2.2. Measurements

Using the eHARS data, the annual county-level viral suppression rate was calculated as the percentage of PWH with a viral load of fewer than 200 copies per ml in each county's last viral load report at each calendar year (excluding those newly diagnosed in that year) (19). Individuals without viral load records in the calendar year (ranging from 30.78% to 33.48%) were excluded from the analysis in that year. Still, they were included in the analysis for other calendar years. This calculation criterion is in line with the US Department of Health and Human Service's calculation (20).

We calculated the racial/ethnic residential segregation using Massy and Denton's formula of isolation index for non-Hispanic black residents (21). The isolation index is suggested to measure racial/ethnic residential segregation regarding infectious disease as it reflects the probability that a minority person shares a unit area with another minority person (22). In this study, the non-Hispanic black isolation index for a county is calculated as follows:

In this calculation, i is the ith census tract in the county, and n is the number of census tracts in the county. The isolation index reflects the probability that non-Hispanic black residents will come across others of the same race/ethnicity, and the index ranges from 0.0 (complete integration) to 1.0 (complete segregation) (23).

We extracted the community health index for each county directly from the US Congress Joint Economic Committee website (16). Community health was calculated based on the registered non-religious non-profits per 1,000, religious congregations per 1,000, and Informal Civil Society Sub-Index (constructed from various state-level sources, such as the share who volunteered, who attended public meetings, and who participated in political activities) (16). The original values of these indicators were standardized, weighted based on principle components analysis, and summed to generate the community health index. The value of community health for the 46 counties in South Carolina ranges from −1.09 to 4.12, with a higher score indicating a higher level of community engagement and volunteerism in the local area.

Based on existing literature on the social and structural determinants of HIV viral suppression (24), we summarized the potential confounders into three categories: (1) population composition (e.g., percent of male, percent of the population who were at least 18 years old); (2) socioeconomic characteristics (e.g., percent of persons with income below poverty level, percent of the population who were unemployed); and (3) healthcare access (e.g., percent of persons with no health insurance coverage, the number of Ryan White HIV centers per 100,000 population within 25 miles radius). We extracted these potential confounders from multiple publicly available datasets, such as the 2014–2018 5-year estimated America Community Survey and the US Congress Joint Economic Committee. The detailed definition and data source of each covariate are displayed in Table 1.

The detailed description of each variable and data source.

We linked the viral suppression rate, residential segregation index, community health index, and potential confounders by the unique Federal Information Processing Standards (FIPS) code of each county.