Geographic variations in racial/ethnic disparities in HIV among injection drug users in the United States

Background: Research suggests that geographic factors are important determinants of racial/ethnic disparities in HIV/AIDS in the United States. However, most of these studies have evaluated social and structural factors at a single geographic unit. This may limit the analytic depth and policy implications of these studies. This descriptive study explores whether variance in HIV infection among people who inject drugs (PWID) is partitioned differently across multiple geographic units and by race/ethnicity.

Methods: Multilevel logistic regression was used to partition variance in the odds of testing positive for HIV across three hierarchical geographic units (i.e., ZIP codes, counties, metropolitan statistical areas [MSAs]) among 9,079 PWID recruited via respondent-driven sampling in 2009 from 19 MSAs as part of the Centers for Disease Control and Prevention National HIV Behavioral Surveillance System. Two multilevel models were created: the first included random intercepts for each geographic level; the second included random intercepts for each geographic level for three racial/ethnic groups (i.e. non-Hispanic White, non-Hispanic Black, Latino). Models did not include other covariates. Stata software was used.

Results: Overall, 19% of the variance in HIV infection was apportioned to geographic units. When race/ethnicity was included in analysis, 31% of the variance was apportioned to geographic units. The total variance apportioned to geographic units was highest for non-Hispanic Whites, and the relative contribution of each geographic unit varied by race/ethnicity (Table 1).

Table 1. Percent variance in HIV by geographic unit and race/ethnicity, among 9,079 PWID

Geographic unit	Percent variance in HIV apportioned to geographic units
	All racial/ethnic groups (N=9,079)	Non-Hispanic Whites (N=2752)	Non-Hispanic Blacks (N=4687)	Latinos (N=1640)
ZIP code	4.3	4.3	0.0	0.0
County	9.2	2.3	6.9	0.0
MSA	17.6	7.5	0.0	10.1
Total	31.1	14.1	6.9	10.1

Discussion:  Racial/ethnic differences in the overall variance in HIV and in the distribution of variance across geographic units were observed. Place appears to be an important factor for HIV among PWID, and the geographic scale that is most influential appears to differ by race/ethnicity. Studies should evaluate social determinants across multiple geographic units, including smaller geographic units, to inform the development of comprehensive strategies that incorporate place-based interventions.