The variation in the proportion of individuals living in a stable HIV sero-discordant partnership (SDPs; i.e., one partner testing HIV seropositive, while the other testing HIV seronegative) [1
], and the potential drivers of variability in HIV sero-discordancy patterns across sub Saharan Africa (SSA) are still not well understood. There is also an intense discussion regarding the potential role of HIV seroconversions among SDPs in the HIV epidemic in SSA, and the prioritization of HIV control interventions focused on HIV discordant couples over other intervention approaches [4
A number of studies have assessed the dynamics of HIV discordancy among SDPs and contribution of SDPs to HIV epidemics, using the cross-sectional Demographic and Health Survey (DHS) data as input data [9
], and found that HIV epidemiology among heterosexual stable couples appears to be a predictable “spill-over” effect of the “core” HIV dynamics in the population [1
]. Moreover, the studies indicated that national-level HIV prevalence explains most of the variation in HIV discordancy measures [1
]. These studies also found that in HIV high-prevalence countries, a considerable proportion of stable couples were affected by HIV, and half of these stable couples were HIV discordant [1
]. In contrast, in low-prevalence countries, smaller proportions of stable couples were affected by HIV, but a higher fraction of these couples were HIV discordant [1
]. These studies concluded that for most countries, HIV incidence among SDPs is unlikely to exceed 50% of new HIV infections in the population at large [8
While these studies included data from 20 countries from SSA, estimations were conducted using aggregated data at the national level [1
]. Recent studies have shown stark geographical variation in HIV prevalence at the subnational level in most countries in SSA, in which a geographical clustered HIV transmission within micro-epidemics of different scales appears to be a common pattern [14
]. This local variation in HIV prevalence within a country may reflect the spatial and temporal dynamics of the HIV epidemic in the country [15
], and estimations using aggregated data at national level could mask local dynamics of the infection. Consequently, an assessment of the geographical patterns of HIV discordancy is of particular relevance.
Against this background, and with the ultimate goal of developing more effective strategies for HIV prevention interventions, this study aimed to (1) examine the geographical distribution of HIV sero-discordancy at the country level and locate clusters of high numbers of SDPs; (2) characterize these clusters by estimating key measures of sero-discordancy; (3) explore descriptively the overlap between SDP clusters and HIV infection clusters; and (4) assess the association between several sero-discordancy measures and HIV prevalence across identified clusters and outside identified clusters. Accordingly, our study aims to clarify HIV epidemiology at the intersection of two topical themes of HIV research today, spatial dimension of the distribution of HIV infection and the role of SDPs in the HIV epidemics in SSA.
We characterized the spatial distribution and clustering of SDPs in high HIV prevalence countries in SSA. We also presented different epidemiologic measures of sero-discordancy within and outside HIV infection clusters and assessed the association between these measures and HIV prevalence. Our results suggest that there are no distinct spatial patterns for HIV sero-discordancy that are independent of HIV prevalence patterns. The spatial clusters of discordancy overlapped with those for HIV prevalence and there were no discordancy clusters that are distinct from those of HIV prevalence.
The variation of the epidemiological measures of discordancy with HIV prevalence across clusters and outside clusters demonstrated similar patterns to those observed at the national level [1
]. The discordancy patterns seen at the national level were mirrored at the sub-national cluster level with no signature for an independent role for geographic variation. This finding is consistent with the picture emerging form a series of studies on discordancy [1
] that HIV epidemiology among heterosexual stable couples appears to be simply a predictable “spill-over” effect of the “core” HIV dynamics in the population, and that stable discordant couples do not constitute a core factor in dictating the dynamics of HIV infection.
were found to significantly increase with higher HIV prevalence at local level, just as they do at the national level [1
]. As the local HIV prevalence increases, the number of HIV infected individuals increases as well, and therefore more discordant partnerships are likely to form in the population, and larger fraction of the sexually active population would be part of a discordant partnership. The observed correlations between both
and HIV prevalence affirm this explanation and that the dynamics of discordancy, whether within or outside clusters, just mirrored the dynamics in the whole population.
Though the trend was not significant for
, both of these measures of discordancy were found to decline with higher HIV prevalence, also just as observed at the national level [1
]. These measures describe the persistence of discordancy among HIV-affected partnerships, that is, the chance that these partnerships will form and will remain sero-discordant as opposed to sero-concordant positive. With higher HIV prevalence, there are more chances for an HIV-affected partnership to be or to become concordant positive as opposed to discordant. Therefore, this pattern further affirms similarity of HIV discordancy dynamics whether within or outside clusters, or in the whole population.
The above results are also consistent with the national analyses indicating that most of the variability in discordancy measures could be explained by HIV prevalence alone [1
]. Kenya provided a stark example to this end where the discordancy measures varied immensely between within and outside clusters, mirroring the stark variation in HIV infection distribution in this country. Our results also affirm that discordant couples constitute the majority of stable couples where at least one partner is HIV infected (
) in geographic areas with low HIV prevalence; conversely, only half of these stable couples are discordant in areas with high HIV prevalence. Moreover, the fraction of couples engaged in an SDP is low in areas with low HIV prevalence, and high in areas with high HIV prevalence.
Our findings suggest that the spatial dimension does not appear to be a fundamental nor independent determinant of the observed patterns of sero-discordancy in high HIV prevalence countries in SSA. While the spatial dimension may not be critical for a proper understanding of discordancy dynamics, the temporal dimension could be an important factor, and may influence discordancy as an HIV epidemic sweeps through a population. Interventions, such as antiretroviral therapy, may also influence the dynamics of discordancy and contribution of SDPs to HIV incidence. Investigating the time variable in discordancy dynamics may provide important insights about drivers of the historical evolution of HIV epidemics, and future projections for discordancy and for SDPs’ role in the epidemic.
Our study has several limitations. We did not find statistically significant SDP clusters in several countries. However, this is likely a consequence of the fact that the number of SDPs is smaller than the number of HIV infections in any region, leading to lack of statistical power to detect these clusters. Moreover, even when there were no statistically significant SDP clusters, there was a trend for SDP clustering. This limitation therefore is not likely to have affected our conclusions. Inclusion of countries in this study was constrained by the availability of a DHS with HIV biomarker information and geographical coordinates of each survey data point. While DHS data are of high quality, they may not be statistically powered to answer research questions at the identified sub-national clusters level. This has limited our ability to control for potential biological or behavioral confounders that could be associated with SDP or HIV clustering besides HIV prevalence. Further analyses including surveys statistically powered in small geographic areas may be necessary to identify the drivers of SDP and HIV spatial clustering in SSA.
Despite the fact that the DHS program has a detailed and well-validated protocol to conduct HIV biomarker data collection, laboratory testing of surveyed specimens is conducted in the host country and may vary from one country to another. Specific assays used in the HIV testing algorithm are generally determined in coordination with the host country, which could be a source of data variation and quality between countries. It has been highlighted recently, based on an analysis of 20 countries, that testing errors associated with false positivity could be present in DHS surveys [24
]. Substantial variation among countries was observed, with greater magnitude of error introduced by false HIV positive results in Malawi, Niger, Sierra Leone, Senegal, and Zambia. Although we conducted our analyses in several countries at the same time, and despite the fact that there are not clear conclusions regarding the impact of this source of error on HIV prevalence estimations and other type of analyses including HIV biomarkers, it is prudent to take into consideration this source of error and interpret our results within the context of this limitation.
Lastly, given the multiple logistical difficulties in conducting DHS rounds, some of our measures could have been influenced by inherent biases in the data such as the variability in response rate to HIV testing [25
]. A potential bias in our study is the geographical position system (GPS) displacement process of the DHS sampling data points, used to preserve the confidentiality of the data points [27
], which could have impacted the precision of the geographical location of the discordancy and HIV prevalence clusters (by few kilometers at most).