Online Program

Using Latent Class Analysis to Explore Social Behaviors Among Children with Developmental Disabilities

*Benjamin Zablotsky, CDC, National Center for Health Statistics 

Keywords: latent class analysis, survey data, mental health

Background: Latent class analysis (LCA) identifies latent subgroups of individuals within a population through the examination of an individual’s response pattern to a set of indicators (McCutcheon, 1987). As individuals with similar response patterns are grouped together in the same class, an LCA can be viewed as a categorical analog to factor analysis, with class membership replacing factor scores at the individual level. An LCA may help in identifying qualitative differences among individuals with similar phenotypic presentations (Yang, 2006).

Objectives: 1) To evaluate the effectiveness of LCA in grouping individuals with various developmental disabilities on a scale designed to measure a child’s social functioning. 2) To determine which covariates, if any, are associated with membership within a subgroup.

Methods: Data are drawn from the Survey of Pathways to Diagnosis and Services which was funded by NIH’s National Institute of Mental Health and conducted as a module of the State and Local Area Integrated Telephone Survey by the Centers for Disease Control and Prevention’s National Center for Health Statistics. Data from approximately 3,000 parents of children with autism spectrum disorders (ASDs), intellectual disabilities (IDs) or developmental disabilities (DDs) were analyzed in the current study. Parents included in the analysis had completed the Children’s Social Behavior Questionnaire (CSBQ) (Hartman et al., 2006). The CSBQ is used to measure a child’s social and behavioral functioning and is divided into six subscales, including the presence of stereotyped behaviors, reduced contact and social interest, difficulty understanding social information, social orientation problems, behavioral problems and fear of and resistance to change. Each of the six subscales were dichotomized at the 75th percentile and subsequently included as indicators in the LCA.

Results: A four-class solution, with subclasses of severe social and behavioral impairment, moderate behavioral impairment, moderate social impairments and mild social and behavioral impairments, was identified. Latent class regressions revealed children diagnosed with ASD to be significantly more likely to belong to the severe social and behavioral impairment class than children with either DD or ID. Children in families with higher education attainment and higher income levels were the least likely to belong to the severe social and behavioral impairment class.

Conclusions: Latent class analysis can be a helpful technique in distinguishing symptom profiles among a heterogeneous population, and may provide insight into why individuals have differing diagnoses.