Study setting
UMlazi Township, which is located in the province of KZN, is the second most populated township in South Africa, with an estimated population of more than half a million people (14). The township falls under the EThekwini Metro, which has the highest HIV prevalence in South Africa (9). Umlazi has ten public clinics serving an average total of more than 50 000 clients per month, and one public hospital. All ten public health clinics participated in the study.
Study design, participants and sampling
An analytic cross-sectional survey was conducted over a period of five months (November 2018 to April 2019). The study sample comprised of 471 women aged 18-48 years who had prior sexual exposure, residing at UMlazi Township and utilising health services at any of the ten UMlazi Township clinics participating in the study. The determination of sample size per each site (clinic) was proportional to the size and volume of patients seen at the facility. The District Health Information System (DHIS) was used to obtain each clinic's monthly headcount for the past six months preceding data collection, in order to determine the average patient volume. Potentially eligible women who utilised services at the time of data collection were approached, introduced to the study and invited to participate. Women below the age of 18 years and those aged 50 years and older were excluded. A convenience sampling technique was applied to enrol participants, due to time limitation. To ensure non-interruption of healthcare services, participants were only enrolled in the study after the health services had been rendered by clinic.
Study instrument and data collection
A structured interviewer administered questionnaire was designed based on literature review (15, 16) and translated in English and IsiZulu (local) languages. The trained and experienced two Research Assistants (RAs), competent in both languages administered the questionnaires, following a refresher training conducted by the principal investigator. The RAs were not part of any clinic staff, therefore, they were external and not associated with any clinic staff or participants. The RAs explained the questionnaires in detail to participants who needed support. A pre-test of study instrument was conducted with 10 participants who did not form part of the actual study. The study instrument collected data on the demographic and socioeconomic characteristics, participants’ awareness of modern contraceptives, the use of contraceptives and information related to sexual behaviour. A series of quality assurance processes were implemented to ensure data quality was not compromised but preserved, including data validation, data cleaning, questionnaire verification, as well as ensuring that questionnaires were tested for consistency.
Measures
Outcome variable
The outcome variable was ‘condom use at last sex’. This variable aimed to determine whether the participant had used a condom during the last sexual encounter. This variable is dichotomous and coded as: 0=no; 1=yes.
Predictor Variables
Selection of the following predictor variables was guided by the reviewed literature and the hypothetical foundation established from the reviewed literature. This study examined the variables at the individual level which included the characteristics of women aged between 18-49 years.
Description of predictor variables and coding
- Number of male sexual partners in the past three months
This variable aimed to determine the total number of sexual male partner(s) the participants had in the previous three months prior to the study. This variable was coded as: 0=none; 1=one; 2=more than one. ‘None’ was a reference category.
This variable aimed to determine whether the current sexual partner is employed or not. This variable was coded as: 1=no; 2=yes. ‘No’ was a reference category.
This variable aimed to determine whether the participant has ever been diagnosed HIV positive. This variable was coded as: 1=no; 2=yes. ‘No’ was a reference category.
- HIV status of sexual partner
This variable aimed to determine whether or not the sexual partner was HIV positive or negative. This variable was coded as” 1= do not know; 2=negative; 3=positive. ‘Do not know’ was a reference category.
- Sometimes hitting/slapping with partner
This variable aimed to determine whether there was sometimes violence perpetration between the participant and the partner. This variable was coded as: 1=neutral; 2=agree; 3=disagree. ‘Neutral’ was a reference category. ‘Neutral’ referred to participants who neither agreed nor disagreed.
- Lots of trust in the relationship
This variable aimed to determine whether there was a lot of trust between the participant and the partner. This variable was coded as: 1=neutral; 2=agree; 3=disagree. ‘Neutral’ was a reference category. ‘Neutral’ referred to participants who neither agreed nor disagreed.
- Partner has control over sex
This variable aimed to determine whether the partner had control over whether or not to have sex. This variable was coded as: 1=neutral; 2=agree; 3=disagree. ‘Neutral’ was a reference category. ‘Neutral’ referred to participants who neither agreed nor disagreed.
- Partner has control over condom use
This variable aimed to determine whether the partner had control over condom use. This variable was coded as: 1=neutral; 2=agree; 3=disagree. ‘Neutral’ was a reference category. ‘Neutral’ referred to participants who neither agreed nor disagreed.
- Ever diagnosed with STI in the past 12 months
This variable aimed to determine whether the participant had been diagnosed with an STI in the past 12 months. This variable was coded as: 1=no; 2=yes. ‘No’ was a reference category.
- Sex under the influence of alcohol
This variable aimed to determine the total number of times the participant had sex under the influence of alcohol in the past three months. This variable was coded as: 0=never, 1=more than one. ‘More than one’ was a reference category.
- Talked about condoms with partner in 12 months
This variable aimed to determine whether the participant had talked about condoms in the past 12 months. This variable was coded as: 1=no; 2=yes. ‘No’ was a reference category.
This variable aimed to determine the age in which the participant had sexual debut. This variable as coded as: 1=12-17 years; 2=18-24 years. Age-group ’12-17 years’ was a reference category.
Ethics
Ethical approval to conduct this study was obtained from the Biomedical Research Ethical Committee (BREC) at the University of KwaZulu-Natal (Ref No: BE424/18). The National Health Research Database (NHRD) from the KwaZulu-Natal Provincial Department of Health (Ref No: KZ_2018009_013), and The EThekwini District’s Ethical Review Committee also approved the study. Gatekeeper permissions were obtained from the participating facilities prior to data collection. To ensure confidentiality of respondents, no personal identifiers were captured in the questionnaires. Likewise, a written informed consent was obtained from the study participants prior to their enrolment. The principal investigator and two research assistants went through the informed consent with the potential participants in a language preferred by the participants. The study adhered to sound ethical standards including confidentiality, voluntariness of participation and full disclosure of the research process.
Data analysis
Data were coded, entered into Epi Data Manager (version 4.6) (17) and exported to Stata (version 15.0) (18) for analysis. Data cleaning was conducted to ensure there’s no discrepancies in data before analysis are carried out. Participants’ background information was analysed using descriptive statistics. The frequency distribution and cross-tabulations of each predictor and outcome variable was carried out for categorical data. Frequency distributions of continuous variables were tested for normality using Shapiro-Wilks test. A Pearson Chi-square test and logistic regression models (bivariate and multivariate) were employed to assess level of significance and the association between the dependent and exposure variables. A stratified, cluster analysis was used for statistical testing. The Pairwise correlation test was used to assess whether correlations between predictor variables existed. The significance level was kept at p<0.05 for all the analyses, and at a confidence interval of 95%.