Methodology
Study design
The research was conducted following a cross-sectional study design. This study was carried out as part of a broader initiative, and the data came primarily from the endline information of a longitudinal cohort study. The longitudinal study was based on a census of young males living in a slum setting and started in 2016. A pre-designed questionnaire was utilized to gather quantitative data on a wide range of different components., including socio-demographics, sexual activity, knowledge level, mental health status, and community ties. Two years later, in 2018, the cohort was followed up with a modified questionnaire version.
Study setting
Bhashantek- one of the oldest and largest slums, situated in the Dhaka North City Corporation area, was the study site. Around 31,000 people lived in the Bhashantek slum, which was separated into four distinct areas. These four areas were different in terms of population, and the one area with the most significant number of people was chosen to conduct this study.
Sample and sampling technique
The research was conducted in three different stages. The process started with identifying potential male participants, aged 18 to 29, enrolling in the baseline survey study, and following up after two years for an endline interview. In the initial stage, households where a man within the desired age range lived, were identified. During this first screening procedure, a total of 942 possible responders from 790 houses were spotted.
In the second stage of the survey, nine trained data collectors worked in the designated region of Bhashantek slum for a total of 26 days in December 2016, after the first household listing. A few names of the potential respondents were excluded from the baseline survey because of some of their migration to elsewhere; meanwhile, a few names were listed mistakenly in the initial list, and two people declined to give written consent.
Ultimately, 824 young males between the ages of 18 and 29 from the research region made up the final group for the baseline survey.
After two years, the endline survey was conducted, aiming to collect data from the same cohort of young men. This time, data collectors were sent to the research area with a modified questionnaire in July 2018. In the endline survey questionnaire, some sections like migration information, general health, equity, and source of entertainment were dropped and replaced with a few new ones like GAD-7, PHQ-9, satisfaction with life, and violence-related questions.
Approximately 19% of respondents were lost from the baseline data during this third part of the research because they relocated elsewhere. With one additional respondent participating in the baseline survey, the total number of answers received during the endline survey was 669.
In this study, those participants who participated both in the baseline and endline survey were included. So, after dropping the respondents who only participated either in baseline or endline survey, the final number of responses was 668.
Ethical consideration
This research received approval from BRAC University’s James P. Grant School of Public Health Institutional Review Board (IRB).
All the respondents had written their consent before participating in this study. Their identity and responses were kept entirely confidential and only used for research by the research team and the partner institutions involved. Participants were fully voluntary to participate in the study, and respondents received no monetary compensation or other rewards for their time. Moreover, no participants were engaged in the study design or the research questions development or analysis or interpretation of the data.
Outcome Variable: Risky sexual behaviour
Risky sexual behaviour (RSB) was characterized as having multiple sexual relationships or sexual activities without wearing a condom, as well as substance abuse that elevates the chance of contracting a sexually transmitted disease such as HIV (Chawla & Sarkar, 2019; Darteh et al., 2020; Muche et al., 2017).
A comprehensive questionnaire was created to elicit replies on the respondents’ sexual behavior in this study. The series of questions were asked to determine whether the participants had multiple sexual partners, used condoms inconsistently, or used sexual stimulants in the recent past. A binary outcome variable of risky sexual behaviour was generated to see if the respondent had reported engaging in any of these three activities or not.
Major determinant 1: Knowledge about STI/HIV: Methodology
Every participant was asked whether they had previously heard of sexually transmitted infections (STIs) to measure their knowledge (Abdul et al., 2018). Whoever said yes was then asked a set of questions to assess their knowledge level about the names of any STI, any symptom, name of any cause, and name of any preventive method, without the interviewer’s any prompting (Abdul et al., 2018; García et al., 2017).
The method of assessing the participants’ knowledge level was mainly adopted from a study conducted by García et al. (2017) and partially adapted from another study done by Abdul et al. (2018). Based on these two studies, four different binary variables were created to determine whether the participants had at least one correct knowledge about (a) STI types, (b) STI signs and symptoms, (c) STI causes, and (d) STI. Later, these four modules were added to create a new variable to see the respondents’ lowest to the highest level of knowledge about STI/HIV.
Major determinant 2: Mental health status: Methodology
Participants’ mental health status was evaluated by measuring their depression and anxiety levels using the Patient Health Questionnaire (PHQ-9) and Generalized Anxiety Disorder (GAD-7) tools (Coyle et al., 2019; Staples et al., 2019).
A total of nine questions were asked on the PHQ-9, while only seven were asked on the GAD-7, and both scales collect Likert-scale data, and the replies are recorded as 0, 1, 2, or 3 (Johnson et al., 2019). For each of the PHQ9/GAD7 questions, respondents were asked whether they had experienced the symptoms in the last two weeks and were scored accordingly (Coyle et al., 2019).
The total score on the PHQ-9 scale varied from 0 to 27 points after the scores of each answer were added together(Johnson et al., 2019). Following a similar procedure, the total score on the GAD-7 scale ranged from 0 to 21 points (Johnson et al., 2019). A total score ≥10 for the PHQ-9 and GAD-7 scales suggested clinical depression and anxiety disorder, separately (Coyle et al., 2019; Johnson et al., 2019). The PHQ-2 and GAD-2 scales were also developed, with the first two elements of the PHQ-9 and GAD-7 scales, correspondingly(Coyle et al., 2019).
The cutoff margin for both PHQ-2 and GAD-2 is ≥3 (Staples et al., 2019). PHQ-9 and GAD-7 have good validity, accuracy, and psychometric properties, as do their shorter versions, PHQ-2 and GAD-2 (Johnson et al., 2019; Staples et al., 2019)
Statistical analyses: Methodology
Exploratory analysis was carried out to obtain summary statistics for the socio-demographic features of young adults. Bivariate logistic regression was first to run to determine the association between the outcome variable, risky sexual behaviour, and two main determinants, namely knowledge of STI/HIV and mental health status.
The multivariate logistic regression model included various control variables such as age, education level, occupation, current marital status, and place of birth to identify any possible confounders. Later, an urban version of the wealth index variable was added to the regression model. The wealth index variable was generated using Equity Tool.
There are seven questions in the tool’s most recent version, which came out in 2014 (Chakraborty et al., 2016). Researchers verified that this Equity Tool/wealth index had been verified earlier for Bangladesh by researchers (Chakraborty et al., 2016). The multivariable logistic regression model incorporated all variables from the simple logistic regression model.
In the multivariate model, a ≤0.05 alpha value was considered to establish the statistical significance of any determinant. All the statistical analysis was done in Stata v17.0.