Privacy is a major concern with HIV-associated data. These data are often aggregated into larger spatial units to preserve privacy. However, the absence of HIV data at finer geographic scales limits the utility of spatial analyses to optimally target HIV interventions. Dasymetric mapping (DASY) is an areal interpolation method where the target polygons are zones of relative homogeneity with the purpose of best portraying the underlying statistical surface of the data being mapped. Here, we developed a cartographic DASY approach coupled with probabilistic reweighting to identify clusters of new HIV infections in San Diego County.
Age, sex, and ethnicity were collected for 657 HIV individuals enrolled in the San Diego Primary Resource Consortium (SDPIRC) across 6 SD Health and Human Services Agency (HHSA) regions. Transforming the data from HHSA region to a high resolution grid involved the following steps (Fig.): Generation of a background 500x500m grid surface combined with residential use data (step 1); DASY to interpolate data on residential land use, U.S. Census demographic data, and HIV prevalence data from Health Department into a 500x500m grid (step 2); finally, probabilistic reweighting was applied to the SDPRIC data to redistribute HIV new infection from HHSA regions to the 500x500m grid (step 3). Constraining variables (data from the SDPIRC cohort and grid cell map) were used to infer which grid cells HIV+ individuals were most likely to reside. A map was generated for each individual, and then aggregated for the full cohort to generate a final grid-based model of the distribution of the SDPIRC cohort.
The resolved grid map shows considerably more details of where clusters of new infections reside throughout the county compared to the map divided into the 6 HHSA regions. While the expected cluster of infection in central San Diego remains, two hot spots that are not visible at the HHSA region level map emerge in north SD County, and in east SD County (Fig., circled in blue). Furthermore, the final grid model shows increased resolution of hotspots of HIV new infections in central and south-central SD.
The ability to identify and predict the spread of transmissible diseases, including HIV, is crucial to optimally target treatment and prevention programs. Downscaling health data without violating privacy and confidentiality restrictions can help to reveal spatial patterns at the local level that are not apparent in aggregated data sets.