Mapping HIV in Kenya: Geospatial variability of HIV diagnoses, treatment, and impact : Implications for HIV epidemic control

Abstract

Background: Kenya is a country in sub-Saharan Africa (SSA) with an HIV prevalence in the adult population is around 5%, which is considered to be a generalized HIV epidemic. The generalized nature of the epidemic makes it difficult to target HIV services and interventions due to misalignment of geographic planning units and finer locations that may need extra resources. This thesis explored geospatial features and their associations with the HIV epidemic with a view of identifying gaps in prevention, care, and treatment. Using a variety of spatial statistics and analytics and mapping, we point out geographic areas that need focussed and intensified HIV interventions. Methods: In Paper I, we conducted a spatial scan statistical analysis to identify hotspots with disproportionate HIV infections using cross-sectional household survey data. In Paper II, we identified disparate geographic regions with high numbers of newly diagnosed HIV infections using routine program data. In Paper III, we conducted spatial-temporal analyses to show impact of prevention of mother to child transmission of HIV (PMTCT) through reduced rates of HIV infections among infants. In Paper IV, we used spatial-temporal analyses and structural equation models to show the covariance relationship of antiretroviral therapy (ART) and viral load suppression (VLS) in reduced HIV positivity over time in Kenya. Results: In Paper I, we have shown that HIV infection in Kenya exhibits localized geographic clustering associated with socio-demographic and behavioural factors, suggesting disproportionate exposure to higher HIV-risk. Identification of these clusters reveals the right places for targeting priority-tailored HIV interventions. The newly diagnosed HIV positives in Kenya are not necessarily, where the HIV burden is high. In Paper II, we identified wide-ranging spatial variation of new HIV diagnoses through cluster and hotspot identification analyses. High HIV-burden sub-National units (SNUs)/counties contain most high yielding sites but some sites are also in low-burden SNUs. Targeting HIV testing services for sites in low-burden regions needs a Geospatial approach. An outcome measure of the success of the PMTCT program through reduction of transmission is highlighted in Paper III. During this period – before universal treatment – the PMTCT program in this region had not reached the target rate of ≤50 cases per 100,000 live births. Using spatial-temporal models with covariates provided better estimates of prevalence and explained the geographically distributed disease burden. In Paper IV we show that over a 3-years period, (2015-2017), improved viral load suppression rates had a direct effect on reduced HIV positivity rates during an era of scaled up ART coverage in Kenya. To assess the trends and impact of implementation of scaled-up care and treatment, spatial-temporal analyses help in identification of geographic areas that need focused interventions. Conclusions: HIV prevalence in Kenya, though generalized, ought to be looked at more critically. Some efforts at epidemic control including ending mother to child transmission (e-MTCT) have born fruits though with geographical disparities. Given the present density of low-yield HTS sites in Kenya, geographic coverage and access to HTS may need better targeting at the spatial level to achieve knowledge of status for at least 90% of the population. Access to HTS is needed everywhere in Kenya, yet, targeting is difficult in low prevalence areas. Gains in reduced number of new HIV diagnoses have been demonstrated where viral load suppression rates are good. This study has demonstrated that geospatial analyses and mapping makes it easier to define refined geographic areas and hotspots in need of enhanced HIV prevention and treatment interventions. We have provided evidence that there are geographic disparities in HIV program impact in Kenya. Micro location-based planning is necessary for improved resource allocation. We recommend clustering analyses to identify areas with disproportionately high number of HIV-infected persons for re-allocation of resources within SNUs and continued use of geospatial analyses for advocacy and planning to help in achieving HIV epidemic control in Kenya.