As a result of the discrimination, homophobic stigma and violence that is pitted against MSM, many tend to hide their identity and sexual orientation because of the unacceptability of same-sex desires amongst most of the South African population. Based on the above-mentioned socio-structural issues facing the study’s participants and respondents a number of assumptions were made. Firstly, it will be difficult to access and enrol the study’s participants and respondents and secondly that constructing a sampling frame would be challenging. The Best Practices Guidance Report for conducting HIV research with same-sex desiring groups asserts that working with MSM presents unique challenges due to stigma, discrimination and dangers that are experienced by same-sex individuals. These issues are further compounded by a lack of community structures which protect same-sex individuals (amfAR et al., 2011). Since the study used a mixed methods approach, when constructing the sampling frame it considered the various objectives that each methodological approach set to achieve. These contemplations had to be cognisant of the population type and the challenges that research focusing on a marginalised group such as BMSM presents for this study.
In quantitative sampling, it is important that a sample that best estimates the characteristics of the population
is selected. This is important because the methods of sampling influence the data collected from which analysis is done and inferences are drawn.
In line with the objectives in this study, the quantitative objective ultimately aimed to explain frequencies and to establish the degree of associations between variables that were representative of the population.
Therefore, a sampling frame that would best approximate the population of BMSM in Msunduzi Local Municipality was selected. This is in contrast to qualitative sampling (selection of participants), whose goal is not necessarily to generate a representative sample of the population but is rather concerned with developing a meaningful understanding about people’s behaviours, beliefs and perceptions or trends (Du Plooy, 2009).
Thus, a qualitative sampling frame does not focus on the resprestitivity of the participants selected but rather, how relevant they are to the research subject. Therefore, this study used acombination of the non-probability sampling techniques, respondent-driven sampling and convenience sampling to address the varying objectives of each methodological approach and to reach the hard to reach a population of BMSM.
The eligibility criteria required individuals to identify as gay, bisexual, straight, curious and to fall into the category of MSM. They had to have reported sexual activity with another man in the past 12 months and be a permanent resident of the Msunduzi Local Municipality with no plans to relocate during the period of the study. Participants’ age range was 18 to 54 years. Crucially, one needed to be willing to volunteer as a participant and be able to communicate in either English or IsiZulu. Non-probability sampling is a non-random process that is subjectively executed based on the participant’s availability and willingness to be involved in the study (Laher and Botha, 2012). The above-highlighted strategies were used at different stages of the recruitment process during either the quantitative (questionnaire), the qualitative (interviews) or in collaboration to reach hard to reach BMSM. This was done to create a more inclusive sample that was representative of the BMSM population in the Msunduzi Local Municipality. Johnston et al. (2008) assert that Respondent Driven Sampling (RDS) is a useful method for studying hidden populations that often do not have sampling frames. This includes HIV risk populations who practice behaviours for which they are stigmatised (e.g., same-sex intercourse). For this reason, it is understandable why these populations are hidden because the public acknowledgement of being a member of the population might place them at risk of discrimination and stigmatisation.
Considering the above, RDS was used to create the quantitative sampling frame by using members of the actual population and their social networks to recruit respondents. This type of decision is closely tied to the epistemological assumptions of the transformative paradigm which encourages the use of a more cyclic model of research (Mertens, 2007). The cyclic model encourages community participation throughout the inquiry process. The goal of a transformative position is to have research that contributes to a sustainable change in the community, hence the need for a cyclic approach. Therefore, an RDS frame allows for the community of interest to play an integral part in the research process. This sampling method was used more prominently to establish the sample for quantitative data generation as the RDS frame posited creating a representative sample, from which data could be statistically tested and for extrapolations to be made about BMSM in the Msunduzi Local Municipality.
A respondent-driven sampling (RDS) strategy uses the chain referral approach but incorporates social network systems to gather a sample that is representative of the target population by mitigating the biases commonly found in chain referrals. In this study RDS was employed in this sequence:
1. Participants were recruited in collaboration with the GLN who helped the researcher to identify a group of key members (seeds) who would be willing to be part of the study and to act as recruiters in the first wave of recruitment. From this assessment, five key members (seeds) of different sexual identifications in different areas were identified and recruited. This was based on the perceived influence that they hold in their communities and the size of their social network which placed them in a favourable position to reach a larger group of potential respondents. The seeds were the first group to complete the questionnaire.
2. Upon their completion of the questionnaire, the seeds were encouraged to recruit five members of the target population from their social network. They were issued with five coded recruitment coupons each to issue to potential respondents. The seeds were advised to provide the potential respondents with the necessary study information and refer them to the study Facebook page for further information. Their recruits presented these coupons upon enrolling for the research and completed the questionnaire making them the first wave of respondents.
3. The first wave of respondents were issued with four coded recruitment coupons to recruit 3-4 respondents from their social networks using the coupons. This became the second wave of respondents.
4. The second wave of respondents were issued with four coded recruitment coupons to recruit 3-4 respondents from their social networks using the coupons. These became the third wave of respondents.
5. The third wave of respondents were issued with four coded recruitment coupons to recruit 3-4 respondents from their social networks using the coupons. This became the fourth wave of respondents.
It was realised that during the fourth wave of recruitment that there was a saturation of social networks amongst the respondents, as the third wave of respondents did not manage to recruit more than one respondent each with some not even recruiting any respondent. Considering cost and time constraints since the researcher was residing out of the municipality at that time, it was necessary to develop a secondary process of recruitment in order to achieve the 100+ respondent rate required for significant quantitative analysis.
A convenience sampling frame was used with the intention to recruit members of the population that had not been reached in the RDS phase. Convenience sampling, which is also referred to as reliance on available subjects, entails selecting participants based on their availability, convenience and accessibility (Babbie and Mouton, 2001). Although this sampling technique does not aim to present statistical results, it was the available strategy taking into consideration the field and the researcher's constraints. Although this was the case, measures were put in place to manage resprestitivity of the population. In this second recruitment, drive participants were recruited through a social media marketing initiative. A Facebook page was then created to provide information to potential respondents and used to invite participants to be part of the study.
An online version of the questionnaire was created using Google Docs (Google, 2018). To protect the anonymity of respondents, there were no email addresses requested, making the documents accessible to anyone using any kind of browser without having to identify themselves. Previous participation assessment was created as the first entrance question to be answered before proceeding to complete the consent form.
On the form, it was ascertained by making the response to this question compulsory, therefore prohibiting the respondent from moving on to the next section if this question was not answered. This was measured through self-reporting to ensure that the researcher was aware if this was a duplication questionnaire.
The similar process to the one described above was used to ensure that informed consent was given. The informed consent form was created after duplication assessment at the beginning of the questionnaire and respondents provided consent for involvement before moving on. A link to the online questionnaire was posted on the study Facebook page and these posts were boasted to garner maximum reach after the necessary parameters (race, gender, location, age) were applied. Several posts with the link were shared on the study’s Facebook Page over a course of two weeks and growth in respondents was managed on the Google Form. Numerous MSM focused Facebook groups were identified for maximum reach and the post with the link was shared on these groups several times over the two-week period. Groups that were identified include UKZN-LGBTI-Forum Chasu PMB, PMB Top Bottom Verse, Ashdown Church and PMB Gay Guys.
A convenience sampling frame was used again for recruiting participants to take part in the semi-structured in-depth interviews after the questionnaires were collected and analysed. Respondents were requested at the end of the questionnaire to stipulate on both the printed and online format if they would like to take part in further research (see appendix 5). To protect anonymity, no names or specific details were required an email address was requested where interested participants could be reached for arrangements. It emerged that there were several participants who because of their SES did not have an active email address, these respondents were asked to supply their contact number(s). Of the 120 questionnaire respondents, 36 indicated interest in taking part in further research and supplied either their email addresses or contact number(s). Before the qualitative data collection commenced, a generic email was sent using the study’s official email address to the respondents. Furthermore, a generic short-message-system (SMS) was sent, where participants had to confirm interest in participation. Of the 36 who had initially indicated interest, only 13 confirmed participation and interviews were arranged. Thirteen interviews were scheduled, and three participants did not turn up for interviews and they could not be reached to facilitate a reschedule, so a sample of 10 interview participants was finalised.