Risk for New HIV Diagnosis Among a Cohort of Persons Receiving a Supportive Housing Intervention — New York City, 2006–2012

Background: Unstable housing is a risk factor for HIV infection, but unknown is whether providing housing decreases HIV incidence. Using a quasi-experimental design, we assessed whether HIV diagnosis rate was lower among... [ view full abstract ]

Background: Unstable housing is a risk factor for HIV infection, but unknown is whether providing housing decreases HIV incidence. Using a quasi-experimental design, we assessed whether HIV diagnosis rate was lower among persons placed into a supportive housing program than among unplaced applicants.

Methods: We studied 18,004 HIV-negative adults who applied for New York/New York III Supportive Housing between November 2006–September 2012. Subjects were followed until they received a diagnosis in the HIV registry or censored when they died or on September 30, 2012. Treatment was defined as having received housing placement. We used propensity score weighting to balance the sample regarding baseline demographic, behavioral, and housing characteristics. We calculated incidence rate ratios (IRRs) for HIV diagnosis by using Poisson regression.

Results: Compared with control subjects (n = 12,757), treated persons (n = 5,247) were more likely to be older, male, abuse substances, and to have spent more days in homeless shelters at baseline. Crude HIV diagnosis incidence was similar between groups (control incidence rate [IR]: 12.2/10,000 person-years; treatment IR: 12.0/10,000 person-years). Propensity score weighting successfully eliminated observed differences in baseline characteristics. In the intention-to-treat analysis, HIV diagnosis incidence was similar between groups (IRR: 1.0; P = .99). Persons who received ≥2 years of placement (n = 1,894) had lower rates of HIV diagnosis (IRR: 0.4, P = .11), although this difference was statistically insignificant.

Conclusions: Propensity score weighting reduced bias between treatment and control groups. Those who received ≥2 years of housing placement had a substantial but nonsignificant reduction of risk for HIV diagnosis. Analysis using a longer follow-up period with more events might clarify association between housing intervention and HIV infection.

Word count: 274

Keywords: Supportive housing, homelessness, HIV, treatment effects, propensity score