Identification and Prediction of Child Behavior Trajectories Among Children Who Have Experienced Maltreatment

Background/ Purpose: Child maltreatment affects almost 700,000 children annually. The consequences of child maltreatment range from physical and mental health issues, at the micro-level, to increased child welfare worker caseloads and overcrowding residential facilities at the mezzo-level, to increased costs and policy implications at the macro-level. These children involved with child welfare services are also at high risk for behavior problems. This study aims to identify internalizing and externalizing behavior paths that this high-risk population follows over a 6-year period. This current study will also identify the predictors of membership in normative and problematic pathways.

Methods:Using data from the National Study of Child and Adolescent Well-being (NSCAW). The sample included 4,997 children and youth who remained home after the investigation. Latent class growth analysis (LCGA) was used to estimate the number, size, and shape of subgroups of children and youth following distinct behavioral pathways.  Raw scores from the CBCL internalizing and externalizing behaviors subscales measured at the 4-time points were entered into a series of unconditional LCGA model using Mplus Version 7.4 with the mixture add-on. Bayesian Information Criterion (BIC) and bootstrap likelihood ratio test (BLRT) were given most credence when selecting the best model fit. Once the best-fitting model was identified, bivariate analyses were conducted to examine the characteristics of each latent trajectory class and to identify correlations between class memberships and theoretically relevant covariates. Multinomial logistic regression was conducted to identify characteristics predicting membership in the behavior trajectory groups.

Results: In estimating mixed models of internalizing and externalizing behavior patterns, both showed a 4-class LCGA model fit the data best as indicated by both the BIC and BLRT tests (BIC_int=5968.718, BLRT_int=22.210, p<.0005; BIC_ext=5965.798, BLRT_ext=57.325, p<.005). The 4-class linear LCGA model includes the following subgroups: (1) persistently high problematic behavior; (2) improving group; (3) worsening group; and (4) low problem group. Some child level and caregiver level variables predicted membership in the trajectory groups. Child level: gender (b=.23, t=-1.54, p<.001); social skills (b=.01, t=-.22, p<.001); exposure to violence (b=.02, t=.14, p<.001); and physical health (b=.09, t=-1.49, p<.001). Caregiver level: age (b=.10, t=-.38, p<.001); number of children in the home (b=.07, t=-0.20, p<.01); physical health (b=.08, t=-.77, p<.001); domestic violence (b=.01, t=.04, p<.001); social support (b=.15, t=-.92, p<.001); and perception of neighborhood (b=.02, t=.17, p<.001). None of the environmental level variables predicted membership in any of the trajectory groups.

Conclusions and Implications: Children and youth differ regarding how they respond to maltreatment and other life events or situations. It is imperative that interventions be individualized to target specific issues and reduce specific behavior problems. Results indicated that improving child social skills and increasing caregiver social support may be key in reducing child behavior problems. Furthermore, identifying early indicators of internalizing and externalizing behavior problems and addressing them with evidence-based interventions to reduce negative behaviors may avert long-term negative outcomes. Both practice and policy implications are discussed as well as recommendations for future research.