Does the ‘inverse equity hypothesis’ explain how both poverty and wealth can be associated with HIV prevalence in sub-Saharan Africa?
- 1Department of Social and Environmental Health Research, Faculty of Public Health and Policy, London School of Hygiene and Tropical Medicine, UK
- 2Department of Infectious Disease Epidemiology Faculty of Epidemiology & Population Health, London School of Hygiene and Tropical Medicine
- Correspondence to Dr James R Hargreaves, Department of Social and Environmental Health Research, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK;
- Received 24 August 2012
- Accepted 30 October 2012
- Published Online First 12 December 2012
Whether it is relative wealth or relative poverty that drives the HIV epidemic in sub-Saharan Africa, is a controversial aspect of HIV/AIDS epidemiology. We suggest that the social epidemiology of HIV in Africa is changing. Previously, new infections were more rapidly acquired by those of relatively higher socioeconomic position (SEP). More recently, those of relatively low SEP are at greater risk. If confirmed, we further suggest in this paper that this pattern would be compatible with Cesar Victora's ‘inverse equity hypothesis’, first articulated in relation to child morbidity and mortality. The hypothesis suggests that those of higher SEP benefit first from new health interventions.1
Reviews draw different conclusions about the association between SEP and HIV infection within sub-Saharan African countries. Some authors stress that poverty is a key driver of HIV, and that poverty alleviation is the only sustainable solution.2 Others show that higher education and greater relative wealth are associated with greater HIV risk, making HIV unusual in this respect.3 A 2010 study suggests that contextual factors are key, and that ‘being poor or being wealthy may be associated with sets of behaviours that are either protective or risky for HIV infection’.4
There might be several methodological reasons for diversity in the association between SEP and HIV infection rates noted in African studies. First, a range of different populations have been studied. Some studies have analysed data from unlinked anonymous HIV testing of samples collected among antenatal clinic attendees, while others have recruited study participants in household surveys of the general population. Both types of study can suffer from biases affecting the reported association between SEP and HIV.5–7
Second, study sampling frames vary. Some studies recruit participants from within small geographic areas while others have sampled nationally. Heterogeneity in SEP among study participants will be larger in national than in small area studies, and this will influence the association estimate.
Third, various methods of SEP measurement have been used. Many studies used educational attainment as the main marker of SEP, others use Demographic and Health Survey (DHS)-style asset indices, and a variety of other approaches have been used.8 This use of different indices of SEP might explain some heterogeneity between study findings, although, while the theoretical basis underlying the use of such indices may be different, they are often highly correlated.9
Fourth, studies take different approaches to adjustment and stratification of analyses. Some studies stratify samples from men/women, urban/rural areas or younger/older populations. Others pool these samples and may (or may not) adjust for these and other factors. Stratification by gender is usually desirable.10 Adjustment for urban/rural setting and age is usually required unless evidence of effect modification is identified. One practice, ‘over-adjustment’ for sexual behaviour factors likely to be on the causal pathway between SEP and HIV, can give misleading results.11 ,12
Fifth, some, though relatively few, studies have examined patterns of incident infections.13 Most have explored prevalent infections. Since the latent period of HIV is long (a median of 8 years from infection to death even in the absence of treatment), prevalence patterns change slowly and are influenced by both patterns of incidence and mortality which can also change over time.
These differences in study characteristics may help explain some differences between study findings. However, there appears to be considerable similarity in the underlying association between SEP and HIV in different African settings, including those with quite different epidemic characteristics and socioeconomic contexts.12 The association between SEP and HIV risk in sub-Saharan African countries appears to be changing over time. Before 1996, studies were more likely to find higher SEP associated with higher HIV risk, while after 1996, studies have been increasingly finding the opposite association.14
In this paper, we do not concentrate on the empirical evidence for this assertion. We focus rather on describing how health-equity theory from another area of public health might be relevant to the changing social epidemiology of HIV infection. We argue that Cesar Victora's inverse equity hypothesis provides a compelling explanation for the apparently contradictory patterns of association between SEP and HIV reported in the literature, and the pattern of change over time.
In its simplest form, Victora's hypothesis suggests that higher SEP groups will benefit first from new health interventions. Victora et al1 describe how this can lead to the finding that improvements in general health outcomes are accompanied by widening of relative social inequalities in health (figure 1A). Victora et al consider a situation where at the start of observation, the relatively poor have worse health outcomes than the relatively wealthy. As the wealthy benefit from new interventions, this gap widens. Subsequently, the poor make up some, but not all, of the deficit, and relative social inequalities in health outcomes are hypothesised to reduce but not to disappear.
A similar phenomenon appears to be occurring in the epidemic of HIV in African countries, with one notable difference. The early distribution of HIV among socioeconomic groups reflected prevailing patterns of sexual networking at the time of HIV's emergence. At the time, higher SEP groups were probably more mobile and had wider sexual networks than lower SEP groups. Given this, it is perhaps unsurprising that in many countries, the early spread of HIV led to higher HIV infection rates among higher SEP groups.12 ,15 Since HIV arrived in different countries at different times, emerging first in East Africa before spreading south and west, different countries have been at different points on the epidemic curve at the same calendar time.
HIV may have been more common among higher SEP groups during the first wave of pandemic spread. However, in later years, the pattern of incident infections likely reflects the more common association between lower SEP and worse health outcomes. During the 1990s, global attention on HIV began to increase. By the time of the 2001 United Nations General Assembly Special Session on HIV/AIDS, huge resources were being spent on HIV prevention. Victora's hypothesis suggests that these scaled-up interventions would benefit higher SEP groups more quickly than lower SEP groups. HIV incidence (and later prevalence) would fall fastest among high SEP groups. Figure 1B expresses this hypothesis in relation to HIV spread in a manner similar to Victora's paper, using output from a highly simplified hypothetical compartmental mathematical model of infection transmission in two socioeconomic groups.
A 2008 review suggested that changes over time predicted by Victora's hypothesis are occurring in a number of African settings.14 However, as discussed above, there is much diversity in the datasets pooled for that analysis, and formal meta-analysis was not possible. However, in several countries, data from methodologically comparable nationally representative household surveys are now becoming freely available. Table 1 shows data from eight African countries for which national surveys at two time points are available in the public domain. In the first survey, only among men from Zimbabwe and women from Lesotho and Benin was HIV prevalence lower among those with secondary education than among those with no education. Between the two surveys, HIV prevalence rose or remained stable in the uneducated group in five of the eight countries among both men and women. HIV prevalence fell among those with secondary education in all cases except women in Malawi and men in Ethiopia. While these analyses are unadjusted and quite crude, the pattern is similar to that predicted by Victora's hypothesis if we make the relatively safe assumption that HIV prevention interventions that were not available in the 1980s or 1990s have been more widely available since approximately 2000. More data of this type are emerging. We are currently engaged in a more formal multicountry analysis of this phenomenon.
There has been much confusion regarding the apparently paradoxical associations between SEP and HIV infection found in sub-Saharan Africa. A guiding theory explaining patterns in the data would help make predictions about the future course of the epidemic, and could guide resource allocation. Available evidence suggests that in the early phase of the epidemic, HIV infections were concentrated among those of higher SEP in many countries. The inverse equity hypothesis suggests that new infections will increasingly concentrate among those of lower SEP. If further analysis confirms this hypothesis, policy responses must be considered. For example, should the roll-out of new HIV prevention strategies (eg, male circumcision, treatment for prevention strategies and female microbicides) be targeted at low SEP groups? Is this feasible and compatible with the development of high-quality services? Does this phenomenon argue for greater investment in structural approaches to HIV control that tackle the social determinants of HIV infection? These strategies, targeting interventions and addressing the social determinants, have been suggested in relation to child health interventions where there is more evidence to support the inverse equity hypothesis. The HIV/AIDS community must also consider these approaches if emerging data confirm the inverse equity hypothesis.
We are grateful to Measure DHS and their country partners for access to the data provided in the Table.
Contributors JRH conceived the paper and wrote the first draft. CD extracted data from DHS reports and contributed to the writing. RGW did the mathematical modelling work for the paper.
Funding JH was supported by STRIVE, a Research Programme Consortium funded by UKaid from the Department for International Development. STRIVE is a collaboration dedicated to tackling the structural drivers of HIV. RGW is funded by a Medical Research Council (UK) Methodology Research Fellowship (G0802414), the Consortium to Respond Effectively to the AIDS/TB Epidemic (19790.01), and the Bill and Melinda Gates Foundation (21675).
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.