Latent Class Analysis of Infants Reported for Maltreatment

Purpose: Infants are at increased risk of being reported for maltreatment relative to older children. This analysis examined a population of children with an initial report to Child Protective Services (CPS) during infancy. The objective was to: (1) analyze family and child characteristics associated with risk of re-report; (2) enhance understanding of relationships among risk factors associated with ongoing CPS involvement, and (3) identify subgroups of infants who are likely to experience multiple reports to CPS.

Methods: The birth records of infants born in California in 2006 were linked to CPS records. A total of 29,135 children (5.2%) were reported during infancy, 81.2% of whom remained in the home. These infants were followed for five years to determine if another CPS report occurred. Re-reporting patterns were analyzed using Latent Class Analysis (LCA), a methodology that models concepts that are empirical and not otherwise observable within a heterogeneous population.  The LCA procedure provided the item-response probabilities based on the individuals’ class membership. Each model was run with 1,000 seeds and 50 starts. Fit was examined using the log likelihood statistic (G²), the Akaike Information Criterion (AIC), the Bayesian Information Criterion (BIC), and entropy. The G², AIC, and BIC continued to drop across all models, indicating the fit of the model continued to improve as more classes were added. Indicators suggested the model with 5 classes was the best fit for the data.

Results: Overall, 60.7% of children reported during the first year of life were re-reported by age 5 and five distinct risk profiles were identified.  Classes were organized based on a gradation of re-report risk: Class 1, had a 35.5% probability of a re-report whereas the probability of a re-report in Class 5 was 77.2%. Families who had multiple risk factors as indicated by birth covariates were more likely to experience a second CPS report. When an infant report was made to CPS, the presence of a sibling CPS report, later prenatal care, having no established paternity, and being low income indicated that there was a 60-80% chance that the child would experience further CPS contact by age 5. Class 2 showed that children with Hispanic, foreign born mothers had a relatively low probability of re-reports (39.1%), although the probability of other risk factors being present was high.  Infants who had teen mothers (Class 3) were at increased risk for re-report (58.6%), regardless of other risk factors.

Conclusions and Implications: The current analysis is the first to use LCA to examine re-reporting patterns, generating insights into characteristics that affect the probability of future CPS involvement. The combination of four specific risk factors appears to signal families at highest risk for future CPS involvement: lack of paternity, later prenatal care, having a CPS family history, and being low income. This analysis demonstrates the power of using LCAs as a tool to examine the relationships between exposures and risks so that vulnerable subgroups can be identified and programs and policies can be developed to target individuals who will benefit most.