Description: In the combined data set three individual data sets were merged; parents/guardians for children to 11 years, children 12 to 14 years, youths and adults 15 years and older.

The data set has modules that contain information on biographical data, orphan status, school attendance, media, communication and norms, knowledge, attitudes, beliefs and values about HIV/AIDS, sexual history, sexually transmitted infections, delivery and care details, male circumcision, HIV testing and risk perception, drug and alcohol use, health status, and violence in relationships. The datasets also contains biological markers such as HIV status, exposure to ARVs, viral load suppression and HIV drug resistance.

The data set contains 1 090 variables and 66 615 cases. Abstract: This is the fifth wave in a series of national cross-sectional surveys that have been undertaken every three years since 2002, by a research consortium led by the Human Sciences Research Council (HSRC). The study design and methods were based on the methods used and validated in the previous four surveys conducted by the HSRC in 2002, 2005, 2008 and 2012. The 2017 survey employed a slightly similar methodology compared to 2002, 2005 and 2008 surveys. In the first two surveys (2002 and 2005), a maximum of three people was randomly sampled in each household, based on pre-determined age categories, namely a child aged 2-14 years, a youth aged 15-24 years, and an adult aged 25 years or older. In 2008, infants aged 2 years and younger were included in the sample as a fourth possible person. In 2012, all household members were included in the survey, and the same approach was used in 2017, making it the fifth national-level repeat survey.

The main objectives of the survey were to undertake the following analyses on a household-level of a nationally representative sample of adults and children in South Africa:

To estimate the HIV prevalence at national, provincial and selected district levels among adults and children in South Africa.

To estimate the extent of exposure to antiretroviral therapy (ART) and the level of HIV drug resistance (HIVDR) at national, provincial and selected district levels among adults and children in South Africa.

To review the progress in reaching UNAIDS 90-90-90 goals for total HIV epidemic control.

Furthermore, to undertake the following analyses at the national and provincial levels and for selected districts:

To assess the prevalence of self-reported TB, related knowledge and attitudes.

To assess the relationship between social and behavioural factors and HIV infection.

To determine the viral load (VL) in HIV-positive individuals and estimate the proportion of persons receiving antiretroviral therapy who are virally suppressed.

The survey targeted 15 000 VPs, of these, 12 435 (82.9%) VPs were approached. Among these VPs, 11 776 (94.7%) were valid VPs. A household response rate of 82.2% was achieved from the valid VPs. In the 9 656 VPs that agreed to participate in the survey, 39 132 individuals were eligible to be interviewed and provide a blood sample. Among the eligible individuals, 36 609 (93.6%) agreed to be interviewed. Among the 39 132 eligible individuals, 61.1% provided a blood specimen for HIV-testing; the samples were anonymously linked to the completed questionnaires.

Clinical measurements

Face-to-face interview

Focus group

Observation

South African population.

As in the previous surveys, a multi-stage disproportionate and stratified cluster sampling approach was used. However, unlike the previous versions of household surveys, the 2017 survey used small area layers (SALs) as the primary sampling unit (PSU) instead of the enumeration areas (EAs) that were used in previous surveys. For the 2017 survey, a new HSRC master sample was developed through the disproportionate stratified sampling of 1 000 SALs from the newly released National Sampling Frame of 84 907 SALs from Statistics South Africa release 2011. The SALs were sampled with a probability proportional to its size with the number of VPs as a measure of size (MOS). The released SALs were updated in 2015 and mapped using aerial photography to create a new master sample, which was used as a basis for sampling households in the 2017 survey. The sampling steps are described below:

Target population: All people living in households in South Africa.

The sampling frame: We used the 2015 national population sampling frame of 84 907 small area layers developed by Statistics South Africa (StatsSA 2017b). From this sampling frame, we drew 1 000 SALs.

Primary sampling units: 1 000 SALs were sampled from the 2015 database of SALs.

The 2011 estimate of visiting points (VPs), MOS was used in sampling 1 000 SALs.

Allocation of the sample: SALs were disproportionately allocated according to province, race group and geographic type (geotype or locality type).

Strata: There were 9 provinces (n=9) and 3 geotypes or locality types (n=3).

Reporting domains: Geotype or locality type was n=3; age groups were n=4; sex was n=2; and race group was n=4.

Secondary sampling units: 15 VPs were systematically sampled from each of the selected 1000 SALs.

Ultimate sampling unit: All individuals living in a household were asked to participate in the survey.

The selection of SALs was stratified by province and locality type. In the four previous surveys, the locality type was stratified into four levels: urban formal, urban informal (informal or squatter settlements), rural informal (tribal area) and rural formal areas (farms). By contrast, in the current survey, only three locality types were identified: urban, rural informal (tribal area) and rural formal (farms). In urban localities, the race category was used as a third stratification variable, based on the predominant race group in the selected SAL (as informed by the 2016 Community Survey).

The survey over-sampled areas that were dominated by Indian, coloured or white race groups as well as the sparsely populated Northern Cape, to ensure that the minimum required sample sizes were obtained for the three minority race groups in South Africa and Northern Cape.

VPs and households were used as secondary sampling units (SSUs). Within each household, all consenting members formed the ultimate sampling unit (USU).

The preselected households for the study were identified using aerial maps, with the aid of GPS instruments if necessary. Up to four visits were made to each selected household to ensure maximum participation. This approach meant that an interviewer visited the household first, explained the purpose of the survey, sought informed consent from the head of the household or a representative, completed the VP questionnaire, and provided information about the study. The household was then informed that the same data collector (possibly together with other collectors) would conduct a survey with individual members of the household who were at home on the day of the first visit, or at a later visit at a pre-agreed time. Electronic tablets were used for data collection. Any household from which a member was absent at the first visit was revisited up to three more times. The revisits were conducted according to people’s availability on certain days of the week and at certain times of the day. The completed VP survey included a list of all the household members. A field supervisor ensured that all individual members of the household were accounted for by appropriately recording persons who refused to participate. The demographic characteristics (age, sex, race) of individuals who refused were obtained from the relevant VP questionnaire.