Data Sources
The main data source used was the 2017 Senegal Demographic Health Survey (DHS), which is a population-based nationally representative survey in which participants were selected using a two-stage stratified cluster sampling design according to DHS sampling methodology (19). For the 2017 Senegal DHS, the following methodology was applied (3): At the first stage, 400 clusters were selected independently with probability proportional to size from the list of enumeration areas (EAs) established during the 2013 General Census of Population and Housing, Agriculture and Breeding (RGPHAE). In the second stage, a sample of 22 households per cluster, both in urban and rural areas, was selected by equal probability systematic sampling. A total of 8,800 households (4,092 in urban areas and 4,708 in middle rural) were selected. In each selected household, a questionnaire was completed to identify women age 15–49, men age 15–59, and children under age 5. Every eligible woman was interviewed with the DHS Woman’s Questionnaire, and in those households selected for the men’s interview, every eligible man was interviewed with the DHS Man’s Questionnaire.
In our study, a secondary analysis of the Senegal 2017 DHS data was performed.Participants from urban and rural areas were selected from all the 14 administrative regions of Senegal. The study focused on sexually active women age 15–49 and men age 15–59. Participants who had sex within 12 months before the survey and those who never had sex were excluded from the datasets, and accounted for 6,709 in women and for 5,320 in men. After weighting and considering missing responses, the sample size for our study was 9,995 women and 3,992 men. The diagram flow of the population of study is represented in Figure 2.
Outcome variable
The outcome variable was: ever tested for HIV in the last 12 months. It was measured on the basis of responses to the survey question asked of sexually active men and women: “Have you ever been tested for HIV in the last 12 months?”
Independent variables
As shown in the conceptual framework, the study considered explanatory variables related to socio-demographic and economic factors, sexual behavior, HIV knowledge, stigma, media exposure, and antenatal care.
Socio-demographic and economic variables included: age (15–24, 25–29, 30–34, 35–39, 40–44, 45–49, 50–54, 55–59); wealth index in quintiles (poorest, poor, middle, rich, richest); and area of residence (urban, rural). Another related variable was residence, with the country’s 14 administrative regions grouped in four geographic zones: West (Dakar, Thiès); North (Saint-Louis, Louga, Matam); Center (Diourbel, Fatick, Kaolack, Kaffrine); South-East (Tambacounda, Kedougou, Ziguinchor, Sedhiou, Kolda). For the analysis, the geographic zone was used rather than the administrative region. Others variables related to socio-demographic factors were: marital status (married, never been in union, divorced, widowed); and level of education (no education, primary, secondary, higher).
Factors related to sexual risk behaviors were: number of lifetime partners (1, 2 and more, don’t know/missing); and history of STI (had any STI in the last 12 months, "Yes/No”).). Factors concerning knowledge of HIV were: sought knowledge of a place to get HIV test ("Yes/No”);); knowledge about efficacy of HART during pregnancy (taking drugs to avoid HIV transmission to baby during pregnancy, "Yes/No”);); and knowledge about HIV and sexually transmitted infections (ever heard of STIs, ever heard of AIDS, knowledge about mother-to-child transmission of HIV). Participants were considered as having a good knowledge of mother-to-child transmission of HIV (MTCT) if they knew the three main transmission routes (pregnancy, delivery, breastfeeding).
To assess perceived HIV-therelated stigma, the variable was: “people hesitate to take HIV test because of the reaction of other people” ("Yes/No”).). For media exposure, the variables were: access to the Internet ("Yes/No”);); and ownership of a mobile phone ("Yes/No”).).
For women, an additional variable was considered: number of ANC visits (none, 1 and more). The modality “none” included women who did not have a birth in the last five years as well as pregnant women who made no ANC visit.
Statistical Analysis
The analysis was performed using STATA/SE 15.1 software. As stated above in the section on data source, a two-stage sampling design was adopted. To account for the survey’s multi-stage sampling design, all data were weighted to adjust for disproportionate sampling and non-response. For women and men, individual weights were applied. SVYSET command was used in Stata to adjust for the effect of complex sample design.
In the descriptive analysis, variables were presented in terms of the frequency and percentage of data for women and men. Intergroup comparisons were made using Chi2 test. The threshold of significance was set at 5%, and 95% confidence intervals (CI) were considered.
To assess the factors associated with HIV testing, two adjusted logistic regressions analyses were performed for women and for men, and adjusted odds ratios were calculated with their 95% confidence intervals. For women, we fit a logistic model of HIV testing with the following 12 independent variables: five-year age groups, zone, educational level, wealth quintile, marital status, knowledge of mother to child transmission of HIV, drugs to avoid HIV transmission to baby during pregnancy, had any STI in last 12 months, lifetime number of sex partners, self-stigma, mobile telephone ownership, and number of ANC visits. For men, a logistic model of HIV testing was fitted with the same independent variables as for women, except the number of ANC visits.
Collinearity was checked and some variables were excluded due to high collinearity with another variable. The variables “residence,” “knowledge of a place to get HIV test,” and “use of Internet” were removed from both the models for women and men. For women, the variable “knowledge of a place to get HIV test” was removed because of empty cells, and collinearity was found between “residence” and “wealth class”. For men, collinearity was found between “residence” and “wealth class,” “knowledge of a place to get HIV test” and “zone,” and “use of Internet” and “educational level”.