Discrimination Due to Sexual Orientation and Associated Factors Among Men Who have Sex with Men in 12 Brazilian Cities: A Respondent-Driven Sampling Survey

Background Discrimination due to sexual orientation (DDSO) has an important association with health outcomes among men who have sex with men (MSM). This study aimed to analyse factors associated with DDSO among MSM in 12 Brazilian cities. Methods A cross-sectional study with 4,176 MSM participants recruited in 2016 through respondent-driven sampling in 12 Brazilian cities. In this analysis, ordinal logistic regression was used to assess associations with previously identied levels of DDSO, which were based on latent class analysis (LCA) on 13 variables from the discrimination section of the questionnaire. Weighted odds ratios and their respective 95% condence interval were estimated using Gile estimator.

There has been an alarming increase in DDSO. In the last decade, studies show high proportion of DDSO among MSM in different countries [19] including Brazil [10,[20][21][22]. Sexual and affective practices among MSM remain illegal in 70 United Nations member states and are punishable by the death penalty in six countries (three in Asia: Iran, Saudi Arabia and Yemen; and three in Africa: Nigeria, Sudan and Somalia) and is a possibility in another ve countries (Mauritania, the United Arab Emirates, Qatar, Pakistan and Afghanistan) [19]. In Latin America, homosexuality is not illegal, nevertheless, the patriarchal, religious and sexist context of countries in this region adds to the impact of DDSO on the physical, mental, social and cultural well-being of MSM [23][24][25].
Brazil's judicial system has protective human rights measures for MSM, including recognizing same sex marriage in 2011 [26], and categorizing discrimination against LGBT individuals as racism, a prosecutable offense, until such time as the Brazilian National Congress approves speci c legislation related to this issue [27]. Despite this, in 2018, 420 LGBT individuals were murdered, targeted for their identities, of which 191 (45%) were MSM [28]. In 2009 and 2016, cross-sectional studies of MSM recorded an increase in the prevalence of self-reported DDSO from 27.7-65% [21,29]. In 2009, the likelihood of reporting DDSOs was higher among younger MSM (below 30 years), with more schooling, a history of sexual or physical violence, suicidal ideation, unprotected receptive anal sex and among those who participated in support organizations (such as an NGO) [29].
This study aims at identifying factors associated with DDSO among MSM in Brazil.

Study design and location
This study utilizes data from the behavioural and biological surveillance survey entitled "The 12 city HIV surveillance survey among MSM in Brazil 2016 using respondent-driven sampling (RDS)" [30]

Participants
The survey sample consisted of 4,176 MSM who reported at least one sexual experience in the 12 months prior to the study; were 18 years of age or older; did not identify as transsexual or transgender women; resided, studied or worked in the selected cities; and signed an Informed Consent Form. To recruit participants, we used Respondent-Driven Sampling (RDS) as recommended for hard-to-reach populations [31]. Formative research using semistructured interviews and focus groups was conducted in each city to prepare for the main study. Six MSM were selected purposively to serve as "seeds". These were individuals of different ages and socio-economic status with relatively large social contact networks. Each of these seeds was asked to recruit three other MSM from their social network by offering a voucher to each recruit. These vouchers provided information about the site and also contained a number linking the recruiter and the recruit. This procedure was repeated with each eligible participant until the desired sample size was reached. Participants received a primary and secondary incentive of BRL 25.00 (US$ 7.40) for each of their recruits to reimburse expenses such as transport and food.

Data collection
Data was collected through computer assisted personal interview (CAPI) in the study o ce in each city, where blood samples were drawn for HIV and syphilis testing. Upon completion of the interviews, these were immediately sent via internet to a central server. The research project was approved by the Research Ethics Committee of the Federal University of Ceará. Further details about the methodology may be found in Kendall [30].

Study variables
The outcome variable, DDSO, was de ned by Latent Class Analysis (LCA) and it was based on thirty questions divided in four DDSO dimensions: i) discrimination in the work sphere (i.e. not selected for a job or red from a job); ii) educational setting (i.e. mistreated or marginalized by teachers and classmates at school/college); iii) private domain (i.e. excluded or marginalized from groups of friends, neighbors, family, or religious environment); iv) and the public sphere (i.e. blackmailed or extorted money, poorly cared for in health services, prevented from donating blood, mistreated in public services, mistreated by police o cers and prevented from entering in a market) as previously described [21]. LCA is a useful method for identifying underlying groups of individuals with similar pro les [32].
The independent (explanatory) variables included in this analysis were: a) sociodemographics: age (< 25 years and ≥ 25 years), self-reported race/skin color (white, black and mixed race -pardo), education (primary or incomplete secondary, post-secondary education), religious a liation (yes, no) and marital status (single or separated or widowed and married or living together or civil union); b) economic: three economic groups were created: A-B: high income; C: average income: and D-E: low income, based on the Brazilian Economic Classi cation Criteria [33]; c) self-reported sexual identity (heterosexual and gay); d) ever experience violence due to sexual orientation (lifetime): sexual (yes, no) or physical (yes, no); e) fear of visiting public places (yes, no); f) family approval of sexual orientation (approves; disapproves/indifferent; and family does not know); g) disclosure of sexual orientation: told friends (yes, no), told mother (yes, no), told father (yes, no); h) participation in LGBT NGOs (yes, no); i) ever tested for HIV (lifetime) (yes, no); j) suicidal ideation in the previous two weeks (yes, no); l) alcohol use classi ed by audit score ( low risk and abstemious; risk and high risk; and dependent) and illicit drug use in the previous 6 months (yes, no).

Data analysis
Gile successive sampling estimator [34] was used to calculate the weighted ratio estimators using RDS Analyst [35]. Data from the 12 cities were merged into a single database in which each city was treated as its own strata. The previously de ned latent variable DDSO [21] and the RDS weights were transferred to the current database for descriptive, bivariate, and multivariate analyses. Stata 15.0 was used with complex survey data analysis tools (Stata Corp, College Station, TX, USA, 2019).
Multivariate analysis included ordinal logistic regression modelling, as the outcome variable was ordinal. And estimators were adjusted with the simultaneous generation of a constant association measure across the categories of the outcome variable [36].
The construction of the logistic model's weighted odds ratio began with a bivariate analysis to assess the factors associated with the outcome variable. Variables associated with DDSO in the bivariate analysis at a p-value of 0.20 were included in the multivariate analysis. Only those with p-value<0.05 remained in the nal model. A review of the literature regarding factors associated with DDSO was also important for the selection of these factors. Weighted odds ratios (OR) with 95% con dence intervals (95% CI) were used as measures of association between explanatory factors and DDSO.

Results
We recruited 4,176 MSM in the 12 Brazilian cities. Most of these MSM were young, 56.1% under 25 years old; 86.3% were single and 53.1% named a religious a liation; 70.3% reported post-primary education; 41.8% of mixed race (Pardo), 31.5% whites and 22.7% blacks. 42.4% percent of MSM reported high income level (A-B), another 41.7% reported average income (C) and 15.9% low income (D-E). The vast majority of participants (92.4%) reported gay sexual orientation (Table 1).  Estimate weighted using Gile-SS c CI: con dence interval; LL: lower limit; HL: higher limit. Table 3 shows the association between explanatory variables and the DDSO categories. Bivariate analysis indicated an increasing odds of DDSO among MSM under 25 years old, of mixed race (pardo) and white, who had post-secondary education, and with higher socio-economic status. In addition, gay sexual identity, sexual and physical violence, family disapproval or indifferent, told mother and father about sexual orientation, and participating in an LGBT NGO were also associated. Finally, a HIV test in lifetime, suicidal ideation experience, risk or high risk of alcohol use and illicit drug use in the previous six months showed increased odds of DDSO. In the nal multivariate analysis, the following variables were independently associated with DDSO: age under 25 years old, white skin color, experience of sexual and physical violence, shared sexual orientation to their father, experienced suicidal ideation in the previous two weeks and used illicit drugs in the previous six months (Table   3).  [29]. Younger men in our study were more likely to report DDSO compared to older men, a trend already observed in 2009 [29] and in other countries [37,38]. Over the past decades, in some countries, there has been a growing tolerance for sexual minorities that might lead to more openness and thus opportunity to become a target of discrimination. In contexts of less tolerance, MSM may adopt masking behaviors. Another explanation may be the rising homophobia engendered by the current and previous government in Brazil, who have implicitly and explicitly granted permission for the expression of homophobia. A third hypothesis is that there could be a desensitization arising from repeated exposure to abuse among older MSM, and the creation of cognitive mechanisms that may partially neutralize the identi cation and effects of discrimination [39]. Part of this desensitization may explain lower reports of experiences of DDSO.
White MSM were more likely to experience DDSO than black men. In Brazil, higher educational levels are found among white men compared to black men [40] and we may hypothesize that the higher education might be associated to the greater perception of discrimination and to more willingness to report [41]. Stigma and discrimination are consequences of social structures and forms of domination used in society to identify and discriminate against all divergence from normative white male identity. In Brazil, the DDSO affects even white and educated men when they express or identify their sexual orientation. On the other hand, for black MSM, DDSO presents in an intersectional manner alongside racial and perhaps class discrimination that could be multiplying health outcome effects [42][43][44].
Other concern relates to the fact that only 26.9% of MSM shared or reported DDSO. Moreover, when they did so, the vast majority did not report to the judicial system, but rather to their friends or relatives. This lack of reporting is problematic in two senses. Firstly, because of underreporting, it underestimates DDSO in Brazilian society, reducing visibility of this problem to public administrators, civil society and legislators. On the other hand, this may affect the physical and mental health of MSM, since not talking about the problem may aggravate internalized homophobia [45,46], which can be de ned as self-directed homophobic attitudes that in turn can lead to low self-esteem and self-hatred [47]. This di culty in reporting discrimination may be aggravated if there is a lack of family acceptance. In this study, we demonstrated that one-third of participants reported that their families do not approve of their sexual identity and disclosing identity to a father was independently associated with greater likelihood of DDSO.
In this study, both physical and sexual violence were strongly associated with DDSO, as in other studies [6,29].
Discrimination and violence may in uence perception of insecurity among MSM: 60% reported feeling fear of circulating in public places, a much higher percentage than found in hostile venues for MSM in Abuja city, Nigeria (17.0%) [48].
Our study also demonstrated an association with suicidal ideation in the previous two weeks and illicit drug use in the previous six months before the survey, respectively. We note the time-frame differences as DDSO was analysed over the previous 12 months. Despite this, other studies showed an association between DDSO and mental health distress [38,49,50], and the use of psychoactive substances among MSM [4,8,11,12]. LGBT Health [51,52] were developed. However, the country has experienced setbacks in LGBT health policies over the last ve years, given the advance of religious groups and conservative social movements in Brazilian society, in Congress, and at the Presidency [53,54]. More recently, the current national government, as of 2019, has further exacerbated this situation [54]. The new president and his party -who elected the largest number of members in Congress -were voted on a platform containing explicit rejection of LGBT individuals and identities, publishing a program vowing to exclude LGBT communities from human rights protections and social policies [53]. The Secretary of Continuing Education, Literacy, Diversity and Inclusion, who was responsible for promoting policy to secure sexual, gender, and ethnic diversity in Brazilian public schools, was extinguished. HIV prevention guidelines for transgender population were also removed, arguing that this document contained "content that was offensive to the family" [53].
The current political context in Brazil is one of increased stigma directed towards LGBT communities that most likely will increase DDSO. Between 2009 and 2016, the persistence and increase in DDSO may constitute a signi cant barrier of access to health services, as well as universal access to antiretroviral therapies to treat and prevent HIV (such as the Pre-and Post-exposure prophylaxis) [55][56][57].
An unfavourable and stressful environment is increasingly associated with physiological effects that make MSM more vulnerable to a range of diseases [58]. In line with Minority Stress Theory [59,60], stressors are portrayed as events and conditions requiring constant change, causing exposed individuals to make intense efforts to adapt to new circumstances. Individuals subject to stressors therefore suffer from constant psychological and physical tension, eroding their capacity to adapt and inducing mental and somatic disorders [61]. As Goffman [62] pointed out, the response to the "deviance" embedded in stigma and discrimination can lead to the problematization of previously uncontested terrain: challenges to masculinity and heteronormativity can be seen as a response to serious problems of gender violence and intimate partner violence -both widespread in Brazil and elsewhere. As with in-group stigma, occupying a stigmatized role can offer opportunities for leadership, not just oppression. In light of this, in relation to strategies that diminish DDSO among MSM, there are several initiatives in Latin America including interventions sensitive to gender and culture, focusing on the formation of new values that allow for an expansion of different forms of masculinity [63].
Although this study presents robust results, it also has some limitations. The original questionnaire was not designed to evaluate DDSO. Moreover, we do not have information about the timing or duration of exposure, the context in which the episode occurred, or the intensity and frequency of DDSO. RDS studies furthermore present limitations regarding sample representativeness, potential biases in estimating and sampling social networks and indicator data, and potential violation of several theoretical assumptions of RDS in implementation. Speci c analyses using the latest estimators for data collected using this method seek to reduce these biases.

Conclusion
Page 16/21 The study reports a high level of DDSO among MSM in Brazil with notably higher levels among younger MSM. The high prevalence of DDSO reported among white and more educated MSM highlights the "iceberg" nature of DDSO, and that much more discrimination exists than what is being documented. There is certainly a need for more in-depth studies on the association between DDSO, sexual and racial minorities, and the social and political context in which these take place. The authors declare that they have no competing interests.

Funding
Ministry of Health through the Department of Chronic Disease and Sexuality Transmitted Infections. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Authors' contributions LM conceptualized, analyzed and interpreted data, and was a major contributor in writing the manuscript the article; ID supervised the eldwork, analyzed and interpreted data and substantively revised it; AFL supervised the eldwork, interpreted data and writing the manuscript; DK supervised the eldwork, interpreted data and writing the manuscript; MDC supervised the eldwork, analyzed and interpreted data and writing the manuscript; XPDB reviewed the manuscript; GMR supervised the eldwork and reviewed substantially the manuscript; MAV supervised the eldwork and reviewed substantially the manuscript; CK supervised the eldwork and reviewed substantially the manuscript; AMB supervised the eldwork and reviewed substantially the manuscript; LK coordinated the research and reviewed substantially the manuscript.