Background
HIV has existed in Vietnam since at least 1990, with 0.44% of the adult population estimated to be infected in 2010 [
1]. As typifies a concentrated HIV epidemic, certain sub-groups of the population, such as injecting drug users (IDU), commercial sex workers (CSW), clients of CSW, and men who have sex with men, have far higher national HIV prevalence estimates than the general population (30%, 9%, 2%, and 2%, respectively) [
1]. Much of the focus of HIV prevention and testing in Vietnam has been in the largest cities and the northeastern coastal provinces that are estimated to have the highest HIV prevalence. Less attention has been paid to the situation of HIV within rural areas, where more than 70% of the population of Vietnam is located, and where there is considerable employment migration to and from the higher HIV prevalence areas [
2,
3].
The close association between HIV infection and HIV-related risk behaviors has been emphasized by Vietnamese media and government information campaigns, where CSW and IDU are often referred to as "social evils" by the government and the wider population alike [
4]. Unfortunately this alienation of risk groups has resulted in a high level of stigma being attached not only to drug use and sex work, which are heavily condemned and illegal in Vietnam, but also to persons who are known or rumored to be living with HIV [
5,
6]. HIV risk is equated with drug use and sex work by much of the population, who show little awareness of the growing transmission of HIV to clients of CSW or to transmission within long-term heterosexual relationships where one member of the couple engages in potentially high HIV risk behavior such as drug use, sex work, or visits to sex workers [
4,
7]. Stigma against members of these key populations and against persons living with HIV are very closely associated and are often referred to as "layered stigma".
Stigma against persons living with HIV has been found to be high overall and considerably higher in rural settings within Vietnam than in urban settings [
8]. Also, when similar measures have been used, persons in Vietnam have been found to express more stigma against PLHIV than persons in settings with generalized HIV epidemics, where HIV epidemiology and stigma dynamics are very different, such as South Africa or Botswana [
9,
10]. The Government of Vietnam has recognized that stigma against PLHIV is in need of attention and in 2006 strengthened legislation and extended protection for PLHIV by promoting their rights to HIV-related confidentiality, medical care, and to integration within the community, as well as prohibiting HIV-related stigma and discrimination [
11,
12]. The law on HIV/AIDS is new and research has yet to be produced on its effects on stigma and discrimination towards PLHIV in Vietnam.
Existing community stigma towards PLHIV, as well as groups thought to be vulnerable to HIV such as sex workers or drug users, impacts the likelihood that those at-risk for or infected with HIV would seek preventative or care services. For PLHIV within and outside of Vietnam, fears of HIV-related stigma and discrimination or the direct experience of it have been described as leading to internal stigma, self-isolation and low self-esteem, including non-disclosure of HIV status and avoidance of contacts with support networks and health care services [
13‐
15]. Evidence from Vietnam shows that stigmatizing attitudes towards PLHIV can act as a major barrier to adherence to HIV treatment [
16] and it has been documented that persons with HIV often present very late for care, thereby reducing their opportunities for treatment success. Also for those who do not know their HIV status, stigmatizing community attitudes towards PLHIV can serve as a barrier to seeking HIV testing as well as to internalizing HIV-prevention messages, because risk reduction strategies such as condom use are associated with negative attitudes and fears about HIV [
17‐
22]. Therefore, stigmatizing community attitudes towards PLHIV affect the quality of life and health outcomes of persons living with HIV, and could potentially result in more transmission of HIV as people delay knowing their HIV status or do not use measures to prevent HIV, such as condoms or needle and syringe programmes, in order to avoid the negative association that these have with HIV.
And so, while it has been documented that stigma against PLHIV has negative effects both on PLHIV and on efforts to prevent HIV in the wider community, interventions to significantly reduce this stigma are few [
23]. Furthermore, when interventions are employed, there is a paucity of information to inform them or to measure their progress [
24]. Recently, increasing effort has been made to develop and validate quantitative stigma scales or indicators for HIV-related stigma [
9,
25‐
27]. This approach brings new possibilities for measuring and comparing HIV-related stigma over time or across populations, which is important in determining whether efforts to decrease stigma towards PLHIV are working or in need of increased attention. However, much of this research has measured stigma as a total community or population score and this mean score may not give the full picture of the complex phenomena of stigma, which is often multi-faceted within individuals and populations. Only a few studies have evaluated the determinants to HIV-related stigma [
9,
22,
28], and these have done so using a variable-centered approach.
Another analysis approach that could enhance understanding of the varying character and levels of HIV-related stigma within a population over time is latent class analysis. Latent class analysis is a "respondent-centered" approach that seeks to group individuals into class groups based on their responses to a set of observed variables (in this case, responses to eight stigma statements which relate to stigma against PLHIV)[
29]. Latent class analysis has been used widely in market research to tailor marketing campaigns to segments of the population as well as to understand patterns of complex health risk behaviors [
30], including substance abuse [
31], mental health [
32,
33], as well as HIV/AIDS risk behavior, knowledge and programming [
10,
32,
34‐
36]. In addition to understanding the pattern of stigmatizing attitudes present in each class by examining the probability of each answer to the stigma statements on which the classes were composed, factors independently associated with the stigma class membership can be identified through latent class regression [
29]. In the present study, we explore whether latent class analysis can be used to analyze the patterns of HIV-related stigma towards PLHIV and to identify predictors of different levels of stigma in a population-based rural sample.
Discussion
Our study is the first in Vietnam that quantitatively examines the patterns of stigmatizing attitudes towards PLHIV within a rural population-based sample with latent class analysis. We found that women and individuals with less formal education, in particular, held the most highly stigmatizing attitudes. There was also evidence that those who had heard of HIV from fewer sources, who did not know anyone with HIV, and whom did not feel at-risk for HIV reported the most stigmatizing attitudes towards PLHIV.
Three discrete groups emerged based on similar within-group responses to statements about PLHIV. Those who held the most stigmatizing attitudes towards PLHIV appeared to differ from those with least stigmatizing attitudes in that they were more likely to be women, to have less education, and not to have migrated out of the area. The inverse relationship between years of education and level of HIV-related stigma was reported in South Africa [
9]. Hong et al [
42], qualitatively describe high HIV-related stigma among Chinese rural-to-urban migrants, and it is likely that those who have not migrated outside of rural areas in our study sample had less likelihood of being exposed to HIV information and stigma reduction campaigns which target urban areas. The relationship between gender and HIV-related stigma is less well-documented, and might be more unique to the Vietnamese context, where women are socially expected to distance themselves from behaviors considered to be "social evils" such as drug use and sex work [
4,
42].
We also saw that those who had heard of HIV from fewer sources were significantly more likely to belong to the
highly stigmatizing group, perhaps indicating that persons in this group had been exposed to or remembered less HIV information. However, the role of media and information campaigns was complex, as those who had heard of HIV from more sources had about twice the odds of belonging to the
ambivalent than the
least stigmatizing group, perhaps indicating that persons in the
ambivalent group had heard of and remembered HIV information sources that were conflicting, confusing or ambiguous in their messages about HIV and PLHIV. Vietnam's National Action Plan for HIV states that HIV stigma reduction campaigns are to be carried out through the media, schools, workplaces and with the help of famous personalities [
6]. Critiques of government-driven information campaigns point to how they closely link HIV to the "social evils" of CSW and IDU, and the evidence from this study does not refute the presence of that link in community attitudes [
4]. Especially among those who were
ambivalent in regards to their attitudes towards PLHIV, further analysis of these information campaigns is necessary to assess the extent to which these are effective or, instead, inadvertently sustain HIV-related stigma by associating infection with "social evils" or by providing incomplete or ambiguous HIV prevention messages [
4].
Those who reported knowing someone with HIV or feeling personally at-risk for HIV were significantly more likely to belong to the
least stigmatizing group. This has been found in Asian settings and elsewhere and points to the potential of putting a "human face" on HIV as a stigma-reduction or a stigma-prevention measure [
9,
25,
43]. However, in a concentrated HIV epidemic setting, such as Vietnam, where HIV prevalence is low, the likelihood of knowing someone with HIV will also be low as compared to the likelihood in a country with a generalized HIV epidemic. We did not assess what it meant to "know" someone with HIV in our study and have not seen this assessed elsewhere. However, these results should be taken into account when assessing the stigma-reducing potential of programs in which PLHIV "go public" and provide community education. In addition to providing HIV prevention information, this could have the potential to reduce HIV-related stigma based on the "human face" associated with HIV [
44].
Our study generally found similar community stigma towards PLHIV as compared to studies conducted in 2005 in other regions of Vietnam [
8,
45] and higher HIV-related stigma than reported outside of Vietnam [
9,
10]. For example, in Southern Vietnam, Nguyen Anh Tuan, et al found that the same percentage of persons in rural Southern Vietnam as in our study (64%) believed that PLHIV should feel ashamed while the urban sample from the same study reported lower stigma (45%) [
8]. Reports on the measure "PLHIV should feel ashamed" from non-Vietnamese settings are much lower ranging from 8-34% in various South African settings [
9]. Fewer persons in Botswana report that they would be reluctant to share a meal with a PLHIV (27%) as compared to 44% in our sample [
10]. There is some evidence of a relationship between the availability of treatment for HIV (ART) and lower attitudes of stigma [
10,
28]. This, in combination with the greater likelihood of knowing someone with HIV, could explain the lower levels of stigma in urban Vietnam, where ART became available in 2004, as compared to this rural district where it was still unavailable in 2007.
In addition to quantitatively describing patterns of stigma within a population, latent class analysis can provide important information for HIV stigma-reduction interventions within communities. A latent class analysis can be used in conjunction with an "audience segmentation" approach [
46], a key concept in commercial and social marketing, to divide the population into sub-groups. Then, intervention strategies can be tailored to the sub-groups' demographic characteristics, knowledge levels and, in the case of stigma against PLHIV, into groups with similar stigma attitudes. In this community, for example, interventions would likely need to approach men and women differently, as gender was an important variable which differentiated the stigma groups. Other factors, such as differences in stigma patterns based on knowing a person with HIV or feeling at personal risk for HIV might also direct intervention and health promotion strategies for this community. It is also of interest that almost 20% of the population was
ambivalent with regard to many of the stigma statements and that these persons appeared to have heard of HIV from more sources. Persons in the
ambivalent group may have the potential to be influenced towards more accepting attitudes and this could be evaluated by applying targeted, high-quality anti-stigma messages and by following-up over time to determine whether the intervention had a sustained impact. Finally, in addition to informing the stigma-reduction needs within the community, latent class analysis could also be used as an evaluation technique to measure community patterns of stigma over time so that changes in group composition or number can be evaluated. This approach might be superior to overall community stigma scores followed over time, which give less detailed information on the distribution of people across stigma classes.
Methodological considerations
Of note, in pilot testing and in subsequent data collection, a sizable portion of the sample responded "Not sure/maybe" to each of the stigma statements (range 8%-25% of respondents, depending on the question). We checked and understanding of the questions was consistently high. Respondents explained that they were ambivalent because they wished to qualify their response based on how sick the PLHIV was or how the person had become infected with HIV. We took the view that these responses were important and typified the reluctance that some community members often have in determining their attitudes towards persons living with HIV and, therefore, kept the "Not sure/maybe" response as a separate category for data collection and analysis. This is an important consideration for those trying to measure stigma in the Vietnamese context and may be an important measure to assess qualitatively and while validating stigma scales in other settings as well.
While our study has the strength of using population-based data, it might be limited due to social desirability bias. However, respondents expressed higher-than-expected stigma against PLHIV and interviews were conducted by well-trained interviewers in a private area to minimize social desirability bias. Women in this study reported more highly stigmatizing attitudes than men and this might have been an effect of the study's use of only female interviewers. If stigma were easier to express to someone of the same gender, then the degree of stigma toward PLHIV among men could have been under-estimated. Since this study was exploratory in nature, we used commonly used stigma items that, in qualitative research that we had conducted, had been linked to less uptake of HIV prevention, testing and care services. Future research could benefit from the use of a validated stigma scale, which can be repeated over time and compared across settings. Finally, often in the analysis of data where cluster sampling is used, such as in this study, the design effect or some other manner of taking sample weights into account is performed. It was not possible to take this into account using R software, nor using other software that were available to us. This may have led to somewhat biased estimates as compared with what would have been achieved if we had used a simple random sample. However, since the number of clusters was large (42 of the total 71) and cluster of origin did not appear to significantly differ with respect to answers on the eight stigma statements in descriptive analysis, we believe that this bias does not seriously influence the study findings.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
AP, AT, NTKC, and NPH conceived of the study and participated in its design. AP, NPH and NTKC coordinated data collection. AP and GM conducted data analysis. AP completed the first draft of the manuscript, and all authors contributed to commenting on the analysis and manuscript. All authors read and approved the final manuscript.