Predicting Recidivism in Adolescents With Behavior Problems Using PCL-SV

Abstract

Studies show that identifying persistent delinquents on the basis of early antisocial conduct yields a significant error rate. However, evaluating childhood or adolescent psychopathic traits is likely to improve matters in this regard. This study seeks to verify the contribution of psychopathic traits in adolescence to antisocial conduct prediction in early adulthood. To this end, a French version of the Psychopathy Checklist -Screening Version (PCL-SV) adapted to adolescents is used to evaluate psychopathic traits in 27 youths aged 15 to 19 years recruited in youth centres and presenting behavioral problems reaching a clinical threshold. The PCL-SV scores contribute significantly above and beyond indices of delinquent behavior to predict self-reported antisocial conduct 2 years later and, specifically, to predict criminal versatility and violent recidivism

Keywords

delinquency prediction psychopathy adolescent evaluation

Research on adults has allowed documenting a robust link between psychopathy and persistent offending and violent behaviors (Hare, 1998, 2003; Hare, Clark, Grann, & Thornton, 2000), thereby fostering a growing interest in applying the measure of psychopathy to children and adolescents (Forth & Burke, 1998; Forth & Mailloux, 2000; Frick, O’Brien, Wootton, & McBurnett, 1994; Lynam, 1996; Seagrave & Grisso, 2002). It appears that the measure of psychopathy in adolescence, like that in adulthood, is associated with different indices of antisocial behavior: presence of violent offences (Forth, Hart, & Hare, 1990; Kruh, Frick, & Clements, 2005), diversity of offences (Lynam, 1997), aggressiveness, and delinquency (Toupin, Mercier, Déry, Côté, & Hodgins, 1996). These links have been observed in adolescent samples in a variety of settings: legal/judicial (Brandt, Kennedy, Patrick, & Curtin, 1997; Campbell, Porter, & Santor, 2004; Corrado, Vincent, Hart, & Cohen, 2004; Forth et al., 1990; Kosson, Cyterski, Steuerwald, Neumann, & Walker, 2002; Kruh et al., 2005; Murrie, Cornell, Kaplan, McConville, & Levy-Elkon, 2004; Salekin, Neumann, Leistico, DiCicco, & Duros, 2004; Spain, Douglas, Poythress, & Epstein, 2004), psychiatric (Langstrom & Grann, 2002; Stafford & Cornell, 2003), clinical (Toupin et al., 1996), and civilian (Lynam, 1997). The meta-analysis conducted by Leistico, Salekin, DeCoster, and Rogers (2008) indicates that psychopathy explains recidivism equally well in samples of adolescents and adults, with a moderate effect (d = .55). In short, results support the relevance of evaluating psychopathy in adolescence to better identify youth at risk to become enmeshed in severe criminal trajectories (Caspi, 2000; Edens, Campbell, & Weir, 2007; Hemphill, Hare, & Wong, 1998; Salekin, Rogers, & Sewell, 1996).

Despite the boom in research in the field of adolescent psychopathy, only about a dozen prospective longitudinal studies have ever been conducted in this regard (Brandt et al., 1997; Catchpole & Gretton, 2003; Corrado et al., 2004; Forth et al., 1990; Frick, Stickle, Dandreaux, Farrell, & Kimonis, 2005; Gretton, Hare, & Catchpole, 2004; Hicks, Rogers, & Cashel, 2000; Langstrom & Grann, 2002; Loney, Taylor, Butler, & Iacono, 2007; Lynam, Miller, Vachon, Loeber, & Stouthamer-Loeber, 2009; Ridenour, Marchant, & Dean, 2001; Salekin, 2008; Toupin et al., 1996; Vincent, Vitacco, Grisso, & Corrado, 2003; Vincent, Odgers, McCormick, & Corrado, 2008). What’s more, there are a number of major methodological differences across these. For example, Gretton et al. (2004) followed up on young offenders over a very long period of time (10 years); Loney et al. (2007) investigated a civilian sample using self-report to evaluate psychopathy; Lynam et al. (2009) and Frick et al. (2005) evaluated psychopathy through informant report (parents and teachers reports), whereas Salekin (2008) focused on a sample of boys and girls, as did Frick et al. (2005). As a result, only nine studies share analogous methodological characteristics: sample of at-risk male adolescents, follow-up over short to medium term (6-27 months), and use of professionals to evaluate psychopathy. For the most part, they also include the use of control variables (demographic data, indices related to criminal history) and examine official criminal recidivism (general, violent, and nonviolent), except for Toupin et al. (1996), who used self-reported delinquency. On the whole, they support the use of psychopathy scores in adolescence to predict antisocial conducts.

Toupin et al. (1996) observed a significant correlation between an adapted version of the Hare Psychopathy Checklist–Revised (PCL-R) to adolescence and self-reported delinquency (r = .42, p < .05) and aggressive behaviors (r = .30, p < .05) a year later in conduct-disordered adolescents receiving services from youth treatment centres. In a study carried out on adolescents incarcerated in a maximum-security establishment, Forth et al. (1990) obtained a correlation of .26 (p < .05) between PCL score (adapted to adolescence) and number of official violent offences at 27-month follow-up. These relationships, albeit modest, illustrate the instrument’s capacity to reveal differences in terms of violent behaviors among youths whose criminal activity is relatively homogeneous. Results obtained by Vincent et al. (2008) are consistent; boys in youth custody who scored one standard deviation (SD) above the mean on the Hare Psychopathy Checklist–Youth Version (PCL-YV) were 3.5 times more likely to have received a conviction for violent offence. However, these studies do not allow establishing whether psychopathic traits are associated with antisocial conduct above and beyond earlier delinquent conduct.

A pressing issue that needs to be verified is the utility of the PCL in predicting criminal behavior when historical or contextual data are also considered. In the study by Ridenour et al. (2001), the score on the PCL-R adapted to adolescents contributed above and beyond criminal history (number of earlier charges) and disruptive disorder symptoms (conduct disorder, oppositional-defiant disorder) to predict official criminal recidivism in 80 youths with adaptation problems. The duration of the follow-up was relatively short, however (1 year), and the independent contribution of the PCL factors was not verified. Evaluating psychopathy in 130 adolescents incarcerated in a maximum-security establishment, Brandt et al. (1997) managed to establish that the PCL-R made a significant contribution above and beyond criminal history (age at first offence, severity of offences, number of earlier incarcerations) and demographic data (educational level, psychoactive substance abuse, age at evaluation) to predict official general criminal recidivism 2 years later (R² = .40, ΔR² = .08, p < .05). The few studies investigating the differential contribution of PCL factors to predict criminal recidivism report that Factor 1 (interpersonal/affective) contributes more to the prediction of violent recidivism (Brandt et al., 1997; Gretton et al., 2004). Factor 2 (social deviance) is mainly associated to general and nonviolent recidivism (Corrado et al., 2004). It should be noted, however, that these studies evaluated psychopathy on the sole basis of institutional records (Brandt et al., 1997; Catchpole & Gretton, 2003; Corrado et al., 2004; Gretton et al., 2004; Langstrom & Grann, 2002). These files do not allow to fully capture the personality traits, which are less documented than behaviors (Côté & Hodgins, 2000; Hare, 2003). As a result, on the grounds of the records alone, Factor 1’s contribution to predict general and nonviolent offending could have been underestimated.

It is also important to emphasize that most of the studies reviewed have focused on the number of offences perpetrated and that those which examined offence versatility only distinguished violent, nonviolent, and, in rare cases, sexual behaviors. Yet, this versatility has been demonstrated to be characteristic of the subgroup of chronic and persistent delinquents, which is why Hare (1991, 2003) and later Forth, Kosson, and Hare (2003) introduced a specific item in this regard in the evaluation of psychopathy. For this reason, it appears relevant to turn our attention not only to predict the number of offences but also offence versatility.

Finally, it needs to be mentioned that, with the exception of Frick et al. (2005) and Toupin et al. (1996), the studies under review evaluated delinquency on the lone basis of official data. Risk assessment studies (Monahan et al., 2001) have demonstrated the reliability of self-report measures; these have been found to yield a better picture of delinquent behaviors than official statistics.

Some authors have also evaluated the PCL’s performance in predicting criminal recidivism using receiver operator characteristics (ROC) analysis. It uses the area under the curve (AUC) as the measure of a test’s global performance from .5 (random guess) to 1 (perfect). This analysis is particularly interesting in that it is not affected by low prevalence rates (Mossman, 1994). In the studies involving adolescents placed under a legal measure, predictions based on the PCL-YV total score range in terms of AUC from .65 to .73 (p < .05) for official violent recidivism (Catchpole & Gretton, 2003; Corrado et al., 2004; Langstrom & Grann, 2002). Where the prediction of nonviolent recidivism is concerned, only one prospective longitudinal study evaluated the PCL’s performance with the ROC curve (Corrado et al., 2004), obtaining a slightly weaker result (AUC = .63, p < .05). However, it is important to mention that these studies used the PCL without controlling previous criminal history.

In light of the above, there was reason to pursue study of the prediction of criminal behavior in adulthood through the evaluation of psychopathy in adolescence and to take into account the previous methodological limitations. Using the Psychopathy Checklist–Screening Version (PCL-SV) with adolescents instead of the PCL-R affords certain practical advantages. For one, because of the relatively low prevalence of psychopathy among adolescents, the method is faster to administer and, therefore, less costly. Moreover, given the frequent absence of detailed information regarding criminal history in juvenile and research records, the PCL-SV is more appropriate than the PCL-R, as it can be completed even if this information is not available (Hart, Cox, & Hare, 1995). Age at first delinquent behavior and number of previous delinquent behaviors had to be controlled variables, in keeping with the methodological strategy generally observed in the studies reviewed. Prediction had to focus on self-reported delinquent recidivism, including frequency of violent and nonviolent recidivism, as well as versatility of recidivism. A particularly interesting contribution could be made by introducing the global prediction index drawn from logistic regression models in a ROC analysis, which is something that none of the studies reviewed did. This procedure offers the advantage of verifying the discriminating power of various indices considered simultaneously, in this case, PCL-SV scores and behavioral indices.

Current Study

The purpose of our study was to establish the contribution of psychopathic traits above and beyond that of behavioral indices in predicting antisocial conduct in early adulthood. More specifically, we sought to achieve the following objectives:

Describe self-reported delinquency in a sample of adolescents followed up into early adulthood;

Correlate data at Time 1 (behavioral indices: number of delinquent behaviors, age at first delinquent behavior; PCL-SV scores: total, Factor 1, Factor 2) and delinquent behaviors 2 years later (violent and nonviolent recidivism, and versatility of delinquent behavior);

Compare the correlation coefficients according to the method proposed by Meng, Rosenthal, and Rubin (1992);

Determine the contribution of PCL-SV scores (total score, Factor 1, Factor 2) above and beyond that of behavioral indices to predict self-reported delinquency at Time 2 (as defined above) using linear regressions; and

Through ROC analyses, evaluate the performance of the models derived from logistic regression (including behavioral indices and PCL-SV scores) in predicting self-reported delinquency at Time 2.

Method

Participants

The participants constituted a subgroup from a broader study (Pauzé, Toupin, Déry, & Mercier, 2000) that aimed at describing adolescents in the care of Quebec youth centres: public institutions responsible for providing specialized assistance to young people experiencing serious difficulties and their families. They originated from three different centres in Montreal, Quebec City, and the Eastern township region, from which they received services by virtue of the Act Respecting Health Services and Social Services (14%), the Youth Protection Act (24%), or the Young Offender’s Act (62%; replaced by the Youth Criminal Justice Act). They were selected (n = 87) according to the following three criteria: (a) male, (b) less than 19 years of age, and (c) presence of conduct disorder (based on youth or parent report), or rated above clinical threshold (T-score of 70) on the Child Behavior Checklist (CBCL) based on parent (Achenbach, 1991) or teacher report (Teacher Report Form [TRF]). Forty-eight of the 87 adolescents or 55.2% of the sample agreed to participate in the psychopathy evaluation. Statistical analyses yielded no significant differences between participants and nonparticipants on age, t(87) = .00, ns; severity of conduct disorder based on interview with adolescent, t(87) = .34, ns; or severity of conduct disorder based on interview with parent, t(87) = .41, ns. Data were missing for 6 participants. The final composition of the sample at Time 1 was 42 participants. They ranged in age from 15 to 19 years (M = 17.6, SD = 1.2); 1 participant was interviewed a few days past his 19th birthday due to delay before meeting him. Nearly all (95%) were Caucasian.

The youths were again contacted in 2003-2004 (Time 2) for the purpose of evaluating presence of delinquent behaviors over the follow-up period (24 months), as per their report. Of the 42 boys, 27 (64.3% of the sample) agreed to take part in this second measurement. At this time, they were 17 to 21 years old (M = 19.7, SD = 1.2). Statistical analyses revealed no significant difference between participants and nonparticipants on age, t(42) = .13, ns; severity of conduct disorder based on adolescent report, t(42) = .75, ns; or severity of conduct disorder based on parent report, t(42) = .11, ns. Similarly, no significant difference emerged between these two groups regarding PCL-SV total score, t(42) = 1.20, ns; Factor 1 score, t(42) = .61, ns; or Factor 2 score, t(42) = 1.37, ns.

Instruments

Diagnostic Interview Schedule for Children-II–Revised (DISC-II-R)

The DISC-II-R (version 2.25; Shaffer et al., 1993) serves to evaluate presence of Axis I mental disorders as per the criteria of the Diagnostic and Statistical Manual of Mental Disorders (3rd ed., rev.; DSM-III-R; American Psychiatric Association [APA], 1987) in youths 9 to 17 years of age. In the course of a structured interview with the principal respondent and the youth, the questions allow establishing presence of one or more mental disorders in the past 6 months. The section on conduct disorders was used to identify eligible participants. The French version of this questionnaire was developed by the Rivières-des-Prairies Hospital research team (Bergeron, Valla, & Breton, 1992; Valla, Bergeron, Bérubé, Gaudet, & St-Georges, 1994). The version administered in our study was adapted to the Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM-IV; APA, 1994) criteria. The internal coherence of the French version is generally satisfactory, albeit weaker and more variable when based on the responses of adolescents relative to those of their parents. The test–retest stability is satisfactory as well (Breton, Bergeron, Valla, Berthiaume, & St-Georges, 1998).

CBCL

There are different versions of the CBCL (Achenbach, 1991) according to child’s age and informant. The scale intended for parent and teacher respondents (TRF) is self-completed and comprises 113 items regarding behaviors problem in youths aged 5 to 18 years. A 3-point Likert-type scale (“does not apply,” “somewhat true,” and “very true”) generates a score for various categories of problems, including internalized and externalized problems. The instrument’s psychometric properties are excellent (Achenbach, 1991; Lowe, 1998). The scale regarding delinquent conduct served to identify eligible participants.

PCL-SV

The PCL-SV (Hart et al., 1995) is administered in the course of a semistructured interview that lasts about 30 to 60 min, made by trained researchers. PCL-SV training and interview procedures for this study have been reported elsewhere (Toupin, Basque, Côté, & Deshaies, 2008). The instrument consists of 12 items rated 0 (characteristic does not represent individual), 1 (characteristic does not represent individual clearly or some uncertainty persists), or 2 (characteristic represents individual) for a total score ranging from 0 to 24. It comprises two factors composed each of six items. Factor 1 covers affective-interpersonal traits and Factor 2 a socially deviant lifestyle. The instrument was adapted for use with adolescents, following the PCL-YV (Forth et al., 2003). Specific instructions were added for Items 3, 7, 9, and 10 to clarify the ratings for the purpose of evaluating adolescents. Greater emphasis was placed on friends, family, and school. The French translation of the instrument adapted to adolescents was validated by Toupin et al. (2008); the instrument has demonstrated good psychometric properties. Regarding internal coherence, coefficients have proved very good (.84 for total scale, .80 for Factor 1, and .87 for Factor 2). As for interrater agreement, the total scale obtained an intraclass correlation of .88 for single ratings and of .96 for average ratings, which is excellent in both cases. Interrater agreement measured by the kappa coefficient (.50) has proved moderate according to the criteria set by Landis and Koch (1977). Finally, in terms of convergent validity, the PCL-SV scores have been found to be related to antisocial personality symptoms, delinquent conduct, and psychoactive substance abuse.

The PCL-SV results were processed on the basis of a continuum, on account of the fact that no cutoff has been clearly established for adolescents on the screening version. Furthermore, recent work (Guay, Ruscio, Knight, & Hare, 2007) using a taxometric analysis of the latent structure of the PCL-R suggests it is dimensional. Note that based on the cutoff for adults, the proportion of psychopaths in the sample was very low (7% with a cutoff of 18). To avoid circular predictions, Items 11 (behavioral problems at age 10 years or less) and 12 (criminal conduct after age 10 years) of the scale were excluded when calculating total score, as these evaluate antisocial conduct directly (Campbell et al., 2004; Marshall, Egan, English, & Jones, 2006).

Delinquency scale

This self-completed scale consists of 27 items that allow establishing frequency of delinquent conduct over lifetime (total score on scale, that is, number of delinquent behaviors manifested) and age at onset of such conduct. Of these items, 20 are drawn from the delinquency scale developed by Le Blanc et al. (1996). This scale presents satisfactory internal coherence and adequate discriminating validity and predictive power (Le Blanc et al., 1996). The seven other questions regard driving without a license or impaired driving, selling drugs, engaging in sexual relations to utilitarian ends, perpetrating sexual abuse, stealing from vending machines, and forging cheques or using stolen credit cards. The behaviors were classified as violent (7 items) or nonviolent (20 items) based on the definition of Statistics Canada. Behaviors were evaluated on the scale from the point of view of a continuum (total score, that is, number of delinquent behaviors) and a dichotomy (low or high frequency of delinquent behaviors along group median). Regarding internal coherence, the alpha coefficient was satisfactory at .74 for the violent behavior scale and at .88 for the nonviolent behavior scale.

Versatility scale

Based on the items of the delinquency scale, specific categories were created to reflect versatility of delinquent behaviors, according to the definition of Item 20 of the PCL-YV (Forth et al., 2003). Ten categories were thus identified: sexual offences, assault, arson-related offences, threats, offences related to driving a motor vehicle, firearms possession, drug-related offences, fraud, theft, and vandalism. Contrary to Item 20 of the PCL-YV, the questionnaire contained no item on the following behaviors: homicide, robbery, kidnapping, obstructing justice, and failure to appear for a judicial hearing. Behaviors were evaluated on the scale from the point of view of a continuum (total score, that is, number of categories of delinquent behaviors) and a dichotomy (low or high versatility of delinquent behaviors along group median). Regarding internal coherence, the alpha coefficient was satisfactory at .74. As for convergent validity, the versatility scale correlated significantly with the delinquency scale (r = .80, p < .05). Despite the strength of the relationship between these scales, both remain relevant to the analyses, given that they constitute different ways of conceptualizing delinquency, that is, in terms of intensity and diversity.

Procedure

Three different respondents (youth’s parent or tutor, youth, and teacher best acquainted with youth) were solicited to complete the questionnaires. The parent or tutor asked to complete the questionnaires had to be the one most frequently in contact with the child in the past year, regardless of who actually had legal custody of the child. Respondents had to consent to take part in the research and had to have sufficient knowledge of French. In this regard, they signed a free and informed consent form. The contribution by the teacher and the parent allowed establishing the list of youths eligible for the study, on the basis of the inclusion criteria mentioned above, and collecting collateral information useful to the PCL-SV evaluation. The project met all requirements of the Ethics Committee of the Université de Sherbrooke.

Results

Descriptive Data

Mean scores on the PCL-SV were 8.9 (SD = 4.6) on the total checklist, 3.8 (SD = 2.1) on Factor 1, and 5.1 (SD = 3.2) on Factor 2. Factor 2 mean score was significantly different from Factor 1 mean score, t(27) = 2.26, p < .05. At Time 1 assessment, participants reported an average of nearly 12 delinquent behaviors lifetime (M = 11.6, SD = 6.4). They also reported having committed more nonviolent offences (M = 8.4, SD = 5.1) than violent ones (M = 2.5, SD = 1.6). It needs to be said, however, that the scale comprised more items of nonviolent behaviors than violent behaviors (20 vs. 7). Regarding versatility, the average number of categories of delinquent behaviors was 4.3 (SD = 2.0.). Mean age at first offence was 9.4 (SD = 3.1). Table 1 presents the data on self-reported delinquency at Time 2. Participants reported an average of 6.2 (SD = 4.7) delinquent behaviors over the follow-up period and an average 3.2 (SD = 1.9) different categories of offences. Mean number of nonviolent delinquent behaviors reported was still higher than mean number of violent behaviors.

Table 1.

Descriptive Data on Delinquent Behaviors at Time 2 (N = 27).

	M (SD)	Minimum-Maximum
Number of delinquent behaviors^a	6.2 (4.7)	0-19
Number of violent delinquent behaviors	1.3 (1.3)	0-6
Number of nonviolent delinquent behaviors	4.6 (3.7)	0-14
Number of categories of delinquent behaviors (versatility)	3.2 (1.9)	0-7

Total score on delinquency scale (behaviors manifested since Time 1).

Bivariate Tests

The strength of the relationship between prediction indices (Time 1) and criterion variables (Time 2) was verified through Pearson’s correlations (Table 2). The correlation coefficients were then compared using a Z test, as proposed by Meng et al. (1992). As expected, age at first delinquent behavior proved negatively associated with criterion variables. All the correlations were statistically significant, except for nonviolent recidivism. However, this may be due to limited statistical power because with the actual number of participants, the detection of significant correlations is possible only if the correlations are greater than .47. Where violent recidivism is concerned, the PCL-SV scores yielded stronger correlations than did the number of offence and age at first offence. In fact, total score and Factor 2 score were more strongly correlated to violent recidivism (r = .83 and .79, respectively, p < .05) than were age at first delinquent behavior (r = −.52; Z = 5.42 and Z = 2.98, respectively, p < .01) and number of delinquent behaviors (r = .43; Z = 3.08 and Z = 2.98, respectively, p < .01). For its part, Factor 1 score (r = .6, p < .05) stood out only relative to age at first delinquent behavior (Z = 4.07, p < .01). Regarding nonviolent recidivism, the number of delinquent behaviors proved to be strongly correlated (r = .55, p < .05). However, it did not differ significantly from the PCL-SV scores (r = .50, p < .05, Z = .31, ns for total score; r = .35, ns, Z = .96, ns for Factor 1 score; r = .49, ns, Z = .43, ns for Factor 2 score). Where behavior versatility is concerned, the PCL-SV scores tended to correlate more strongly than did indices of delinquent behavior. The difference proved significant only with age at first delinquent behavior (r = .73, p < .01, Z = 4.21, p < .01 for total score; r = .62, p < .01, Z = 3.66, p < .01 for Factor 1 score; r = .62, p < .01, Z = 3.48, p < .01 for Factor 2 score). Factor 2 score did not demonstrate significantly different correlations with delinquency indices at Time 2 than did Factor 1 score (Z = 1.4, .80, and 0, ns, for violent and nonviolent recidivism, and behavior versatility, respectively).

Table 2.

Correlations Between Prediction Variables and Criterion Variables (N = 27).

	Violent recidivism^a	Nonviolent recidivism^b	Versatility scale
Behavioral indices (Time 1)
Number of delinquent behaviors	.43*	.55**	.58**
Age at 1st delinquent behavior	−.52**	−.37	−.42*
PCL-SV scores
Total	.83**	.50**	.73**
Factor 1	.62**	.35	.62**
Factor 2	.79**	.49	.62**

Note: PCL-SV = Psychopathy Checklist–Screening Version.

Number of violent delinquent behaviors (behaviors manifested since Time 1).

Number of nonviolent delinquent behaviors (behaviors manifested since Time 1).

p < .05. **p < .01.

Predicting Recidivism

Linear regressions were carried out to establish the contribution of the PCL-SV scores to predict recidivism in terms of violent and nonviolent delinquent behavior and behavior versatility, when behavioral indices are controlled statistically (Table 3). Owing to limited statistical power, PCL-SV total score, Factor 1 score, and Factor 2 score were tested in separate models.

Table 3.

Prediction of Violent and Nonviolent Recidivism (N = 27).

	Violent recidivism					Nonviolent recidivism
Equations	β	SE	β (standardized)	Total R²	ΔR²	β	SE	β (standardized)	Total R²	ΔR²
Step 1				.28*	.28*				.30*	.30*
Age at 1st delinquent behavior	−.19	.11	−.44			.09	.29	.07
Number of delinquent behavior	.02	.05	.11			.35*	.14*	.60*
Step 2				.72*	.44*				.37	.07
Age at 1st delinquent behavior	−.10	.07	−.25			.17	.29	.15
Number of delinquent behavior	−.03	.04	−.16			.29	.14	.50
PCL-SV total score	.29*	.05*	.79*			.31	.20	.31
Step 2				.55*	.28*				.36	.06
Age at 1st delinquent behavior	−.17	.09	−.41			.10	.29	.09
Number of delinquent behavior	.00	.04	.06			.32*	.14*	.56*
PCL-SV Factor 1 score	.33*	.09*	.54*			.40	.29	.24
Step 2				.63*	.36*				.35	.04
Age at 1st delinquent behavior	−.06	.08	−.15			.20	.31	.17
Number of delinquent behavior	−.04	.04	−.20			.29	.15	.50
PCL-SV Factor 2 score	.51*	.11*	.82*			.47	.40	.28

Note: PCL-SV = Psychopathy Checklist–Screening Version.

p < .05.

Where the prediction of violent recidivism is concerned, the PCL-SV total score made a markedly significant contribution above and beyond that of behavioral indices (ΔR² = .44, p < .05), with the model explaining 72% of the variance. However, its contribution was not significant to predict nonviolent recidivism (ΔR² = .07, ns); here, behavioral indices alone managed to explain 30% of the variance. As for the PCL-SV factors, Factor 2 score made a greater contribution than did Factor 1 score to predict violent recidivism (ΔR² = .36, p < .05 vs. ΔR² = .28, p < .05). However, where the prediction of behavior versatility is concerned, Factor 1 score made a stronger contribution (ΔR² = .25, p < .05 for Factor 1 score vs. ΔR² = .11, p < .05 for Factor 2 score; Table 4).

Table 4.

Prediction of Versatility of Delinquent Behaviors (N = 27).

Equations	β	SE	β (standardized)	Total R²	ΔR²
Step 1				.34*	.34*
Age at 1st delinquent behavior	.01	.15	.02
Number of delinquent behavior	.18*	.07*	.59*
Step 2				.60*	.26*
Age at 1st delinquent behavior	.10	.12	.16
Number of delinquent behavior	.11	.06	.38
PCL-SV total score	.32*	.08*	.61*
Step 2				.59*	.25*
Age at 1st delinquent behavior	.03	.12	.04
Number of delinquent behavior	.15*	.06*	.49*
PCL-SV Factor 1 score	.45*	.12*	.51*
Step 2				.45*	.11*
Age at 1st delinquent behavior	.11	.14	.18
Number of delinquent behavior	.12	.07	.42
PCL-SV Factor 2 score	.41*	.19*	.46*

Note: PCL-SV = Psychopathy Checklist–Screening Version.

p < .05.

ROC Analysis

The ROC analysis was used to verify the performance of the PCL-SV in predicting violent and nonviolent recidivism, as well as versatility of delinquent conduct (dichotomous variables), above and beyond the contribution of behavioral indices. To this end, the variables were introduced in the logistic regression models as follows: Model 1 included the behavioral indices (number of delinquent behaviors and age at first delinquent behavior), whereas Model 2 included these and the PCL-SV scores (total score, Factor 1 score, Factor 2 score; introduced in separate models on account of the small sample size). The probabilities yielded by the logistic models were then introduced in a ROC analysis (Hosmer & Lemeshow, 2000).

As shown in Table 5, all the AUC proved statistically significant. Regarding the prediction of violent recidivism, the results supported the contribution of the PCL-SV. The predictive power of the model including the PCL-SV total score (AUC = .97) was 19 percentage points better than that of the model with only the behavioral indices (AUC = .78). Similarly, when it came to predict delinquent behavior versatility, the model with the PCL-SV total score offered an excellent performance (AUC = .92, p < .05). Regarding nonviolent recidivism, however, results were less conclusive; adding the PCL-SV total score to the model did not improve the performance of the behavioral indices (AUC almost unchanged at .80). As for the contribution of the PCL-SV factors, Factor 1 score improved the model’s performance in predicting behavior versatility slightly more than did Factor 2 score (AUC = .94 vs. AUC = .85), whereas the predictions were almost the same for violent and nonviolent recidivism.

Table 5.

ROC Analysis—Prediction of Violent and Nonviolent Recidivism and of Versatility of Delinquent Behaviors.^a

	Violent recidivism		Nonviolent recidivism		Versatility
	AUC	95% CI	AUC	95% CI	AUC	95% CI
Model 1
Number of delinquent behavior	.78*	[.59, .97]	.78*	[.61, .96]	.85*	[.70, .99]
Age at 1st delinquent behavior^b
Model 2
Number of delinquent behavior	.97*	[.89, 1.0]	.80*	[.63, .97]	.92*	[.82, 1.0]
Age at 1st delinquent behavior^b
PCL-SV total score
Number of delinquent behavior	.92*	[.81, 1.0]	.80*	[.62, .97]	.94*	[.86, 1.0]
Age at 1st delinquent behavior^b
PCL-SV Factor 1 score
Number of delinquent behavior	.92*	[.77, 1.0]	.79*	[.61, .97]	.85*	[.71, .99]
Age at 1st delinquent behavior^b
PCL-SV Factor 2 score

Note: ROC = receiver operator characteristics; AUC = area under the curve; CI = confidence intervals; PCL-SV = Psychopathy Checklist–Screening Version. Variables dichotomized along group median.

Variables dichotomized along group median.

Inverse relationship.

p < .05.

Discussion

The aim of our study was to verify the contribution of psychopathic traits in adolescence to predict delinquency in early adulthood. Results support this contribution as PCL-SV scores make a significant contribution above and beyond that of behavioral indices to predict versatility in delinquency and violent recidivism. These results indicate the interest to pursue this line of inquiry as it could lead to clinical applications.

First, our analyses demonstrate that the PCL-SV scores are related to self-reported delinquent behavior 2 years later and, even more markedly, to versatility and violent behavior. In this regard, the PCL-SV total score generates significantly stronger correlations than do indices of previous delinquency. Regarding the prediction of violent recidivism, the PCL-SV total score provides a clearly significant contribution above and beyond that of number and age at first delinquent behavior. Its contribution to predict nonviolent recidivism is nonsignificant, thus confirming the instrument’s specificity in identifying a subgroup of delinquents with severe criminal behaviors (Hare, 1998, 2003; Hare et al., 2000). Along the same line, the PCL-SV total score proved relevant to predict versatility in delinquent behavior over behavioral indices alone.

Contrary to the results obtained by Brandt et al. (1997) and Gretton et al. (2004), the ability of the PCL-SV to predict violent recidivism seems to be mainly attributable to Factor 2 scores. It should be noted, however, that owing to the small sample size, both factors could not be tested in the same model and, therefore, their independent contribution could not be clearly established. Moreover, the participants in our study obtained lower scores on Factor 1 than on Factor 2, which may have made it more difficult to grasp the contribution of Factor 1. Interestingly, however, it emerged that this factor made a greater contribution to predict delinquent versatility. Consequently, the results do not allow reaching a clear verdict regarding the contribution of Factor 1 to predict delinquency when psychopathy is evaluated on the basis of a clinical interview.

Second, our study stands out from a methodological viewpoint through its use of values predicted by logistic models through ROC analyses, which is something that none of the studies reviewed had sought to do. This procedure allowed demonstrating that the PCL-SV total score improves on the performance of behavioral indices in identifying participants with considerable violent recidivism, boosting the AUC to a “near perfect” performance according to the criteria set by Hosmer and Lemeshow (2000). Similarly, the PCL-SV total score illustrates the added value of this measure in identifying participants with a marked versatility of delinquent behaviors. The PCL-SV’s performance in predicting behavior versatility appears to be more strongly associated to Factor 1 than to Factor 2, whereas no difference is observed with respect to violent and nonviolent recidivism. The use of a versatility scale in the study proved relevant as it helped shed new light on the differential contribution of the factors.

In sum, this study allowed establishing the added value of the measurement of psychopathic traits in adolescence in predicting criminal behavior in early adulthood. As it is one of the rare studies to have examined self-reported criminal recidivism, it should however be replicated. The study is not without certain limitations. First, its statistical power is limited, which may have hampered the detection of associations with nonviolent recidivism. Second, it needs to be underscored that a higher number of participants would have allowed integrating the two PCL-SV factors in the same regression equations to verify their independent contributions, as other authors had previously done (Brandt et al., 1997; Corrado et al., 2004; Gretton et al., 2004). Third, our participants are mostly Caucasian adolescents with antisocial behaviors (diagnosis of conduct disorder or high score on delinquency scale) referred to youth centres. As the sample is specific, results cannot be generalized to adolescents from the general population.

Future studies should investigate other indices of adaptation in adulthood for adolescents with psychopathic traits, including employment, marital relationships, and peer relations. As pointed out by Gretton et al. (2004), the absence of criminal recidivism is not a guarantee of good social adaptation; it merely represents one facet of this adaptation. Furthermore, a longer follow-up period would allow verifying whether the predictive power of the PCL-SV persists over time. Finally, it would be interesting to verify the prediction of the same scale with female adolescents. Vincent et al. (2008) highlight clear differences between boys and girls in the prediction of recidivism using PCL-YV. Moreover, research by Forouzan (2003) and Forouzan and Cooke (2005) on female psychopathy suggest notable gender differences in the expression of certain traits and in the type of delinquency manifested in adolescence. It would be worthwhile to pursue this line of investigation.

Conclusion

Taking into account clinically specific personality-related indices such as psychopathic traits allows improving prediction of violent recidivism and versatility of delinquent behavior. Hence, in the interest of the protection of society, the evaluation of psychopathy in adolescence appears justified, as it could prevent youths from becoming ensnared in a severe criminal trajectory. Moreover, in the light of the heterogeneousness among delinquents, the evaluation of psychopathic traits would allow identifying more specific typologies and, in turn, establishing differential interventions (Forth et al., 2003).

Footnotes

Acknowledgements

We thank Francis Lafortune for manuscript revision.

Authors’ Note

This study was conducted at the labs of the Groupe de recherche sur les inadaptations sociales de l’enfance (GRISE), Université de Sherbrooke.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The study benefited from the financial support of the Social Sciences and Humanities Research Council of Canada (SSHRC), the Fonds Québécois de recherche sur la société et la culture (FQRSC), and the Université de Sherbrooke.

References

Achenbach

T. M.

(1991). Manual for the Child Behavior Checklist and 1991 profile. Burlington: Department of Psychiatry, University of Vermont.

American Psychiatric Association. (1987). Diagnostic and statistical manual of mental disorders (3rd ed., rev.). Washington, DC: Author.

American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author.

Bergeron

Valla

J.-P.

Breton

J.-J.

(1992). Pilot study for the Quebec Child Mental Health Survey: Part II. Correlates of DSM-III-R criteria among six to 14 year olds. Canadian Journal of Psychiatry, 37, 381-386.

Brandt

J. R.

Kennedy

W. A.

Patrick

C. J.

Curtin

J. J.

(1997). Assessment of psychopathy in a population of incarcerated adolescent offenders. Psychological Assessment, 9, 429-435.

Breton

J.-J.

Bergeron

Valla

J.-P.

Berthiaume

St-Georges

(1998). Diagnostic Interview Schedule for Children (DISC-2.25) in Quebec: Reliability findings in light of the MECA study. Journal of the American Academy of Child & Adolescent Psychiatry, 37, 1167-1174.

Campbell

M. A.

Porter

Santor

(2004). Psychopathic traits in adolescent offenders: An evaluation of criminal history, clinical, and psychosocial correlates. Behavioral Sciences & the Law, 22, 23-47.

Caspi

(2000). The child is father of the man: Personality continuities from childhood to adulthood. Journal of Personality and Social Psychology, 78, 158-172.

Catchpole

R. E.

Gretton

(2003). The predictive validity of risk assessment with violent young offenders: A 1-year examination of criminal outcome. Criminal Justice and Behavior, 30, 688-708.

10.

Corrado

R. R.

Vincent

G. M.

Hart

S. D.

Cohen

I. M.

(2004). Predictive validity of the Psychopathy Checklist: Youth version for general and violent recidivism. Behavioral Sciences & the Law, 22, 5-22.

11.

Côté

Hodgins

(2000). Remords, culpabilité et confession: La pratique à la lumière de l’échelle de psychopathie de Hare [Remorse, guilt and confession: Practice in the light of the Hare Psychopathy Checklist]. 25th Anniversary Congress on Law and Mental Health, Siena, Italy.

12.

Edens

J. F.

Campbell

J. S.

Weir

J. M.

(2007). Youth psychopathy and criminal recidivism: A meta-analysis of the psychopathy checklist measures. Law and Human Behavior, 31, 53-75.

13.

Forouzan

(2003). Psychopathy among women: Conceptualisation and assessment problems. Third Annual International Association of Forensic Mental Health Services (IAFMHS) Conference, Miami, FL.

14.

Forouzan

Cooke

(2005). Figuring out la femme fatale: Conceptual and assessment issues concerning psychopathy in females. Behavioral Sciences & the Law, 23, 765-778.

15.

Forth

A. E.

Burke

H. C.

(1998). Psychopathy in adolescence: Assessment, violence, and developmental precursors. In Cooke

D. J.

Forth

A. E.

Hare

R. D.

(Eds.), Psychopathy: Theory, research and implications for society (pp. 205-229). Dordrecht, Netherlands: Kluwer.

16.

Forth

A. E.

Hart

S. D.

Hare

R. D.

(1990). Assessment of psychopathy in male young offenders. Psychological Assessment, 2, 342-344.

17.

Forth

A. E.

Kosson

D. S.

Hare

R. D.

(2003). The Hare Psychopathy Checklist: Youth Version. Toronto, Ontario, Canada: MHS.

18.

Forth

A. E.

Mailloux

(2000). Psychopathy in youth: What do we know? In Carl

(Ed.), The clinical and forensic assessment of psychopathy: A practitioner’s guide (pp. 3-54). Mahwah, NJ: Erlbaum.

19.

Frick

P. J.

O’Brien

B. S.

Wootton

J. M.

McBurnett

(1994). Psychopathy and conduct problems in children. Journal of Abnormal Psychology, 103, 700-707.

20.

Frick

P. J.

Stickle

T. R.

Dandreaux

D. M.

Farrell

J. M.

Kimonis

E. R.

(2005). Callous-unemotional traits in predicting the severity and stability of conduct problems and delinquency. Journal of Abnormal Child Psychology, 33, 471-487.

21.

Gretton

H. M.

Hare

R. D.

Catchpole

R. E. H.

(2004). Psychopathy and offending from adolescence to adulthood: A 10-year follow-up. Journal of Consulting and Clinical Psychology, 72, 636-645.

22.

Guay

J. P.

Ruscio

Knight

R. A.

Hare

R. D.

(2007). A taxometric analysis of the latent structure of psychopathy: Evidence for dimensionality. Journal of Abnormal Psychology, 116, 701-716.

23.

Hare

R. D.

(1991). The Hare Psychopathy Checklist–Revised: Manual. Toronto, Ontario, Canada: MHS.

24.

Hare

R. D.

(1998). Psychopaths and their nature: Implications for the mental health and criminal justice systems. In Millon

Simonsen

Birket-Smith

Davis

R. D.

(Eds.), Psychopathy: Antisocial, criminal, and violent behavior (pp. 188-212). New York, NY: Guilford.

25.

Hare

R. D.

(2003). The Hare Psychopathy Checklist–Revised (2nd ed.). Toronto, Ontario, Canada: MHS.

26.

Hare

R. D.

Clark

Grann

Thornton

(2000). Psychopathy and the predictive validity of the PCL-R: An international perspective. Behavioral Sciences & the Law, 18, 623-645.

27.

Hart

S. D.

Cox

D. N.

Hare

R. D.

(1995). Psychopathy Checklist: Screening Version. Toronto, Ontario, Canada: MHS.

28.

Hemphill

J. F.

Hare

R. D.

Wong

(1998). Psychopathy and recidivism: A review. Legal and Criminological Psychology, 3, 139-170.

29.

Hicks

M. M.

Rogers

Cashel

(2000). Predictions of violent and total infraction among institutionalized male juvenile offenders. Journal of the American Academy of Psychiatry & the Law, 28, 183-190.

30.

Hosmer

D. W.

Lemeshow

(2000). Applied logistic regression (2nd ed.). New York, NY: Wiley.

31.

Kosson

D. S.

Cyterski

T. D.

Steuerwald

B. L.

Neumann

C. S.

Walker

M. S.

(2002). The reliability and validity of the Psychopathy Checklist: Youth Version (PCL: YV) in nonincarcerated adolescent males: The clinical assessment of children and adolescents. Psychological Assessment, 14, 97-109.

32.

Kruh

I. P.

Frick

P. J.

Clements

C. B.

(2005). Historical and personality correlates to the violence patterns of juveniles tried as adults. Criminal Justice and Behavior, 32, 69-96.

33.

Landis

R. J.

Koch

G. G.

(1977). The measurement of observer agreement for categorical data. Biometrics, 33, 159-174.

34.

Langstrom

Grann

(2002). Psychopathy and violent recidivism among young criminal offenders. Acta Psychiatrica Scandinavica, 106, 86-92.

35.

Le Blanc

McDuff

Fréchette

Langelier

Levert

Trudeau-Le Blanc

(1996). Mesures de l’adaptation sociale et personnelle pour les adolescents québé Manuel et guide d’utilisation, 3e édition [Measures of personal and social adapfor Quebec adolescents: Manual and user’s guide (3rd ed.)]. Montréal: Université de Montréal.

36.

Leistico

A.-M. R.

Salekin

R. T.

DeCoster

Rogers

(2008). A large-scale meta-analysis relating the Hare measures of psychopathy to antisocial conduct. Law and Human Behavior, 32, 28-45.

37.

Loney

B. R.

Taylor

Butler

M. A.

Iacono

W. G.

(2007). Adolescent psychopathy features: 6-year temporal stability and the prediction of externalizing symptoms during the transition to adulthood. Aggressive Behavior, 33, 242-252.

38.

Lowe

L. A.

(1998). Using the Child Behavior Checklist in assessing conduct disorder: Issues of reliability and validity. Research on Social Work Practice, 3, 268-301.

39.

Lynam

D. R.

(1996). Early identification of chronic offenders: Who is the fledgling psychopath? Psychological Bulletin, 120, 209-234.

40.

Lynam

D. R.

(1997). Pursuing the psychopath: Capturing the fledgling psychopath in a nomological net. Journal of Abnormal Psychology, 106, 425-438.

41.

Lynam

D. R.

Miller

D. J.

Vachon

Loeber

Stouthamer-Loeber

(2009). Psychopathy in adolescence predicts official reports of offending in adulthood. Youth Violence and Juvenile Justice, 7, 189-207.

42.

Marshall

Egan

English

Jones

R. M.

(2006). The relative validity of psychopathy versus risk/needs-based assessments in the prediction of adolescent offending behaviour. Legal and Criminological Psychology, 11, 197-210.

43.

Meng

X.-L.

Rosenthal

Rubin

D. B.

(1992). Comparing correlated correlation coefficients. Psychological Bulletin, 111, 172-175.

44.

Monahan

Steadman

H. J.

Silver

Appelbaum

P. S.

Robbins

P. C.

Mulvey

. . . Banks

(2001). Rethinking risk assessment: The MacArthur study of mental disorder and violence. New York, NY: Oxford University Press.

45.

Mossman

(1994). Assessing predictions of violence: Being accurate about accuracy. Journal of Consulting and Clinical Psychology, 62, 783-792.

46.

Murrie

D. C.

Cornell

D. G.

Kaplan

McConville

Levy-Elkon

(2004). Psychopathy scores and violence among juvenile offenders: A multi-measure study. Behavioral Sciences & the Law, 22, 49-67.

47.

Pauzé

Toupin

Déry

Mercier

(2000). Portrait des jeunes inscrits à la prise en charge des youth centres du Québec et description des services reçus au cours des premiers mois [Portrait of the youths registered under the care of Quebec’s youth centres and descripof services received in the first months]. Sherbrooke: Université de Sherbrooke.

48.

Ridenour

T. A.

Marchant

G. J.

Dean

R. S.

(2001). Is the revised Psychopathy Checklist clinically useful for adolescents? Journal of Psychoeducational Assessment, 19, 227-238.

49.

Salekin

R. T.

(2008). Psychopathy and recidivism from mid-adolescence to young adulthood: Cumulating legal problems and limiting life opportunities. Journal of Abnormal Psychology, 117, 386-395.

50.

Salekin

R. T.

Neumann

C. S.

Leistico

A.-M. R.

DiCicco

T. M.

Duros

R. L.

(2004). Psychopathy and comorbidity in a young offender sample: Taking a closer look at psychopathy’s potential importance over disruptive behavior disorders. Journal of Abnormal Psychology, 113, 416-427.

51.

Salekin

R. T.

Rogers

Sewell

K. W.

(1996). A review and meta-analysis of the Psychopathy Checklist and Psychopathy Checklist–Revised: Predictive validity of dangerousness. Clinical Psychology: Science and Practice, 3, 203-213.

52.

Seagrave

Grisso

(2002). Adolescent development and the measurement of juvenile psychopathy. Law and Human Behavior, 26, 219-239.

53.

Shaffer

Schwab-Stone

Fisher

Cohen

Piacentini

Davies

. . . Regier

(1993). The Diagnostic Interview Schedule for Children–Revised (DISC-R): I. Preparation, field testing, interrater reliability, and acceptability. Journal of the American Academy of Child & Adolescent Psychiatry, 32, 643-650.

54.

Spain

S. E.

Douglas

K. S.

Poythress

N. G.

Epstein

(2004). The relationship between psychopathic features, violence and treatment outcome: The comparison of three youth measures of psychopathic features. Behavioral Sciences & the Law, 22, 85-102.

55.

Stafford

Cornell

D. G.

(2003). Psychopathy predicts adolescent inpatient aggression. Assessment, 10, 102-112.

56.

Toupin

Basque

Côté

Deshaies

(2008). Adaptation de la PCL-SV à l’évaluation des adolescents suivis en Centre jeunesse: Une étude préliminaire. [Adaptation of the PCL-SV in the evaluation of adolescents treated in Youth Center: A preliminary study.]. Revue Canadienne de Criminologie et Justice Pénale, 50, 83-103.

57.

Toupin

Mercier

Déry

Côté

Hodgins

(1996). Validity of the PCL-R for adolescents. In Cooke

D. J.

Forth

A. E.

Newman

J. P.

Hare

R. D.

(Eds.), Issues in criminological and legal psychology: No. 24, International perspectives on psychopathy (pp. 143-145). Leicester, UK: British Psychological Society.

58.

Valla

J.-P.

Bergeron

Bérubé

Gaudet

St-Georges

(1994). A structured pictorial questionnaire to assess DSM-III-based diagnoses in children (6-11 years): Development, validity, and reliability. Journal of Abnormal Child Psychology, 22, 403-423.

59.

Vincent

G. M.

Odgers

C. L.

McCormick

A. V.

Corrado

R. R.

(2008). The PCL-YV and recidivism in male and female juveniles: A follow-up into young adulthood. International Journal of Law and Psychiatry, 31, 287-296.

60.

Vincent

G. M.

Vitacco

M. J.

Grisso

Corrado

R. R.

(2003). Subtypes of adolescent offenders: Affective traits and antisocial behavior patterns. Behavioral Science & the Law, 21, 695-712.