Society for Social Work and Research 18th Annual Conference: Research for Social Change: Addressing Local and Global Challenges

January 15 - 19, 2014

Background and Purpose: In this symposium, we describe findings from three studies testing the feasibility of latent class analyses (LCA) in identifying specific subpopulations of individuals. LCA techniques model heterogeneity by using a latent variable (categorical or continuous) to represent a mixture of subpopulations differentiated by their particular patterns on multiple indicators (McCutcheon, 1987). Identifying specific subpopulations of individuals has clear relevance for child welfare policy and practice, which has often been criticized as taking a ‘one-size-fits-all approach’ (Perez, 2003). In spite of innovative approaches in child welfare to increase engagement of families (U.S. DHHS, 2005), coordinate services (Ryan, et al., 2006), and deliver evidence-informed interventions, these efforts have done little to ‘move the dial’ in terms of addressing some of the most vexing problems facing child welfare systems.  However, identification of subpopulations of individuals may help to reveal important heterogeneity that could inform our efforts to develop and target interventions. In spite of these benefits, the application of this technique has been relatively uncommon in child welfare research (Roesch et al., 2010). In this symposium, we take a closer look at these issues through the unique lens of three studies employing slightly different applications of LCA with child welfare populations.

Methods: We present results from three LCA studies, each of which applies LCA to an area in child welfare. Two of the papers address a fundamental question: Can distinct profiles of complex experiences involving foster children be identified? The proposed papers take two slightly different approaches to this question. One study applies LCA to the maltreatment histories of a sample of 801 foster youth who reached the age of majority in Illinois.  Another identifies patterns of placement for two groups of foster children: those with a diagnosis of autism spectrum disorder (ASD) and those without. The third paper presents two examples of child welfare research that make use of recently-developed extensions of traditional LCA -- repeated-measures LCA and multi-level LCA

Results: The results from the study employing LCA to identify naturally occurring subgroups of aging out foster youth based on their child welfare maltreatment records finds five distinct profiles showing moderate to high levels of maltreatment. In the second study, the authors identify distinct placement patterns between those with ASD and those without. The two studies described in the third paper demonstrate the utility of using LCA to specify the type of complex, multi-dimensional, multi-level phenomenon that often characterizes public child welfare operation and outcomes.

Conclusions and Implications:  The application of LCA to identify distinct subgroups of foster children holds promise for both identifying and addressing specific vulnerabilities in high risk groups. Different profiles of individuals served in child welfare may assist in guiding policy makers and administrators in strategically allocating resources. Research that identifies heterogeneity in populations may also assist in facilitating the translation of intervention research to real world practice.