Target population and sampling

The target population was defined as MSM within feasible reach of prevention activities. Feasibility was related to meeting place attendance. In order to estimate the size of the target populations, a mapping exercise was undertaken in each city. Members of the populations under survey visited the selected cities and, with the help of the local population and other key informants, mapped the sites where MSM were concentrated, while simultaneously providing numerical estimates of the populations in each of these areas.

Total sample size was estimated, based on the expected differences in condom use rate at the end of the project, and adjusting for the design effect of cluster sampling, since the unit of randomization was cities.

Results of the mapping were applied to assign sample size by city, although it should be highlighted that mapping data primarily disclosed numbers of easily accessible MSM, that is, individuals frequenting common meeting spots for these groups.

Data collection

The questionnaire was developed by an international multidisciplinary and inter-institutional team. The design process was coordinated by the lead researcher responsible for the FPP evaluation, and included an active collaboration of local researchers, as well as key local stakeholders, particularly members of the populations under survey.

This questionnaire was designed to provide a comprehensive image of MSM and their contexts, including socio-demographic information (age, education, work situation, assets), data on sexual behaviour (particularly, types of sexual partners, kinds of sexual practice, condom use, and other details of their last three sexual partners), information on existing regular partners, knowledge about HIV/AIDS and STDs in general, and attitudes towards people living with HIV.

To ensure suitability of language and context, a series of focus group discussions and in-depth interviews with MSM were performed. Data collection took place between July and September 2003. Anonymous face-to-face interviews were conducted with participants. Most interviews took place in locations frequented by MSM when away from work, while others were conducted in work places. Since names were not registered, data cannot be linked to specific individuals. Each hot-spot was visited once and before starting an interview, potential respondent was asked to confirm that he had not been interviewed before for this project.

Before the commencement of each interview, the aim of the study was explained to potential candidates. Written informed consent was obtained from all participants. The participants were asked to sign or mark the consent form to indicate their consent. No attempt was made to validate the signature against any formal identification; the signature or mark was not scanned or otherwise captured in the electronic database; and the paper versions of the consent forms and questionnaires were secured in locked storage after data entry. A copy of the informed consent form was provided to each individual, which included contact information for the researchers.

Data management

Data from the questionnaires were digitalized with a data entry interface using LSD software (Sistemas Integrales, Santiago de Chile, Chile) previously used for complex surveys. As a way of guaranteeing data quality – specifically, conformity to information in the printed questionnaires – all surveys were entered twice, and inconsistencies in the first digitalization were revised against the printed forms and corrected. Variables included in the analysis were reviewed to eliminate outlier values. The resulting database was transferred to STATA 8.0 for analysis.

Variables & statistical analysis

Descriptive statistics were generated for the socio-demographic characteristics of MSM and reported rates of condom use. To examine associations between condom use with last each of the 3 partner and socio-demographic variables, logistic regression models were estimated, with condom use with each of the 3 last partners as a dependent variable, using an encounter panel allowing up to 3 observations per MSM. A significance level of 95% was applied for all analyses.

For descriptive statistics as well as regression, the design effect of cluster sampling was taken into account to estimate robust errors. The multivariable model applied to identify association with condom use was estimated using town fixed effects and random effects by MSM.

To facilitate inclusion in the multivariable model and interpretation of estimates, continuous variables were categorized into age, years of sexual life, years of schooling, and price paid for condoms. In each case, between three and four groups were created, according to the observed variable distribution. Additionally, three indices were generated and included in the model, specifically, socio-economic level, social capital, and life skills.

A socio-economic index was created using a polyserial correlation model with net income, possessions, and level of education as principal components. The resulting scores obtained with the principal components methodology were categorized in quartiles.

The social capital indicator was established using variables indicative of the MSM perception of the support they could receive in specific situations, while the life skills indicator was generated by taking into consideration variables focusing on self-perception of empowerment for condom use. The methods used for these two indices were analogous to those applied for a similar study in India [11].

All variables were introduced simultaneously in the model. Interactions between variables were explored, in particular those between social capital and life-skill indexes, and socioeconomic level and condom price categories.

Since it is argued that in the interpretation of odds ratios there is a tendency to over-estimate the risk (or protection) attributed to a variable [12], prevalence ratios were estimated indirectly from odds ratios according to a formula developed by Zhang and Yu [13]. While a number of alternative direct methods may be employed to assess prevalence ratios, most do not allow the estimation of two-level clustering.

The outcome variable for multivariate analysis was condom use, defined as whether condoms were used during intercourse with a given partner, that is condom use with each of the last 3 times he had penetrative sex (regardless of whether the last three times were with three different partners, two different partners or a single partner. Data were collected for up to 3 sexual encounters (last, penultimate, and antepenultimate).

Correlates of condom use

The prevalence ratio for condom use among unmarried in relation to married subjects was 5.45 (95% CI 4.236–6.37), and 0.9 for a female partner (95% CI 0.81 – 0.99). Data on age and education obtained from multivariate or bivariate analyses were not statistically significant.

The reported condom price was strongly correlated with condom use, suggesting that a higher cost significantly reduces use. It is important to highlight that this occurs in a model including the socio-economic level, and thus the issue is not the absolute cost of the condom, but rather its relative price.

MSM who reported a higher number of partners in the preceding week (five or more) used a condom more often (RP 1.06; 95% CI 1.05 – 1.07), although the effect was minor.

Alcohol consumption was not associated with condom use, either positively or negatively, and neither was family awareness of sexual preference.

MSM are more likely to use a condom with new partners than those they already know. Life skills also correlated strongly with condom use. However, the social capital indicator employed is not associated with the use of condoms.

Finally, the socio-economic level plays an important role, and is directly proportional to the use of condoms. In this case, a dosage effect is not clear, but MSM belonging to the higher half of the socio-economic divide use condoms more often. Significantly, this result was obtained by controlling for education level and condom price.

MSM in Ecuador

The population included in the MSM category is a very heterogeneous group. The width of the confidence intervals reflects important differences with regard to city, which need to be integrated in the design of interventions. For example, the percentage of MSM who had never used a condom varied from 4% to 38%, depending on the city.

The participants of the survey turned out to be a young and reasonably well-educated population, with an average of almost 10 years of schooling. Consistent with the profile of the group in the region, bisexuality was very common. In fact, a fifth of participants were married, and almost a quarter were parents. This specific profile needs to be considered for prevention, since messages focusing on men having sex exclusively with other men may not be adequate to target this population. Importantly, meeting places identified in this survey provide an avenue for reaching a broad spectrum of MSM, who are important channels for communicating information. The fact that more than 90% of this heterogeneous population is literate also provides an opportunity to use more complex and targeted messages.

While the interpretation of "community" is currently ambiguous, 87% of participants disclosed that their sexual preference was known in the community, while only 29% reported the same with respect to their family. This suggests a double life, which also has implications for prevention. It is important to distinguish prevention messages in environments where these individuals are in their MSM life from those in their heterosexual life. It also highlights the importance of providing information within the context of an environment that can be hostile to same-sex sexual behaviour.

It is important to acknowledge the limitations of these results, which are due, in part, to the type of sample as well as the analysis performed. Firstly, although every effort was made to include a representative sample of all MSM in the cities, it was, at best, a model of the most accessible MSM at the moment of the survey. Since we performed a cross-sectional analysis, causality was not attributed to proposed independent variables. Although we believe it unlikely that some men were interviewed twice (if they told the interviewer that they had already responded to the survey then the interview was aborted and there were no financial incentives for participating), no systematic effort was made to exclude repeat interviews.

With any stigmatized population there are concerns about how potential stigma and discrimination may have introduced participation bias. Despite assurances of anonymity, potential respondents who are more worried about possible breaches of confidentiality may have been less likely to participate in the survey. It is difficult to speculate about the likely direction of such bias. On the one hand it may be that those most worried about disclosure are men who lead heterosexual lives and rarely venture into MSM venues. Or, one could argue that men at highest risk, those selling sex or using drugs, may be most wary of law enforcement and may be less likely to participate in the survey.