Introduction The design and evaluation of national HIV programs often relies on aggregated national level data. However, there is often substantial geographical heterogeneity in the HIV epidemic within countries, which could be used to tailor programmatic responses to the specific local needs. Given this background, the study attempts to answer the question: Could environmental and socio-behavioural factors be used to generate sub-national estimates of HIV prevalence in sub-Saharan Africa?

Methods Data were obtained from Demographic and Health Surveys (DHS) conducted in Tanzania, Malawi, and Kenya. Associations between covariates and HIV prevalence were assessed using non-spatial logistic regression models. Covariates included life time number of sexual partners, percentage of condom use, wealth index, percentage of male circumcision (MC), distance to urban settings, distance to roads, and normalised difference vegetation index (NDVI). The prediction formula generated from the covariate analysis along with kriging interpolation techniques were used to produce high resolution continuous surface maps of HIV prevalence.

Results Condom use, MC, distance to main roads, NDVI, poverty and wealth index were associated with HIV prevalence in Tanzania and Malawi, and were included to generate HIV prevalence maps in these countries. Conversely, only NDVI, poverty life time number of sexual partners and MC were associated with HIV prevalence in Kenya. All three maps illustrated substantial geographical variation of HIV prevalence within the country, and localised areas where HIV prevalence is concentrated.

Conclusion The HIV prevalence maps generated highlight the stark spatial disparities in the epidemic within a country, and localise areas where both the burden and drivers of the HIV epidemic are concentrated. These maps present an opportunity to apply differential local approaches to maximise impact by informing both planning and delivery of defined packages of services. Additional work is needed for routine incorporation of such strategies into existing national HIV estimation processes.

Disclosure of interest statement Nothing to declare.