Child Well-Being Effects of a Supportive Housing Program for Families Involved with Child Welfare

Background: Homelessness is a critical barrier to child well-being, no less so for those involved with child welfare. In 2012, the Administration for Children, Youth and Families (ACYF) funded five demonstration sites across the country to design and test supportive housing programs for high-need homeless families involved with child welfare. The theory was that a “housing first” approach could reduce the need for foster care and promote child and family well-being by way of stabilizing housing. The San Francisco site provided permanent housing vouchers and intensive support services to homeless families beginning a child welfare case. In a randomized controlled trial, children from families receiving the intervention versus usual services were expected to exhibit greater improvement in well-being over time, as measured by the Child and Adolescent Needs and Strengths (CANS) assessment. 

Methods: 108 children (58 treatment, 50 control) received an initial CANS assessment following randomization into the experiment. Of those, 93 received one or more subsequent assessments over a period of up to four years after randomization. Five CANS indicators of child well-being were common to these opening and closing CANS documents: anxiety, depression, adjustment to trauma, development, and interpersonal functioning. Each was scored on a four-point scale. We tested for differences by condition in child well-being over time in two ways using the five CANS indicators: (1) a factor score extracted from a confirmatory factor analysis (CFA), and (2) a mean score. We entered each measure as the outcome in a multilevel model with fixed main and interactive effects for condition and time since randomization, and random effects at the child and family level. 

Results:  For half of all children, their first CANS was completed within 154.5 days of randomization (mean=231.56, sd=227.57), and their last CANS was completed within 772.5 days of randomization (mean=707.37, sd=393.66). The median time between CANS was 137 days (mean=166.43, sd=148.27). The CFA, which loaded all five CANS indicators onto one latent factor, demonstrated excellent fit statistics (chi2(5)=7.37, p=0.20; RMSEA=0.034, CFI=0.98, TLI=0.97), and good measurement invariance over time. Using a factor score extracted from the CFA as the outcome, multilevel models indicated that treatment children experienced a quarter of a standard deviation improvement in the well-being factor each year over control children (b=-0.27, p=0.02). Using the mean score as the outcome, models indicated that treatment children experienced a 0.1-point improvement in mean CANS score each year over control children (b=-0.13, p=0.02). 

Implications: Children in one demonstration site experienced somewhat greater improvements in well-being over time compared to those receiving usual services, as indicated by the CANS. While the magnitude of the effect is clinically small, it suggests that, notwithstanding mental health treatment which was not part of the intervention, participation in a multicomponent housing first program may help to improve aspects of child well-being, including social and emotional functioning and development.