Theoretical framework
Based on socio ecological model (SEM), HIV risk perceptions or and behaviours leading to HIV infections can take place at different levels, including, individual, relational, community or social groups, and national policy (enabling environment) levels[38–40]. However, realising that SEM is flexible and that 'no one model is sufficient to describe factors that influence individual behaviour across the diverse domain', Baral and colleagues proposes for a "modified social ecological model" (MSEM) which incorporates the 'stages of HIV epidemic[41]. This modification underscores that apart from the fact that HIV infections and transmissions occur in diverse social cultural contexts, stages of HIV epidemics informed not only HIV risk perceptions and behaviour for HIV acquisitions and transmissions[41,42], but also public discourses and strategies regarding HIV preventions. Although, the modified version of SEM was formulated to analyze people's risk behaviour for HIV infections, we find it to be a relevant framework to understand FSWs’ perspectives toward HTS including HIVST, particularly regarding public discourses on risk perceptions and contribution of KPs in HIV transmissions.
Program setting
The study was conducted to inform a comprehensive HIV combination prevention program which delivers biomedical, behavioural and structural interventions to FSWs and other key populations in Tanzania (the Sauti program). Sauti is a five-year program funded by PEPFAR through the US Agency for International Development (USAID), administered by Jhpiego in consortium with other implementing partners and in conjunction with Tanzania’s Ministry of Health, Community Development, Gender, Elderly and Children (MoHCDGEC).
This study was conducted in 2017 in Dar es Salaam municipality, and Iringa, Mbeya and Shinyanga regions, all of the regions where the Sauti program implemented interventions in that year. With exception of Dar es Salaam (4.7%), the other three regions have HIV prevalence higher than the national average (5%) [43]. Dar es Salaam is the largest city in Tanzania located at the eastern part of Tanzania. Iringa and Mbeya regions are in the Southern highlands, and Mbeya borders Zambia and Malawi. Shinyanga is located around Lake Victoria. Iringa and Shinyanga regions have timber/tea plantations and mining sites respectively, which attract a high number of seasonal migrants including FSWs and labour forces. Dar es Salaam and Mbeya attract heavy traffic of travellers entering and leaving the country.
Study Design
We conducted a qualitative study utilizing two methodologies for capturing views and experiences. In-depth interviews (IDI) were used to capture experiences / opinions which were sensitive or of a personal nature, and participatory group discussion(s) (PGD) were used to capture less sensitive community or societal norms towards HIVST.
Sampling and recruitment of study participants
We conducted IDIs with FSWs from seven wards across the four regions, with three participants per ward (overall, 21 IDIs). We anticipated that by the time we conduct 21 IDIs, we shall have reached data saturation. In case the saturation was not reached by the 21 IDI we planned to continue recruiting new batches of FSWs and interview them until the saturation point is reached. Eligible participants for this study were; females, aged 18 years or above, self-identified as a sex worker, consenting to participate in the study. The Sauti Program definition of a sex worker is someone whose primary source of income (i.e. over half of their income) derives from payment for sex.
Both purposive and snowball (i.e. peer referral) sampling strategies were used to recruit IDI participants. Community gatekeepers/ stakeholders, including owners of entertainment venues and civil society organisations (CSOs) provided us an initial list of FSWs to contact. The first batch of FSWs were influential, outspoken or peer leaders of FSWs. The first batch of FSWs group was asked to recruit up to three FSWs counterparts (second batch of FSWs).
We also conducted PGD sessions. Similar to the procedures used to recruit participants for IDI, we employed snow ball sampling to recruit participants for PGD sessions. Initial list of PGD participants were drawn by researchers in consultations with the CSO providing HIV control services to FSWs. A total of 23 PGD sessions involving 227 participants were conducted across the four regions.
Data collection
Data collection was conducted by six interviewers trained in social science disciplines and experienced in qualitative study methods. During a two-week training, interviewers were trained on research ethics, consenting procedures, confidentiality and techniques to elicit sensitive information as well as being oriented to study tools and processes. This included a practical period piloting study tools and conducting interviews.
IDIs were conducted in a private location, such as guest house rooms, or in the offices of civil society organizations serving FSWs. IDI were conducted by a single interviewer in Swahili. Since at the time of this study, HIVST had not yet been introduced in Tanzania, prior to the interview, we read to the participant a description of HIVST. After obtaining consent to participate as well as to audio record the interview, the interviewer started with general questions on awareness and access toward HIV testing services. This was subsequently followed by questions about HIV risk perceptions, sexual behaviour and treatment seeking for sexual and reproductive health problems issues. On average, each IDI took about 45 to 60 minutes.
Similar to IDIs, PGDs sessions were conducted in private rooms in the community setting, including offices of CSOs or guest house rooms. PGD took a form of modified focus group discussion with participants ranging from between 8 and 12 persons. Prior to the PGD sessions, participants filled in a form to collect socio demographic data (but no identifying information was collected).
Three members of the study team facilitated the PGD: a facilitator, a moderator and a note taker. Alongside audio recording PGD sessions, the interviewers took notes which were typed into a laptop computer for further analysis. Two senior social science researchers fluent in both Swahili and English (SN and DN) reviewed the notes from PGD session and provided feedback in real-time to the research team. On average, a PGD session took 90 to 120 minutes.
Data Management and Analysis
On a daily basis, audio files were transferred to a secure server using terms described by a File Transfer Protocol. All audio files were transcribed verbatim in Swahili and translated into English. The written notes from PGDs were typed and merged with corresponding transcribed texts.
The two senior research scientists (SN and DN) were responsible for analysis, including development of themes and codes. A three-stage data analysis process was conducted. In the first analysis stage, a narrative including verbatim transcripts from each PGD and IDIs were written. In this narrative, the researchers used both inductive and deductive methods to identify emerging themes of interest. The predetermined (a priori) themes were coded using the codes developed during the study design stage as well as new ones that emerged during content analysis. The second analysis stage consisted of transferring the compiled data by themes into the qualitative software (Nvivo 11) which allowed to identify regularities and patterns. The conclusions were drawn based on the predetermined and emerging themes, regularities, patterns, and causal flows towards study objectives.