Socio-economic Status and Associated Factors in the Uptake of HIV Testing: Findings from the South African Population-based National Household Survey Conducted in 2017

Background: Improved understanding of barriers to HIV testing is important for reaching the rst of the UNAIDS 90-90-90 targets, which states that 90 percent of HIV positive individuals ought to know their HIV status. This study examined socioeconomic status (SES) factors and related covariates associated with HIV testing uptake in the general population in South Africa. Information on SES and HIV testing may impact the HIV treatment and care cascade and prevention of HIV transmission. Methods: This study used data obtained from a national cross-sectional, population-based household survey conducted in 2017 using a multi-stage stratied random cluster sampling design. A composite SES score was created using multiple correspondence analyses of household assets; households were classied into wealth quintiles and dichotomized into low SES/poorest (lowest 3 quintiles) and high SES/less-poor (highest 2 quintiles). Bivariate and multivariate logistic regression models were tted to analyse associations between the uptake of HIV testing and a set of socio-demographic and HIV-related variables. Results: HIV testing uptake was 73.8% and 76.7% among low and high SES household, respectively, both of which were below the rst 90 target. The ndings showed increased uptake of HIV testing was associated with being female and having a higher educational level independent of SES. Decreased uptake of HIV testing was associated with being from White and Indian/Asian race groups, residing in rural formal areas, alcohol consumption, and high self-perceived risk of HIV infection. Conclusions: We found that HIV testing uptake was similar for people from both low and high SES. However, interventions aimed at improving HIV testing in order to reach the rst 90 target, should target males, all race groups, those with low education attainment, and those residing in rural formal areas. Reaching the rst target will require efforts to improve community knowledge, and awareness about the negative impact of alcohol use and address HIV risk perception.


Background
Sub-Saharan Africa (SSA) bears the largest burden of the HIV epidemic with 53% of the world's 36.9 million [31.1-43.9 million] people living with HIV in 2018 [1]. The HIV epidemic in South Africa is the largest globally, with a national prevalence of 14%, which translated to an estimated 7.9 million people living with HIV in 2017 [2]. HIV testing uptake serves as the foundation for the Joint United Nations Programme on HIV/AIDS (UNAIDS) strategic framework in the ght against HIV [3]. This framework speci es that 90% of HIV-positive people should be aware of their status, 90% of those diagnosed should receive sustained antiretroviral therapy (ART), and that 90% of those on ART achieve viral suppression [3]. This framework was adopted in December 2014 by the South African government as the basis of its national response to the HIV/AIDS epidemic in the country.
Achieving the UNAIDS and South African government targets hinges on reaching the key goal of identifying 90% of people living with HIV.
HIV testing is crucial as it provides a diagnosis for people living with HIV for linkage to care and treatment [4]. Globally, progress is being made, and at the end of 2017, three-quarters of people living with HIV knew their status [1]. Despite the excellent strides made with HIV testing in SSA, awareness of HIV status remains lower than the rest of the world. Findings from 10 population-based studies conducted in 2012, identi ed that the average percentage of people living with HIV who know their status was below 40% [5].
In Southern Africa, despite the progress South Africa has towards reaching the 90-90-90 goals, there are still barriers to HIV testing [3]. Several barriers to HIV testing have been identi ed in studies that have been done in SSA. In the 2005 Demographic and Health Survey in Cote d'Ivoire, a low socio-economic status was related to lower proportions of individuals testing for HIV testing, which included lack of or lower educational attainment and employment status [6,7]. Furthermore, cross-sectional surveys conducted as part of the Multi-country African Testing and Counselling for HIV (MATCH) study in Burkina Faso, Kenya, Malawi, and Uganda found that a higher SES was an indicator of the probability of HIV testing [8].
Other barriers identi ed from various studies included a perceived low risk of infection [5,7,9,10] as well as fear of nding out ones' status and perceived stigma [10][11][12][13]. A cross-sectional study in 2011 in Ethiopia found that HIV testing was lower in rural communities with lower SES [14]. A household survey conducted in 2007 in ten countries in Southern Africa found that people living in urban areas are more likely to be tested [15].
Knowing one's HIV status should be independent of one's SES and where one resides [16]. Unless these barriers to testing are addressed, the nal goal of the UNAIDS targets where 90% of people who are on treatment are virally suppressed will not be attained [3]. Improved understanding of testing barriers is important to allow interventions focused on reaching the rst 90 target. This study examined SES factors and related covariates associated with self-reported history of ever testing for HIV using a national survey in South Africa.

Data
This analysis used data from a national cross-sectional, population-based household survey "South African national HIV prevalence, incidence, behaviour and communication survey" conducted in 2017 using a multistage strati ed random cluster sampling approach, described in detail in the survey report [2]. In summary, a total of 1 000 small area layers (SALs) were sampled using the 2015 national population sampling frame of 103 000 SALs developed by Statistics South Africa [17]. The selection of SALs was strati ed by province, locality type (urban formal, rural formal, and rural informal/tribal areas) and race groups in urban areas. A total of 15 visiting points (VPs) were randomly selected from each of the 1 000 SALs, targeting 15 000 VPs. Of these, 12 435 (82.9%) VPs were approached. Among these VPs, 11 776 (94.7%) were valid and a household response rate of 82.2% was achieved from the valid VPs. All members of the selected households were eligible to participate in the survey.
The survey collected data using a household questionnaire and three age-appropriate questionnaires were administered to consenting individuals. For those younger than 18 years of age, consent was given by parents/guardians and assent by the participant. The interview instruments solicited information among others on socio-demographic characteristics, HIV-related knowledge, attitudes, and behaviours, including questions on HIV testing. Questionnaire data were collected digitally using electronic tablets. Fieldworkers also collected dried blood specimen samples from participants using a nger prick. Samples were sent to a centralised laboratory for HIV antibodies. The focus of the present study was on youth and adults 15 years and older who responded to the question on HIV testing.

Primary outcome variable
The dependent primary outcome variable 'HIV testing' was ascertained by the question: "Have you ever been tested for HIV? Yes=1 and No=0." Ever been tested was de ned as having accessed HIV testing services at least once before the survey.

Explanatory variables
Independent variables included socio-demographic characteristics and socio-economic characteristics. Sociodemographic characteristics comprised age (15 to 24 years, 25 to 49 years, and 50+ years) sex (male and female), race (Black African, White, Coloured, and Indian/Asian), and current marital status (married and not married; which included divorced/separated and widowed/widow). Socio-economic characteristics included highest educational level completed (no education, primary, secondary, and tertiary), employment status (not employed and employed), locality type (urban formal, rural informal, rural formal), and asset-based SES which was constructed using multiple correspondence analyses (MCA), based on questions on the availability of essential services and ownership of a range of household assets [18]. MCA is a data reduction technique for categorical data, which calculates a composite indicator score computed by adding up all the weighted responses. The predicted score for each household was used to compute ve quintiles (1st lowest, 2nd lower, 3rd middle, 4th higher and 5th highest) representing a continuum of household SES from the poorest to the least poor. The quintiles were then dichotomized into low SES or poorest (lowest 3 quintiles) high SES or lesspoor (highest 2 quintiles).
HIV-related risk factors included age at sexual debut (less than 15 years and more than 15 years), age-disparate partnerships (partner older by 5 years, partner younger by 5 years, a partner within 5 years), multiple sexual partners in the last 12 months (one partner, and two or more sexual partners), condom use at last sex (no and yes), alcohol use risk score (abstainers, low, high, and hazardous risk drinkers) based on the Alcohol Use Disorder Identi cation Test (AUDIT) scale [19,20], correct HIV knowledge and myth rejection (yes and no) based on responses from the following questions (Can AIDS be cured? Can a person reduce the risk of HIV by having fewer sexual partners? Can a healthy-looking person have HIV? Can a person get HIV by sharing food with someone who is infected? Can a person reduce the risk of getting HIV by using a condom every time he/she has sex?), self-perceived risk of HIV infection (no and yes).

Statistical analysis
All statistical analysis was done in STATA 15.0 (Stata Corporation, College Station, Texas, USA) software. The 'svy' command was used to introduce weights which take into account the complex design of the survey. Data descriptive statistics (unweighted frequencies and weighted percentages) were used to summarize sociodemographic characteristics and HIV-related risk factors of the study sample. The Pearson's chi-square test was used compare differences in proportions between categorical variables. Bivariate logistic regression was used Page 5/18 to assess the relationship between HIV testing uptake and the explanatory variable. Statistically signi cant variables were entered into a multivariate logistic regression model to examine the independent effects of covariates associated with the uptake of HIV testing. The analysis was strati ed by asset-based SES (low and high). Crude and adjusted odds ratios (aORs) with 95% con dence intervals (CI) , and p-values less than 0.05 were reported for all statistically signi cant associations. Coe cient plots were used to display the results of the nal models [21].
There were signi cant differences in characteristics between participants with low and high socio-economic status with regards to age, sex, race, marital status, level of education, employment status, and locality type (p < 0.001).  Table 2 shows the reported uptake of HIV testing among the population aged 15 years and older, by assetbased socio-economic status and socio-demographic characteristics of the study sample. Overall, people with a high SES reported signi cantly higher HIV testing uptake than those with a low SES, 76.7% vs 73.8% (p < 0.001).
There were signi cant differences in HIV testing by SES among people aged 25-49 years, males, Black African, employed, and those residing in rural informal areas (all p < 0.001). HIV testing uptake for those aged 50 years and older, not married, residing in urban and rural formal areas were also signi cantly different between low and high SES (all p < 0.05).  Table 3 shows the reported uptake of HIV testing among the population aged 15 years and older by asset-based socio-economic status by HIV-related risk characteristics of the study sample. There were signi cant differences found in reported HIV testing between the people from low and high SES with for those with no self-perceived HIV risk and survey HIV status result (both p < 0.001). Differences were also found among people that ever had sex and those who did not have sex, differences with people that have a sexual partner ve years and older than themselves, those that do not drink alcohol and risky alcohol drinkers, and correct knowledge of HIV prevention and transmission (all p < 0.05).

Discussion
This nationally representative population-based study undertaken to identify risk characteristics showed that people with a high socio-economic status reported signi cantly higher HIV testing uptake than those with a low SES, 76.7% vs 73.8%. The differences in HIV testing between high and low SES were found in certain demographic and HIV-related risk characteristics. Bivariate analysis showed that people from high SES that were aged 25 years and older, not married, females, Black African, employed, and residing in all locality types were more likely to have tested for HIV compared to those with a low SES. However, the 90% target for HIV testing has yet to be achieved. In addition, HIV-related risk characteristics that were signi cantly associated with HIV testing were self-perceived HIV risk perception, knew survey HIV status result, sexual activity, differences in age of sexual partner, alcohol consumption, and correct knowledge of HIV prevention and transmission.
Findings from the multivariate logistic regression models showed increased uptake of HIV testing was associated with being female and having a higher educational level, independent of SES. Higher levels of HIV testing among females is consistent with global statistics [22,23]. The limited male participation in HIV testing observed in our study is worrisome since HIV positive men who are unaware of their status may continue to engage in unsafe behaviour [24]. These results highlight the importance of developing interventions to reach men and address gender disparities in HIV testing.
Our ndings are consistent with other studies conducted in sub-Saharan Africa where HIV testing was associated with educational attainment [25][26][27]. Those with higher educational attainment may be more knowledgeable about HIV and understand the bene ts of testing and the importance of knowing one's HIV status and have greater control over the decision to test [28]. Interventions are needed to improve testing rates among the less educated population group.
Similar to other studies, HIV testing uptake was higher in the older age group than among adolescents [29].
Young people report signi cant psychological barriers to HIV testing including a lack of community support and perceived negative attitudes of health care workers [29,30]. For adolescents in high SES households, addressing their age-speci c concerns is important. HIV testing is the rst step into both the HIV prevention and treatment cascade, and can positively in uence one's sexual risk behaviour and reduce the likelihood of future HIV acquisition or further transmission [31]. Addressing this age gap in HIV testing is crucial for improving the awareness of HIV status among young people and to ensure that HIV positive youth receive sustained treatment and are virally suppressed to save lives and curtail the transmission of HIV.
In agreement with other studies, our ndings showed that uptake of HIV testing in the rural informal areas was low compared to the urban areas [32]. This demonstrates that this rural-urban gap in testing persists in the modern era of effective HIV therapy when early diagnosis and linkage to care as part of the UNAIDS 90-90-90 targets is even more essential. Some overarching factors include poverty in many rural areas, limited resources, and structural barriers that pose challenges to accessing HIV testing services [33]. This accentuates the need for scaling up effective models for delivering HIV testing in remote areas, such as mobile and home-based testing [34]. Provider-initiated modes of testing may also be more accessible to lower socio-economic groups than traditional HIV counselling and testing. Testing through the prevention of mother-to-child transmission services has proven to minimise the socio-economic hurdles faced [8]. Context-based strategies are therefore crucial to improve access to HIV testing services [4,14], and cognisance needs to be taken of those who do not interact with the health system often [15].
Consistent with current ndings, other studies also showed that married women were more likely to undergo HIV testing than unmarried women [35,36]. This may be re ective of marriage as a supportive relationship that motivates couples to seek HIV testing [36]. Couples are also more likely to undergo HIV testing when they are planning their marriage to know their HIV serostatus, than those who have never been married [35]. Targeted interventions are needed to increase HIV testing among those that are not married.
The results showed that Whites and Indians/Asians were less likely to test for HIV compared to Black Africans who form the majority in the South African context. This is because of the underlying evidence-based narrative that suggests Black Africans are at greater risk of HIV infection for historical reasons, which account for both the economic inequality and the poverty that pervade this racial group. Elsewhere, racial differences in HIV testing have been attributed to the health services utilization model [37]. HIV testing was associated generally with the self-perceived risk of HIV among minority groups. There is a need to improve current practices to reduce health disparities generally, and disparities in HIV testing in particular. [37]. In line with other studies, the ndings also revealed that not being married was associated with low uptake of HIV testing [25,35]. This is worrying given that not knowing one's HIV status is a major impediment to accessing appropriate HIV prevention, care, and treatment [38]. Individuals in non-marital relationships may be more fearful of HIV positive test results. This raises serious public health concerns when considering that the majority of these individuals may be engaging in unprotected sex and would bene t from being aware of their HIV status [38]. There is a need to devise alternative approaches to create demand for the uptake of HIV testing services among nonmarried individuals with no prior exposure to HIV testing.
Low and high risk alcohol drinking was associated with low HIV testing uptake, independent of SES. Elsewhere, alcohol consumption has also been observed as a barrier to prior HIV testing in a population-based study [39].
Alcohol use has been consistently associated with risky sexual behaviour [40,41]. This highlights the importance of addressing social risk behaviours such as alcohol consumption, towards improving uptake of HIV testing and the prevention of HIV transmission.
High self-perceived risk of HIV infection was associated with low uptake of HIV testing. This implies that individuals might refuse HIV testing even if they know that they are at high risk of HIV infection. This differs from the view that people who refuse HIV testing commonly do so because they do not perceive themselves to be at risk [42]. Either way self-perceived risk is an important factor in the uptake of HIV testing. This emphasizes a need to assess whether an individual's self-perception of their HIV risk is accurate -to encourage testing, make the diagnosis, and link patients to care and treatment services [43].
The study has some limitations. Due to the cross-sectional design of the current study, causal inferences cannot be drawn, and the analysis can only identify associations. Additionally, the strength of the association was limited by wide con dence intervals in some instances. Furthermore, the analysis is based on self-report of HIV testing and risk factors, which are prone to social desirability and recall bias. Despite these limitations, this study used a nationally representative sample, and the ndings are generalizable to the entire country.

Conclusions
The ndings from this study show that socio-demographic factors such as gender, race, locality, and including