Background

Sub-Saharan Africa (SSA) has by far the largest HIV epidemic in the world, with 68%of the global HIV disease burden and 1.9 million new infections in 2010 [1]. The complex epidemiological context of this infection has prevented todate a clear identification of the specific drivers that have led to such largegeneral population HIV epidemics in SSA, and nowhere else [2].

The ‘Know your epidemic, know your response’ is a Joint United NationProgramme on HIV/AIDS (UNAIDS) focused on HIV prevention [3], which has become one of the first calls to modify the current strategyby recognition of the fact that there is not a single global HIV epidemic. Thisconceptual approach emphasizes the necessity to identify populations at higher riskof HIV infection, and to implement an effective prevention strategy byacknowledgement of the substantial variation in which HIV spreads throughpopulations and communities [4, 5].

This strategy also highlights the significant role that geographical space plays inthe identification of populations at higher risk. This fundamental characteristic ofan epidemic, however, has been poorly explored in the context of HIV. Measures ofdisease occurrence are frequently available only by large geographicaladministrative units. These large scales (national or regional) could hide thenatural scale of the HIV transmission process.

To overcome the current gap of knowledge about the spatial structure of the HIVepidemic, particularly in SSA, we explored and described the geographicalheterogeneity of the HIV epidemic in SSA. We aimed to identify geographical settingswhere the risk of HIV infection is higher or lower. Studying the epidemics at their“microscopic” scale, that is through spatial mapping of the clusteringof HIV infection at the local level, may provide a fresh look into the dynamics ofthese epidemics [6, 7], and informed insights about their drivers in this part of the globe.

Methods

Data sources

The main source of data for our study were the Demographic and Health Survey(DHS) [8] databases. Countries were included for analysis based upon theavailability of DHS HIV serological biomarker survey and the geographicalcoordinates of each survey data point. For each country, we only considered themost recent DHS where HIV data were collected. As a result, a total of 20countries in SSA were included: Burkina Faso (2010), Burundi (2010), Cameroon(2004), Congo Democratic Republic (2007), Ethiopia (2011), Ghana (2003), Guinea(2005), Kenya (2008–2009), Lesotho (2009), Liberia (2007), Malawi (2010),Mali (2006), Mozambique (2009), Rwanda (2010), Senegal (2010–2011),Swaziland (2006–2007), Sierra Leone (2008), Tanzania (2007–2008),Zambia (2007), and Zimbabwe (2010–2011).

Spatial clustering detection

We identified the spatial clusters with high and low numbers of HIV infections ineach country through a Kulldorff spatial scan statistics analysis [9]. This methodology has become the most widely used test for clusteringdetection in epidemiology [10–12], and its efficiency and accuracy has been well documented [13, 14]. A spatial scan statistics is a cluster detection test able to findthe location of areas with higher or lower numbers of cases (for instance HIVinfections) than expected under spatial randomness, and then evaluate theirstatistical significance by gradually scanning a circular window that spans thestudy region. The radius of the circle is changed continuously so that it cantake any value from 0 up to a pre-specified maximum value. A maximum circularwindow of 100 Km radius was used for scanning potential clusters with high orlow numbers of HIV infections.

Statistical analysis

For each potential cluster, a likelihood ratio test was computed assuming thatthe number of HIV infections in each circular window is an independent Bernoullirandom variable. The numbers of observed and expected HIV infections within andoutside the circular window were then compared with the likelihood L 0 under the null hypothesis of spatial randomness. The circular windowswith the highest likelihood ratio values were identified as potential clusters.An associated P-value of the statistics was then determined throughMonte Carlo simulations and used to evaluate whether HIV infections are randomlydistributed in space or not.

Cluster characterization

After a cluster was identified, the strength of the clustering was estimatedusing the relative risk (RR) of HIV infection within the cluster versus outsidethe cluster. The fraction of the population, and HIV prevalence were alsoestimated for each cluster. Furthermore, the general RR of HIV infection for allindividuals belonging to (high or low) clusters was also estimated by combiningall (high or low) clusters identified in a particular country. All geographicinformation system (GIS) analysis and cartographic displays were performed withthe software ArcGIS version 9.2 [15].

Mathematical modeling

A deterministic compartmental mathematical model was constructed based onextension of earlier models [16–18] to describe the heterosexual transmission of HIV in a givenpopulation [18]. The model consists of a system of coupled nonlinear differentialequations, and stratifies the population according to HIV status, stage ofinfection and sexual risk group. Our model incorporates 10 sexual risk groups inthe population, starting from lower to higher levels of sexual risk behavior.The level of sexual risk behavior was parameterized by the effective partnershipchange rate in each risk group, and in essence it is a measure of the risk ofexposure to the HIV infection. Further details about the model structure can befound in the Additional file 1.

Results

Our analysis identified 38 clusters with high HIV prevalence, and 45 clusters withlow HIV prevalence. The locations of these clusters are illustrated inFigure 1 for countries with HIV prevalence largerthan 4%, and in Figure 2 for countries with HIVprevalence lower than 4%.

Figure 1
figure 1

Spatial distribution of the clusters with high and low HIV prevalence incountries with national HIV prevalence higher than 4%. Geographicallocalization of the clusters with high (red-solid circles) and low(blue-dashed circles) HIV prevalence.

Full size image
Figure 2
figure 2

Spatial distribution of the clusters with high and low HIV prevalence incountries with national HIV prevalence lower than 4%. Geographicallocalization of the clusters with high (red-solid circles) and low(blue-dashed circles) HIV prevalence.

Full size image

From the 20 countries included in the study only two, Guinea and Swaziland, did nothave clusters with either high or low HIV prevalence, possibly as a consequence ofthe small geographic size of these countries. Clusters with only high HIV prevalencewere identified in Burkina Faso, Burundi, Democratic Republic of Congo and Mali.Clusters with high and low HIV prevalence were identified in the remaining 14countries. General description of the clusters for each country is summarized inTable 1, and a more detailed characterization of eachcluster is included in Additional file 1: Table S1.

Table 1 General description of the clusters with high and low HIV prevalenceidentified in the 20 countries included in the study
Full size table

HIV prevalence within the clusters with high prevalence ranged from 1.9% in a clusterin Senegal (RR = 3.33, P < 0.001) to 30.8% in acluster in Zimbabwe (RR = 1.91, P < 0.001), witha median of 11.5%. Likewise, HIV prevalence within the clusters with low HIVprevalence ranged from 16.4% in a cluster in Lesotho (RR = 0.71,P = 0.034) to clusters with 0% HIV prevalence in Liberia,Ethiopia, Ghana, Tanzania, Kenya, Mozambique, and Malawi, with a median of 1.7%.

The fraction of the population within clusters with high HIV prevalence had a medianof 14.4%, and this fraction increased with the national HIV prevalence(P = 0.005). This association, however, did not explain most ofthe variation (adjusted R-squared = 0.41). The fraction of thepopulation within clusters with low HIV prevalence had a median of 15.5%. However,we found no statistically significant trend in the association between the fractionof the population within clusters with low HIV prevalence and the national HIVprevalence (P = 0.73).

The RR of HIV infection for individuals within clusters with high HIV prevalence wasnegatively associated with the national HIV prevalence(P < 0.001) (Figure 3B).Conversely, the RR of HIV infection for individuals within clusters with low HIVprevalence was positively associated with the national HIV prevalence(P < 0.001) (Figure 3D). The HIVprevalence in clusters with both high and low HIV prevalence was positivelyassociated with the national HIV prevalence (P < 0.001).

Figure 3
figure 3

Strength of HIV clustering. (A) Relative risk of HIV infection inclusters with high HIV prevalence (each bar represents the relative risk ina single cluster). (B) Association between relative risk of HIVinfection in clusters with high HIV prevalence and the national HIVprevalence. (C) Relative risk of HIV infection in clusters with lowHIV prevalence (each bar represents the relative risk in a single cluster).(D) Association between relative risk of HIV infection inclusters with low HIV prevalence and the national HIV prevalence. Countriesare shown in order of increasing national HIV prevalence.

Full size image

The mathematical model constructed to assess the behavior of the epidemic around theepidemic threshold showed that a modest change of 10% in sexual risk behavior(illustrated by the red dots in Figure 4), correspondingfor example to a 10% increase in the number of sexual partners per year (or a 20%reduction in male circumcision coverage among adult males), leads to 250% increasein HIV prevalence. Beyond the region of epidemic threshold, the same increase insexual risk behavior (illustrated by the green dots in Figure 4) leads to only 8% increase in HIV prevalence, a factor about 30 timessmaller than the increase near the epidemic threshold.

Figure 4
figure 4

Generic relationship between sexual risk behavior (or HIV risk ofexposure) and HIV prevalence in an HIV epidemic. Three dynamicalregions can be discerned: 1) Below the epidemic threshold (blue zone)representing the HIV dynamics among the general population outside ofsub-Saharan Africa (SSA). 2) Just above the epidemic threshold (red zone)representing the hypothesized HIV dynamics among much of the generalpopulation of SSA. 3) Well above the epidemic threshold (green zone)representing the HIV dynamics among high-risk populations globally(including SSA). These results were generated using a conventionaldeterministic HIV epidemic model [16–18].

Full size image

Discussion

The results of our analysis indicate stark geographical variation in HIV prevalencein most of the countries. The observed spatial variation in HIV prevalencehighlights a clustered HIV transmission across SSA within micro-epidemics ofdifferent scales. The map of HIV clustering reflects a landscape with‘valleys’ (areas with high HIV prevalence), ‘dams’ (areaswhere HIV found barriers to propagate efficiently), and ‘islands’ (smallisolated areas with characteristically either very high or very low HIVprevalence).

Our results indicate that only ~14% of the population across the countries resideswithin clusters of high HIV prevalence. The strength of the clustering tended to behigher in countries with low national HIV prevalence. For instance, the strongestclustering (highest RR) is found in a cluster in Senegal (RR = 6.69, HIVprevalence = 4.3%); the country with the lowest national HIV prevalence(0.7%). Our study revealed similar settings with localized epidemics at high HIVprevalence hidden in a map of low national HIV prevalence, such as in Burkina Faso,Congo, Sierra Leone and Ethiopia.

The strength of the clustering was smaller in countries with high national HIVprevalence, indicative of more diffusive epidemics. For instance, in Zambia andLesotho, the strength of the clustering was fairly small (RR = 1.74, andRR = 1.28, respectively). In Swaziland, no clusters with high HIVprevalence were identified. This result underlines how the HIV epidemics in thesehigh prevalence countries had percolated throughout much of the demography andgeography of these countries.

We also identified clusters with low HIV prevalence in most of the countries includedin our study. These clusters appear to reflect ‘dams’ where somebehavioral or biological protective factors appear to have slowed HIV transmissionin such populations, in contrast to their neighboring populations. In fact, weidentified settings with very low HIV prevalence even in countries with substantialHIV epidemics such as in Tanzania, Kenya and Malawi.

The topography of this infection poses a question about the drivers of such starkheterogeneities even at micro-geographic scales within countries. Male circumcision [19], the presence of other sexually transmitted infections (STIs) [20], tropical co-infections increasing HIV viral load [17], hormonal contraception [21], viral factors [22], and host genetics and immunology [22] vary across SSA, and are believed to influence HIV transmission risk.Behavioral factors such as concurrency [23], number of sexual partners [24], commercial sex [25], and coital frequency [26] appear also to vary across SSA, and may contribute to explaining theheterogeneities in prevalence. Preliminary statistical analyses of the DHS databases(not shown) indicated that it is challenging, if not a formidable task, todisentangle the contribution of the different factors in the clustering of theinfection. This is a consequence of the complex array of independent variables toconsider, and also because of the population sizes of the clusters’sub-samples which are not large enough to power meaningful multivariate regressionanalyses. Nevertheless, these preliminary analyses suggest that the scale anddistribution of the differences in the biological and behavioral factors, withinversus outside the clusters, may not be sufficient to explain the observed sharpcontours in the topography of HIV infection at the local level. We suspect thatthere is an additional dynamical factor that has strongly influenced the localecology of this infection even when the differences in the biological and behavioralfactors may not have been markedly large.

We hypothesize that the HIV epidemic among the general population in much of SSA isnot far from its epidemic (or sustainability) threshold. A generic feature of aninfection epidemic is that near the epidemic threshold, the prevalence dependsnon-linearly on the determinants of infection transmission, and that small changesin the epidemiological context can drive much larger changes in the prevalence ofthe infection [27]. Figure 4 illustrates this dynamical effectfor the case of HIV infection. As can be seen in the figure, modest changes insexual risk behavior near the epidemic threshold could generate a substantialincrease in HIV prevalence. Conversely, beyond the region of epidemic threshold, thesame increase in sexual risk behavior could generate only a modest increase in HIVprevalence.

Accordingly, we hypothesize that an essential driver of the stark variability in HIVinfection transmission in SSA is that the epidemiology of this infection is not farfrom its epidemic threshold in the general population outside of conventionalhigh-risk groups. The conspicuously large clustering of HIV infection may notstrictly reflect conspicuously large variations in sexual risk behavior or thepresence (or absence) of specific biological co-factors in HIV transmission. Thevariability in sexual risk behavior or biological co-factors within the populationhas driven a much larger variability in HIV prevalence, thanks to the non-linearepidemic dynamics near the infection sustainability threshold. This hypothesis mayalso contribute to explaining the global variability in HIV infectious spread whereonly in SSA massive general-population HIV epidemics have occurred. In SSA, butnowhere else, the epidemiology of HIV infection has crossed, though not by far, theepidemic threshold of sustainability in the general population (Figure 4). That was enough to spark localized epidemics in the generalpopulation; and these epidemics, not far above the sustainability threshold,exhibited consequently high diversity in size at the micro-geographic scale(Figures 1 and 2).

Several study limitations could have affected our results. First, the selection ofthe DHS round for the different countries was constrained by the availability of HIVbiomarker information and geographical coordinates of each survey data point at anyparticular DHS round. This limited our ability to consider more countries in SSA foranalysis with more recent DHS rounds. Small clusters of HIV infection could havebeen missed if there is not enough sampling within their geographic setting. Giventhe multiple logistical difficulties in conducting the DHS, some of our measurescould have been influenced by inherent biases in the data such as the variability inresponse rates to HIV testing [28, 29].

Mobile individuals and high-risk subpopulations such as female sex workers, injectingdrug users, and men who have sex with men, may have been undersampled by the DHS.Clusters of HIV infection among such subpopulations may have been missed in ouranalysis. It is not clear though whether undersampling of such populations couldnecessarily affect our findings or not. Epidemics among high-risk subpopulationsshould lead to some infection onward transmission such as among spouses and clientsof female sex workers which are less likely to be undersampled in the DHS. Lastly,due to the cross-sectional nature of the data used in this study, some of theclusters identified here could reflect epidemics at different stages, rather thangenuine differences in epidemic sizes. The HIV epidemic in SSA, however, is in amature stage [25, 30], and therefore this potential limitation is probably not influencing ourresults. Moreover, we analyzed the clustering of the infection in four countriesthat had more than one DHS serological biomarker survey at different years, and noconsequential changes in the distribution of the clusters were observed (notshown).

Conclusions

In sum, our study provides evidence for a striking geographic clustering of HIVinfection across SSA. The exact drivers of such rich and complex infectiontopography are not well understood. However, the clustering possibly reflectsdifferences in specific behavioral and biological factors between sub-populationsthat have been amplified as larger differences in HIV prevalence, as a consequenceof the infection epidemiology being not far from its epidemic threshold. Ourfindings support the need for spatially-targeted prevention strategies in SSA, andour results delineate the map of the high disease-burden areas. If indeed HIVepidemiology in SSA is not far from its epidemic threshold in the generalpopulation, this would indicate that even modest intervention-driven changes in riskbehavior, or risk of HIV acquisition, through behavioral or biomedicalinterventions, may have considerable impact in reducing the sizes of the Africanepidemics. The recently observed rapid declines in HIV prevalence in SSA [1, 31, 32] may suggest that this transition is already taking place.