Multidimensional Aspects of Social Networks: Implications for CPS Recurrence

Background and Purpose: Supportive social connections have been identified as a key protective factor in nationally recognized frameworks for preventing child maltreatment. Prior studies have focused primarily on either the protective effects of positive network ties or the risk associated with negative network ties, however, most social networks are comprised of both positive and negative dimensions. This study examines both positive and negative social network characteristics of mothers reported to and investigated by the child protection system (CPS). Specifically, we assessed having a recurrent CPS investigation within one year of an index investigation (that did not result in an ongoing CPS case) as a function of positive social network ties, negative social network ties, and perceived neighborhood support. We then used cluster analysis to identify distinct combinations of network characteristics among mothers with recent investigations and examined whether different cluster types were predictive of CPS re-investigations within one year.

Methods: Study data were derived from a randomized control trial designed to evaluate an intervention, Getting Access to Income Now (GAIN), to reduce CPS recurrence. Sample members included the subgroup of respondents who identified as female and as the biological or adoptive parent of one or more children in the home (N = 670). Key predictors included positive social network ties, negative social network ties, and perceived neighborhood support. The outcome of interest was a CPS investigation that occurred within one year of randomization into the GAIN study, which directly followed the closure of a CPS investigation. We controlled for a range of demographic and known child maltreatment risk factors.

Data analysis

First, we ran descriptive statistics on the outcome and all covariates. We then conducted a cluster analysis using our three key predictors of interest. We determined the optimal number of clusters using the Ward method and generated the final clusters using the kmeans procedure in SPSS. We then ran one-way ANOVA tests to examine how the other covariates and the outcome in our models varied by cluster. Finally, we used logistic regression to predict CPS recurrence within one year of the index CPS investigation.

Results: Across clusters, mothers reported average, high, or very high positive social ties. Clusters differed significantly on perceived levels of both positive and negative interpersonal ties as well as perceived neighborhood support and were associated with different levels of known child maltreatment risk factors. Controlling for sociodemographic characteristics and the treatment assignment group, Cluster1 respondents were 20% more likely than Cluster 3 respondents (the reference group) to have a CPS investigation within one year of their index investigation. Clusters with lower levels of perceived neighborhood support were more likely to be associated with future CPS investigations, however, this association became statistically insignificant when controlling for mothers’ depressive symptoms.

Conclusions and Implications: Study results suggest that a more multi-faceted view of social networks can be helpful to understand the social contexts of mothers as they experience contact with CPS and raises questions about how these contexts interact with parental mental health in relation to CPS recurrence.