Evaluation of the Parent-Report Inventory of Callous-Unemotional Traits in a Sample of Children Recruited from Intimate Partner Violence Services: A Multidimensional Rasch Analysis

Background: Callous-unemotional (CU) traits in children refer to a general lack of remorse, shallow affect, callousness, insensitivity to others’ feelings, and/or hyper-regulated emotions. These traits are used as a clinical specifier for the course and treatment of conduct disorder (CD) in the DSM-V.  There is some evidence to suggest that exposure to adverse rearing environments, such as homes where intimate partner violence (IPV) occurs, increases children’s risk of developing CU traits. It is hypothesized that early abnormal affective experiences may interfere with the development of morality and appropriate emotional responses to distress cues. A variety of instruments have been used to measure CU traits in children; however, the Inventory of Callous-Unemotional Traits (ICU; Frick, 2004) is a measure that has garnered increased attention as a clinical screening tool during the past decade. The current study evaluated the utility of the ICU as a measure of CU traits among school-age children exposed to IPV.  Specifically, we tested multiple competing latent factor structures of the Spanish and English language versions of the parent-report ICU using multidimensional item response theory (MIRT).

Method: Maternal caregivers (N= 291) were recruited from community-based IPV services and reported on their children (77.9% ethnic minority; 47% female), who ranged in age from 7 to 12 years. MIRT was conducted using a multidimensional random coefficients multinomial logit model in ConQuest software. We compared nine models that were based on prior psychometric evaluations of the ICU. Dimensionality, item difficulty, response category use, and differential item functioning (DIF) were examined. Convergent validity of the best-fitting model was tested using Pearson correlations between the ICU scales and six Child Behavior Checklist (CBCL 6/18; Achenbach & Rescorla, 2001) symptom scores (e.g., conduct problems).

Results: An 18-item version with two subscales provided the best fit to the observed ICU data. Our findings are consistent with recent work reporting that 2-factor models, comprised of CU and Empathic/Prosocial (EP) scale scores, best fit parent-report ICU data. Our results support the convergent validity (r = .46 ~ .63) and incremental value of the CU (R²⁼ .14 to .37) and EP (R²⁼.13 to .21) scales. The scales showed no DIF for child age or gender. However, DIF was found for five of eight EP items when comparing Latina and non-Latina respondents, and four of eight EP items when comparing English and Spanish language versions. 

Conclusion: Our findings support recent assertions that EP and CU traits are distinct dimensions of behavior. Thus, the absence of EP behaviors, while associated with CU behaviors, should not be considered indicative of CU traits in youth. This is of particular importance given recent changes to the DSM’s diagnostic criteria for CD, which includes a limited prosocial emotions specifier for CD that treats high CU and low EP behaviors as conceptually equivalent. In addition, our results suggest that Latino cultural factors may impact the experience and/or expression of CU traits, and, relatedly, item endorsement on the ICU. Thus, researchers and clinicians should interpret the results of the ICU with caution when assessing Latino children.