Homelessness is a substantial problem among families under the supervision of the child welfare system (Culhane, et al., 2003). Children in homeless families who come to the attention of the child welfare system are at increased risk for placement (Fowler, et al., 2013), and children from homeless families who need to be placed are less likely to reunify than maltreated children who are not homeless (Courtney, McMurtry, & Zinn, 2004). In an effort to demonstrate the effectiveness of housing interventions for improving child welfare outcomes, the Children’s Bureau funded five sites nationally for five years to design and test models that provided permanent housing along with supportive services (U.S. Department of Health & Human Services, 2017). This paper includes long-term child welfare outcomes from the San Francisco demonstration site, Families Moving Forward (FMF).
A total of 261 children in 154 families were randomized from 2013 through 2016 (133 treatment group children in 79 families and 128 control group children in 75 families). A longitudinal database using administrative data was developed to follow preservation and reunification outcomes for at least two years post-randomization, addressing the primary research questions on the prevention of out-of-home placement, likelihood and timing of reunification, recurrence of maltreatment, and likelihood of reentry.
The initial method was an Intent-to-Treat (ITT) approach. However, because approximately one-third of treatment group families either never engaged with FMF or left before they were housed, we also report a Complier Adjusted Causal Effect (CACE). Probit regressions estimated dichotomous outcomes; Cox-proportional hazards estimated durations. Models controlled for child age, gender, and race. All analyses were conducted at the child level with standard error adjustments for the clustering of children within families.
For children in preservation cases at randomization, we found no significant impact of FMF on the time to case closure, or the probability of out-of-home placement up to two years after randomization. We also did not find any significant impact of FMF on any of the recurrence metrics (re-report, re-substantiation, subsequent case openings, or reentry into foster care).
For reunification cases, treatment had a marginally significant impact on reunification within six months using the ITT model (B=0.195, p=0.082), and a similar result using the CACE (B=0.241, p=0.073). While descriptively more children in treatment group families reunified at 12 and 24 months than children in control group families, the difference was not statistically significant. Measuring time to reunification adds nuance to the marginal six-month finding. Children in treatment group families were nearly twice as likely to be reunified within the first six months after randomization (B=1.78, p=0.049).
The presenter will discuss the findings in the context of important practical challenges with implementation, namely that in San Francisco the median time to house families was nearly ten months. Coupled with the finding that reunification tended to happen within six months, a perplexing result emerged: 85 percent of reunifications preceded housing. Early and intensive supportive services might have provided the motivation for families, workers, and the court to expedite reunification.