Research That Matters (January 17 - 20, 2008)

Purpose: Children who are maltreated have been shown to experience adverse emotional, social, and behavioral consequences. Evidence suggests a mediating role for maladaptive emotions on the link between child maltreatment and maladaptive behavior (Agnew, 2001). However, gender may condition this link in that boys' may externalize while girls may internalize their responses to maltreatment (Broidy & Agnew, 1997; Mazerrolle, 1998). The present study explores the mediating influence of internalizing symptoms on the relationship between child maltreatment and externalizing behavior among girls and boys. As hypothesized, internalizing symptoms exerted a mediating influence that was conditioned by gender. Only girls' internalizing symptoms mediated the link between child maltreatment and externalizing behavior.

Methods. A secondary data analysis (Waves 1-4) of the “Longitudinal Pathways to Resilience Among Maltreated Youth” was conducted. Three hundred youth aged 7-11 (n=168,maltreated youth) attended a yearly one week summer camp. Internalizing and externalizing behavior was measured using the Child Behavior Checklist (Achenbach, 1991). Information on child maltreatment was based on child protective services cases. Data was analyzed using structural equation modeling to test for longitudinal mediation and moderation effects. Findings. Sobel analysis found a significant (p<.05) mediation model where abuse history directly effects Internalizing Symptoms Time 1 (TSI-T1) and Externalizing Symptoms-Time 1 (TSE-T1) while also having an indirect effect on TSE-T1 through TSI-T1. We hypothesized that the mediation model would also have residual effects on the latter measurements of TSE-T2, TSE-T3, and TSE-T4. Therefore, we integrated the mediation model with an autoregressive model of TSE1-4. The model produced excellent fit (chi-square=5.14, p=.16, CFI=.99, RMSEA=.06) with all regressions (beta weights) positive and highly significant (p<.001). In the end, the model accounted for between 23.5% and 37.1% of the variance for measures of TSE at time points 1-4. Of note is that the model only account for 3.8% of the variance for TSI1.

We next conducted a moderation model to see if gender moderates the influence of the above mentioned model. We ran the same model separately for males and females. Although both the male (chi-square=3.68, p=.30, CFI=.99, RMSEA=.04) and female models (chi-square=1.41, p=.70, CFI=.99, RMSEA=.00) provided excellent model fit, the female model appeared a better overall fit for the following two reasons. First, all the regression (beta weights) in the female model were significant (p<.01). This is dissimilar to the male model where the beta weight between abused status and TSI1 was not significant (p=.14), indicating that abuse status did not significantly influence TSI1 score. As this link represents a key piece of the original model this gender difference appears quite significant. Second, the female model stands out as the better model when comparing the AIC scores. The female AIC(49.41) was lower than the male AIC(51.68). Consequently, the original model appears only relevant to the female data.

Implications: These findings suggest the development or improvement of gender and age specific intervention strategies, especially among young children. Expanding the use of nonverbal creative arts interventions that address emotional, psychological, and behavioral responses to trauma are presented.