In the community sample, our findings suggest that the proposed CU specifier would designate between 10% and 32% of children with CD. Further, the base rates of the CU specifier and the individual CU symptoms were fairly low in children without CD. In the clinic-referred sample, between 21% to 50% of children and adolescents with CD showed the CU specifier, depending on the informant. However, unlike in the community sample, a significant number of clinic-referred children without CD showed high rates of CU traits and this was especially true for children with an Oppositional Defiant Disorder diagnosis. Youth with CD and the CU specifier, as compared to CD youth without the specifier, exhibited significantly higher rates of aggressive behavior in both samples, as well as higher cruelty ratings within the clinic sample. These findings were present across different informants for assessing the specifier. In addition, youth with CU traits also demonstrated higher rates of comorbidity in the community sample. Within the clinic sample, there was no difference in the level of impairment between CD youth with and without CU traits. Thus, in clinic referred youth, children with CD appeared to be significantly impaired, irrespective of their level of CU traits.
The individual items that comprise the CU specifier showed very similar prevalence rates and predictive power. The one exception is the low prevalence rate of the item ‘does not show feelings or emotions’, which also showed the lowest positive predictive power in the community and clinic-referred sample. Although the performance of this item was not low enough to recommend against including it in the specifier at this time, future research should continue to evaluate this item and whether its wording could be enhanced to increase its predictive power.
One important issue in interpreting these comparisons between the community and clinic-referred samples is that the two samples not only differed on referral status but they also differed on their ethnic composition, with the community sample being predominantly Caucasian (77%) and the clinic referred sample being predominantly African-American (88%). Past research has suggested that, while the association between CU traits and severity of antisocial behavior is found in both Caucasian and ethnically diverse samples, the strength of the association is often weaker in samples with high rates of ethnic minority youths (Edens, Campbell & Weir, 2007
). Thus, although it is important that we did find some utility of the new specifier in a predominantly African-American sample, we are unable to determine whether the differences between samples were due to referral status or the ethnic composition of the sample. Interestingly, one potential explanation that has been given for the weaker predictive utility of CU traits in ethnically diverse samples is that minority individuals may be rated as higher on these traits by non-minority raters. This is not likely an explanation for findings in our clinic-referred sample, since the ratings of CU traits were from parents and from self-reports. Also, although we did not obtain the ethnicity of the teachers who rated students in the community sample, most teachers in the participating school systems were Caucasian. Further, in this community sample, both parents and teachers were more likely to rate African-American students as being more likely to meet the CU specifier and, more importantly, the association between race and meeting the specifier was somewhat lower (phi-coefficient=.16, p
< .001) for teacher ratings than for parent ratings (phi-coefficient=.19, p
<.001). Thus, there was no evidence of an ethnic bias for teacher ratings relative to parent ratings. However, this is a critically important area for future research, which should continue to evaluate whether the CU specifier is valid across ethnic groups, as well as examine the most appropriate assessment methods to use across ethnically diverse individuals.
There were other limitations in the study that need to be considered when interpreting these results. Methodological differences between the studies make it challenging to isolate potential sources of any differences in the pattern of findings. For instance, the community sample used teacher and parent ratings to assess CD, while the clinic sample relied on semi-structured interviews of the youth and caregiver. By using multiple informants, it allowed us to study the effects of different informants on the prevalence and validity of the CU specifier. For example, in the community sample teachers rated a higher percentage of CD youth with CU traits (24%) than parents (10%). In the clinic sample, parents rated a higher percentage of CD youth with CU traits (31%) than youth themselves (21%). Despite somewhat different levels of endorsement, the validity across informants was quite similar. Thus, these findings support the utility of gaining information from multiple informants in the assessment of CU traits. Unfortunately, it was beyond the scope of this paper to compare the different ways of combining informants but this is another important direction for future research, given that different combinatorial strategies may yield varying results (for example see Youngstrom, Findling, & Calabrese, 2003
Another limitation within the clinic-referred sample was that the measures of impairment were general indices related to health and daily life activities. Measures of impairment more specifically related to CD, such as police arrests, school suspensions, or substance use may have showed clearer differences between CD groups with and without the CU specifier. On the other hand, inclusion of different methods and sampling strategies enhances confidence in the generalizability of findings that were consistent across both samples. Further, the measurement of the proposed specifier in both samples was based on standardized rating scales. It is unclear how likely clinicians are to use such assessment methods for making diagnostic decisions in typical practice. Also, the study was correlational and, as a result, the association between CU traits and higher rates of aggression and cruelty should not be considered causal. It is possible that persons with CU traits lead someone to act more aggressively and cruel, but it is also possible that a history of aggression and cruelty could make someone callous to the pain and suffering of others. Finally, data were not available from each informant in all cases. Thus, the single-informant groups and combined-informant groups were based on somewhat different samples. We used this method because it utilized the most available data and approximated typical clinical practice, addressing concerns about clinical generalizability that are central to the DSM-5 revision process. To determine how this methodology may have influenced our results, we repeated all analyses in both samples for only those with both informants; these analyses led to very similar results.
In addition to considering the validity of the proposed specifier, an additional consideration is the potential harmful consequences of using a pejorative label like “callous-unemotional” as part of a diagnosis. Several issues are important in considering this concern. First, although there is no research directly testing the effects of the label “callous-unemotional”, there is an empirical literature studying the negative effects of the use of the term “psychopathic traits” when applied to children and adolescents. A review of these studies suggests that the term ‘psychopathic traits’ does affect ratings of treatability by clinicians, as well as legal decisions made by judges and juries; however, it does not have more negative effects than the term “conduct disorder” (Murrie, Boccaccini, McCoy & Cornell, 2007
). Thus, these results suggest that any diagnosis related to antisocial and aggressive behavior could have pejorative effects. Results also highlight one potential problem in not recognizing that only a small percentage of antisocial youths or children and adolescents with CD show characteristics associated with psychopathy; specifically, these terms can become viewed as being interchangeable. Second, previous attempts to capture CU traits in the DSM used names designed to reduce the potential stigmatizing effect of the label (e.g., “undersocialized”) (American Psychiatric Association, 2000
), but the lack of clarity in this term led to great variability in how the construct was conceptualized and assessed by researchers and clinicians. Third, there is a danger that if terms that seem to connote a less severe disturbance (e.g., “uncaring”) are used for the specifier, this could actually be more harmful by resulting in many children and adolescents with less severe disturbances being diagnosed by clinicians.
Further, it is important to note that recent research examining the predictive utility of CU traits has shown these traits alone are associated with overall conduct and emotional problems, as well as general psychiatric difficulties in a large sample of children and adolescents followed over three years (Moran et al. 2009
). These findings suggest the presence of CU traits in youth, without conduct problems, could also be a clinically relevant group and future research should continue to examine this unique subgroup. This is particularly relevant for the clinic-referred children, given that a significant number of youth without CD met criteria for the CU specifier in our clinic sample. Also, there needs to be more research testing ways to improve the measurement of these traits for both clinical diagnoses and research (Johnstone & Cooke, 2004
; Poythress et al., 2006
). Finally, the current study examined the prevalence rate of the newly proposed DSM-5 specifier in an outpatient clinic-referred and community sample. Thus, the impact of this specifier among detained youth, inpatient youth, and other samples which may include a large number of children with CD is warranted.