Study setting

The study was conducted in Cali, Colombia, a city with a population of over two million people located in southwestern Colombia. Cali is one of the cities with the fastest growing economy in Colombia and it is a key center for rural-urban migration, as well as a receptor of populations displaced by rural violence. A variety of race/ethnic groups populate Cali and approximately 23% are racial minorities (Afro-Colombians), most of whom are concentrated in low and middle-income neighborhoods. By 2016, Cali had high unemployment (15%) and casual employment (35%) rates and a violent death rate of (70/100,000) ²³.

Recruitment

Between 2012 and 2015, participants were recruited from individuals who attended a local HIV counseling and testing campaign. This campaign started by approaching community leaders from the targeted populations: MSM, transgender women, and non-MSM, who were asked to refer individuals whom they thought could benefit from HIV testing/counseling. Interested individuals were instructed to attend community sites where HIV counseling and testing were provided at no cost. At these sites, potential participants were invited to join the study. Informed consent was obtained from all individuals who agreed to participate.

This study focused on adult participants who self-identified as MSM, transgender women, and non-MSM. The study activities were conducted at twenty-two community centers, three sex work venues, and one HIV clinic. At each site, the research personnel applied a paper-based survey before providing HIV counseling and testing following a protocol as reported elsewhere ²⁴. The survey was anonymous and was applied in a private room by a trained research assistant. The survey typically took 30 minutes to complete and collected data on sociodemographic characteristics, sexual risk behavior, other risky behaviors, HIV testing, history of sexually transmitted infections, and early sexual experience.

Study measures

Psychosocial conditions. 1) History of childhood sexual abuse: We defined the history of childhood sexual abuse as the reporting of sexual activity at age 13 years or younger because in Colombia adults engaging in sexual activity with children younger than 14 years of age commit an abusive sexual act considered a criminal offense ²⁵; 2) Recent history of forced sex was defined as any experience of sex without consent with any sexual partner in the last 12 months; 3) Frequent alcohol use was defined as the consumption of five or more alcoholic drinks in one single occasion more than once in a month, and 4) Use of illicit drugs, which was defined as the consumption of any drug in the last six months, injected or inhaled, from a list of seven options including marijuana, cocaine, basuco (cocaine-containing paste) poppers, hallucinogens, heroin, amphetamines, or stimulants. In total, 18% of the sample reported the use of one type of drug, 8% reported two drugs, and 5% reported three or more. Thus, drug use was analyzed as no drugs vs. one or more drugs.

Main outcomes. 1) Unprotected sex with regular partners was defined as the frequency of condom use (“most of the time”, “sometimes”, or “never”) with a partner of any sex living permanently with the participant; 2) Unprotected sex with transactional partners was defined as sex not always accompanied of condom use with a partner of any sex with whom they exchanged money or other benefits for sex; 3) Unprotected sex with occasional partners was defined as sex not always accompanied of condom use with a partner of any sex who was not regular or transactional. Transactional partners were infrequent among MSM and non-MSM; we analyzed this variable as a dichotomous outcome (having ever vs. never had a transactional partner in the last 12 months).

Covariates. Educational level (highest level of education completed), having health insurance, personal income (based on the minimum monthly wage for Colombia, i.e., USD$ 308 at the time of the survey), history of incarceration and homelessness (defined as currently living on the streets). We also asked the participants to self-report their race/ethnicity. The definition of gender/sexuality was based on men’s self-report of MSM, transgender women, or non-MSM (men who reported themselves as heterosexual).

Statistical analysis

We followed the analytical recommendations from Tsai, et al.²⁶. First, we explored the relationship of all psychosocial variables with each outcome in one model to test if their effect was independent of the presence of the other. Next, we tested if the ‘accumulation’ of psychosocial conditions was related to each of the outcomes for which we created an index summarizing the number of psychosocial problems experienced by each respondent. This “syndemic” index was analyzed as a dummy variable ²⁶ with the reference point of having one psychosocial problem (given the low frequency of zero psychosocial conditions in transgender women). Next, we calculated the interaction effect of two psychosocial conditions on each outcome. We assessed the additive interaction using the relative excess risk due to interaction (RERI) or part of the total effect due to interaction (null value=0) ²⁷ and we calculated the multiplicative interactions using the common approach of exposure1*exposure2 on the outcome.

The significance of the additive interactions was obtained using the ICP program in Stata^™ (StataCorp LP, College Station Texas, USA). A Poisson analysis with robust variance was performed for all analyses ^28).

Discussion

Our study helps better understand the effects of accumulation of and the interactions between psychosocial conditions on unprotected sex and transactional sex in a diverse sample from Colombia. For the first time in a sample of MSM, transgender women, and non-MSM from Colombia, we tested the “syndemics” framework using 1) the traditional analytic approach of adding the presence of psychosocial conditions and examined their effects on HIV risk behaviors, and 2) two analytical approaches to assess interactions of combinations of psychosocial conditions. Our results highlight the differences across the sample that have important implications for future interventions to address the HIV epidemic in Colombia, as summarized below.

Findings in MSM

Our findings add to the existing research demonstrating that psychosocial conditions increase the possibility of engaging in HIV risk behaviors in MSM ^3,29. Drug use was highly frequent in the sample related to unprotected and transactional sex. The importance of illicit drug use in driving the epidemic of HIV in MSM is well established and our study reinforces these findings ³⁰. The role of early life adversities, particularly violent experiences, its high frequency, and its effects on unprotected and transactional sex in MSM in our sample have also been reported in Colombia and other populations ^22,31. Forced sex, on the contrary, was reported by a low proportion of the sample with no differences between the groups; it was not related to unprotected sex but was more likely to be present in those who engaged in transactional sex similarly to reports from other authors ^22,32. Alcohol abuse has been related to less protected sex in other studies including some from Colombia ^6,33 and in our sample, it was related to greater engagement in transactional sex among MSM.

Given the importance of all psychosocial conditions on outcomes in MSM, it is not surprising that the accumulation of psychosocial conditions increased the risk of unprotected sex with occasional partners, which is consistent with most of the published literature in these populations ^34,35. The risk of engagement in transactional sex also increases with the accumulation of psychosocial conditions, a novel finding of our study for MSM populations. Approximately 10% of MSM have engaged in transactional sex, consistent with other estimations in Latin American MSM (7%) ²².

Of special relevance is the finding of synergistic (or “syndemic”) effects of combinations of psychosocial conditions on unprotected sex and engagement in transactional sex. The risk of unprotected sex with occasional partners increases when both CSA and frequent alcohol use occur together. Chakapri, et al. also reported multiplicative interaction effects of the experience of violence and alcohol use on unprotected sex in MSM in India ¹⁴. We also found synergistic effects for all combinations with childhood sexual abuse, drug use, and frequent alcohol use on transactional sex. We are unaware of other studies reporting similar findings. The fact that some of these psychosocial conditions act synergistically in increasing risky behaviors suggests that even a single intervention, in this case addressing childhood sexual abuse, could greatly reduce engagement in unprotected sex and transactional sex in MSM, even if the other factors are not considered ¹⁵. The “syndemics” framework claims that clustering and interaction between psychosocial conditions are tied to structural factors. Our findings, then, should be understood as a consequence of structural violence (lack of legal protection) ³⁶, social inequities (i.e., unemployment, poverty) ²², and minority stress (greater exposure to discrimination and prejudice) ³⁷ suffered by MSM populations.

From a public health perspective, our findings emphasize the harmful effects of sexual abuse during childhood ³¹, a critical period for the development of sexual and other health behavior preferences that, if adverse, as is the case of abuse, could lead to higher alcohol abuse, illicit drug use, and violence. This and other experiences of violence throughout life should be addressed and prevented by public health programs and child protection policies. If others confirm the synergistic effects found in our study, acting on experiences of abuse and violence could help prevent HIV cases through their effects in reducing unprotected and transactional sex.

Findings in transgender women

In our sample of transgender women, 62% engaged in unprotected sex, a figure consistent with another report in a larger Colombian sample ³⁸. Childhood sexual abuse, illicit drug use, and alcohol misuse were reported frequently by transgender women, factors that have been related to greater HIV risk behaviors in transgender women elsewhere ^19,39. We found only one factor, forced sex, to be related to unprotected sex. Transgender women could have less power to negotiate condom use in abusive relationships. Having transactional partners was also frequently reported (71%) and, in this case, none of the psychosocial conditions was related to this behavior. Our findings contrast with those of other studies in transgender women from North America, India, and Jamaica ^19,35,40 relating the accumulation of psychosocial adversities to greater HIV risk behaviors.

Two factors may have contributed to this lack of results in the direction expected. One is the small sample size. Recruitment of transgender women has always been difficult in Cali as the level of trust in health organizations is poor there (Héctor Mueses, personal communication). We have used community leaders to reach transgender women but even with their enthusiastic assistance, our recruitment was low. In another sample of Latin American transgender women, sample sizes have been over 100 participants while our sample size was 59 ³⁹. The other aspect is related to the psychosocial experiences inquired in our survey. Ours was an exploratory study conducted in the context of HIV testing/counseling campaigns. Hence, we felt that in this scenario it was not feasible to inquire more in detail about violence, discrimination, or mental health conditions. In future studies, it would be important to focus specifically on mental health problems, criminalization, social exclusion, and violence displacement as additional psychosocial problems driving HIV risk among transgender women in Colombia ^2,18,41.

Findings in non-MSM populations

In the case of non-MSM, this study reached a high-risk population not usually targeted for HIV testing or other prevention campaigns. This population lives in poverty, is poorly educated, and one-third of them are racial minorities (Afro-Colombians and aboriginals). Approximately 68% engage in occasional relations with no condom use, the accumulation of psychosocial conditions was even higher than among MSM and apparently associated with higher engagement in transactional sex. As non-MSM could serve as a bridge for HIV and sexually transmitted infection transmission to heterosexual women or bisexual men ¹¹, we suggest that non-MSM with low socioeconomic status should be also prioritized for HIV prevention interventions ¹¹. It seems that targeting alcohol abuse would be an effective strategy to reduce the risk of unprotected sex in non-MSM.

Strengths and limitations

Our study is one of the few using the “syndemics” theory to compare HIV risk in groups of diverse gender and sexual preference ⁴². The regular collection of data on HIV and risk behaviors among high-risk populations is not done in Colombia ⁴³. The lack of surveillance of behavioral data among high-risk groups may prevent the optimal development and implementation of public health interventions. Our group has taken this initiative and our results point out important implications for the development of HIV prevention programs. We recognize that this study has limitations. The sample was not probabilistic and estimates may not be generalized; participants were a self-selected sample likely based on their perception of risk, knowledge, and experiences of stigma and discrimination related to HIV testing. The MSM sample was captured through leaders of the MSM community and may reflect a selected group with more favorable psychosocial conditions. Our estimates are, however, close to those found using respondent-driven sampling methods with larger samples in Colombia ³.

Unfortunately, our sample was underpowered to compare experiences of transgender women and larger samples will be needed for this purpose. Also, our analyses rely on cross-sectional data. In this regard, there is a need for longitudinal studies and intervention trials that allow testing the temporality of psychosocial condition effects, mediation effects, and interactions on the incidence of HIV and risk behaviors ⁴⁴.

Implications of the results

From a research perspective, our study strongly argues that the “syndemics” theory is a more suitable approach to test the synergistic effects of psychosocial conditions on HIV risk given the multifactorial and complex nature of this problem. Expanding the range and quality of the psychosocial conditions assessed and identifying populations at risk of HIV through drug and alcohol-related venues is also required. Our findings revealed the importance of identifying and addressing mental health (illicit drug use, alcohol use) among MSM, transgender women, and non-MSM ³¹. Interventions using community-based or empowerment theories that have been effective in other populations could be adapted in Colombia and other Latin American countries with similar problems.