Symbolic and anticipated HIV stigma are associated with mental health and education in South Africa

ABSTRACT

HIV stigma continues to act as a barrier to HIV care in South Africa, necessitating further research on the intersections of socioeconomic factors and the anticipation and expression of stigma surrounding HIV. We measured the prevalence of HIV-related stigma and evaluated factors associated with symbolic and anticipated stigma in Umlazi Township, South Africa from 2013 to 2019, using a validated HIV stigma scale, before undergoing HIV testing. Among 7,724 people evaluated, 1,318 (16.9%) reported symbolic stigma and 2,396 (30.8%) anticipated HIV stigma. Prevalence of symbolic and anticipated stigma were significantly more common among both women and people living with HIV, compared to men and those who tested negative for HIV. In multivariable analyses, higher education and depressive symptoms were the strongest correlates with both symbolic stigma and anticipated stigma. Younger age, not being married, and having a partner who was not living with HIV appeared to be important correlates with anticipated stigma, but not symbolic stigma. Overall, the anticipation of experiencing stigma because of infection with HIV continues to be an important factor in the testing and management of HIV.

KEYWORDS:

• HIV/AIDS
• HIV stigma
• anticipated stigma
• symbolic stigma

Introduction

In South Africa, an estimated 7.5 million people are living with HIV (AIDSinfo, 2021), nearly 40% of whom are worried they will experience prejudice or discrimination because of stigma surrounding their HIV status (Peltzer & Pengpid, 2019). Stigma is a complex social construct that results in devaluation and discrediting of individuals in the eyes of others (Schaay et al., 2008). This process of labeling and distinguishing individuals allows for multiple types of stigma experiences, including both experienced discrimination such as being denied housing, and stigma experiences falling outside the purview of the law such as being gossiped about (Stangl et al., 2019). Additionally, patients may experience hurtful stigmatization through non-verbal cues, looks, and changes in body language while interacting with people in their environment (Davey, 1999). Of concern is that virtually all health-related behavior of people living with HIV is influenced by stigma, from engaging in preventative measures to being tested, disclosing for treatment and support, and maintaining required adherence (Peltzer & Pengpid, 2019; Stangl et al., 2019; Visser, 2018; Wolfe et al., 2008).

Two common forms of HIV stigma include symbolic stigma anticipated stigma.

Symbolic stigma is based on individuals’ negative perceptions or moral judgements about people living with HIV (Haffejee et al., 2018; MacPherson et al., 2011; Peltzer & Pengpid, 2019; Schaay et al., 2008; Treves-Kagan et al., 2015; Wolfe et al., 2008). Endorsing the statement “A primary school pupil with HIV should not be allowed to continue going to school.” would be an example of symbolic stigma.

Anticipated stigma is the expectation of experiencing discrimination or social judgment after testing HIV positive (Haffejee et al., 2018; MacPherson et al., 2011; Peltzer & Pengpid, 2019). Manifestations of anticipated stigma include worries about “being gossiped about” and worries about “being verbally insulted/harassed/threatened” (Peltzer & Pengpid, 2019). Endorsing the statement “From what you have seen in your community, and you were HIV positive and people found out, do you think that you would be sidelined by friends?” would be an example of anticipated stigma.

Recent research on stigma has focused on how broader social and structural forces influence stigma and create intersectional stigma (Gilbert, 2016; Stangl et al., 2019). A recent study by Stangl, et al developed a framework for studying and addressing health related stigma (Stangl et al., 2019). The framework identifies “drivers” and “facilitators” of health stigma as the basis for the process of health stigma. Drivers include elements of symbolic and anticipated stigma and facilitators include factors such as cultural norms, the presence or absence of occupational health standards and health policies (Stangl et al., 2019). These drivers and facilitators lead to what the authors call “stigma marking” in which stigma is applied to groups or individuals based on health-related conditions, race, gender, sexual orientation, and other socioeconomic factors (Stangl et al., 2019). Previous studies have described socioeconomic correlates of stigma: Peltzer and Pengpid found that patients of younger age, those with a lower wealth status and those who were not part of an HIV support group were more likely to report stigma, among several other factors (Peltzer & Pengpid, 2019).

Our study evaluates factors correlated with HIV stigma and assesses the relationship between drivers of HIV stigma and socioeconomic factors that are thought to intersect with HIV stigma. Additionally, we report cross-sectional data on responses to the stigma scale over the seven years the parent study was conducted. This was an incidental finding that we found to be interesting in light of the arrival of Universal Test and Treat in 2016, a program providing access to HIV testing for all patients and treatment with antiretroviral therapy for all patients who test positive for HIV regardless of their CD4 count (Mukora et al., 2022). Among a cohort of patients presenting for HIV testing across several years in a high HIV-prevalence urban township of South Africa, we measured the point prevalence of stigma, and the sociodemographic and clinical factors correlated with symbolic and anticipated HIV stigma.

Methods

We conducted a cohort study between September 2013 and February 2019 at iThembalabantu People’s Hope Clinic in Umlazi township, Durban, South Africa. The study described in this paper is a cross-sectional analysis of the parent study at the time participants were enrolled.

Setting & participants

This site was chosen because the iThembalabantu People’s Hope clinic provides free HIV testing for approximately 10,000 people each month, and treatment at no cost for over 10,000 people living with HIV. People who presented to this clinic for HIV testing and were 18 years or older were approached and asked to participate in the study. A research assistant, trained in the administration of the stigma scale, administered the questionnaire to each participant in their preferred language (English or isiZulu) before HIV testing was performed. Exclusion criteria included pregnancy at time of testing; the parent study required the exclusion of participants with a history of anti-fungal drug use in the three months prior to the clinic visit.

Measure of stigma

Our primary outcome, stigma, was assessed using a brief, validated scale designed to measure stigma at the time of HIV testing in sub-Saharan Africa, and assesses both symbolic and anticipatory stigma (MacPherson et al., 2011). The stigma scale includes seven items assessing symbolic stigma and five items measuring anticipated stigma and asks respondents to choose whether they “totally agree”, “somewhat agree”, or “don’t agree” with twelve statements in total. Questions evaluating symbolic stigma included how much the respondent agreed with the statement “People with HIV are immoral.” and items evaluating anticipated stigma included whether the respondent agreed that “From what you have seen in your community, if you were HIV positive and people found out, do you think that you would be fired from work or lose your job?” Verbatim questions are shown in Supplementary Table 2. Responses to each question were then tallied and summed to provide an overall stigma score, which ranged from 0–24 and categorized participants into one of two categories: “no stigma” (score = 0) or “stigma present” (score ≥1). We calculated separate symbolic stigma and anticipated stigma scores for each participant. We chose to dichotomize stigma scores based on a study which used a similar stigma scale in Botswana and had a similar study population to ours (Wolfe et al., 2008) and for simplicity. Symbolic and anticipated stigma was assessed at baseline, before participants were informed of their HIV status.

Conceptual model

We used a conceptual model developed by Stangl et al. (2019) to guide our a priori choice of independent variables (Stangl et al., 2019). Applying this theory, we chose possible explanatory variables including sociodemographic indicators that have been found to be associated with stigma (Peltzer & Pengpid, 2019; Wolfe et al., 2008), and those that we hypothesized to be possible “stigma markings” or intersectional stigmas (Stangl et al., 2019) with HIV stigma.

The predictive outcomes included age (in years), sex, marital status, educational attainment (categorical), employment status (whether unemployed/employed part- or full-time at time of enrollment), monthly income (categorical), number of children, whether the participant attends church (weekly/occasionally/not at all), whether the participant has visited a traditional healer (ever), whether their partner has been tested for HIV, depressive symptoms measured with a validated survey instrument, and finally, HIV status, which was revealed to the participant following completion of baseline assessments. Sociodemographic characteristics were coded as binary questions except for age, education and partner HIV status which were coded with multiple categories as described in the tables. We used the Patient Health Questionnaire (PHQ-9) to measure depressive symptoms in participants. Scores were summed and patients are classified into five groups (0–4 = none, 5–9 = mild, 10–14 = mild – moderate, 15–19 = moderate, 20–27 = severe). Following a simplified version of the PHQ-9 scoring system, we classified participants dichotomously as having “no depressive symptoms” (score = 0–4) or “mild to severe depressive symptoms” (score = 5–27) (Kroenke et al., 2001; Kroenke & Spitzer, 2002).

Statistical analysis

We used descriptive statistics to summarize participant characteristics and estimate the point prevalence of stigma reported among participants during the study period. We used Fisher exact tests to determine if the differences in the responses between the groups (e.g., men compared to women, or those who tested HIV negative compared to those who tested HIV positive, those who reported any stigma compared to no stigma) were statistically significantly (alpha <0.05). Cronbach’s alpha was used to measure the internal consistency of the questions.

We used logistic regression modelling to identify statistically significant associations between sociodemographic characteristics and presence of self-reported stigma. Our dichotomous outcome was the presence of stigma as measured by each respective (symbolic and anticipatory) stigma scale and operationalized in the analysis with each individual reporting “no stigma” or “stigma present” as described above. If a variable within a category of variables (i.e., age 18 −25) was significantly associated with outcome (p ≤ 0.05) then all the variables in that category were included in the multivariable analysis. We used a Homer-Lemeshow goodness of fit test to assess model fit and a link test to ensure model specification. A p-value of ≤0.05 was considered a statistically significant difference, and we used Stata/IC 16.1 and for data analyses.

Ethical considerations

This study was approved by the University of Kwazulu-Natal (REF:BF052/13) and the University of Washington (STUDY00000124) ethics committees. All data was de-identified to ensure that participants remained anonymous, and all participants provided written and informed consent. All experiments were performed in accordance with relevant guidelines and regulations.

Results

Of 7,877 participants enrolled in the study, 7,724 (98.0%) completed the stigma scale (Table 1). Among those who completed the stigma questionnaire, 3,034 (39.3%) tested HIV positive. Of the 153 participants who did not complete the scale the only statistically significant difference in demographic characteristics compared to participants who completed the scale was education level, with non-respondents reporting less completed years of education (p = 0.025). The full results can be found in supplementary table 1. Overall results from the stigma scale (anticipated and symbolic stigma combined) indicated that 5,210 (67.4%) respondents reported no stigma, and 2,514 (32.5%) respondents’ surveys indicated stigma present. There were small but statistically significant differences between males and females in response to questions on four statements related to symbolic stigma (Supplementary Table II). (Table 2)

Table 1. Participant characteristics.

Variable	Respondents to Simple Stigma Scale N (%)
	N = 7,724
Sex
Male	3,932 (50.9%)
Female	3,792 (49.1%)
Age (years)
18–25	2,254 (29.2%)
26–35	3,344 (43.3%)
36–45	1,305 (16.9%)
> 45	821 (10.6%)
Marital status
Married	525 (6.8%)
Never Married (Single)	7,129 (92.3%)
Widowed/Divorced	70 (0.9%)
Education
Higher Degree (University)	308 (4.0%)
Some High School or Matric (completed High School)	6,202 (80.3%)
Primary School	126 (1.6%)
None (primary school not completed)	1,087 (14.1%)
Missing values	1 (<0.1%)
Employment status
Not employed	4,670 (60.5%)
Employed	3,054 (39.5%)
Income
< 2,000 ZAR/month (<$143 USD/month)	5,819 (75.3%)
≥ 2,000 ZAR/month (≥ $143 USD/month)	1,833 (23.7%)
Missing values	72 (0.9%)
Number of children
0	2,302 (29.8%)
≥ 1	5,382 (69.7%)
Missing values	40 (0.5%)
Church attendance
Yes	5,698 (73.8%)
No	1,536 (19.9%)
Missing values	490 (6.3%)
Visit to a traditional healer w/in last six months
Yes	479 (6.2%)
No	7,239 (93.7%)
Missing values	6 (0.1%)
Partner tested for HIV
Yes, HIV (+)	1,348 (17.4%)
Yes, HIV (-)	2,364 (30.6%)
No or unknown status	3,967 (51.4%)
Missing values	45 (0.6%)
Depressive symptoms (by PHQ-9)
None	6,547 (84.8%)
Mild-Severe	1,098 (14.2%)
Missing Values	79 (1.0%)
HIV status
HIV Negative	4,690 (60.7%)
HIV Positive	3,034 (39.3%)

ZAR = South African Rand; PHQ-9 =  The Patient Health Questionnaire

Table 2. Unadjusted and adjusted odds ratios of reported symbolic stigma by sociodemographic characteristics

	n/N (%)	Unadjusted (N = 7,724)		Adjusted (N = 7,524)
Sociodemographic characteristics		OR (CI)	P – value	aOR (CI)	P – value
Sex
Female	680/3,792 (17.9%)	1.15 (1.02–1.30)	0.020	1.11 (0.95–1.29)	0.175
Male	627/3,932 (16.0%)	1.0	–	1.0	–
Education (N = 7,723)
Higher Degree (University)	152/308 (49.4%)	7.64 (5.70–10.22)	<0.001	6.99 (4.87–10.03)	<0.001
Some High School or Matric	979/6,202 (15.8%)	1.47 (1.20–1.79)	<0.001	1.34 (1.05–1.70)	<0.001
Primary School	53/126 (42.1%)	5.69 (3.87–8.49)	<0.001	3.17 (1.89–5.33)	<0.001
None	123/1,087 (11.3%)	1.0	–	1.0	–
Employment Status
Employed	640/3,054 (21.0%)	1.59 (1.41–1.79)	<0.001	1.47 (1.22–1.77)	<0.001
Not employed	667/4,670 (14.3%)	1.0	–	1.0	–
Income (N = 7,652)
≥ 2000 ZAR/month	419/1,833 (22.9%)	1.67 (1.47–1.90)	<0.001	1.16 (0.94–1.43)	0.171
< 2000 ZAR/month	876/5,819 (15.0%)	1.0	–	1.0	–
Visit to a Traditional healer (N = 7,718)
No	1,242/7,239 (17.2%)	1.37 (1.04–1.80)	0.024	2.13 (1.52–2.98)	<0.001
Yes	63/479 (13.2%)	1.0	–	1.0	–
Partner tested for HIV? (N = 7,679)
No	743/3,967 (18.7%)	1.11 (0.94–1.30)	0.213	1.12 (0.92–1.37)	0.262
Yes, HIV (-)	328/2,364 (13.9%)	0.77 (0.64–0.93)	0.006	0.91 (0.73–1.15)	0.446
Yes, HIV (+)	232/1,348 (17.2%)	1.0	–	1.0	–
Depressive symptoms (N = 7,645)
Mild – severe	710/1,098 (64.7%)	19.30 (16.60–22.44)	<0.001	19.46 (16.58–22.85)	<0.001
None	567/6,547 (8.7%)	1.0	–	1.0	–
HIV status
HIV positive	599/3,034 (19.7%)	1.38 (1.23–1.56)	<0.001	1.02 (0.87–1.19)	0.810
HIV negative	708/4,690 (15.1%)	1.0	–	1.0	–

^a OR = odds ratio; aOR = adjusted odds ratio; CI = 95% confidence intervals. Age, marital status, children, and church attendance were not significant in the unadjusted model and were not included in the adjusted model.

Variability and reliability of the stigma scale

Cronbach’s alpha was 0.93 for the stigma scale, 0.90 for the questions about symbolic stigma and 0.86 for the questions on anticipated stigma. The Hosmer-Lemeshow goodness-of-fit test for the symbolic stigma model was a good fit (P = 0.35) and was decently fit for anticipated stigma model (P = 0.05). Correlation matrices for the individual scale items can be found in supplementary tables.

Symbolic stigma

1,307 (16.9%) of participants reported at least one expression of symbolic stigma.

The strongest correlates with expressing symbolic stigma included having depressive symptoms (odds ratio [OR] =  19.46, 95% confidence interval [CI]: 16.58–22.85) and having attended university (OR = 6.99, CI: 4.87–10.03) (Table 3). Other factors associated with symbolic stigma included employment status, having completed primary or high school compared to no formal education, and having visited a traditional healer within six months of the clinic visit (Table 3).

Table 3. Unadjusted and adjusted odds ratios of reported anticipated stigma by sociodemographic characteristics.

	n/N (%)	Unadjusted (N = 7,724)		Adjusted (N = 7,093)
Sociodemographic characteristics		OR (CI)	P – value	aOR (CI)	P – value
Sex
Female	1,224/3,792 (32.3%)	1.16 (1.05–1.28)	0.003	1.22 (1.08–1.38)	0.001
Male	1,146/3,932 (29.2%)	1.0	–	1.0	–
Age (years)
18–25	748/2,254 (33.2%)	1.57 (1.31–1.89)	<0.001	1.49 (1.16–1.91)	0.002
26–35	1,031/3,344 (30.8%)	1.41 (1.18–1.68)	<0.001	1.42 (1.13–1.80)	0.003
36–45	394/1,305 (30.2%)	1.37 (1.12–1.67)	0.002	1.33 (1.03–1.72)	0.027
> 45	197/821 (24.0%)	1.0	–	1.0	–
Marital status
Single	2,253/7,129 (31.6%)	2.01 (1.61–2.52)	<0.001	1.87 (1.40–2.51)	<0.001
Divorced/Widowed	19/70 (27.1%)	1.62 (0.92–2.87)	0.096	1.11 (0.51–2.42)	0.783
Married	98/525 (18.7%)	1.0	–	1.0	–
Education (N = 7,816)
Higher Degree (University)	193/308 (62.7%)	7.54 (5.71– 9.94)	<0.001	7.10 (5.10–9.87)	<0.001
Some High School or Matric	1,914/6,202 (30.9%)	2.00 (1.70–2.36)	<0.001	2.01 (1.65–2.45)	<0.001
Primary School	65/126 (51.6%)	4.78 (3.26–7.01)	<0.001	3.93 (2.37–6.53)	<0.001
None	198/1,087 (18.2%)	1.0	–	1.0	–
Employment
Employed	1,021/3,054 (33.4%)	1.24 (1.12–1.36)	<0.001	1.24 (1.10–1.40)	0.001
Not employed	1,349/4,670 (28.9%)	1.0	–	1.0	–
Income (N = 7,744)
≥2000ZAR/month	720/1,833 (39.3%)	1.69 (1.51–1.88)	<0.001	_b
<2000 ZAR/month	1,613/5,819 (27.7%)	1.0	–
Children (N = 7,740)
No	757/2,302 (32.9%)	1.15 (1.04–1.28)	0.009	1.05 (0.92–1.20)	0.464
Yes	1,608/5,382 (29.9%)	1.0	–	1.0	–
Church Attendance (N = 7,290)
No	519/1,536 (33.8%)	1.40 (1.24–1.58)	<0.001	1.40 (1.22–1.60)	<0.001
Yes	1,522/5,698 (26.7%)	1.0	–	1.0	–
Partner tested for HIV? (N = 7,771)
No	1,313/3,967 (33.1%)	1.51 (1.31–1.74)	<0.001	1.79 (1.50–2.12)	<0.001
Yes, HIV (-)	706/2,364 (29.9%)	1.30 (1.11–1.51)	0.001	1.65 (1.37–2.00)	<0.001
Yes, HIV (+)	333/1,348 (24.7%)	1.0	–	1.0	–
Depressive symptoms (N = 7,736)
Mild – severe	807/1,098 (73.5%)	9.16 (7.91–10.59)	<0.001	9.52 (8.02–11.29)	<0.001
None	1,522/6,547 (23.2%)	1.0	–	1.0	–
HIV status
HIV positive	988/3,034 (32.6%)	1.16 (1.05–1.28)	0.004	1.07 (0.94–1.21)	0.330
HIV negative	1,383/4,690 (29.5%)	1.0	–	1.0	–

^aOR = odds ratio; aOR = adjusted odds ratio CI = 95% confidence intervals. Visiting a traditional healer within the past six months was not significant in the unadjusted model and was not included in the adjusted model.

^b Income excluded in final model due to poor model fit.

Anticipated stigma

Nearly one-third (30.7%) of respondents anticipated a negative experience because of their HIV status if they were to test positive.

In the multivariate regression, the factors that most strongly associated with anticipatory stigma were having attended university (OR =   7.00, CI: 5.05–9.72) and having depressive symptoms (OR =   9.34, CI: 7.89–11.05) (Table 4). Other factors associated with anticipated stigma included female sex, younger age, single marital status, having a primary or high school education, current employment status, not attending church, and having a partner that either did not test for HIV or tested negative (Table 4). Income was dropped from the anticipated stigma model due to poor model fit.

Table 4. Frequency of participants reporting symbolic and anticipated stigma by year.

Year	Symbolic Stigma Reported	Anticipation of Stigma Reported
	n/N (%)	n/N (%)
2013	189/448 (42.2%)	300/448 (67.0%)
2014	692/1,426 (48.5%)	673/1,426 (52.8%)
2015	375/1,134 (33.1%)	421/1,134 (37.1%)
2016	35/1,884 (1.9%)	143/1,884 (7.6%)
2017	11/1,427 (0.8%)	250/1,427 (17.5%)
2018	5/1,245 (0.4%)	431/1,245 (34.6%)
2019	0/161 (0.0%)	72/160 (45.0%)

Discussion

In a cohort of HIV status-naïve participants in urban township of South Africa, symbolic and anticipated stigma were commonly reported, and were significantly more common among women and people living with HIV. The strongest correlates with symbolic and anticipated HIV stigma were higher education and depressive symptoms. Anticipated stigma was more commonly reported within the cohort than symbolic stigma. The point prevalence of stigma at time of enrollment within the population changed year to year with the point prevalence of symbolic stigma decreasing dramatically over the six-year study period, from 42.2% in 2013–0.4% in 2018. Conversely, the point prevalence of anticipated stigma declined by 2016, after Universal Test and Treat was introduced, and has been increasing since then.

Unfortunately, we cannot draw conclusions about trends in stigma overtime because we were only able to measure the point prevalence, however we found these results interesting in the context of increasing availability of highly active antiretroviral therapy. It has been anticipated that expanded access to anti-retroviral therapy would substantially reduce HIV-associated stigma (Gilbert, 2016; Treves-Kagan et al., 2015). While some reduction in levels of HIV stigma have been observed following expanded access to antiretroviral therapy (Treves-Kagan et al., 2015; Visser, 2018; Wolfe et al., 2008), HIV-associated stigma remains pervasive throughout sub-Saharan Africa. Furthermore, research in South Africa and other settings has found that levels of anticipated and/or perceived stigma remain high even in communities where symbolic stigma is decreasing or low (Chan & Tsai, 2016; Chan et al., 2015; Visser et al., 2009). Further research should be done to assess how treatment of HIV at no cost regardless of CD4 count impacts HIV stigma in communities.

Correlates with anticipated stigma in our study differed somewhat from previous research. Previous studies have demonstrated a positive dose response relationship between anticipated stigma and psychological distress (Parcesepe et al., 2018). Our study found that patients who were experiencing mild to severe depressive symptoms were more likely to endorse anticipated HIV stigma. However, since our study is cross-sectional in nature, we are unable to determine any directionality or causation. We found that female sex was associated with anticipated HIV stigma, which is different from previous studies in South Africa (Treves-Kagan et al., 2017; Yan, et al., 2019). We also found that being single and not attending church were associated with anticipated stigma, and while our study was not designed to evaluate causation, our finding is in line with other evidence in suggesting that greater social support may protect against stigma (Lipira et al., 2019; McDowell & Serovich, 2007; Peltzer & Pengpid, 2019).

Previous research has investigated the relationship between symbolic stigma and depression. A study that surveyed 1,268 adults in Botswana in 2004, which used a different stigma scale than the one used in this study, found no association between depression and symbolic stigma (Wolfe et al., 2008). Our study found a statistically significant association between experiencing depressive symptoms and having symbolic stigma. Further research should elucidate the relationship between holding symbolic stigma and experiencing depressive symptoms. Determining the directionality between stigma and depression is important because of the significant stigma surrounding both mental health conditions and HIV (Stangl et al., 2019), thus creating “layered” or intersectional stigma (Gilbert, 2016).

The strong association we observed, between higher education and stigma, has been contrary to findings of other studies (MacPherson et al., 2011; Visser et al., 2009; Wolfe et al., 2008). However, only one of these studies (MacPherson et al., 2011) grouped participants who had attended university separately from participants who had completed or attended some high school (Visser et al., 2009; Wolfe et al., 2008). This may be an important distinction as university attendance is heavily influenced by socioeconomic status and other cultural factors (Temple, 2009). Given the low rate (4%) of higher education (education beyond high school) among our study population, it is possible that “stigma marking” of those with lower education may have played a role in the outsized levels of symbolic stigma and anticipated stigma seen in participants with higher education.

Theoretically, the differences found in our studies compared with previous studies on drivers of HIV stigma may indicate that the drivers of stigma vary based on factors such as average level of educational attainment and rates of employment. This suggests that interventions for stigma reduction should tailored to the specific community for with the intervention is designed to reach. General stigma reduction programs utilized across multiple communities may not be effective if these communities are culturally and/or economically different. Further research needs to be done to develop specific interventions to reduce stigma in this community. Research on how the implementation of universal access to antiretroviral therapy has impacted stigma in the community will be an important step to guide these interventions as well. Despite the introduction of Universal Test and Treat, stigma is still persistent in HIV-endemic settings and should be addressed to achieve the end of the HIV epidemic.

There were several limitations to this study. First, because we only interviewed participants at one clinic site, our findings may not be generalizable. Secondly, although frequently used to measure HIV stigma within communities, measures of symbolic stigma and anticipated stigma obtained via surveys may not fully capture the extent of stigma within communities (Wolfe et al., 2008). We were not able to measure enacted stigma (people’s lived experiences of discrimination) because questions about lived experience were not included in the survey, nor were we able to measure unconscious bias against people living with HIV. Finally, Because Universal Test and Treat guidelines came out during the time this study was conducted, it is possible that some participants knew their HIV status prior to testing and were coming back to get retested to start ART. We were not able to collect data on participants’ awareness of their HIV status prior to testing. However, strengths of this study include our large sample size, we evaluated people before revealing the result of their HIV testing, and enrolled participants over a six-year period.

In conclusion, despite a decline in symbolic stigma, perceived HIV stigma persisted among a cohort of newly enrolled, HIV status-naïve participants in an urban township of South Africa. Depressive symptoms and higher levels of education appear to be important drivers of HIV stigma in this cohort and our findings were suggestive that greater social support could be protective against anticipated HIV stigma. Fears of prejudice because of HIV positive status may be greater than the prejudice experienced within the community.

Ethics approval

This study was authorized by the University of Kwazulu-Natal (REF:BF052/13) and the University of Washington IRB (STUDY00000124).

Consent to participate

Informed consent was obtained from all individual participants included in the study.

Authors’ contributions

PKD was the principal investigator for the study. PKD and M-YM conceived and designed the study; MM, SRG, SG, BG, M-YM conducted the clinical research; MM conducted the biostatistical analyses; MM wrote the main text of the manuscript; all authors reviewed and approved the final manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Harvard University Center for AIDS Research: [Grant Number P30 AI060354]; National Institute of Allergy and Infectious Diseases: [Grant Number K23 AI108293]; Infectious Disease Society of America Education & Research Foundation and National Foundation for Infectious Diseases: [Grant Number]; Massachusetts General Hospital Executive Committee on Research: [Grant Number].