Socio-economic determinants of HIV testing and counselling: a comparative study in four African countries
Research indicates that individuals tested for HIV have higher socio-economic status than those not tested, but less is known about how socio-economic status is associated with modes of testing. We compared individuals tested through provider-initiated testing and counselling (PITC), those tested through voluntary counselling and testing (VCT) and those never tested.
Cross-sectional surveys were conducted at health facilities in Burkina Faso, Kenya, Malawi and Uganda, as part of the Multi-country African Testing and Counselling for HIV (MATCH) study. A total of 3659 clients were asked about testing status, type of facility of most recent test and socio-economic status. Two outcome measures were analysed: ever tested for HIV and mode of testing. We compared VCT at stand-alone facilities and PITC, which includes integrated facilities where testing is provided with medical care, and prevention of mother-to-child transmission (PMTCT) facilities. The determinants of ever testing and of using a particular mode of testing were analysed using modified Poisson regression and multinomial logistic analyses.
Higher socio-economic status was associated with the likelihood of testing at VCT rather than other facilities or not testing. There were no significant differences in socio-economic characteristics between those tested through PITC (integrated and PMTCT facilities) and those not tested.
Provider-initiated modes of testing make testing accessible to individuals from lower socio-economic groups to a greater extent than traditional VCT. Expanding testing through PMTCT reduces socio-economic obstacles, especially for women. Continued efforts are needed to encourage testing and counselling among men and the less affluent.
La recherche indique que les personnes dépistées pour le VIH ont un statut socioéconomique plus élevé que celles qui ne sont pas dépistées, mais on en sait moins sur la façon dont le statut socioéconomique est associé aux modes de dépistage. Nous avons comparé les individus qui ont été testés à travers l'initiative du prestataire pour le conseil et dépistage (PITC), ceux qui ont été testés à travers le conseil et dépistage volontaire (CDV) et ceux qui n'ont jamais été testés.
Des surveillances transversales ont été menées dans des établissements de santé au Burkina-Faso, au Kenya, au Malawi et en Ouganda, dans le cadre de l’étude MATCH (Conseil et Dépistage du VIH Multi-pays Africain). 3.659 personnes ont été interrogées sur le statut du test, le type d’établissement pour le test le plus récent et le statut socioéconomique. Deux mesures de résultats ont été analysées: jamais testé pour le VIH et le mode de dépistage. Nous avons comparé le CDV dans les établissements autonomes et le PITC, qui comprend des établissements intégrés où le test est fourni avec des soins médicaux, et les établissements PTME (prévention de la transmission de la mère à l'enfant). Les déterminants pour n'avoir jamais été testé et pour l'utilisation d'un mode particulier de dépistage ont été analysés en utilisant la régression de Poisson modifiée et des analyses logistiques multinomiales.
Le statut socioéconomique plus élevé a été associé à la probabilité de tester à travers le mode CDV plutôt qu’à travers d'autres établissements ou ne pas tester. Il n'y avait pas de différences significatives dans les caractéristiques socioéconomiques entre ceux testés à travers le PITC (établissements intégrés PTME) et ceux non testés.
Les modes de dépistage à l'initiative du prestataire permettent l'accès au dépistage aux personnes appartenant aux groupes socioéconomiques inférieurs dans une plus grande mesure que le CDV traditionnel. Etendre le dépistage par la PTME réduit les obstacles socioéconomiques, en particulier pour les femmes. Des efforts continus sont nécessaires pour encourager le conseil et le dépistage chez les hommes et les moins nantis.
Investigaciones previas indican que los individuos que se realizan la prueba del VIH tienen un mayor estatus socioeconómico que aquellos que no se la hacen, pero se conoce menos acerca de cómo están asociados el estatus socioeconómico con las diferentes formas de hacerse la prueba. Hemos comparado a individuos que se han realizado la prueba a través de un programa de aconsejamiento y prueba iniciado por el proveedor (APIP), aquellos que se la han hecho siguiendo un aconsejamiento y prueba voluntarios (APV), y aquellos que nunca se han hecho la prueba.
Se realizaron estudios croseccionales en centros sanitarios de Burkina Faso, Kenia, Malawi y Uganda, como parte del estudio MATCH (por sus siglas en inglés “Multi-country African Testing and Counselling for HIV”). Se preguntó a 3,659 personas si se habían realizado o no la prueba de VIH, el tipo de centro en el que se habían realizado la última prueba y su estatus socioeconómico. Se analizaron dos parámetros: si se habían sometido a no alguna vez a la prueba del VIH y el tipo de prueba al que se habían sometido. Comparamos el APV en centros independientes y el APIP, incluyendo a centros sanitarios integrados, en donde la prueba se ofrece con cuidados médicos y prevención de la transmisión vertical (madre-hijo). El análisis de los determinantes de haberse realizado la prueba alguna vez y de utilizar una forma particular de prueba se hizo mediante la regresión modificada de Poisson y la regresión logística multinomial.
El tener un mayor estatus socioeconómico estaba asociado con la probabilidad de realizarse la prueba mediante APV comparado con utilizar otros centros o no hacerse la prueba. No había diferencias significativas entre las características socioeconómicas de aquellos que se realizaban la prueba en APIP (centros integrados y centros de prevención de la transmisión vertical) y quienes no se habían realizado nunca a la prueba.
La opción de APIP hace que la prueba sea asequible a individuos pertenecientes a grupos con un menor estatus socioeconómico, mucho más que el APV tradicional. Expandir la realización de pruebas mediante programas de prevención de la transmisión vertical reduce los obstáculos socioeconómicos, especialmente para las mujeres. Se requiere de esfuerzos continuados para promover el aconsejamiento y prueba entre los hombres y los menos adinerados.
The rapid increase in the availability of testing and counselling services globally and in sub-Saharan Africa has taken place against the background of debates regarding how to encourage testing, particularly among groups at higher risk of infection (De Cock et al. 2006; WHO, UNAIDS, UNICEF 2010). Recognising that late diagnosis of HIV resulted in part from the missed opportunities to test health facility users (McDonald et al. 2006; Nakanjako et al. 2007), major efforts were launched to scale up provider-initiated testing and counselling (PITC) through the routine offer of HIV testing (CDC 2006; WHO, UNAIDS 2007). Policies and programmes were initiated in sub-Saharan Africa to expand HIV testing beyond traditional centres for voluntary counselling and testing (VCT), through facility-based and outreach efforts including campaigns and home-based testing (Bassett & Walensky 2010; WHO, UNAIDS, UNICEF 2010). The remarkable diversification of modes of testing raises important research and ethical questions regarding the extent to which different approaches encourage testing among different population groups (Hensen et al. 2012; Obermeyer et al. 2012).
There is growing evidence about obstacles to HIV testing: some result from factors related to health services, such as shortages of staff, resources and infrastructure; others from the difficulty of paying for tests or obtaining transportation; and yet others from insufficient awareness about HIV and concerns about stigma and discrimination (Matovu & Makumbi 2007; Obermeyer & Osborn 2007). In general, however, the evidence about how socio-economic factors influence testing comes from disparate studies using different measures and does not distinguish among different components of socio-economic status, which may operate in different ways: although education and wealth are associated with higher infection in some settings, educational attainment may increase uptake of testing through increased recognition of the importance of knowing one's HIV status and greater control over the decision to test (Jukes et al. 2008), whereas wealth may be associated with greater awareness of risks and with reduced financial barriers to testing (Mishra et al. 2007; Parkhurst 2010). Moreover, most studies simply measure whether individuals have been tested or not, and only a few studies have examined the impact of particular ways of providing testing (Helleringer et al. 2009; Menzies et al. 2009). To generalise about the extent to which different modes of testing increase uptake among those with different socio-economic characteristics, it is necessary to go beyond studies conducted in individual sites and to undertake systematic comparisons of testing in settings that vary by mode of testing and among groups that vary by socio-economic characteristics.
In this study, we compare four sub-Saharan African countries and examine how different approaches to the provision of testing vary in the extent to which they reach different socio-economic groups. We use standardised measures of several dimensions of socio-economic status and compare those tested through PITC, those tested through VCT and those not tested. By comparing groups in the population who use health facilities, it is possible to gain insights into the factors that hinder or facilitate testing among facility users, and this complements analyses of the population at large which would be conducted through other means such as household surveys.
The objectives of the study were as follows: to compare the socio-economic status of health facility users who had tested and those who had not tested for HIV in four sub-Saharan African countries; to compare the socio-economic status of those tested at different types of facilities – VCT, integrated and PMTCT – in those countries; and to draw the implications of the results for programmes designed to expand access to HIV testing in low-resource settings.
Study design and sample
The Multi-country African Testing and Counselling for HIV (MATCH) study was conducted in 2008–2009 in Burkina Faso, Kenya, Malawi and Uganda, four countries with different HIV prevalence levels, policies and programmes. As in many other settings, levels of testing and knowledge of HIV status remain low. Testing at health facilities has increased in the four countries as a result of efforts by governments, international donors and non-governmental organizations (NGOs), but the history of testing programmes, the role of NGOs and the specific guidelines for HIV testing differ (Burkina Faso 2010; Kenya 2010; Malawi 2010; Uganda 2010).
A cross-sectional survey of clients was conducted at health facilities. To facilitate the logistics of the survey and ensure that the sample included both testers and non-testers, the research teams drew up a list of testing facilities in each country, designed to include different modes of testing (VCT and PITC) and facilities that were the major providers of HIV testing services. Another list was drawn, of facilities or services that did not provide HIV testing. About 20 facilities were selected for inclusion, with the goal of interviewing a total of about 900 clients per country, two-thirds testers and one-third non-testers. To achieve geographical variability, we included facilities located in the capital city and in one province in each country. Adult clients (aged 18 and above) who were present at the selected facilities on the appointed days were invited to participate; the fraction of clients to be approached was based on the expected numbers at the facility: all clients were invited at small facilities, whereas every nth client was invited at larger/busier facilities. Testing status was ascertained at the start of the interview for all respondents, regardless of whether they were recruited at a testing or non-testing facility.
The study was well received, with high response rates in Malawi and Uganda (over 90%) and Burkina Faso (80%); the lower response rate in Kenya (about 50%) reflects the difficult political and security situation prevailing in 2008, which resulted in reluctance to participate and cancellation or interruption of interviews. The instrument included closed- and open-ended questions. All respondents were asked about socio-demographic information and whether or not they had ever tested for HIV; testers were asked about experience with HIV testing, pre- and post-test counselling, disclosure, stigma and follow-up care. The interview lasted approximately 30 min on average. Further details on the MATCH study have been reported in a previous publication (Obermeyer et al. 2012).
Two outcome measures were used in this analysis: whether the respondent had ever tested for HIV (asked of all respondents) and the mode of testing of the respondent's most recent test (asked of testers). Both measures were self-reported. Modes of testing were grouped into three types: (i) stand-alone sites for VCT, which usually provide testing at the initiative of the client; (ii) hospitals and medical facilities where testing is provided along with medical services (referred to as integrated testing); and (iii) testing through prevention of mother-to-child transmission (PMTCT) programmes, which include prenatal clinics and facilities offering care to pregnant women. Women who reported testing at integrated facilities because of pregnancy were coded as PMTCT testers, while 12 men who tested as a result of their wives’ PMTCT were coded as having tested at integrated facilities. Stand-alone sites correspond to what is usually referred to as VCT, whereas integrated and PMTCT correspond to provider-initiated testing and counselling (PITC).
Age, sex, education and wealth were the key covariates of interest. Age was self-reported, measured in years and entered into the analysis as a categorical variable. We included covariates based on both wealth and education in our analyses because they measure different dimensions of socio-economic status and may have independent effects on testing behaviour. Education was specified as no schooling, incomplete or completed primary education, or at least some secondary education. Wealth was measured with an assets quartile method that is widely used in surveys in low- and middle-income countries and has been shown to be a valid proxy for household wealth (Filmer & Pritchett 2001; Rutstein & Johnson 2004; Clements & Pritchett 2008). It was based on a principal components analysis of a respondent's household assets (television, electric or gas stove, telephone, land or animals) and amenities (tap water, flush toilet and electricity). Indicators of household assets and amenities were converted into z-scores, and factor loadings for a single wealth factor were calculated. For each respondent, values of the indicator variables were multiplied by the factor loadings to obtain a wealth score. Respondents within each country were grouped into quartiles by wealth score.
We analysed the determinants of ever testing for HIV and the determinants of using a particular mode of testing (integrated, VCT, PMTCT), compared with not testing. Because women's options and decisions regarding testing were assumed to be different from men's, regression analyses are reported separately for women and men. Testing was not rare among respondents (P > 10%), so odds ratios based on logistic regression analysis would not provide a reasonable estimate of the relative prevalence of testing. We used a modified Poisson regression analysis with robust standard errors to estimate the relative risk of testing, a technique that has previously been used in a variety of observational and experimental settings and on clustered data (Zou 2004; Yelland et al. 2011). Multinomial logistic analysis was used to estimate the relative prevalence of testing at a particular type of facility. All regression analyses were adjusted for country, using a fixed effect (fixed effect parameters and constants not shown), and for clustering of responses at the interview facility, using a generalised estimating equation (GEE). We report 95% confidence intervals for all parameters and exact P-values from Wald tests of significance for all tests where P ≥ 0.001. The results of these analyses estimate the cross-sectional association between testing behaviours, socio-economic status (SES) and other covariates. All analyses were completed in Stata SE 10.1 (StataCorp 2009).
The study was approved by the Ethics Review Committee of the World Health Organization and by an institutional review board (IRB) in each of the four countries (Burkina Faso's Comité d'Ethique pour la Recherche en Santé of the Ministries of Health and Higher Education; Kenyatta's National Hospital's IRB in Kenya; the National Health Science Research Committee of the Ministry of Health and Population in Malawi; and the IRB of Makerere University and the National Council for Science and Technology in Uganda). The WHO Ethics Review Committee gave specific approval at the start of the study and through continuing reviews annually. Informed consent was obtained from all respondents who were invited to participate in the study. As approved by the IRBs, in Burkina Faso and Uganda, consent was in writing for virtually all respondents, except for a few illiterate respondents who provided a thumbprint. In Malawi and Kenya, oral consent was obtained for most respondents and noted by interviewers, consistent with local practices, and in the case of Malawi, with higher illiteracy. Where the documentation of consent was incomplete (24% of respondents in Kenya), special permission to use the data was obtained from the IRB.
A total of 3659 respondents were interviewed at health facilities. All respondents reported their testing status, except for 18 who were excluded. We also excluded those testing before 2007 or missing a test date (367); respondents missing age, education or wealth information (60); those with incomplete or unclear data on place of most recent test (33); those missing information on facility of interview (32); and those interviewed at facilities with fewer than six respondents (24), for a final total of 3125 respondents. The distribution of respondents by testing status and mode of testing in each country (see Supporting Information Table S1 and S2) shows that about two-third of respondents were testers, that stand-alone VCT testing was more common in Burkina Faso and Uganda, and that integrated testing was more frequent in Kenya and Malawi. The study population included more women (57%) than men and tended to be well educated, with only 18% of women and 10% of men reporting no formal education (data not shown).
Table 1 presents bivariate analyses of socio-economic differences between non-testers and those tested through different modes of testing. Non-testers tended to have lower SES and educational attainment than testers. Among women, non-testers were more likely than testers to have no formal education, although this result was only marginally significant (31.7% of those with no formal education had not tested, compared with 25.9% of those with secondary education). Men with higher education were significantly more likely have tested (36.8% of men with no education had not tested, compared with 38.1% of those with secondary education) as were men in the highest assets group (36.8% compared with 52.3% in the lowest asset group).
Older women were more likely to have tested at integrated facilities, and younger women to have tested for PMTCT (39.4% of women aged 45 years and older tested in integrated facilities and 2.4% in VCT, compared with 20.2% and 27.7% in the 18- to 24-year-old group). Women in the highest educational group were more likely to have tested in VCT sites (28.5% compared with 18.6% of women with no formal education). Men tested at integrated facilities also tended to be older than VCT testers (46.8% of men aged 45 years and older group were tested in integrated facilities, compared with 20.2% in the younger group). Men in the lower socio-economic categories were less likely to have tested at VCT facilities (18.7% of men in the lowest educational category and 14.4% of men in the lowest assets group, compared with 30.3% of men in the highest education category and 33.2% of those in the highest assets category).
Table 1. Age, education, wealth and country by sex, testing status and mode of testing
|No formal education||31.7||27.7||18.6||22.0||49.3||32.1||18.7|
|Secondary or more||25.9||26.4||28.5||19.2||38.1||31.5||30.3|
|Household assets index (quartiles)b|
Table 2 presents estimates of the relative prevalence of testing compared with not testing among women and men. The association between educational attainment and testing status found in unadjusted analyses was no longer statistically significant after adjustment for wealth, age and country. The adjusted prevalence ratio (APR) comparing those with secondary education to those with no formal education was 1.29 for men (95% CI: 0.97, 1.73) and 1.16 (95% CI: 0.99, 1.37) for women.
Table 2. Mutually adjusted effectsa of age, educational attainment and wealth on testing compared with not testing, between men and women (adjusted prevalence ratios (APR), confidence intervals (CI) and P-values)
|(1.1, 1.41)||(0.96, 1.74)|
|P = 0.001||P = 0.091|
|(1.07, 1.35)||(1.15, 2.3)|
|P = 0.003||P = 0.006|
|(0.76, 1.16)||(1.08, 2.27)|
|P = 0.55||P = 0.018|
|No formal education||Reference||Reference|
|(1, 1.24)||(0.84, 1.4)|
|P = 0.061||P = 0.53|
|Secondary or more||1.16||1.29|
|(0.99, 1.37)||(0.97, 1.73)|
|P = 0.073||P = 0.083|
|Household assets index (quartiles)c|
|(0.95, 1.16)||(0.8, 1.33)|
|P = 0.35||P = 0.81|
|(0.87, 1.22)||(0.81, 1.44)|
|P = 0.77||P = 0.60|
|(0.84, 1.19)||(0.91, 1.55)|
|P = 0.92||P = 0.90|
Table 3 compares testers by mode of testing with non-testers as the reference category. The socio-economic characteristics of non-testers and those tested at integrated facilities did not differ significantly among women or men, and among women, there were no notable differences between non-testers and PMTCT testers. But among both men and women, higher educational attainment was associated with a greater likelihood of testing at VCT facilities after adjustment for covariates. Men with secondary education were three times more likely to test at stand-alone VCT sites than those with no formal education (APR = 3.01, 95% CI: 1.55–5.83). After adjustment for age and wealth, women with secondary education were over three times more likely to test at stand-alone VCT sites than those with no formal education (APR: 3.45, 95% CI: 1.82–6.52). After adjustment for educational attainment and country, wealth was not significantly associated with testing at different types of facilities compared with not testing.
Table 3. Mutually adjusted effects of age, education and wealth on mode of recent testa APR, confidence intervals and P-values
|Integrated testing and medical|
|25–34||3.2 (2.22, 4.61)||2.24 (1.32, 3.82)|
|P < 0.001||P = 0.003|
|35–44||3.15 (2.23, 4.45)||4.03 (2.2, 7.38)|
|P < 0.001||P < 0.001|
|45+||1.67 (0.87, 3.19)||3.58 (1.68, 7.62)|
|P = 0.12||P = 0.001|
|No formal education||Reference||Reference|
|Primary complete||1.4 (0.88, 2.22)||1.07 (0.63, 1.83)|
|P = 0.16||P = 0.80|
|Secondary or more||1.51 (0.77, 2.93)||1.21 (0.65, 2.26)|
|P = 0.23||P = 0.55|
|Poorer||1.43 (0.93, 2.2)||1.21 (0.7, 2.09)|
|P = 0.099||P = 0.50|
|Wealthier||1.25 (0.64, 2.42)||1.19 (0.63, 2.23)|
|P = 0.52||P = 0.60|
|Wealthiest||1.12 (0.58, 2.16)||1.27 (0.73, 2.22)|
|P = 0.73||P = 0.40|
|25–34||2.36 (1.46, 3.8)||1.4 (0.68, 2.86)|
|P < 0.001||P = 0.36|
|35–44||3.61 (2.27, 5.74)||3 (1.4, 6.43)|
|P < 0.001||P = 0.005|
|45+||1.45 (0.75, 2.82)||2.12 (0.96, 4.67)|
|P = 0.27||P = 0.062|
|No formal education||Reference||Reference|
|Primary complete||1.52 (0.89, 2.59)||1.2 (0.65, 2.22)|
|P = 0.12||P = 0.55|
|Secondary or more||3.45 (1.82, 6.52)||3.01 (1.55, 5.83)|
|P < 0.001||P = 0.001|
|Poorer||0.92 (0.48, 1.76)||0.84 (0.37, 1.91)|
|P = 0.80||P = 0.67|
|Wealthier||0.88 (0.31, 2.51)||1.2 (0.39, 3.73)|
|P = 0.81||P = 0.75|
|Wealthiest||0.94 (0.31, 2.83)||1.89 (0.62, 5.73)|
|P = 0.92||P = 0.26|
|25–34||1.84 (1.28, 2.64)||–|
|P = 0.001||–|
|35–44||0.52 (0.29, 0.94)||–|
|P = 0.032||–|
|45+||0.06 (0.02, 0.26)||–|
|P < 0.001||–|
|No formal education||Reference||Reference|
|Primary complete||1.46 (0.93, 2.27)||–|
|P = 0.099||–|
|Secondary or more||1.08 (0.58, 2)||–|
|P = 0.813||–|
|Poorer||1.17 (0.77, 1.78)||–|
|P = 0.45||–|
|Wealthier||1.1 (0.56, 2.17)||–|
|P = 0.78||–|
|Wealthiest||0.83 (0.4, 1.74)||–|
|P = 0.63||–|
A major result of this analysis is that higher educational attainment is significantly associated with the likelihood of testing at VCT sites. Among both men and women, secondary education is associated with a threefold increase in the prevalence of VCT testing. Both education and wealth are significant in bivariate models, but educational attainment is more consistently significant in bivariate analyses and remains independently significant after adjustment for wealth. These results are consistent with those of other studies conducted in sub-Saharan Africa where VCT was associated with knowledge of HIV and education (Hutchinson & Mahlalela 2006; Tenkorang & Owusu 2010; Venkatesh et al. 2011), and with an analysis of survey data from 13 countries of sub-Saharan Africa, showing that prior to the availability of treatment, VCT testing was associated with secondary education (Cremin et al. 2012). Because in this study, we used the same measures of socio-economic status across the four sites, the consistency of the results is compelling. Moreover, because the measure most strongly associated with mode of testing was educational attainment, independent of wealth, this suggests that knowledge and motivation play an important role, over and above resources, in influencing HIV testing behaviours.
Another important result of the study comes from comparing provider-initiated modes of testing to VCT in terms of their associations with socio-economic status. Unlike VCT, there were no significant differences in the socio-economic characteristics of those tested through PITC modes compared with those not tested: we found no significant socio-economic differences between women tested through PMTCT and those not tested, nor between male and female respondents tested at integrated facilities and those not tested. This suggests, first, that programmes offering HIV testing to pregnant women are generally successful in reducing socio-economic obstacles to testing, and hence in improving equity of access to HIV tests. This result is consistent with global statistics (WHO, UNAIDS, UNICEF 2010; WHO 2011). Secondly, it suggests that PITC expands testing to disadvantaged health facility users to a greater extent than VCT. Other analyses of MATCH data show that this expansion does not appear to take place at the expense of the quality of the services provided or the protection of human rights (Obermeyer et al. 2012). These results support efforts at international and national levels to scale up HIV testing at health facilities (CDC 2006; WHO, UNAIDS 2007; WHO, UNAIDS, UNICEF 2010; WHO 2011).
A strength of this study is that it used the same protocol, sampling and instruments in the four countries, and developed measures of socio-economic status based on both education and wealth to make comparisons across settings. The limitations of the study are related to sampling. The selection of facilities and respondents was systematic rather than random, and thus, samples are not representative of all health facility users in a given country. Results such as the relative percentages of women and men, differences in ages or socio-economic status, or differences among countries should not be interpreted as estimates of population-level statistics. A related limitation is the recruitment of testers and non-testers at health facilities. This choice is justified by the fact that most health facility users have never tested, and hence, it is important to examine obstacles to testing among facility users. The disadvantage of such a choice is that the results cannot be generalised to the population at large – this would require other study designs, such as household surveys, and considerably greater resources. However, the inclusion of a comparison group of health facility users who never tested makes it possible to compare socio-economic factors among both groups and assess whether these factors influence testing.
Because respondents’ testing status was not known in advance and represented potentially sensitive information, it could not be used to screen respondents before they consented to participate. Testing status was therefore ascertained through self-report at the start of the interview. As self-reported data, testing status and socio-economic characteristics may be subject to recall errors or to social desirability bias, but we think these are probably low. Taking an HIV test is a known experience, easily remembered, and respondents appeared willing to provide this information, with only 18 respondents missing data (who were excluded from the analysis). Social desirability bias is also likely to be low, as the interviewers were not connected to the health facilities and were trained to ask questions in a non-judgemental way. Thus, the potential for misclassification of testing status in negligible, and there is no reason to think that it would vary systematically in a way that confounds the results.
Comparisons with national surveys in the four countries show that age and marital status were similar among MATCH and Demographic and Health Survey respondents, but MATCH respondents are more urban, more educated and have higher standards of living than respondents in national surveys, except for Malawi where their socio-economic characteristics are similar (Institut National 2003; National Statistics 2005; Uganda Bureau 2007; Central Bureau 2004). The higher socio-economic characteristics of MATCH respondents are consistent with the fact that health services users are generally better off than general populations. Consequently, our results likely underestimate socio-economic differentials between testers and non-testers in the general population: had the study been conducted through a household survey, there would probably have been a stronger positive association between socio-economic status and testing.
The results of this study underscore the importance of outreach. Users of health facilities with less-favourable socio-economic characteristics are less likely to take the initiative to test through VCT and could be encouraged by providers, but other means should also be deployed outside of health facilities to increase uptake among those who do not use health facilities: mobile units, home-based approaches and campaigns can encourage testing among disadvantaged groups, and also among men who use health facilities less frequently than women. In general, support for different approaches to testing is desirable, because individuals’ needs vary by socio-economic status and over time, and choice should be provided for both practical and ethical reasons (April 2010).
This analysis found that higher socio-economic status was associated with the likelihood of HIV testing through VCT; that lower socio-economic status was associated with the likelihood of testing at integrated facilities; and that PMTCT and integrated testers were similar to non-testers and had lower levels of educational attainment compared with VCT testers. These results have implications for the implementation of programmes designed to ensure access to testing in low-resource settings (Bassett & Walensky 2010; Obermeyer et al. 2013). They suggest that provider-initiated modes of testing can increase uptake among socio-economically disadvantaged strata to a greater extent than traditional VCT at stand-alone facilities. Secondly, the lack of socio-economic differentials for PMTCT is consistent with the notion that expanding testing through PMTCT has reduced socio-economic obstacles for women (WHO, UNAIDS, UNICEF 2010; WHO 2011). It is important to develop comparable ways to reach men and address the gender dimension of HIV testing, which has been recognised in global documents. Thirdly, given low levels of testing worldwide and the persistence of socio-economic obstacles to the uptake of testing, continued efforts are needed to encourage testing among the less affluent through multiple means.
The project was supported by a grant from the National Institutes of Health. The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.