Study design: We conducted a cross-sectional study of 600 young female sex workers recruited by the respondent-driven sampling, RDS strategy. The study aimed to examine recruitment approaches, risk behaviors and willingness of young female sex workers to participate in an HIV vaccine trial.
Study site and population: The study took place in Dar es Salaam region at the Infectious Disease Clinic, IDC. The region has an estimated 4.4 million people [13]. It is one of the major cities with business, touristic and administrative importance in Tanzania. In this region, the HIV prevalence is at 4.7% which is slightly lower than the national prevalence estimated at 5% among adults ages 15 to 64 years [3]. In this region, various HIV preventive services are in place such as HIV counseling and testing, care and treatment, provision of contraceptive services and sexual reproductive health education to young people. IDC is a government facility located in the Ilala Municipality that provides youth-friendly health care services such as screening and treatment of sexually transmitted infections, contraceptives, counseling, testing, care and treatment for HIV, as well as general information about adolescent sexual and reproductive health [14]. The targeted population was young female sex workers aged between 18 and 25 years; who high-risk behaviors such as the history of any STI in the past 1 year, who had transacted sex in exchange for money, drugs, goods or gifts in the past 6 months before participation in this study. The young women who lived in Dar es Salaam for at least the past 6 months before the study and presented at the clinic with valid coupons. We excluded the participants who were either less than 18 or more than 25 years old and those without or with invalid coupons.
Sample size, sampling technique, and procedure: In this study, the participants were recruited from February to mid-May 2017 using respondent-driven sampling, RDS, a form of peer-driven chain referral sampling. This method reduced biases in recruitment by producing samples that were independent of the initial subjects from which sampling began [15]. Furthermore, it facilitated the recruitment of hard to reach participants by using a dual incentives system that compensated a participant for being interviewed and recruiting others in the study [15]. The findings of a formative study conducted among 15 informants between November and December 2016 using in-depth interviews set the foundation for the current study. The interviews gathered information on hotspots, scheduling of visits at trial centers, perceived stigma, and barriers to health-seeking behaviors. Initially, the researchers identified seeds or initial recruiters from 10 of 15 key informants. Eventually, the number of seeds increased to reach 20 because some initial recruitment chains died early by not referring their peers to the study site. The seeds recruited the first wave of study participants. Each of the seed and subsequent recruiters received three coupons with unique serial numbers. The researchers gave eligibility criteria to all recruiters as follows to recruit eligible female peers/friends, who were also exchanging sex for goods, gifts or money, whom they knew where they resided, to observe deadline for the validity of coupons given to recruited participants, of compensation as a recruiter after bringing eligible participant for enrolment in the research.
The sample size, n = 572 for the current study was calculated using the following formula n = [DE*Np (1-p)]/ [(d2 / Z2 1– α/2*(N-1) + p*(1-p)] by an OpenEpi, Version 3, open-source calculator. Definition of the formula is as follows: N = 6000, is estimated size of female sex workers population and p, 26.6% is the proportion of female sex workers infected with HIV in Dar es Salaam, d is standard of error, corresponding to 5%, and DE is design effect =2. For studies using respondent-driven sampling, the design effect of 2 is recommended to make a sample size twice as large as would be needed under simple random sampling. For this study, in order to account for seeds, missing data (drop-outs), non-response, and other sources of bias, the initial sample size above had been increased from 572 to 600 participants. In order to attain the required sample size, the research team distributed 1800 coupons, out of which a total of 650 participants were screened for eligibility. The study recruited 600 eligible participants out of 650 people screened for eligibility, including 20 participants who served as seeds during initial recruitment. Fifty participants, 8% were not included in the study because some came with expired coupons, were off age limit and others came after we stopped recruiting because we had reached the needed sample size.
At the clinic, each referred respondent voluntarily presented herself with a valid coupon, signed a consent form and filled the questionnaire. They have reimbursed a total of 18 United States of America dollars for being both study participants and recruiters. In order to minimize repeat enrolment among the study participants, the researchers recorded identifying information. That information, however, was not linked to questionnaires.
Data collection: Data collection was conducted using a self-administered questionnaire among 600 young female sex workers aged 18–25 years. In the current study, the questionnaires were prepared in English, translated to Kiswahili, and then back-translated to English for consistency. After signing of consent forms, the research participants filled in a questionnaire for assessing social network size, tracking of recruiter-recruit relationship and reasons for accepting the coupons. That process was followed by filling in of standardized behavioral assessment questionnaire for assessing risk behaviors, contraceptive uses, willing to participate in vaccine trials, utilization of health services, power dynamics within sexual networks concerning participation in vaccine trials and practices of using HIV preventive methods. For those who could not read and write, were asked to bring someone they trusted at the study site for assistance. After the participants submitted the questionnaires, research assistants reviewed the questionnaires to ensure the participants filled all questions correctly. Furthermore, the assistants provided clarified all questions that were not clear to the participants.
In our study tracking of recruiter-recruit relationship/recruitment patterns was done by documenting a unique identifying number of returned coupons, linking them with recruiter’s coupon unique identifying number. Like in RDS study conducted in United States of America, USA, the link between the recruiter and the recruit was documented by matching the serial numbers of the recruitment coupons given to each respondent with the serial numbers of the coupons returned to the project by the recruits [16]. All participants were counseled and tested for HIV only. The HIV testing followed The National HIV Testing algorithm that follows a ‘serial’ testing strategy. That is, a blood sample is first tested using a highly sensitive test, followed by a second highly specific test. A second test is only done when the first HIV test revealed an HIV- reactive result. HIV rapid testing can be performed in the laboratory or outside the laboratory setting by health service providers trained to performed HIV rapid tests [16]. The counselors referred to HIV positive participants to care and treatment centers of their own choice. The clinician provided a syndromic screening of other STIs to those requested. The counselors offered the participants condom, contraceptive pills and referral to the nearby health center as part of providing routine services.
Data management and analysis: In RDS, the sample is used to make estimates about social network connecting the population, then using information about social network size estimates about characteristics of the population are generated [17]. Estimates of weighted population proportions and 95% confidence intervals for select variables were calculated using the RDS-SS Analytical package. RDS-SS is a successive-sampling based estimator for population means based on respondent-driven sampling data when the estimated size of the target population is known. It addresses challenges of sampling without replacement in RDS researches [18]. Under successive sampling, larger units tend to be sampled earlier. This approach leverages the information in the decreasing size of sampled units overtime to make inferences about population size [19].
In the current study, we double entered the data to minimize entry error. We cleaned the data by comparing the unique RDS and registration numbers and other variables between the two data sets, data discrepancies identified were resolved by going back to the source documents (questionnaires). After that, coding of some variables into binary and categorical variables was done, with analytical outputs of weighted estimates of the population proportions of selected variables of interest. Our proposed estimator is based on estimated sampling weights [10], which are based on the sampling probability given by the successive sampling procedure, applied to nodal sampling units and sizes given by nodal degrees [19]. Furthermore, STATA analyzed weighted data, taking account of the weighting and clustering. The associations of willingness to participate in HIV vaccine trials using selected independent variables were determined through computing weighted data by STATA 13/SE package in regression models [20]. The variables with a P-value of ≤0 .2 were entered into the final model to look for factors associated with willingness to participate in trials, WTP. The odds ratios with 95% confidence intervals used to present the results of logistic regression models. After that, all variables that had a P value of ≤0.05 were regarded as statistically significant in the final model.