Recurrence and Service Matching with Substantiated Child Welfare Cases

Background and Purpose:  Discourse about factors associated to maltreatment recurrence is ongoing, but neglect is consistently listed as a predictor. With an ultimate goal of preventing children from being exposed to chronic maltreatment it is vitally important to look at the effectiveness of services provided and to understand which services are most essential to successful prevention of future maltreatment.  Thus, the objective of this study is to explore the impact of matching services to their respective risk factors on recurrence.  We hypothesize that the levels of service matching (identified child and family risk factors are matched with the provided services) will impact the probability of recurrence.  Also of interest was the general question on: What other factors contribute to recurrence? Are there other factors that interact with matched services that would account for recurrence?

Methods:  Data used is from the NCANDS Child Files for FFY2010 and 2011. All except one state, District of Columbia, and Puerto Rico submitted data for the Child File- FFY2010.

 FFY 2010 and 2011 Child files datasets were used to develop a recurrence dataset that included only substantiated, indicated, and alternate response victim for the calendar year 2010.  A dummy variable was developed to capture recurrence within 6 months of the inclusion disposition for all cases in the 2010 calendar year.  Number of children who were included in the dataset who met the disposition criteria totaled to 1,058,176.  Of those, 94.7% (n=1,002,143) did not recur while 5.3 (56, 033) did within 6 months of disposition.

Variables of interest included service matching, Child factors (age, gender, prior victim), Caretaker factors (head of household, prior abuse perpetrator, age), neglect only (type of abuse) were examined for recurrence using bivariate and logistic regression statistical analyses. 

Results:  Bivariate analyses showed all child, caretaker, and neglect only factors to be statistically significantly related to recurrence at the p = 022 level or lower. 

 In the logistic regression analysis, hierarchical (blockwise entry) was utilized because of our particular interest in service matching and neglect as well as finding the most parsimonious model to predict recurrence.  The most parsimonious model included child’s age, perpetrator’s age, child as a prior victim, perpetrator as a prior perpetrator, and interactions between service matching and child’s age, perpetrator’s age, child as prior victim, and perpetrator as prior perpetrator also improved the model (X²=17.964, p = .000, Cox & Snell R Square = .024, Nagelkerke R Square = .067).    All variables showed statistically significant contributions to recurrence with the exception of service matching alone (p=.336) and the interaction term between perpetrator’s age and service matching (p=.128). 

Implications for practice:  Service matching as shown in the results contributes to a better understanding of recurrence.  However, the relationship is in a direction that wasn’t expected.  One of the reasons for this might be due to issues with information that is available on the dataset.  Further implications for practice will be discussed as they relate to our findings.