Small Area Estimation of County-Level U.S. HIV Prevalent Cases

Abstract Purpose Recent trends of HIV in the United States have indicated that the epidemic is no longer an urban issue; however, HIV data in rural settings is incomplete. Our objective was to estimate HIV prevalence in U.S. counties using small area estimation techniques (SAE) to better assess the burden of HIV nationally. Methods We performed SAE modeling to predict the reported number of HIV cases across the continental U.S., including unreported counties using source data from the CDC National HIV Surveillance System. Our model borrowed strength from auxiliary HIV risk-indicator data, including geospatial information. Cross-validation was conducted to identify and assess the precision of the estimates. Results Our findings showed that the majority of the 677 unreported counties had low HIV prevalence levels (quintiles 1-2). Estimates in the South had high levels of HIV (quintile 4- 5). Cross-validation techniques indicated good precision of the estimates, as 42% of the residuals were within +/- 10 HIV cases. Conclusions HIV was highest along the coastlines and in the U.S. South. Cross-validation techniques provided sufficient support of our model. Our study provides a more complete picture of the burden of HIV across the U.S. and identifies communities in need of future targeted interventions.