Planning area-specific prevention and intervention programs for HIV using spatial regression analysis

The study was conducted to inform area-based prevention intervention programs and plan resource allocation to reduce new infections in the District of Columbia (DC), United States of America.

The analysis used spatial regression to evaluate the spatial heterogeneity of the new HIV rate and its association with sexually transmitted infection repeaters (STIREPs) and socio-economic as well as demographic characteristics. The HIV and STIREP data were obtained from the DC Department of Health surveillance data (2010-2016). Other covariates were obtained from the American Community Survey, 2016.

Ordinary least squares (OLS) and geographically weighted regression (GWR) were used to compare global and local relationships. GWR-computed robust results were compared with other spatial regression methods such as spatial lag or spatial error methods.

For the OLS model, age, high school dropouts (NHSD), and the black population had an association with new HIV diagnoses (HIVDV_i). The results from the GWR model demonstrate spatial variations of association of STIREPs; mean age of each block group; and percentage of female population, NHSD, unemployment, and poverty with HIVDV_i. Akaike information criterion (AICc) value for the global model was 2770.99, and R² was 0.54 (54%). The R² and AICc of the GWR model was 0.81 (81%) and 2580.84, respectively, where the latter showed a 0.27 (27%) increase in R² and a decreased AICc.

These results will assist in planning HIV prevention and intervention strategies. These results will also be used for targeted testing, planning pre-exposure prophylaxis, and access to health care. The results will help plan resource allocation to community-based providers for prevention intervention programs and fund public health programs such as condom distribution, mobile vans, and youth-based sex education.