A county-level HIV prevention gap index in the US Deep South using publicly available proxy indicators.

BACKGROUND

The US Deep South bears a disproportionate human immunodeficiency virus (HIV) burden. Mismatch between HIV burden and measurable proxies of biomedical prevention and care performance, pre-exposure prophylaxis (PrEP) utilization, viral suppression (treatment as prevention), and density of listed HIV testing locations may contribute to geographic variation in outcomes, but regional patterns are poorly described.

METHODS

We conducted an ecological county-level analysis across the US Deep South states (Alabama, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, and Texas). Using public datasets, we constructed a Prevention Gap Index (PGI), defined as standardized HIV prevalence minus the mean of three standardized prevention/care proxies aligned with Ending the HIV Epidemic (EHE) domains: pre-exposure prophylaxis (PrEP) (Prevent), viral suppression (Treat), and testing site listing density (Diagnose/access). Indicators were z-standardized, and counties were classified into PGI quartiles to flag potential gaps for planning and resource allocation. Spatial weights used queen contiguity. Multivariable ordinary least squares (OLS) and spatial regression models examined associations between PGI and structural determinants (ethnicity, income, education, and social association measures). Convergent validity was assessed by comparing high PGI counties with EHE Phase I priority counties.

RESULTS

The dataset included 877 counties. Testing site density averaged 7.40 per 100,000 [standard deviation (SD), 8.71]; PGI mean was 0.01 (SD, 1.12). Notably, 65% (13/20) of regional EHE Phase I counties were high PGI (top quartile), and PGI moderately differentiated EHE designation [area under the curve (AUC), 0.761; p < 0.001]. Among counties with available new diagnosis rates (n = 512), PGI correlated with diagnoses (Spearman coefficient, ρ = 0.51; p < 0.05) and classified high-diagnosis counties (AUC, 0.819; p < 0.001). In bivariate analyses, PGI correlated positively with the percentage of non-Hispanic Black (r = 0.362; p < 0.001) and negatively with log income (r = -0.171; p < 0.001), social associations (r = -0.095; p = 0.005), and some college education (r = -0.085; p = 0.012). In adjusted OLS models, the percentage of non-Hispanic Black remained positively associated with PGI (β = 0.024; 95% confidence interval [CI], 0.020-0.027). Residual clustering supported spatial modeling; spatial lag and error models improved fit, with the spatial error model best capturing remaining dependence.

CONCLUSION

HIV burden and prevention/care proxies vary spatially across the Deep South. PGI may support geographically targeted planning and monitoring toward national HIV goals and evaluation locally over time, but should be complemented with local program data that capture additional prevention domains.