Psychopathy and violent recidivism among young criminal offenders. Criminal offenders with severe or persistent anti- social behaviour often fulfil diagnostic ...
Copyright ª Blackwell Munksgaard 2002
Acta Psychiatr Scand 2002: 106 (Suppl. 412): 86–92 Printed in UK. All rights reserved
ACTA PSYCHIATRICA SCANDINAVICA ISSN 0065-1591
Psychopathy and violent recidivism among young criminal offenders La˚ngstro¨m N, Grann M. Psychopathy and violent recidivism among young criminal offenders Acta Psychiatr Scand 2002: 106 (Suppl. 412): 86–92. ª Blackwell Munksgaard 2002. Objective: To investigate the predictive validity of psychopathic personality traits (assessed with the revised psychopathy checklist, PCL-R; Hare, 1991) for violent criminal recidivism among young offenders. Method: The relationship between PCL-R psychopathy and violent reoffending was studied in 98 young (M ¼ 18.40, range 15–20 years) violent and sex offenders subjected to forensic psychiatric evaluation in Sweden during 1988–95. Subjects were followed during detainment and for 24 months in the community to first reconviction for a violent offence. Results: We found a modest but significant association between PCL-R scores and violent recidivism, almost exclusively accounted for by behavioural criteria. Among 13 possible confounders tested, conduct disorder before age 15 and a young age at first conviction eliminated the relationship between psychopathy and violent recidivism in pair-wise logistic regression models. Conclusion: PCL-R psychopathy may be a less valid predictor for violent criminal recidivism among severe youthful offenders than among adult offenders.
Introduction Psychopathy and violent recidivism among young criminal offenders
Criminal offenders with severe or persistent antisocial behaviour often fulfil diagnostic criteria for antisocial personality disorder (DSM-IV; 1) or dyssocial personality disorder (ICD-10; 2). However, the DSM and ICD conceptualizations have both been criticized for being vague and unspecific and, in the case of the DSM, for relying too heavily on behavioural aspects and criminal conduct criteria (for example, 3). A recent review of the literature showed that the average prevalence of antisocial personality disorder was 46% among inmates in correctional settings (4) and the prevalence among inmates convicted of violent crimes is considerably higher. These operationalizations may result in that the personality disorder is seen as synonymous with severe criminal offending that, after all, is not a psychiatric condition per se (5). 86
N. L ngstrm, M. Grann Division of Forensic Psychiatry, Karolinska Institutet, Stockholm, Sweden
Key words: risk assessment; violence; psychopathic personality; youth; forensic psychiatry Niklas L!ngstr"m, Karolinska Institutet, NEUROTEC, Division of Forensic Psychiatry, PO Box 4044, SE-141 04, Huddinge, Sweden This paper, in its preliminary version, was read at the 2nd European Congress on Violence in Clinical Psychiatry symposium 'Violence in public health and victims' perspective', Stockholm, Sweden, 28–30 June 2001
In contrast, a more narrowly defined subgroup of offenders with severe and persistent antisocial behaviour has been described as being grandiose, egocentric and manipulative and lacking the ability to feel empathy or remorse. On the behavioural level, impulsivity, sensation seeking and a failure to fulfil social obligations are common characteristics. This socially devastating constellation of affective, interpersonal and behavioural characteristics is termed psychopathic personality disorder or psychopathy. Since individual characteristics associated with the ability to inhibit aggression and violence are less pronounced in the psychopathic individual, psychopathy is currently the most important personality construct for the study of antisocial behaviour and criminal offending in adults. The state-of-the-art conceptualization of psychopathy is reflected in a 20-item symptom construct rating scale, the psychopathy checklist revised (PCL-R; 6). The PCL-R is a clinical tool for diagnosing psychopathy and defines 20 traits and characteristics rated on a three-point scale from 0
Psychopathy and violent recidivism for ÔabsentÕ via 1 for Ôpartially or possibly presentÕ to 2 for ÔpresentÕ. The item sum, ranging from 0 to 40, is used as a continuous indicator of the extent to which a person matches the clinical construct of psychopathy. A PCL-R score of 30 or more indicates the presence of psychopathy, although lower cut-off levels have been suggested for European settings (for example, 7). The instrument thus allows for both categorical and dimensional measures of psychopathy. The PCL-R has excellent psychometric properties in terms of inter-rater reliability, internal consistency and test-retest reliability (6, 8). PCL-R psychopathy has generally been shown to have two correlated underlying factors. Factor 1 captures a callous, selfish, remorseless use of others, whereas Factor 2 taps characteristics related to a chronically unstable and antisocial lifestyle (9). However, the two-factor conceptualization of the psychopathy construct has been challenged. Using item-response analyses of large sets of data from various North American and European settings, Cooke and Michie (10) recently presented evidence that PCL-R psychopathy may more appropriately be broken down into three factors rather than two: one ÔaffectiveÕ (purely intrapersonal), one interpersonal and one behavioural. PCL-R psychopathy in adult offenders. The validity of PCL-R psychopathy as a robust predictor of violent and general criminality in adult offenders, even over and above variables such as previous criminality and substance use disorder, is well established (11, 12; for a meta-analysis, see 13). Although originally developed for use with adult males in North American forensic (prison, probation and forensic psychiatric) settings, good crosscultural validity has been demonstrated in studies from several Western European settings (for example, 12, 14). Moderate predictive validity for criminal recidivism in specific offender subgroups, e.g. offenders with schizophrenia (15–17) and spousal assaulters (18), has been reported. At present, we know considerably less about psychopathy among adult female offenders, though at least one study suggested promising predictive ability for criminal recidivism (19). Psychopathic traits in adolescents. High emotional, humanitarian and societal costs and the association with substance abuse and criminality, has led researchers to look for precursors or correlates to psychopathy in adolescence and childhood (20–22). PCL-R psychopathy has been found to correlate with intrainstitutional violence among incarcerated juvenile delinquents (23–26), early onset, aggressive, violent or pervasive conduct
disorder or criminal offending (23–27) and substance abuse disorder (25, 28) in adolescent offenders. In one study (25), psychopathic young offenders more often self-reported having committed arson, being coercive and controlling with girlfriends or having used force or pressure to obtain sexual contact with a female against her will. The predictive validity of PCL-R psychopathy for criminal recidivism in non-adult offenders has been addressed in a small number of North American studies, e.g. for violent criminal recidivism among adolescent offenders in general (23, 24, 29, 30). In addition, in two recent studies of adolescent sex offenders, psychopathy was a powerful predictor of general criminal recidivism, but not sexual recidivism (31), and violent re-offending including sexual offences (32), respectively. However, Murdock Hicks et al. (33) recently presented data that suggested that the PCL:SV used with incarcerated adolescent offenders did not add incremental predictive validity concerning intrainstitutional infractions in multiple regression over that obtained with selected scales of they Minnesota multiphasic personality inventory for adolescents (MMPI-A). In addition, in univariate analyses, the PCL:SV appeared less valid when used with Caucasian and Hispanic offenders as compared with offenders of African descent. We explored the relationship between psychopathy as assessed with Hare’s PCL-R slightly adapted for use with youth and violent criminal recidivism in a group of young offenders subjected to forensic psychiatric assessment in Sweden. The potential confounding effect for 13 variables on the relationship between psychopathy and recidivism was tested in logistic regression models. We hypothesized that psychopathy would be positively related to violent recidivism among adolescents and young adults similar to the effect previously demonstrated for adult offenders. Material and methods
Two samples of young offenders were combined for the present study. Both samples consisted of young (defined as 15–20 years of age at the time of the index crime) offenders subjected to a major forensic psychiatric evaluation (FPE) in Sweden. Since the age of criminal responsibility is 15 years in Sweden, younger offenders were not included. An FPE is initiated in conjunction with court proceedings when a defendant is suspected of having committed a severe crime under the influence of mental disorder. The FPE consists of an extensive evaluation during a 4-week stay in a specialized forensic psychiatric unit. A team with a 87
La˚ngstro¨m and Grann forensic psychiatrist, a clinical psychologist and a social worker obtains a comprehensive history covering the defendant’s entire life span, and performs psychiatric examinations, psychological testing and personality assessments. Ward observations supplement the examinations. Data are collected from multiple sources, such as past institutionalization and medical records, school and employment reports, social service notes, police reports and court testimonies. Subjects
Young violent (non-sexual) offenders. All young male offenders subjected to FPE during 1988–90 and diagnosed with an ICD-9 (2) personality disorder with or without concomitant substance use disorder were included (n ¼ 44). ÔViolent crimeÕ was defined as a conviction of homicide (33%), assault (44%) or robbery (31%) (because some subjects were convicted on several counts the sum adds up to more than 100%). The average age at the time of the index crime was 18.79 years (SD ¼ 1.05, range 17–20). The young violent non-sexual offenders constituted approximately 10% of a sample of personality disordered offenders among whom the predictive validity of PCL-R psychopathy has been reported previously (14, 34). Young sex offenders. All male young sex offenders subjected to FPE in Sweden during 1988–95 (n ¼ 54) constituted the second subsample. The most severe sex offence in each case was rape (63%), contact sex offences involving less threat or perceived danger, typically child molestation (28%), and non-contact offences like indecent exposure (9%). The average subject age at the time of the index offence was 18.06 years (SD ¼ 1.45, range 15–20). The predictive validity of PCL-R psychopathy specifically for sexual recidivism in the young sex offender subsample was described as part of a prior study by La˚ngstro¨m and Grann (31). However, though including two samples of offenders partly described earlier, this study expands on previous investigations in specifically testing the associations between PCL-R two- and three-factor conceptualizations and violent criminal recidivism among 15–20-year-old offenders while controlling for various empirically derived possible confounders of such associations. Procedure
Separate methodological studies indicated that the quality of Swedish FPE files permitted retrospective 88
data collection with high reliability for a number of subject background variables and risk factors (35), including PCL-R psychopathy (36). PCL-R psychopathy was rated from files only, taking into account the minor modifications of coding procedures suggested for adolescents by Forth et al. (24). Psychopathy and other baseline data (including potential confounders) were coded from sources explicitly devoid of any data on outcome during follow-up. In addition, the collection of follow-up data concerning criminal reconvictions for violent crime from official registers was done blindly. All subjects were initially followed for reconvictions for violent offences already during detainment (i.e. intrainstitutional violence). Hereafter, they were followed from release from prison, discharge from compulsory inpatient forensic psychiatric treatment and first day of probation, respectively, until first event of violent recidivism, end of follow-up, emigration or death. Mean timeat-risk in the community after release, discharge or onset of probation was 48.97 months (SD ¼ 34.64, range 0.03–133.40). To standardize for varying lengths of time-at-risk, we chose to study reconvictions up to 24 months of time-at-risk in the community (corrected for incarcerations, prison sentences or forensic psychiatric care because of non-violent offences). To explore factors potentially confounding the relationship between psychopathy and recidivism, a number of logistic regression models were built with any reconviction for a violent offence (homicide, assault or robbery) within 24 months at risk in the community as the dependent variable. PCLR total score (continuous variable) was entered alone in the first model. Next, a series of 13 factors identified in the literature (e.g. 37, 38) as potential risk factors for antisocial behaviour were entered in consecutive models, alone and together with PCL-R psychopathy. The potential confounders investigated were: first generation immigrant status (i.e. not born in Sweden) (yes ⁄ no); low socioeconomic status (yes ⁄ no); separation from parents or primary caregiver before adulthood (yes ⁄ no); low (under class average) grades in school (yes ⁄ no); history of institutional placement (treatment or youth home) (yes ⁄ no); any previous psychiatric treatment (child- and adolescent psychiatry, adult civil or forensic mental health services) (yes ⁄ no); fulfilled criteria for DSM-IV conduct disorder before age 15 years (yes ⁄ no); a clinical diagnosis (ICD-9) of substance abuse or dependence at FPE (yes ⁄ no); and signs of intellectual deficiency (yes ⁄ no) as extracted from the psychometric testing at FPE. Several variables
Psychopathy and violent recidivism Table 1. Crude and adjusted ORs with 95% CI derived from logistic regression for associations between PCL-R psychopathy, 13 potential confounders and violent recidivism among 81 young violent and sex offenders Violent recidivism within 24 months of time-at-risk in the community (n ¼ 30, 37%) Adjusted OR (95% CI) Variable PCL-R total score PCL-R factor 1 PCL-R factor 2 First generation immigrant status Low SES Separation from parents Low grades in school Any institutional placement History of psychiatric treatment DSM-IV conduct disorder before age 15 Low IQ Substance use diagnosis No. of previous convictions No. of previous violent convictions Age of first conviction Age of release or onset of probation
Crude OR
95% CI
Model I*
Model II
Model III
Model IV
1.07 1.09 1.13 3.21
1.01–1.14 0.95–1.24 1.02–1.26 1.01–10.21
1.07 (1.01–1.14)
1.05 (0.97–1.13)
1.05 (0.98–1.13)
1.07 (1.01–1.15)
2.67 0.65 0.88 1.47
0.85–8.42 0.26–1.65 0.34–2.25 0.59–3.65
1.92
0.72–5.14
5.77
2.00–16.68
0.77 3.34 1.13 1.25
0.28–2.11 1.29–8.68 0.99–1.30 0.80–1.95
0.72 0.82
0.54–97 0.64–1.07
3.07 (0.92–10.16)
4.23 (1.13–31.70)
3.34 (1.8–25.92)
0.79 (0.58–1.07)
* Models I–IV included PCL-R total score and any of the five variables exhibiting crude ORs (in univariate analyses) significantly different from 1 in pair-wise logistic regression models.
related to previous and current offending were also studied: number of previous registered convictions for any criminal offence excluding minor traffic offences (continuous variable); number of previous convictions for violent offences (continuous variable); age (years) at first registered conviction for any crime (continuous variable); and age (years) at release, discharge or onset of probation (continuous variable). Statistics
Odds ratios (ORs) with 95% confidence interval (95% CI) derived from univariate and multivariate logistic regression models were used to describe the relationship between PCL-R psychopathy and registered violent recidivism within 24 months of time-at-risk. Finally, receiver operating characteristic (ROC) analysis was used to test the predictive validity of the full-scale PCL-R factor 1 and factor 2 scores, respectively. All calculations were made with the Statistical Package for the Social Sciences (SPSS) version 10. Results
PCL-R psychopathy was possible to rate for 97 subjects. The mean total PCL-R score (according
to the adaptation suggested by Forth et al. 1990) was 19.76 (SD ¼ 7.13, range 5–34). When using a score of 26 or higher as a cut-off, the prevalence of psychopathy as a dichotomous construct was 24% (n ¼ 24). Eighty-one young offenders had at least 24 months of time-at-risk in the community after release from prison, discharge from compulsory inpatient forensic psychiatric treatment or onset of probation. Thirty (37%) of these were reconvicted for a violent reoffence during follow-up. Among subjects fulfilling criteria for PCL-R psychopathy, 33% were reconvicted of a violent crime within 24 months of time-at-risk, as compared with 22% among the others, corresponding to an OR of 1.82 (95% CI ¼ 0.66–5.00). When the violent offender and sex offender subgroups were analysed separately, the corresponding ORs for violent recidivism were 1.35 (95% CI ¼ 0.37–4.92) and 1.50 (95% CI ¼ 0.23–9.61), respectively. A ROC analysis with a reconviction of a violent crime within 24 months of time-at-risk as a function of the total PCL-R score yielded an area under the curve of 0.65 (95% CI ¼ 0.53–0.77). Corresponding values for PCL-R factor 1 and factor 2 were 0.58 (95% CI ¼ 0.46–0.71) and 0.66 (95% CI ¼ 0.54–0.78), respectively. Thus, PCL-R factor 2 essentially accounted for all of the association between PCL-R psychopathy and violent recidivism. 89
La˚ngstro¨m and Grann A separate analysis of the three-factor solution of the PCL-R (10) yielded no significant relationships between the affective (0.54, 95% CI ¼ 0.42–0.67), interpersonal (0.62, 95% CI ¼ 0.50–0.74) and behavioural factors (0.60, 95% CI ¼ 0.48–0.73), respectively, and violent recidivism within 2 years of time-at-risk. We also tested the associations between PCL-R psychopathy and intrainstitutional violent offending resulting in reconvictions during detainment. Among the 72 subjects that received sentences including incarceration or inpatient forensic psychiatric care, seven (10%) were reconvicted for a violent crime during the sentences following the index offence. Four of these subjects later recidivated within 24 months of time-at-risk following release or discharge. Areas under the curve in ROC analyses with a reconviction during detainment as the dependent measure were 0.69 (95% CI ¼ 0.55–0.83) for total PCL-R scores. Estimates for PCL-R factor 1 and factor 2 were 0.61 (95% CI ¼ 0.46–0.76) and 0.72 (95% CI ¼ 0.58–0.86), respectively. A separate analysis of the three-factor solution of the PCL-R (10) yielded a significant association between the behavioural factor (0.76, 95% CI ¼ 0.63–0.90), but not the affective (0.68, 95% CI ¼ 0.50–0.85) and interpersonal (0.59, 95% CI ¼ 0.42–0.75) factors, and a reconviction during detainment. Due to low statistical power, we did not further study the independent effects of PCL-R psychopathy and other risk factors on intrainstitutional violent offending in multivariate analyses. Crude ORs for associations between PCL-R psychopathy and 13 other risk factors, potential confounders of the relation between psychopathy and violent recidivism, are presented in the leftmost part of Table 1. For the continuous predictor variables, the ORs are to be interpreted as the risk increase (>1) or decrease (
