Does neighborhood matter? An analysis of HIV prevalence in Sub-Saharan African countries

ABSTRACT

Over two-thirds of the population living with HIV were concentrated in Eastern, Southern, Western, and Central Africa in 2021. This paper employs data from the Demographic and Health Survey to assess the relationship between HIV prevalence and its socio-economic and demographic drivers at the neighborhood (macro-cluster) level. Additionally, the study examines the existence of differences in such relationships among countries. The results of the fractional logistic regression models highlight that highly educated neighborhoods are less likely to be affected by HIV. A greater average number of children, potentially due to programs that promote access to preventive antenatal care and prevention, is associated with a lower likelihood of residents living with HIV. Notably, HIV testing coverage is prevalent in neighborhoods with a high prevalence of HIV. It is also evident that there are notable differences between countries, which demonstrate national context plays a crucial role in the association between education, number of children, testing coverage, and HIV prevalences.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Data Availability Statement

The datasets are available in the DHS repository at https://dhsprogram.com/.

Notes

¹ The HIV datasets contain biomarker data and results of ELISA (enzyme-linked immunosorbent assay) tests to determine respondents’ serostatus.

² The geographic datasets contain the GPS coordinates of the centroids of the clusters of households involved in the survey.

³ This statistic must be considered as an underestimation since no data about HIV is available in official statistics for Mozambique (Baffes, Hyland, and Blankespoor 2022; UNAIDS 2022).

⁴ Clusters in the DHS are census units. For further details see Croft, Marshall, and Allen (2018).

⁵ Centroids are defined as single points representing the barycenter of a geographical area. Since, in this case, clusters represent groups of households, their GPS coordinates refer to the location of the centroid of each territorial area (i.e., cluster). Hence, macro-clusters have the barycenter of the clustered geographical coordinates as their mean point.

⁶ Macro-clusters with missing data and those in the top and bottom 5% for population size were dropped from the analysis.

⁷ Country dummies were included as covariates to account for regional specificity, common to macro-clusters in the same country while different among countries.

⁸ Angola, having the median HIV prevalence among the analyzed countries, was chosen as the reference in all four models.

⁹ Wealth is only significant in Model 2 and Model 3.

Additional information

Funding

Anna Maria Parroco was supported by the University of Palermo under [FFR2024] research grant. Micaela Arcaio was supported by the National Operatiional Programme (PON) on Research and Innovation 2014-2020 of the Italian Ministry of University and Research.