Equity in utilization of antiretroviral therapy for HIV-infected people in South Africa: a systematic review

Introduction About half a million people in South Africa are deprived of antiretroviral therapy (ART), and there is little systematic knowledge on who they are – e.g. by severity of disease, sex, or socio-economic status (SES). We performed a systematic review to determine the current quantitative evidence-base on equity in utilization of ART among HIV-infected people in South Africa. Method We conducted a literature search based on the Cochrane guidelines. A study was included if it compared for different groups of HIV infected people (by sex, age, severity of disease, area of living, SES, marital status, ethnicity, religion and/or sexual orientation (i.e. equity criteria)) the number initiating/adhering to ART with the number who did not. We considered ART utilization inequitable for a certain criterion (e.g. sex) if between groups (e.g. men versus women) significant differences were reported in ART initiation/adherence. Results Twelve studies met the inclusion criteria. For sex, 2 out of 10 studies that investigated this criterion found that men are less likely than women to utilize ART, while the other 8 found no differences. For age, 4 out of 8 studies found inequities and reported less utilization for younger people. For area of living, 3 out of 4 studies showed that those living in rural areas or certain provinces have less access and 2 out of 6 studies looking at SES found that people with lower SES have less access. One study which looked at the marital status found that those who are married are less likely to utilize ART. For severity of disease, 5 out of 6 studies used more than one outcome measure for disease stage and reported within their study contradicting results. One of the studies reported inconclusive findings for ethnicity and no study had looked at religion and sexual orientation. Conclusion It seems that men, young people, those living in certain provinces or rural areas, people who are unemployed or with a low educational level, and those being unmarried have less access to ART. As studies stem from different contexts and use different methods conclusions should be taken with caution.


Introduction
South Africa is home to the largest HIV-infected population worldwide, with 6.1 million people living with HIV/ AIDS in 2012 [1]. The country also has the largest antiretroviral therapy (ART) program worldwide: with domestic investments amounting to US$1.9 billion in 2011 [2], it provided treatment to about 80% (2.0 million people) of all eligible people in 2012 [1]. Current South African guidelines state that all those with CD4 cell counts of ≤350 cells/μL are eligible for ART [3].
Nevertheless, a significant treatment gap of about half a million people remains between those who receive treatment and those in need according to the eligibility criteria [1]. There is little knowledge on which people are deprived from treatmente.g. by severity of disease, sex, age, socio-economic status (SES) and area of living [4], limiting the development of policy measures to specifically target and improve treatment coverage among these groups. This is illustrated in South Africa's 'National Strategic Plan on HIV, STIs and TB 2012-2016' which flags the importance of inequalities in treatment utilization but is not specific on which marginalized groups should be targeted [5].
It is clear that ART not only improves a patient's health and survival [6,7], but also substantially reduces their infectiousness [8,9]. As a result, ART can play an important role in controlling the epidemic in South Africa [10][11][12]. The World Health Organization (WHO) recently released new consolidated guidelines, taking both the prevention and treatment benefits of ART into account [13,14]. The new guidelines state that ART should be provided for HIV infected people with a CD4 cell count of ≤500 cells/μL, who are in a serodiscordant relationship, and/or pregnant [14]. In addition, the WHO also states that guidelines should be expanded when universal access for those with CD4 cell counts of ≤350 cells/μL has already been achieved [14]. As treatment programs continue to expand, identifying and targeting hard-to-reach populations will be increasingly important.
We determined the current quantitative evidence-base on equity in utilization of antiretroviral therapy (ART) among HIV-infected people in South Africa. This information may provide insight into methods used for equity research and may help policy makers to identify and target hard-to-reach populations.

Methods
We performed a systematic review on the basis of the Cochrane Handbook for Systematic Reviews of Interventions, Version 5.1.0.4 [15]. Our search was performed on 18 February 2013 using Pubmed, Embase, Central and Psychinfo database. Our search syntax consisted of search terms in four categories (ART, HIV, South Africa and Equity), that were combined using AND. The search strategy is presented in summary in Table 1 and in detail in Additional file 1.

Conceptual model
Following the WHO's guidance on monitoring equity in AIDS treatment programs [16], we distinguished five domains of coverage: 1) availability of resources; 2) physical and financial accessibility; 3) acceptability; 4) use of service; and 5) effective coverage (defined as the proportion of the population in need of an intervention who fully comply with the recommended treatment program). This review focuses on the latter two domains as the other domains feed into these. We included studies on both ART initiation and adherence and this was together labeled as ' ART utilization'. We acknowledge that an individual's health care utilization can be explained by a function of predisposing factors (e.g. education, culture, health beliefs, age and sex), enabling factors (income, health insurance, waiting time, genetic factors) and need factors (perceived need to seek and adhere to care and professional's judgment about people's health status) [17]. We used the terms 'equity' and 'inequity' to reflect differences in utilization of ART by criteria such as severity of disease, age, or SES [18].

Inclusion and exclusion criteria
A study was included if it: 1) compared for different groups of HIV infected people (by sex, age, severity of disease, area of living, socio-economic status, marital status, ethnicity, religion and/or sexual orientation (i.e. equity criteria (World Health Organization, Guidance on Priority Setting in Health Care (GPS health) in preparation) [19]) the number initiating/adhering to ART with the number who are not); 2) was performed in South Africa; and 3) reported in English. Although some equity criteria are the social determinants of health, severity of  [19]. A study was excluded if it: 1) focused on prevention of mother to child transmission (PMTCT), death during follow up, barriers of accessing care or tuberculosis (TB) services for HIV infected patients; 2) was a qualitative study, comment, editorial, economic evaluation or conference abstract; 3) was a duplicate reference from different databases; and 4) reported only differences in groups by a simple comparison with the gross number of people initiating or adhering to ART. We only included studies that take into account the underlying need of a group for ART. For example, the mere fact that more women than men have access to ART does not necessarily indicate an inequity as more women than men may be infected in the country. There was no restriction for publication date for inclusion of studies. Following the Cochrane guidelines grey literature was excluded due to expected low methodological quality of studies [15].

Study selection, data extraction and quality evaluation
Two independent reviewers (CM and EM) assessed if the studies from the database search satisfied the inclusion criteria. First, all studies were screened on the basis of title and abstract, and subsequently on the basis of full-text. Reference lists of the retrieved articles were screened for additional studies (snowballing). The reviewers used a data collection form (Additional file 2) to extract relevant information (study characteristics, results per equity criteria, and study limitations) from the articles. Both reviewers evaluated the quality of studies using a quality-grading protocol (Additional file 2) adapted from existing protocols [15,20,21]. The protocol covers 20 indicators and for each item 0-2 points are given and added up to get an overall quality score (ranging from 0 to 40 points). Studies were categorized as low-quality (<20 points), medium-quality (20)(21)(22)(23)(24)(25)(26)(27)(28)(29) or high-quality (≥30). During the study selection, data extraction and quality assessment, disagreements were resolved through discussion with a third researcher (NT) until consensus was reached.

Data synthesis and analysis
A matrix was developed containing the study results per investigated equity criterion. We established the following categories to summarize the results for each equity criteria investigated in a study: 1) associated, differences reported in ART utilization between groups (e.g. men versus women for sex) were significant (p < 0.05, or when 1.0 does not fall in 95% confidence interval (95% CI)); 2) not associated differences reported in ART utilization between groups were not significant (p value >0.05 or 1.0 falls in 95% CI; contradicting results, within one study contradicting results were reported for differences in ART utilization between groups due to the use of multiple outcome measures for an equity criterion (e.g. CD4 cell count levels and WHO disease stages for the equity criterion severity of disease); and inconclusive results, differences in ART utilization between groups was investigated but the authors drew no conclusions due to small sample sizes. We adhered to the PRISMA guidelines for reporting of this systematic review [22].

Study inclusion
From the initial search (801 articles), 268 studies were duplicates, 483 studies were excluded on the basis of title/abstract and 39 on the basis of full-text screening. Screening of the references of the remaining 11 studies resulted in one extra article and added to a total of 12 studies that are included in this review ( Figure 1, Table 2).

Characteristics of included studies
Seven studies assessed inequities in ART initiation (Table 3) and five studies in ART adherence ( Table 4). All studies were based on primary data analysis from observational surveys, except for one study using secondary data [23] and one review [24]. Studies defined ART initiation differently, like 'at least 14 days on ART' [25] or 'visited the ART clinic at least once after testing HIV positive' [26]. Definitions of non-adherence were also varied, and were measured in terms of patients' absence at the clinic for more than one [27] three [28,29] or six [30] months, or in terms of the number of pills not taken and brought back to the clinic (clinic-based pill counts) [31]. The outcome measure used for equity criteria varied widely among studies. For severity of disease, some compared the differences in utilization of ART by WHO disease stages [26][27][28][29]31], while others used CD4 cell count levels [26][27][28][29][30][31] or viral load [28,31]. For age, many different age categories were used. Fatti et al. [29] only included children in the study population, and the oldest age group in that sample is younger than the youngest age group in for example Govindasamy et al. [26] (who compared people below and above 30 years of age). Six studies, all using different databases, investigated urban and rural areas of the Western cape province and two studies reported at the national level [23,24]. More than half of the articles (seven) [25][26][27][29][30][31][32] were of high-quality, three had medium-quality [28,33,34] and two were of low-quality [23,24]. Table 5 gives an overview of the quality scoring per study.
For severity of disease, five [26][27][28][29]31] out of six studies [26][27][28][29][30][31] reported contradicting results. In four [26][27][28]31] out of these five studies an association was found between ART utilisation and a person's CD4 cell level while no association was found with a patient's WHO status. Of these studies, one ART initiation [26] and one on adherence [27] (both high quality) reported that higher CD4 cell counts are associated with lower utilization of ART. On the contrary, two other studies on adherence (one high-and one medium-quality) reported that lower CD4 cell count is associated with less utilization [28,31]. In one other study (high quality) that reported contradicting results for severity of disease among children, an association was found with WHO stage but not with CD4 cell count level [29]. The sixth study (high quality) reporting for severity of disease, only looked at CD4 cell count levels and found that patients with a higher CD4 cell count level adhered less to ART [30].
For area of living, three [23,29,34] out of the four studies [23,29,32,34] that reported on this criterion found an association between area of living and ART utilization. Two studies (high-and medium-quality) reported that people in certain provinces have lower utilization of ART (see Table 3) [23,34]. One of the studies (high-quality) reported that children living in rural areas and who visit ART clinics in urban areas, have lower utilization than children that visit clinics in their own area of living (urban or rural area) [29]. The fourth study (high-quality) that reported on area of living found no association between ART utilization and area of living (peri-urban, urban or rural area) [32].
Socioeconomic status was found to be associated with ART utilization in two [28,33] (both medium-quality) out of the six studies [25,26,28,[31][32][33] that reported on this criterion, which showed that those unemployed have lower utilization of ART. One of these two studies also reported that those with lower education utilize less [33]. Of the four studies that found no association, one (high-quality) found no differences on the basis of employment and education [26]. The other three (all high-quality) found no differences in ART utilization between those with differences in SES [25,31,32]. One of these also found no association between educational level and ART utilization [32].
For marital status only one study (medium-quality) was included in this review and reported that being Full-text articles included after snowballing reference list of 11 included full-text articles (n = 1) Full -text articles assess ed for eligibility (n=50)   defined as a WAZ score of < −3 (severe underweight) or a WHO stage ≥3. 3 rural/urban, urban/urban, rural/rural = first term indicates place of residence and second the area of accessing ART clinic. 4 no significance was reported and authors concluded that ART utilization was different among groups that were compared. * = low quality, ** = medium quality, *** = high quality.   CI = Confidence Interval, aOR = adjusted odds ratio, ART = antiretroviral therapy, WHO = world health organization. * = low quality, ** = medium quality, *** = high quality.  b. Children on ART that do not loss to follow up (n = 2153). Both from retrospective cohort of children, (<16 yrs, ART naïve), enrolled in 44 routine public healthcare facilities (7 rural, 33 urban/12 secondary level hospitals, 32 primary health care clinics) supported by a NGO, used electronic data collection systems for patient monitoring. Children were divided in 3 groups a) urban residence and urban ART facility attended (urban group, n = 1727); rural residence and rural facility attended (rural group, n = 228); and rural residents attending urban facilities (rural/urban group, n = 377) CI = confidence interval, HR = hazard ratio, ART = antiretroviral therapy, WHO = world health organization, LTFU = loss to follow up, VL = viral load. * = low quality, ** = medium quality, *** = high quality. Cooke [32] Fatti [29] Govindasamy [26] Boyles [30] Cleary [25] Orrell, Cornell [28] Tsai [33] Adam [34] Muula [24] Nattrass [23] Total score (out of 40 points)  37  34  34  33  32  31  30  26  26  26  19  12 1 Study design (peer reviewed = 2, other = 0) unmarried is associated with lower ART utilization [33]. For ethnicity only one study (high-quality) was found, and it reported inconclusive results due to a small sample size [25]. None of the included studies looked at the ART utilization by religion or sexual orientation.

Discussion
This is the first systematic review that examines equity in utilization of ART in South Africa and identified 12 studies. It seems that men, young people, those living in certain provinces or rural areas, people who are unemployed or with low educational level, or those who are unmarried have less access to ART. For severity of disease, most studies used more than one outcome measure for disease stage and reported within their study contradicting results. No evidence of inequity in ART utilization by ethnicity, religion and sexual orientation was found. There were large heterogeneities in both context (study area, type of program, time period) and methodology of the studies in this review.
Only one high-and one low-quality study reported a significant difference in utilization of ART among men and women, and eight other studies found no differences. Although it is encouraging that access to ART seems mostly equal for both genders, the studies in our review failed to take the timing of ART initiation into account. Observational studies from South Africa recently showed that case-fatality rates among HIV-infected men were substantially higher compared to women in South Africa, most likely related to late entry into care [35,36]. Late entry by men can be explained as ART is mainly provided through primary health care services, and its antenatal care services frequently serve as an entry-point for HIV treatment for women.
The findings in some studies which showed that young age is associated with low utilization raises concerns. Young people may face more barriers to treatment (like lack of knowledge about treatment possibilities and benefits and fear for stigma and discrimination) [32]. Yet, this relationship may be confounded by eligibility, as older people are more likely to be eligible because of more advanced disease stages. In addition, many studies did not cover all ages. As the HIV epidemic in South Africa is ageing [37,38] it will become increasingly important to determine ART utilization among elderly, a group previously neglected in research on ART utilization.
Both area of living and SES did not seem to be associated with ART utilization. However, the studies looking at area of living were mostly of low-quality. The studies by Nattrass et al. [23] and Adam et al. [34] reported coverage levels for different provinces. However, these studies used a simple Markov-model to estimate the need for ART, and it is difficult to determine whether the model projections are valid. The study by Fatti et al. [29] reports on children in four different areas. Lower utilization for children living in rural areas and accessing clinics in urban areas can be explained by financial and non-financial barriers such as the monetary cost of transportation or the opportunity cost of accessing health care services [33]. Nevertheless, more research is needed in order to generalize these findings to other areas and population groups. Finally, Tanser et al. [39] showed that self-reported visiting of health clinics in a rural South African area was significantly associated with the distance between the clinic and home, with greater distance resulting in lower utilization, yet we did not include this study because it didn't specifically concern ART utilization.
Studies on SES and area of living will likely measure the same inequities as people in deprived areas might have lower SES. Tsai et al. [33] found significant evidence of socioeconomic inequities in the uptake of ART services within a rural and deprived part of South Africa during the early years of the public sector scaling up of ART (2003ART ( -2005. Poorer households in South Africa and in sub-Saharan Africa generally have less access because they face various barriers like cost for transport to the clinic, knowledge of the benefit of ART treatment and a lower propensity to seek formal sector treatment for illness [40,41]. Cleary et al. [25] reported no differences in SES distribution between those in need and those accessing ART in urban areas in 2008. This is in line with the 'inverse equity hypothesis' which predicts a paradoxical worsening of health inequities as effective new public health interventions first diffuse among the well-to-do but later also among the poor. Last years ART has been scaled-up drastically (and now reaches about 80% of those in need) barriers to access might have been reduced or removed and those least able to overcome those initial barriers are now able to use the services [25]. Yet, still about 20% lacks access to treatment and this group likely faces most barriers. In addition, if South Africa adopts the new WHO guidelines and further expands its ART program new inequities might appear.
We found contradicting results for severity of disease as within studies differences in ART utilization were reported for HIV-infected people with different CD4 cell count levels but not for different WHO disease stages. Also, some studies reported lower utilization for healthier patients while other studies for the most severely ill. One of the studies by Govindasamy et al. [26] addressed ART initiation and concluded that those with a CD4 cell count of >350 are less likely linked to care after testing HIV positive than those ≤350. This can be explained by the fact that these patients were not yet eligible for ART and only needed to enrol in the clinic to monitor their CD4 level, or because they feel less need for care as they do not suffer from symptoms. The other five studies addressed ART adherence. Boyles et al. [30] and Kranzer et al. [27] both found that those with higher CD4 cell count (CD4 > 200) adhere less to ART and this may also be explained by the fact that individuals who default do so because they feel better on treatment [42,43]. In contrast, Fatti et al. [29] and Orrell et al. [31] found that most severely ill patients were more likely to lost of follow-up. One explanation could be that patients perceived a lack of effectiveness of treatment when ill or not being able to take the medicine because of symptoms [42]. However, the status of patients who are lost to follow-up is difficult to assess, and it is also likely that many of those are unregistered deaths, thus explaining the higher rates among those with advanced disease.
Only one of the studies looked at marital status and reported less access for unmarried people. However, this study was of medium quality as it compared socioeconomic characteristics of a community sample with a clinic sample which were taken from different areas. For ethnicity, religion and sexual orientation no evidence was available and more research is needed to determine inequities in ART utilization by these criteria. It is likely that inequities exist on the basis of ethnicity, as the history of apartheid caused differences in access between black and white South Africans [44]. Also, among black Africans differences in access between ethnic groups like Zulu-speakers, French speaking Cameroonians and Xhosa speakers likely exist, partly due to differences in language barriers that they may face when accessing care [45,46]. Although HIV-prevention services for men who have sex with men (MSM) are expanding across the country, there are still several gaps [47,48]. This group may face barriers in ART access due to fear of provider stigma and social isolation [49,50]. Low HIV testing rates are reported among Muslim people in predominantly Muslim residential areas in Cape Town [51] and different religions might face different levels of HIV-related stigma which might cause inequities in ART utilization [52].
After analyzing the findings of the included studies we found no patterns of equities or inequities that may be explained by differences in program design (e.g. NGO or university supported, public program, availability adherence counselor), time period (e.g. before or after scale up of ART), target population (e.g. indigent populations, children) and study area (e.g. townships, rural areas). On the other hand, patterns might have been identified if the number of studies included in this review were higher.
We found only 12 studies which looked at equity criteria for ART utilization, and two of these were of low-quality. In addition, all studies differed in context (year of study, area, study population), methodology, and outcome measured. Access to ART in South Africa has evolved quickly over the past decade [53] and inequities that were reported at the start of the ART scale-up might no longer be relevant now. Given the incomplete and mixed evidence base, we call for more rigorous analysis on equity of ART treatment in South Africa, and beyond. We flag three important domains. First, reviewed studies were based on different samples and this made any comparison or generalisation difficult to achieve. A national monitoring system on ART initiation and adherence, which also registers key criteria such as severity of disease, gender, age, SES and area of living could fill in this gap. To measure those in need for ART we recommend using the definition 'eligible for ART on the basis of the country guidelines' as not all HIV-infected people might be already eligible for ART. Yet, the challenge remains to identify HIV-infected patients who are in need of treatment but have not yet been linked to care. Second, most studies only assessed a few equity criteria. This could be explained by the emphasis in strategic ART plans worldwide to reduce gender, SES and area of living inequities [1]. In addition, the recent health equity monitor launched by the WHO uses a list of indicators to present a country's equity profile, but recommends to differentiate groups on the basis of SES, gender, area of living and education level only [54]. We therefore recommend getting similar insights in inequalities between groups that differ in age, severity of disease, marital status, ethnicity, sexual orientation and religion for ART utilization. Third, studies employed a variety of definitions of both ART initiation and adherence measures, but also of equity criteria measures, indicating the need to develop standardized measures in this area of study.

Conclusions
On the basis of 12 studies identified in this review it seems that men, young people, those living in certain provinces or rural areas, those who are unemployed or with a low educational level, and those who are unmarried are disadvantaged from utilization of ART. For severity of disease, most studies used more than one outcome measure for disease stage and reported within their study contradicting results. For ethnicity, religion and sexual orientation there was no evidence available to draw conclusions. As studies stem from different contexts and use different methods, findings cannot be generalized and conclusions should be taken with caution. In order to better inform policy makers, we call for improved guidance in equity research on ART, addressing the need to develop national monitoring of inequity of utilization of ART and employing standardized measures of utilization and equity criteria.