Data were obtained from the longitudinal “Swiss Study for Clarification and Goal-Attainment in Youth Welfare and Juvenile Justice Institutions” (Modellversuch Abklärung und Zielerreichung in stationären Massnahmen [MAZ.]), which was conducted between 2007 and 2012. The aim of the MAZ. study was to examine mental health, psychosocial problems, and delinquent behavior of children, adolescents, and young adults in residential child welfare and juvenile justice institutions throughout Switzerland [35]. All institutions accredited by the Swiss Federal Ministry of Justice were invited to participate, of which 64 institutions (35%) agreed to participate (20 institutions in the French-speaking, 38 in the German-speaking, and 6 in the Italian-speaking part of Switzerland). Youth were admitted to these institutions through criminal law, civil law, or voluntary placement. Youth who had lived in the facility for more than 1 month prior to the assessment and who were able to complete the French, German, or Italian assessment instruments (sufficient language skills and IQ > 70) were asked to participate. Assessments consisted of clinical interviews conducted by trained psychologists, computerized self-report measures, and ratings by institutional social workers. Prior to participation, the youth, their legal guardians, and social caseworkers received verbal and written information about the study and were asked to provide informed consent. The study procedure was approved by the Ethics Committees for Research Involving Human Subjects of the Universities of Basel and Lausanne (Switzerland) and by the Institutional Review Board of the University of Ulm (Germany). For further details on the methodology, see Schmid et al. [35].

For the current paper, data were obtained from 136 adolescents and young adults aged 12–25 years (Mage = 16.41; SD = 2.30; 80.2% < 18 years). The sample consisted of 66.2% (n = 90) males and 33.8% (n = 46) females (See Table 1). All participants were examined for mental disorders using the Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present and Lifetime Version (K-SADS-PL) [26] and completed the Youth Psychopathic traits Inventory (YPI) [7], from which we derived the 15 items for the LPE specifier in accordance with Jambroes et al. [24]. In addition, participants completed the Youth Self-Report (YSR) [1] to assess internalizing and externalizing mental health problems, and were assessed for Personality Disorders (PDs) using the Structured Clinical Interview for DSM-IV-TR Axis II Personality Disorders (SCID-II) [18]. Finally, information on officially recorded convictions up to the end of 2017 was obtained from the Swiss Federal Statistical Office.

Schedule for Affective Disorders and Schizophrenia for School-Age Children—Present and Lifetime Version (K-SADS-PL) [26]. The K-SADS-PL is a standardized, semi-structured clinical interview for the assessment of mental disorders in children and adolescents aged 6–18 years according to the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) [4]. Individual responses are scored on a 4-point Likert scale (0 = no information available, 1 = not present, 2 = subthreshold level, 3 = threshold level). The psychometric properties of the K-SADS- PL have been found to be good [25]. For the current study, only CD diagnoses were included.

Youth Psychopathic traits Inventory (YPI) [7]. The YPI is a 50-item self-report questionnaire designed to assess the core personality traits of psychopathy in adolescents. The YPI was developed in accordance with a three- dimensional conceptualization of psychopathy [15]: an arrogant and deceitful interpersonal style (Grandiose-Manipulative Dimension), an inadequate affective experience (Callous-Unemotional Dimension), and an impulsive and irresponsible behavioral style (Impulsive-Irresponsible Dimension). The Grandiose-Manipulative dimension has four subscales: Dishonest Charm, Grandiosity, Lying, and Manipulation; the Callous-Unemotional dimension has three subscales: Callousness, Unemotional, and Remorselessness; the Impulsive- Irresponsible dimension has three subscales: Impulsiveness, Irresponsibility, and Thrill Seeking. Items are scored on a 4-point Likert-type scale ranging from 1 = does not apply at all to 4 = applies very well [7]. The psychometric properties of the instrument are generally good. A more detailed overview can be found in Boonmann et al. [9].

Only the CU dimension of the YPI was used to assess the LPE specifier; the subscale remorselessness (five items) was used to assess the “lack of remorse or guilt” LPE specifier criterion, the subscale callousness (five items) was used to assess the “callous–lack of empathy” LPE specifier criterion, and the subscale unemotional (five items) was used to assess the “shallow or deficient affect” LPE specifier criterion. The “unconcerned about performance” LPE specifier criterion cannot be assessed with the YPI, meaning that only three of the four LPE specifier criteria can be assessed with this instrument. Importantly, previous research has suggested that this criterion was not critical to test the utility of the specifier, and that the YPI can be used as an indicator of the LPE specifier [12]. If at least one item on a subscale was rated “applies very well”, the LPE specifier criterion was met. If at least two LPE specifier criteria were present, the youth was assigned to the LPE group (consistent with the methodological preceding of [24].

Next to the categorical classification of the LPE specifier, we build an LPE dimensional score summing up all items related to the dimensions of LPE, resulting in a score ranging from 15 to 60 (This score is in fact the same as the CU score).

Youth Self-Report (YSR) [1] / Child Behavior Checklist (CBLC) [2]. The YSR is a self-report questionnaire and the CBCL a third-party questionnaire both designed to assess internalizing and externalizing mental health problems. The questionnaires list approximately 120 behavioral and emotional difficulties commonly found in children and adolescents. Items are rated on a 3-point Likert scale (0 = does not apply to, 1 = somewhat or sometimes applies, 2 = very true or often applies). The YSR/CBCL provide three broadband scales: total problems (TOT), internalizing problems (INT), externalizing problems (EXT). The psychometric properties of the instruments were found to be good [3].

Structured Clinical Interview for DSM-IV-TR Axis II Personality Disorders (SCID-II) [18]. The SCID-II is a semi-structured interview designed to assess DSM-IV and DSM-IV-TR PD diagnoses (i.e., paranoid, schizoid, schizotypal, histrionic, borderline, antisocial, narcissistic, avoidant, dependent, obsessive–compulsive, depressive, and passive-aggressive PDs). The interview consists of 134 items rated on a 3-point Likert scale (1 = absent, 2 = subthreshold, and 3 = threshold). Categorical diagnoses are provided according to the specific diagnostic thresholds of the PDs.

Conviction data Conviction data (both juvenile and adult conviction data) were obtained from the Swiss Federal Statistical Office until the end of 2017, up to 10 years after the initial assessment of the study. In accordance with the publications of the Swiss Federal Statistical Office, we assessed convictions for the two most serious types of offenses (felonies, misdemeanors), excluding the most minor category of offenses (contraventions). Violent crimes were classified according to the definitions used by the Federal Statistical Office and included all crimes involving actual or threatened harm to persons, such as all forms of assault, robbery, or coercion.

First, we present descriptive statistics for various sociodemographic characteristics, placement-specific variables, and prior delinquency for the total sample, as well as for those with and without the LPE specifier. Differences were examined using χ2 tests for categorical variables and t-tests and Wilcoxon-tests for dimensional variables. Wilcoxon tests were used as nonparametric alternative when assumptions for t-tests were violated. Second, we tested for differences in self-reported and professional caregiver-reported psychopathology (total, internalizing, and externalizing) and in the prevalence of PDs between participants with CD with and without the LPE specifier using t- or χ2 tests. Third, we used hierarchical logistic regression analyses to model general and violent offending (binary), including first the categorical or the dimensional LPE specifier/score as a predictor, then age and gender, and finally prior offending as covariates. Fourth, we fitted hierarchical negative binomial regression analyses to model the number of total and violent offenses committed, including first the categorical or the dimensional LPE specifier/score as a predictor, then additionally age and gender, and finally also prior offences as covariates. Fifth, we used Cox proportional hazards regression to model the time to subsequent general and violent offending, again including first the categorical or the dimensional LPE specifier/score as a predictor, then age and gender, and finally prior offences as covariates. The time scale of the Cox model represented the time in years from the first measurement point to the first general or violent offense or to the last measurement point in which no offense was committed (right censoring).

The statistical software used was R (Version 4.2.2) via RStudio (Version 2022.12.0, Boston, MA, USA). Descriptive analyses and model performance were analyzed using the “easystats” ecosystem for R [28,29,30]. Regression models were based on complete case analyses, as no missing data were identified for these study variables of interest was apparent. For sociodemographic and mental health data, with missing data, we used complete case analyses assuming Missingness at Random and report the exact number of cases included for each of these analyses. All p-values were two-tailed, and p-values < 0.05 were marked as statistically significant, exact p-values for all tests are reported in the tables.