Exploring Intersectionality Among Latinos Living with HIV: A CHAID Analysis of Health Outcomes and Care Engagement

Latina/o/x (Latino) populations are disproportionately affected by HIV, with intersecting social and structural factors shaping outcomes. To capture these complex intersections, we applied a decision tree approach, Chi-squared Automatic Interaction Detector (CHAID), to examine engagement in care and viral suppression among adult Latinos enrolled in the DC Cohort, a longitudinal electronic health record-based cohort of people with HIV across 15 Washington, DC, clinics (2011-2019). Socio-structural variables included age, gender identity, HIV risk category, country of birth, mental health diagnosis, substance use, housing, employment, and insurance. Engagement in care was defined as ≥ 2 visits ≥ 90 days apart with laboratory work in the prior year; viral suppression as < 200 copies/mL. Analyses included chi-square tests and CHAID decision trees. Among 541 participants, most were cisgender male (85%) with a median age of 47 years; 48% were foreign-born, 31% had substance use disorder, and 49% a mental health diagnosis. Overall, 80% were engaged in care, and 88% were virally suppressed. CHAID identified site of care (academic vs. community clinic) as the strongest predictor of engagement, with higher engagement in community clinics. Within community sites, employment, housing stability, and age further differentiated outcomes. For viral suppression, housing was the most important predictor, followed by HIV transmission risk, substance use, and age. Decision tree analysis enabled an intersectional, non-additive approach to HIV outcomes. Housing emerged as the most significant factor, interacting with employment, substance use, and age. Findings highlight the utility of CHAID for identifying at-risk subgroups and guiding targeted support for Latinos with HIV.