Introduction
It is estimated that in Mexico there are around 56,000 people who inject drugs (PWID), half of which are allocated in the cities of Tijuana, Ciudad Juárez, and Hermosillo. In this population, it is estimated that 8% are women (UNAIDS, 2014). Although injected drugs do not have epidemic proportions, PWID are specially exposed to HIV and hepatitis C, viruses that represent a high burden of disease for the Mexican health system.
Within PWID, several studies have shown the greater vulnerability women have for HIV and other health conditions, compared to their male counterparts. Brouwer et al., (2012) report an HIV prevalence among women who inject drugs (WWID) of 10.2%, in contrast to 3.4% among men. Therefore, it is important to understand the role of gender in risk behaviors that facilitate the transmission of HIV and other blood-borne virus.
The higher HIV prevalence among WWID could be partly explained by the overlap between sex work and drug use networks (Sherman, Latkin, & Gielen, 2001; Evans et al., 2003; Strathdee, Magis-Rodriguez, Mays, Jimenez, & Patterson, 2012), in conjunction with the structural poverty conditions women face in Mexico (Patterson et al., 2006; Strathdee et al., 2008; Strathdee et al., 2013), and the highest exposure of WWID to violence (Beletsky et al., 2013; Syvertsen et al., 2014; West et al., 2016). In addition, the biologically transmission of HIV is more efficient from man to woman than vice versa (Strathdee & Stockman, 2010).
There are few studies in Mexico that investigate the differentials in risk environments for HIV between men and women who inject drugs. The lack of data on the characteristics of the risk environments that are specifically addressed by the WWID and the processes through which these characteristics operate in the probabilities of getting infected make it difficult to incorporate a gender approach in the design of policies and harm reduction programs.
This article describes differences in the characteristics that build differentiated risk environments for HIV between men and women who inject drugs in Ciudad Juárez and Hermosillo.
Method
Study design
Data were collected between January and June 2012 in places that PWID were known to frequent (hotspots), following time-location-sampling methodology. The survey was supported by the Global Fund to Fight AIDS, tuberculosis and malaria (Round 9), and its aim was to evaluate HIV risk among PWID in Mexico’s northern region.
Participants
Potential respondents were screened to verify if they met the eligibility criteria, being above 18 years old, reported having injected drugs at least once during the last three months, and reported being permanent residents of the selected cities. Screening began with the provision of general information about the study aims and procedures.
Measurements
We included in the analysis variables that correspond to the different domains outlined by the risk environment framework (Rhodes, 2002) which posits the need to shift from individualistic approaches of risk behavior towards an understanding of how structural and environmental conditions shape an individual’s vulnerability to HIV acquisition. The focus of the analysis from this perspective is to identify social situations and places in which harm is produced and reduced. The risk environment is structured by interactions between domains (physical, social, economic, and political) and its different levels of influence (macro and micro) (Rhodes, 2002).
Following the risk environment framework, we included the city of residence and the gender as measures of the macrophysical and macrosocial domains, respectively. Age, migration status (born in the current city of residence vs. born anywhere else); level of educational attainment (completed secondary or more vs. partial secondary or less), and having a commercial sex partner during the last six months were included as measures of microsocial domain. We also included drug related variables within the microsocial domain of the risk environment such as type of fist illegal drug used. We used the variable CODAR, created by the Pan American Health Organization, to denote those substances that increase the risk of HIV such us cocaine, heroine, and methamphetamine (CODAR vs. no CODAR). Other drug related variables included within the microsocial domain were receiving assistance at firs drug injection (yes vs. no), frequency of drug injection (daily vs. less than daily), crystal-meth use during the last 12 months, and frequency of alcohol and drug use during their sexual encounters (less than half of the times vs. half of the times or more). Monthly income (less than 212 USD vs. more) was added for assessing the micro-economic domain. Micropolitical environment was assessed by variables related to interactions with criminal justice or public health institutions. These included: having experienced syringe confiscation by police during the last six months (yes vs. no), and perceived syringe/needle availability (based on responses to the question “In the last six months, how easy or hard was it for you to get new, unused syringes when you injected drugs?;” responses were dichotomized as “hard/very hard” vs. “easy/very easy”).
For the outcome variable participants were considered to have participated in receptive needle sharing if they responded “sometimes,” “often,” or “always” to the question: “How often did you use a syringe that you knew or suspected had been used before by someone else?”
Procedures
Data were collected by trained interviewers who also had experience working with PWID. The survey includes a range of questions related to sociodemographic, sexual, and drug use risk behaviors, trajectories of drug use and treatment, exposure to health facilities, and other aspects of the physical and social risk environments that may increase the probability of HIV acquisition among participants.
Statistical analyses
Bivariate analyses were conducted to examine associations of gender and sociodemographic, behavioral, and context-related characteristics. We used chi square tests to compare proportions among men and women and Wilcoxon test for continuous variables. We then built generalized linear mixed models with random effects (GLMMIX) to identify factors independently associated with the odds of having receptive-sharing syringes practices in the past six months. As it is possible that observations within each city are correlated among each other, violating the assumption of independent observations, we use city of residence as a cluster variable in each model and we included random effects in the linear predictors as a mean to adjust the variance within and between cities.
As we want to test the hypothesis that independent variables play different roles on receptive-sharing needle, according with gender, interactions terms between all the predictors and sex were included in a second model assessed.
Then we built a we fitted a third GLMMIX stratified by sex and compared the outcomes between the models. Models were developed using an a priori design whereby variables with a significance level of p < .1 in univariate analysis were considered for inclusion in a multivariable model using the forward method. We analyzed multicollinearity between the predictor variables using the Pearson Correlation Test and we only included variables with a correlation lower than .3. We used the Hosmer and Lemeshow test to analyze the model’s goodness of fit. The statistical analysis was run using STATA version 14.
Ethical considerations
Potential participants were asked for verbal consent before beginning the screening interview. Ineligible individuals were offered free condoms, information and referrals for HIV testing. All participants signed an informed consent form. The Institutional Review Boards of the UCSD School of Medicine and the Mexico’ Instituto Nacional de Salud Pública (INSP) approved the study (Robertson et al., 2014).
Results
There were 824 PWID surveyed, among which 74.0% were men and 26.0% were women. The sociodemographic profile of participants shows conditions of greater vulnerability for women in terms of migration, schooling, occupation, and income compared to men (Table 1). While almost 39% of the women reported being migrant to the current city of residence, the proportion was about 23% among men (p < .001). Most women reported a schooling attainment lower than completed secondary, while the proportion among men was almost 34% (p < .001). Women also reported much lower access to formal employment than men (7.0% vs. 27.3%, p < .001), even though Hermosillo and Ciudad Juárez have an important presence of maquilas, which hired mainly women. This could express that these women face an especially difficult situation to get incorporated into the local labor markets. Consequently, sex work appears as an important source of income for women but not for men (35.6% vs. 0.15%, p < .001); however, their income level is lower than the one reported by men (p < .05).
When comparing trajectories of illegal drug use (Table 2), data show that women initiated their drug use with cocaine, heroin, or methamphetamine in a higher proportion than men, which is associated with a faster transition to drug injection (Ospina-Escobar, 2018). Regarding the first injection of an illegal drug, most women (90%) declared that someone had injected them, while the proportion among men was 68% (p < .001).
We found no differences on the type of drugs participants injected during last 12 months by sex, though women reported a slightly higher frequency of injection, which posit them in higher risk of blood infections and vein collapses (Table 2). On the other hand, women also reported a higher frequency of alcohol and drug use during their sexual encounters.
Regarding access to services and interaction with police, women reported a higher access to HIV testing, which could be associated to the local regulation of sex work in Ciudad Juárez and Hermosillo, which impose them regular HIV testing at a cost. We found no significant differences by sex on the experience of syringe confiscation by police and access to treatment for drug abuse (Table 3).
These descriptive findings suggest that men and women face differential structural characteristics and distinct barriers to access health care services. Therefore, it is possible to think that predictors for receptive needle sharing have differential effects and weights on men and women.
To test this hypothesis, we fitted generalized linear mixed models with random effects (GLMMIX). The first one includes sex as a predictor of receptive needle sharing, the second one includes interactions between sex and selected predictors and the third one is stratified by sex (Table 4).
As shown in Table 4, factors related with characteristics of drug use and drug injection initiation have a strong positive association with receptive needle sharing. For example, those who were injected for someone else at first drug injection were 60% more likely to get engaged in receptive syringe sharing than those who injected themselves (adjusted odds ratio [AOR] = 1.60, 95% confidence interval CI [1.11, 2.25], p < .05). As expected, those with higher frequency of drug injection are 80% more prone to share syringes than those that inject less than once a day (AOR = 1.80, 95% CI [1.17, 2.70], p < .05). Those who have used alcohol or drugs at least half of the time at their sexual encounters have 1.64 times more likelihood of sharing needles than those with lower frequency of drug or alcohol use while having sex (AOR = 1.64, 95% CI [1.16, 2.47], p < .05). On the other hand, duration between first illegal drug use and first injection shows a protective effect on the likelihood of syringe sharing as such, for each year of delay in first drug injection the odds of syringe sharing is reduce in 4% (AOR = .96, 95% CI [.93, .98], p < .05).
Political domain factors also shown significant and positive association with receptive syringe sharing. As such, participants who had experienced syringe confiscation by police during the prior six months are 54% more likely to share syringes sharing than those who haven’t faced this police practice (AOR = 1.54, 95% CI [1.13, 2.19], p < .05). Perceiving syringe availability as hard or very hard is also associated with higher proness to syringe sharing (AOR = 2.29, 95% CI [1.49, 3.32], p < .01).
Gender appears as a significant factor associated with syringe sharing. The analysis indicates women have 43% less likelihood of having receptive syringe sharing practices compared to men (AOR = .57, 95% CI [.34, .89], p < .01). Even if this finding is interesting, it doesn’t unravel which characteristics associated with the living conditions of gender play a role in the likelihood of syringe sharing among men and women. For that reason, we fitted a model with interaction terms between sex and selected characteristics of the risk environment.
The results from the second model, which includes interaction terms between sex and independent variables, indicates that adding interactive terms doesn’t adds much to the models, as none of the selected variables were significantly associated with the likelihood of syringe sharing. However, as sex appears as a driver for syringe sharing, as in the first model, we run a stratified logistic model to test the hypothesis about the different effects that same predictors could have in the likelihood of syringe sharing. The stratified logistic analysis shows that while having an illegal occupation is not significantly associated with syringe sharing among women (AOR = 2.97, 95% CI [.56, 15.59], p = .2), it is for men (AOR = 3.67, 95% CI [1.47, 9.20], p < .05). In terms of the political environment, the analysis shows that syringe confiscation by police is a factor associated with syringe sharing for men (AOR = 1.74, 95% CI [1.20, 2.50], p < .05) but not for women (AOR = .91, 95% CI [.38, 2.17], p = .8).
Contrary to expected, having have commercial sex partner and having used alcohol or drugs during their sexual encounters are variable significantly associated with syringe sharing among men (AOR = 1.88, 95% CI [1.06, 3.33], p < .05; AOR = 1.83, 95% CI [1.22, 2.76], p < .05 respectively), but not for women (AOR = 1.68, 95% CI [.71, 3.94], p = .2; AOR = 1.33, 95% CI [.42,4.25], p = .6, respectively).
Duration time between first drug use and first drug injection, perception of syringe availability, and having received assistance at first drug injection have the same positive association with the likelihood of syringe sharing for men and women, but their weights seems to be higher among women, which could mean that for them a poor perception of syringe availability, shorter time between first drug use and first drug injection, and receiving assistance at first drug injection could have a stronger effect on the odds of receptive syringe sharing.
Discussion and conclusion
The differential HIV prevalence reported in previous studies among men and women who inject drugs (Brouwer et al., 2012; Patterson et al., 2006) could respond to differentiated risk environments enabled by the social construction of gender in Mexico.
In this paper, we explored some characteristics that could build different risk environments for men and women. Particularly, we found that institutional settings such as syringe availability and police practices are important drivers for syringe sharing but have differential effects on the likelihood of sharing needles among men and women. We also found that social settings such as sex work and dynamics of injection drugs initiation have different impact among men and women.
In terms of institutional settings, for women the variable that has a stronger significant association with the likelihood of receptive needles sharing is syringe availability, while for men it is police practices. Findings show that those women with the poorer perception of syringe availability had three times more likelihood of syringe sharing than those with more positive perception. Consequently, our data suggest that increasing access to sterile syringes is critical to addressing HIV risk among PWID in Mexico’s northern region, especially among women.
Among men the combination of structural living conditions with police harassment and perception of insufficient syringe availability builds a risk environment that enables receptive syringe sharing. For example, syringe confiscation by police increases 82% the likelihood of receptive needle sharing, while working on an illegal context increases almost four times the likelihood of involvement in this risk behavior. Decriminalization of drug use and training of police forces are harm interventions that could address these components of risk environment among men.
In terms of social settings, involvement in sex work and using alcohol and drugs during sexual encounters are important predictors of syringe sharing among men but not among women. These could be associated with a higher stigmatization of male sex work and how it the limits uptake of healthcare services and their mental health (Crowell et al., 2017). Nevertheless, dynamics of transition to injecting drugs have stronger effect in the likelihood of sharing needles among women compared to men, as men often initiate women into injection and women used to have less skills about injection practices (Robertson et al., 2010).
Understanding how gender differences are associated with the dynamics of drug use is crucial to design specific interventions targeting men and women, but also transgender populations (Rusch et al., 2009). As women who inject drugs are more invisible, designing specific interventions for them is needed. This means to identify places where women who inject drugs gather and to penetrate in their social and sexual networks to distribute sterile equipment, rapid HIV test, and information, as well as addressing gender-based violence, as it is a factor associated with a higher prevalence of problematic drug use and risky injection practices (Jain et al., 2018).
In this paper, we show the importance of run differentiated analysis by sex in order to highlight how structural, biographical, institutional, and situational variables operate differentially for men and women who inject drugs. To recognize these differences and collect better data to follow up these findings is the first step to offer better services of harm reduction to people who inject drugs in Mexico aiming to reduce the gaps between men and women in a context of extreme vulnerability.
This paper has some limitations. Firstly, it is difficult to know to what extent the samples are representative of the broader population of PWID in the selected cities. As data were collected where PWID gathered, it is possible that the samples had higher levels of homogeneity among participants compared with the overall populations of PWID in each city. As such, findings cannot be generalized to the broader population of PWID in Mexico. Secondly, the sample was composed largely by men and it was not possible to oversample females as data were collected at places where PWID gather and these places use to be mainly inhabited by men, with scarce presence of women. To get more and better data for WID, it is needed to considerate specific sample strategies and try to oversample them to have a more robust comparison analysis with men data. We address these sample issues using a generalized mixed model, which allow us to take into account variability within-person and across-person variability, and thus, having more accurate estimations between men and women.
Thirdly, the cross-sectional nature of the survey data also limited our capacity to detect causal associations. As such, this study is restricted to highlighting the associative relationships between variables. More research is needed to identify the causal pathways and the complex set of relationships that influence syringe sharing and HIV prevalence among sampled PWID. Given the scarcity of data available, there is an urgent need to better understand the mechanisms of HIV spread among PWID in Mexico’s northern region. These findings provide an important preliminary insight about how gender construction shapes different risk environments for men and women who inject drugs.