Abstract Body

Background

Novel HIV prevention methods such as injectable lenacapavir (LEN) will have the greatest impact by reaching populations at the greatest HIV risk in the highest incidence areas. However, incidence is not directly measurable at small geographic scales. We explored relative HIV incidence proxy measures, calculated from routine service delivery data, to identify subdistricts for prioritised HIV prevention in South Africa.

Methods

We used aggregated DHIS2 data (2023-24) on HIV status ascertainment in routine antenatal care (ANC) from 242 subdistricts, and district-level estimates from the Naomi small-area estimation model. Three incidence proxy metrics were constructed: HIV diagnosis rate at first ANC visit, proportion seroconverting between HIV tests during pregnancy, and district-level HIV incidence derived from viremia prevalence. Each proxy provided relative weights to distribute the national number of new infections among adults aged 15–49-years across 242 subdistricts. Consensus subdistrict incidence was estimated as the average of the three proxies. To identify optimal geographic LEN allocation, we constructed concentration curves by ordering subdistricts by descending consensus incidence rate and calculating the cumulative proportions of new HIV infections, and HIV-negative adults.

Results

Across 242 subdistricts in 2023-24, a median 4.7% (IQR: 3.4%–6.1%) of women tested positive for HIV at first ANC visit, 0.55% (IQR:0.39%–0.84%) of women HIV-negative at first ANC tested positive at retesting, and median Naomi incidence among 15-49-year-olds was 5.0 per 1,000 (IQR:3.4–6.1). HIV retest diagnosis exhibited most variation (coefficient of variation [CV]=0.60), followed by HIV diagnosis at first ANC (CV=0.41) and district viremia prevalence (CV=0.37). Correlation was highest between HIV diagnosis at first ANC and modelled district incidence (r2=0.62). Median subdistrict consensus incidence was 4.7 per 1,000 adults aged 15-49 years (IQR:3.6-5.8).16% of new infections were in 29/242 subdistricts, representing 10% of HIV-negative adults, and 41% in 93/242 subdistricts with 30% of HIV-negative adults (Area Under the Curve =0.59; Fig 1).

Conclusions

Routine data can provide actionable and easily calculable proxy indicators for population relative HIV incidence. When focused on subdistricts with 30% of the South African population, geographic prioritisation of limited LEN, guided by incidence proxies, could avert around one-third more new infections compared to uniform geographic allocation.

Table or Figure