Abstract
Background
Homelessness is a risk factor for family involvement with child welfare services (CWS). Housing interventions are promising—but reasons for this are not well understood, and housing resources could be better targeted to families at risk of increased CWS involvement.
Objective
We sought to better understand the relationship between homelessness and CWS involvement and examine whether homeless shelter data could combine with CWS data to enhance intervention targeting.
Participants and setting
For 4 years, we followed 2063 families investigated by the San Francisco Human Services Agency in 2011.
Methods
Matching CWS data to homeless shelter data, we fit Cox models to examine the relationship between shelter use and subsequent CWS outcomes and produced ROC curves to judge model accuracy with and without shelter information.
Results
Absent CWS covariates (family demographics, CWS history, and family safety and risk), past shelter entry predicted repeat maltreatment referral (HR = 1.92, p < .001), in-home case opening (HR = 1.51, p < .05), and child removal (HR = 1.95, p < .01), but not child reunification. With CWS covariates, past shelter use no longer predicted case opening and child removal, but still predicted referral (HR = 1.58, p < .01). Shelter data did not contribute to models’ predictive accuracy.
Conclusions
We find mixed evidence that shelter use independently leads to CWS involvement. Housing interventions might help by addressing present housing problems and family experiences correlated with past shelter use. However, we find no evidence that data matches with shelter systems could enhance targeting.