The Effects of a Supportive Housing Program on Reunification from Foster Care

Background: Homeless families are more likely to have child welfare investigations (Font and Warren 2013) and experience an out of home placement (Culhane, Webb, Grim, & Metraux, 2003). Emerging evidence suggests that housing assistance improves child welfare outcomes. Housing vouchers may shorten the time to reunification (Pergamit, Cunningham, & Hanson, 2017) , and families that receive permanent housing subsidies may be less likely to have a subsequent foster care placement (Shinn, Brown, & Gubits, 2017).

In line with this evidence, the Administration for Children, Youth and Families (ACYF) funded Partnerships to Demonstrate the Effectiveness of Supportive Housing for Families in the Child Welfare System for five demonstration sites across the country between 2013-2017. The San Francisco site implemented an intervention comprised of permanent housing and supportive services for high-need homeless families beginning a child welfare case. A randomized controlled trial (RCT) design was used to test, among other things, the following questions about whether or not the intervention: 1. improved the likelihood of reunification; and 2. reduced the time to reunification.

Methods: Administrative child welfare data were used to build a longitudinal out-of-home placement file and link it to treatment and control children randomized into the experiment described above. The entry cohort randomized in calendar year 2014 was observed through December 2017. Sample characteristic frequencies were analyzed for differences in race, gender, age, and prior placements between the two groups (N=88). A bivariate Pearson chi-square test and a logistic regression tested the relationship between treatment assignment and reunification likelihood. Survival analysis assessed differences in the time to reunification.

Results: Of the children in the intent to treat group who were randomized and began an out of home placement spell in 2014 (N=48), 69% percent reunified compared to 38% of control group children (N=40, p=0.005). The multivariate logistic regression indicated that treatment assignment was positively associated with the likelihood of reunification (estimate=1.3, p=0.02). Log-rank and Wilcoxon tests for equality indicate that the median durations to reunification were not significantly different from each other. The findings are preliminary as this only represents the 2014 entry cohort, since nearly half of children randomized after 2014 were still in care at the time of analysis. Final results with an additional six months of observation time (through June 2018) will bolster the power of the analysis.

Implications: Permanent housing and supports may help parents reunify with their children but may not accelerate the process. The similar duration in care for both groups highlights the fact that this demonstration project intentionally targeted especially high-need families for randomization. Families in the experiment had substantial co-morbidities that, together with homelessness, may have extended treatment trajectories beyond what other child welfare involved families require. Furthermore, housing in San Francisco was far from an immediate inoculation; the median time for treatment group families to find housing was eleven months. Implementation challenges and their policy implications will be discussed.