encounter actual enactments of stigma to fear its social consequences. Stories communicated norms about HIV and shaped behavior.
Although striking, the low levels of enacted stigma and the relatively higher levels of felt normative stigma are consistent with prior research. Early stigma theories classified scorned conditions into those that create discredited
identities (Goffman, 1963
). The distinction focuses on a condition's visibility. Those that cannot be hidden create discredited identities, in which a person's inferior status is known in all social situations and the individual must develop strategies for coping with the resulting prejudice and discrimination. By contrast, conditions that can be hidden create discreditable identities. In this case, the fundamental challenge becomes the management of stigma information. A person with a discreditable condition tries to control who does and does not know about a scorned condition to minimize the likelihood of being assigned a discredited status.
Our data clearly align with the latter category. Participants experienced low levels of enacted stigma, which is to be expected given that HIV is often not readily apparent (unless a person develops symptoms stereotypically associated with the disease). By contrast, participants were more acutely aware of the potential for stigma and employed strategies such as disclosure avoidance, that they thought would help protect against being discredited (i.e., stigmatized). Unfortunately, these strategies come with a cost. In our second study, greater efforts to limit disclosure of HIV status were associated with higher levels of depression. In addition, disclosure-avoidance techniques did not always accomplish their intended goal. For women specifically, disclosure avoidance had no relationship to the proportion of people in their social networks who knew of their infections. This observation may reflect how gender influences HIV care. Prior research has shown that the consequences of HIV stigma are particularly harsh for women, including rejection from husbands' families and fewer financial resources for managing the disease (Pallikadavath et al., 2005
). These consequences may place women at the mercy of extended relatives or public assistance programs, making it more challenging to maintain medical privacy.
By design, our felt normative and internalized stigma instruments paralleled each other. Although responses on these measures were correlated, it was clear that personal endorsements of stigmatizing beliefs were driven only partially by perceptions of stigma in the local community. The finding reinforces the importance of keeping perceptions of stigma levels (felt normative) and endorsement of stigma (internalized) conceptually distinct (Herek, 2008
). Furthermore, the consequences of felt normative and internalized stigma are somewhat different. Perception of high levels of stigma was linked to depression exclusively through its influence on disclosure-avoidance behaviors. By contrast, internalized stigma directly correlated with depression, even while also influencing disclosure-avoidance behaviors. Our data do not allow us to understand why
some people believe HIV stigma is valid whereas others do not. But, given the strong link with depression, it is clear that a major need for people with high internalized stigma is helping them challenge their negative beliefs about the disease. It also points to the importance of reducing negative stereotyping in popular media, which could reinforce internalized stigmatizing beliefs.
Our findings offer several potential avenues for developing stigma interventions in India. First, the critical mediating role of disclosure-avoidance indicates that programs are needed to help participants disclose their infection. Interestingly, the link between disclosure-avoidance and depression held true for men and women, even though there were gender differences in the effectiveness of the strategies at limiting knowledge of a participant's serostatus. This suggests that it may be the act itself (and not the success of its intended outcome) that makes disclosure avoidance psychologically damaging. Indeed, prior research has shown that disclosure of one's HIV status sometime has direct health benefits (Cole, Kemeny, Fahey, Zack, & Naliboff, 2003
; Cole, Kemeny, Taylor, Visscher, & Fahey, 1996
; Smyth & Pennebaker, 2001
; Strachan, Bennett, Russo, & Roy-Byrne, 2007
). Thus, a critical first step may be providing opportunities for disclosure, even if it is to people who do not provide critical forms of social support in daily life.
Establishing opportunities for disclosure and social support—even ones as straightforward as hospital-based social groups or peer counselors—holds significant challenges. First, hospitals that provide HIV care attract patients from large geographical areas. PLWH from rural areas often travel long distances to receive treatment. As such, their access to support services may be limited, particularly if they rely on family for transportation and housing when seeking care. Second, PLWH may not always be able to communicate with one another. In southern India alone, there are four major languages (in addition to English). If people are unwilling to seek services locally, one must implement programs in urban centers where PLWH do not all speak a common language. Fortunately, many Indian NGOs already are tackling these and similar challenges as they develop services for PLWH.
A second area for intervention development is challenging PLWH's own stigmatizing beliefs about the disease. How this is best accomplished remains an area in need of urgent study. Prior research has suggested that accurate knowledge about the disease is a major determinant of attitudes (Ambati, Ambati, & Rao, 1997
), highlighting a potential role for education. But there may also be culturally-specific beliefs that complicate education-based strategies. For example, traditional Hindu theology places importance on accepting one's fate. Endorsement of this tenet may make people hesitant to challenge beliefs about the disease.
Third, it is critical to consider how interventions can create structural change. As suggested earlier, many women may not have—by circumstances, family, or finances—the ability to exert agency over how they respond to their own infection. Social support groups and educational programs will have little influence if a person's attitudes, beliefs, and behaviors are driven primarily by environmental constraints. Finally, future study is needed to understand how interventions should address related forms of stigma, such as prejudices against sex workers, migrant laborers, injection drug users, and men who have sex with men. Our current study doesn't allow us to determine to what degree these stigma drive beliefs about HIV, nor how they might modify the kind of interventions needed for PLWH. But it is notable that many of the better-performing items in our felt normative and internalized scales related to morality and shame. This suggests that attitudes about the disease are still driven by judgments about how it is acquired.
Our findings are limited by several aspects of the study design. First, the scales were examined in a cross-sectional assessment, precluding definitive conclusions about causality. We currently are collecting longitudinal data to determine how stigma, disclosure avoidance, and psychological reactions relate over time. Second, our sample consisted of PLWH on antiretroviral treatment. As such, it over-represents individuals in later stages of the disease when immune system impairments may produce more visible symptoms. (That said, it should be noted that the primary symptom of HIV in India is tuberculosis, a condition that also occurs commonly among people uninfected with HIV. Characterizing one's HIV illnesses as being due solely to tuberculosis was one of the more common disclosure-avoidance strategies employed by our participants.) Finally, the findings may be affected by selection bias. It is possible that individuals who experienced very high levels of stigma were unwilling to participate in an HIV-specific study.
Stigma is an important consideration in the response to HIV in India. Interventions and policies are needed to ensure the PLWH have the safety and security to disclose their serostatus and to promote more positive opinions about the disease to ensure psychological well-being among people living with the disease.