Bridging Disciplinary Boundaries (January 11 - 14, 2007)

PURPOSE: (1) establish a system for classifying foster care cases and (2) identify associations between case types and outcomes. METHOD: (1) an extensive literature review, (2) a survey of state AFCARS administrators from 12 purposively selected states (for example, New York, California, Illinois, Pennsylvania, Texas, Michigan), (3) focus groups of workers and supervisors in four counties from one state, and (4) a cluster analysis of AFCARS case data in each of 10 states. The cluster analysis used case socio-demographics, reasons for entry and other case descriptive information. FINDINGS: The survey found great variability among states with respect to the types of cases included in the AFCARS data system. For example, some states did not include children with emotional disabilities, others did not include voluntary placements. In focus groups, workers and supervisors identified types of cases seen in child welfare and likely associations between types and case outcomes. The qualitative analysis, while not broadly representative, yielded rich data on case types identified in all or most all focus groups: parents' inability to care for children due to parental substance abuse, parental mental health problems, child's mental health problems, child's medical reasons, child's behavior, physical abuse, sexual abuse, different types of neglect, abandonment, and prior unsuccessful child welfare intervention. The AFCARS cluster analysis resulted in 10 major groupings of variables that were organized around child's race, occurrence of neglect, child's medical condition, child's disability and only one removal from the home. Associated variables important in forming the clusters were age, source of placement funding, child identified as emotionally disturbed, behavior problems, voluntary removal, type of placement and single female parents. Nineteen hierarchical case types were established based on these clusters and knowledge from the literature, administrator survey and focus groups. These case types were combined with other variables in logistic regressions for each of 10 states to predict two federal outcome measures: placement stability and re-entry into foster care. The case types were significant in predicting both outcome variables in most states. In addition, certain case types were systematically associated with higher odds of an undesirable outcome. Compared to “children under one year of age from all race/ethnic groups,” children with behavior problems ages 11 and older from all race groups had the greatest likelihood of re-entry for most states. Other case types that had high re-entry rates were those who were emotionally disturbed in all ages, and children ages 11 and older for all race groups. Tables and graphics will be used to depict statistical and analytic findings. RECOMMENDATIONS: The literature is replete with predictive models identifying various case characteristics combined in a single model to predict case outcomes. There is very little literature on the actual case types that are represented by these statistical models. This study takes the analysis to the next step, clearly identifying these case types using multiple sources of information and suggesting further models for intervention and outcome measurement be based on the case type and case mix in each state or locality reviewed.