Methods: We used data from the 2005, 2011, and 2016 Ethiopia Demographic and Health Survey program (DHS). The spatial-temporal distribution of HIV was estimated using the Kuldorff spatial scan statistic, which determines the likelihood ratio of HIV within possible circular clusters across the country. Significant clusters (p<0.05) were identified and compared based on known HIV risk factors using descriptive statistics to compare them to the non-cluster area of the country. All analyses were conducted in SaTScan and R.
Results: Data from 11,383, 29,812, and 26,753 individuals with HIV were included in the 2005, 2011, and 2016 DHS, respectively. Four HIV clusters were identified consistently over the three time points, with the clusters representing 17% of the total population and 47% of all HIV cases. The four clusters were centered on the Addis Ababa, Afar, Dire Dawa/Harare, and Gambella regions, respectively. Cluster 1 is characterized by higher levels of unsafe injections (4.9% vs. 2.2%, p<0.001) and high-risk occupations, such as truck drivers (5.7% vs. 1.7%, p<0.001), when compared to non-cluster regions, but by lower levels of transactional sex (18.6% vs. 23.0%, p<0.001). Cluster 2 is also characterized by higher levels of high-risk occupations (2.8% vs. 1.7%, p<0.01), whereas cluster 4 is characterized by a lower prevalence of circumcised men (59.1% vs. 91.3%, p<0.01). No cluster had significantly higher levels of having more than one sexual partner in the last 12 months, although cluster 3 had a significantly lower level (0% vs. 1.7%, p<0.001).
Conclusion: HIV in Ethiopia is composed of heterogeneous clusters of HIV transmission that appear to be driven by different risk factors. Further decreasing the HIV burden will likely require targeted and prioritized interventions in specific regions rather than uniform national policies.
B. R. Schackman, None
S. Verguet, None