Meta-analysis and reliability generalization of the Psychopathy Checklist applied to youth

This study applied meta-analysis and reliability generalization procedures to examine the reliability and validity of the Psychopathy Checklist (PCL and PCL-R) and its derivatives (PCL-SV and PCL-YV) within youthful samples (average age range: 13--18 years, M = 16 years). Data were extracted from studies published in academic journals, dissertations/theses, and book chapters published and circulating prior to May 2002. Reported in 31 sources (journal articles, dissertations/theses, and book chapters) were effects from 35 independent youthful samples analyzed to describe (a) the internal consistency and interrater reliability of the PCL total score and (b) PCL total score correlations with recidivism, undesirable treatment outcomes, and violent behavior. These reliability estimates and outcomes were analyzed using the conditionally random-effects meta-analysis approach described in Hedges and Vevea (1998). The results of the reliability generalization study indicated good internal consistency and interrater reliability of the Psychopathy Checklist in youthful samples. The results of the meta-analysis of recidivism effects produced a statistically significant, yet surprisingly small magnitude mean-effect correlation between the Psychopathy Checklist and two types of juvenile recidivism (violent and generally-defined/non-violent). Only two studies reported correlations between the Psychopathy Checklist and juvenile sexual recidivism the resulting mean effect size was non-significant and small.The meta-analyses of PCL correlations with all types of youth treatment outcomes (both violent and non-violent undesirable behaviors and outcomes) produced statistically significant, but surprisingly small magnitude mean effect correlations. The meta-analysis of PCL correlations with juvenile violent criminal histories and aggressive behavior produced both significant and medium sized mean effect correlations. Moderator analyses were conducted to examine the variability of the individual effects. Supplementary analyses included funnel-plot analysis to check for publication bias and fail-safe N analyses to examine the potential impact of the file-drawer effect on both the significance and magnitude of mean effect sizes. The results are discussed in regard to the use of the PCL to address school safety assessment needs as well as the use of the PCL as an assessment for the justice system that identifies, classifies, and predicts youth at risk for serious and violent youth crime. Two main conclusions can be drawn from the results of this study: (a) PCL total scores were reliable in the observed samples but this conclusion was reached on the basis of PCL data obtained from almost exclusively male, mostly Caucasian, offender samples from North America, and consequently this conclusion should not be generalized to unobserved populations at this time, and (b) the PCL total score was not a valid predictor of recidivism and undesirable treatment outcomes in the youthful samples observed in this study. This conclusion is based on the small magnitude of the average correlation observed between the PCL total score and all recidivism and undesirable treatment outcomes.Moreover, the paucity of research investigating the relationship between the PCL and recidivism in youth suggests the instrument has not gained general acceptance as a method of predicting recidivism in youth. The implication of these findings for the application of the instrument in applied settings is that the PCL in youth should be restricted solely to research applications that have no bearing on life, liberty, or public safety.