Abstract
Research suggests that irritability and defiance are distinct dimensions of youth oppositionality that are differentially associated with internalizing and conduct problems, respectively. Because much of this evidence has emerged with limited psychometric evaluation, we conducted the first multi-informant examination of selected Child Behavior Checklist (CBCL) and Youth Self Report (YSR) items for measuring irritability and defiance in a large clinical sample. Clinically referred youths (N = 1,030; ages 6–15; 43% female, 42% ethnic minority) were assessed prior to treatment using multi-informant rating scales and diagnostic interviews. Analyses examined factor structure, invariance, internal consistency, multi-informant patterns, and convergent, discriminant, and criterion validity with internalizing and externalizing problems/disorders. A correlated 2-factor model of irritability (stubborn/sullen/irritable, mood, temper) and defiance (argues, disobeys-home, disobeys-school) fit well for both informants. Adequate measurement invariance and scale consistency was consistently found for parent-report but not youth-report. With both informants, all hypothesized convergent and discriminant validity associations were supported: irritability and defiance with internalizing and conduct scales, respectively. However, hypothesized criterion validity associations were largely found only by parent-report: irritability with anxiety and depressive disorders, defiance with conduct disorder, and both with oppositional defiant disorder. Results consistently supported the reliability and validity of the CBCL irritability and defiance scales, with somewhat less consistent support for the YSR scales. Thus, CBCL items may provide psychometrically sound assessment of irritability and defiance, whereas further research is needed to advance youth-report and multi-informant strategies. Results also provide further support for a two subdimension model of oppositional defiant disorder symptoms that includes irritability and defiance.
Acknowledgments
We thank the youths, caregivers, staff, and administrators who participated in this research.
Disclosure statement
The authors report that they have no potential conflicts of interest.
Supplemental Material
Supplemental data for this article can be accessed on the publisher’s website.
Correction Statement
Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/hcap.
This article has been republished with minor changes. These changes do not impact the academic content of the article.
Notes
1 With minor variations in content and terminology, nearly all studies identify the dimensions of irritability and defiance; some identify a third area related to spiteful/vindictive behavior (Evans et al., Citation2017). However, this third factor is typically only identified using alternative measures and item sets (e.g., DAWBA). ODD criteria include only one indicator of spiteful/vindictive behavior, and one indicator is not sufficient for identifying a latent construct. The bulk of the evidence supports two ODD dimensions, irritability and defiance. For example, Burke et al.’s (Citation2014) large multi-sample factor analysis found evidence for a two-factor structure, which was recently replicated and validated by Waldman et al. (Citation2018). Regarding the CBCL/YSR, there are zero items in the oppositional/aggressive item set that correspond to a spiteful/vindictive area. We found only one study of CBCL/YSR oppositionality (Aebi et al., Citation2013) that included such a third factor; this was done by pulling an item from the DSM-oriented scale for conduct problems, widely regarded as distinct from oppositionality (e.g., Lahey et al., Citation2008). For all of these reasons, we focus on the two-factor model best supported by the available evidence; however, we also acknowledge that the number and nature of these dimensions remains an important question for further investigation.
2 WLSMV is appropriate for ordered-categorical data such as the CBCL/YSR three-point response scale. Because CBCL/YSR data are usually scored and interpreted as continuous measures, we also estimated all models using robust maximum likelihood estimator (MLR) appropriate for interval data. Results from MLR and WLSMV models were quite similar in terms of overall model structure, fit, and invariance.