Construct Validity and Measurement Invariance of the Children's Scale for Hostility and Aggression: Reactive/Proactive (C-SHARP)

Background:  Aggression is a serious concern for some children with intellectual and developmental disabilities (IDD), including those with autism spectrum disorder (ASD). Aggression confers greater risk of classroom removal, psychotropic treatment, is a leading cause of residential placement (Tsakanikos et al., 2007), and is also a significant source of parental and caregiver stress (Hodgets et al., 2013). For these reasons, the study of aggression is important to improve the lives of children with ASD/IDD and their families. A major impediment to this research is a dearth of detailed instruments; the most popular instruments measure general oppositionality (e.g., NCBRF) or devote only a short subscale to aggression (e.g., CBCL). The Children’s Scale for Hostility and Aggression: Reactive/Proactive (C-SHARP; Farmer & Aman, 2009) is a detailed measure of aggression. The original factor structure was derived in a heterogeneous sample of children receiving special education services (Farmer & Aman, 2009), but has since been used to compare aggression in children with and without ASD (Farmer & Aman, 2011; Farmer et al., 2014). 

Objectives:  The objective of this study was to assess the psychometric characteristics of the C-SHARP by replicating the factor structure, assessing factorial invariance, and exploring its convergent validity in a sample of developmental clinic-referred children with and without ASD.

Methods:  The C-SHARP is a parent-report instrument comprising 48 items on five subscales: Verbal Aggression, Bullying, Covert Aggression, Hostility, and Physical Aggression. Data were collected from six sites across the United States. The sample consisted of clinic-referred children who were diagnosed with ASD (n=414) and those who were not (n=243). We conducted confirmatory factor analysis on the mean structure of the C-SHARP across groups (ASD versus non-ASD), and then sequentially evaluated configural, strong, and strict factorial invariance using the lavaan package for R (Rosseel, 2012). Convergent/divergent validity was assessed by correlation of the C-SHARP raw subscale scores with selected CBCL Syndrome Scale T-scores. We hypothesized that the externalizing subscales, especially Aggression, would be more strongly correlated with the C-SHARP than the internalizing subscales.

Results:  All models [configural, weak, strong, strict 1 (equal residuals), and strict 2 (equal residuals and latent variance/covariance)] had a significant Satorra-Bentler Chi Square (p<.001). However, a combination of fit measures (AIC, BIC, RMSEA, and TLI) except SRMR favored the more restricted model, where the factor loadings, item intercepts, residuals, and latent variance/covariance were fixed to be equal but group means were freely estimated (see Table 1 for factor loadings, intercepts, and residuals). The C-SHARP and the CBCL were significantly (p<.01) correlated across most subscales. The strongest correlations were with CBCL Aggressive Behavior. Consistent with expectations, the Internalizing subscales were generally weakly correlated (r<.30) with the C-SHARP subscales, with the exception of CBCL Anxious/Depressed and C-SHARP Hostility (r=.42).

Conclusions:  The original five-factor structure of the C-SHARP was confirmed, and construct validity was supported. Aggression, as conceptualized by the C-SHARP, is invariant across the clinic-referred population, making the C-SHARP an ideal measure for children suspected of having an IDD.