Prediction of Youth and Adult Recidivism Among Spanish Juveniles Involved in Serious Offenses

Abstract

This study analyzes the predictive validity of the Youth Level of Service/Case Management Inventory (YLS/CMI) for youth and adult recidivism in a Spanish juvenile sample. Participants’ age ranged between 14 and 18.09 years old (N = 264) and 82% were boys and all had been sentenced to probation and custody centers. Data on juvenile and young adult recidivism were collected for the sample with mean follow-up periods of 13.74 and 20.19 months, respectively. The area under the curve, Kaplan–Meier and Cox regression survival analyses were each conducted to check for predictive validity. The findings demonstrated that the YLS/CMI is able to predict recidivism in both the juvenile period and the emerging adult period in a different cultural context. Prior Offenses and Education/Employment emerged as significant predictors for youth and young adult recidivism. The entire YLS/CMI is therefore an effective tool for risk classification in a different cultural sample.

Keywords

YLS/CMI prediction recidivism juvenile risk factor

Introduction

Risk assessment has been defined as the evaluation of how the characteristics of both the youth and the situation relate to a relevant outcome (Bonta & Andrews, 2017), in this case, future youth reoffending. An accurate risk assessment in youth recidivism has become a major priority in criminological research, and serves important functions, mainly the promotion of public safety, assistance in sentencing decisions, recommendations for services, and targets for treatment (Olver et al., 2009). Moreover, youth risk assessment is crucial, as it is the baseline for analyzing the possible continuity between juvenile and adult crime (Basto-Pereira et al., 2015; Vincent et al., 2019). The use of instruments with long-standing empirical support and proven predictive value is therefore essential (Guy et al., 2012; Henggeler, 2004; Taxman, 2018).

The Youth Level of Service/Case Management Inventory (YLS/CMI) is one of the most widely used risk assessment instruments for predicting recidivism for youth. It is a Fourth-Generation risk assessment that emphasizes the link between assessment and case management. There have been many field studies concerning the reliability and predictive validity of the YLS/CMI (e.g., Olver et al., 2009; Schmidt et al., 2005). In this context, predictive validity can be considered the ability of risk assessment tool score to correctly assess the likelihood of reoffending (Singh, 2013). Various studies have effectively demonstrated that this instrument accurately evaluates the likelihood of reoffending and is one of the best options for assessing the risk of recidivism among minors (Gendreau et al., 1996; Olver et al., 2014). Moreover, the results of different studies have shown that the YLS/CMI works equally well for different subgroups, including the type of recidivism (e.g., violent/nonviolent), recidivism outcome (e.g., how the variable was measured; number of new files; presence/absence of recidivism), country, and sample characteristics (e.g., high-risk minors; Onifade et al., 2008; Pusch & Holtfreter, 2017; Thompson & McGrath, 2012). In summary, this Inventory was selected among other predictive instruments mainly due to its cutoff scores which allow to classify the minor in a range of risk recidivism levels, the importance given to dynamic risk factors in the assessment and its good reliability and predictive validity values.

In addition, the YLS/CMI measures the eight dimensions of risk factors, which remain useful to understand, predict, and prevent recidivism (Bonta & Andrews, 2017; Campbell et al., 2019; Flores et al., 2004; Rennie & Dolan, 2010). These eight dimensions are as follows: Antisocial Attitudes, Antisocial Friendships, an Antisocial Personality Pattern, a History of Previous Offenses, Deficient Family Circumstances, Education and Employment, Substance Abuse, and free time for Leisure and Recreation. The first four dimensions are called “the Big Four,” given their stronger predictive power in relation to recidivism in comparison to the rest of factors. Nonetheless, not all the dimensions are equally predictive in all type of populations (Bonta & Andrews, 2017). In this regard, researchers must take into account the social context and culture when analyzing predicting factors of recidivism. The purpose of this study is to examine the predictive validity of the YLS/CMI Inventory in Spain.

The Predictive Validity of the YLS/CMI in Spanish Populations

Few studies have analyzed the validity of the YLS/CMI in populations outside the Anglo-Saxon countries. Nonetheless, these few studies showed strong predictive validity for the YLS/CMI according to Rice and Harris’s (2005) classification in Asian (Benuto et al., 2014; Chu et al., 2015; Mori et al., 2017), Spanish (Cuervo & Villanueva, 2015, 2018), Portuguese (Pimentel et al., 2015), or Dutch populations (De Ruiter & Hildebrand, 2009). Nevertheless, some specific dimensions, such as Prior Criminal History and Substance Abuse, lacked content representativeness and predictive validity and the risk score means were lower than those on Anglo-Saxon countries (Takahashi et al., 2013). Cultural and legal differences may account for the low content representativeness and predictive validity.

There are relevant legal and cultural differences between the Spanish and Anglo-Saxon cultures that may affect the validity of the YLS/CMI, such as the different legal systems and the collectivism versus individualism cultural differences. For instance, according to the Spanish legal system, minors between 14 and 17 years old can be judged under the juvenile system, and subsequently assigned a wide range of educational measures (Cuervo & Villanueva, 2015; Ortega-Campos et al., 2017). Educational measures are sanctions received by youth according to their risk level. These measures include different levels of rights and freedom restriction, all of which are included in Spanish Criminal Law for minors. In this context, the use of risk assessment tools such as the YLS/CMI helps to assign the most appropriate intervention to each minor. Meanwhile, the juvenile age of jurisdiction in the majority of states in the United States is age 17. Only four states set the age of juvenile jurisdiction at age 16.

With regard to cultural and social differences, the collectivism versus individualism dimension also shows interesting nuances between the Spanish and Anglo-Saxon cultures. In a study carried out by Hofstede (1980), 40 countries presented an index value on the individualism/collectivism dimension. Lower scores on this index were indicative of collectivism, while higher scores were indicative of individualism. This is one of the dimensions that try to explain the national culture of a country, that is, the national character. Whereas individualism refers to individuals taking care of themselves and their immediate families only; in collectivism, the emphasis is on belonging to clans or organizations which protect individuals in exchange for loyalty (Hofstede et al., 2010; Markus & Kitayama, 1991). In this sense, Spain (classification: 51; Scale: 0–100, according to Hofstede, 1980) is far more collectivist than the United States of America (classification: 90) or Canada (classification: 80), hence Spaniards are more likely to take into account the group decisions or the importance of the social context. Those cultural factors might have an impact on the dimensions predicting recidivism. For instance, in Spain, the factors that emerge as the most discriminative for recidivism prediction were mostly collectivists, such as Education/Employment, Leisure/Recreation (Cuervo & Villanueva, 2015), and falling outside of the “Big Four” (Bonta & Andrews, 2017; Papp et al., 2019). Despite these differences, adequate levels of recidivism prediction for the YLS/CMI have been reported in the few studies of Spanish populations that have been conducted, with area under a receiver operating characteristic (ROC) curve (AUC) values ranging from .64 to .83 (Cuervo & Villanueva, 2015; López et al., 2016), albeit with an abridged version of the YLS/CMI (Cuervo & Villanueva, 2018). However, these studies included general populations in contact with the Criminal Justice System rather than focusing solely on serious trajectories.

Juveniles committing serious crimes are minors obtaining the highest risk levels of recidivism in risk assessment tools and therefore assigned the more restrictive educational measures in the Spanish system. Probation and custody centers are the most restrictive common educational measures in the Spanish system. Although a minority in the judicial system, these types of youth tend to have long crime trajectories and higher rates of recidivism, and, hence, tend to commit most of the crimes (Moffitt, 1993, 2007). For this reason, it is crucial to focus the research on this target group. Moreover, these studies did not include any follow-up period into emerging adulthood. Both aspects will be addressed in this study using a Spanish population.

The Predictive Validity of the YLS/CMI Into Emerging Adulthood

According to several authors (Basto-Pereira et al., 2015; Brame et al., 2018; Mowen & Boman, 2018; Schmidt et al., 2011), a comprehensive assessment of juvenile risk can predict not only juvenile recidivism but also criminal persistence into emerging adulthood. In fact, developmental research has shown that several outcomes (e.g., risk-taking behavior, impulsivity, or emotion regulation) of youth up to an age of 25 are more similar to younger adolescents than to adults (Arnett, 2000; Farrington et al., 2012). That is, there is a number of skills that continue to mature through the mid-20s, making possible a continuous valid prediction of youth risk assessments into emerging adulthood. Some studies with long follow-up periods (Olver et al., 2012; Schmidt et al., 2011) found that the YLS/CMI had moderate-to-high predictive validity for youth and young adult recidivism, with AUC values ranging from 0.66 to 0.77 (Olver et al., 2012). The specific risk factors that predicted adult recidivism were Prior offenses, Peers, and Personality (Cox et al., 2016; Olver et al., 2012). Again, these studies were carried out in Anglo-Saxon countries, mainly Canada and the United States.

Although the YLS/CMI is designed to predict general criminality, it was also predictive of both general and violent recidivism (Catchpole & Gretton, 2003). In other words, no differences were found between violent and nonviolent recidivism predictions (Schmidt et al., 2011). This similar predictive validity was also observed in high-risk youth populations (Olver et al., 2009). Nonetheless, Kalvin and Bierman (2017) found that self-reported peer deviancy is highly predictive of nonviolent adult crime but not for violent crime during young adulthood. This study included only two risk factors (deviant peer affiliation and parental detachment), and not the wide range of criminogenic needs assumed in a risk inventory.

Thus, there is a need to study the predictive validity of risk assessment tools after the age of 18. Moreover, studies of adult recidivism do not usually take the minor’s level of risk into account; a valid and reliable inventory for predicting the level of risk, the YLS/CMI (Hoge & Andrews, 2006), is therefore applied to all participants in this study. Finally, most studies on this topic have been carried out in the United States and Northern Europe. Consequently, as the cultural context must be taken into account when predicting recidivism, more studies with Spanish youth involved in the Justice system are needed. All these aspects, together with the inclusion of juveniles committing serious crimes and a long follow-up period for recidivism, are the original contributions of this study.

Aim and Hypotheses

The main aim of this study is to determine the validity of YLS/CMI to predict different types of recidivism in a sample of high-risk Spanish adolescents involved in the Justice system. Thus we expect the following: (a) The YLS/CMI will present an adequate predictive validity in Spanish juveniles committing serious crimes for youth and adult outcomes; (b) The YLS/CMI will not present prediction differences between violent and nonviolent youth recidivism; (c) Youth risk factors would present similar values when predicting youth recidivism than when predicting young adult recidivism; and (d) The type of risk factors would be the same for violent and nonviolent recidivism.

Method

Participants

All youth who had served probation (76.9%) or been confined to a juvenile detention center (23.1%), in a Spanish province between 2009 and 2015, participated in the study (N = 264). Both are the most restrictive common educational measures in the Spanish system. Custody centers respond to the peculiarly serious nature of the acts committed. Offenses committed by these juveniles are characterized by violence, intimidation, or danger to people. The priority objective of custody centers is to confine the minor to a controlled environment that provides the appropriate educational conditions so that they can redirect those deficiencies that have characterized their behavior. In this study, 83% (n = 51) of the youth assigned a custody center, engaged in external activities such as leisure activities, academic qualifications, and so forth (semi-open modality). The rest were in close confinement with no external activities (16.4%, n = 10). No significant differences between both groups could be found for youth or adult recidivism rates (p = .11; p = .06, respectively), in Mann–Whitney independent tests.

Meanwhile, probation is one of the most commonly applied penal sanctions in juvenile justice in Spain (40%), according to the Spanish Statistics Institute (INE, 2018). In this measure, the minor is subject to supervision by specialized personnel, to acquire the skills, abilities, and attitudes necessary for a correct personal and social development. During the time of probation, the minor must also comply with the obligations and prohibitions that the judge can impose.

The youth ages ranged from 14 to 18.09 years old, with a mean of 16.49 years (SD = 1.00); 216 were boys (82.40%). They all had Spanish nationality. Most of them were born in Spain from Spanish parents (74.63%), and the rest were second- and later-generation Spanish, that is, born in Spain or moved there in very early childhood or preadolescence, accompanied by their families who were not Spanish (25.37%). In the latter group, their origins were the following ones: 16.53% Romanian and Russian, 6.34% North African, and 2.43% from other countries in Europe. This demographic distribution matches the one shown by the Spanish population (INE, 2019). All participants spoke regular Spanish, so no difficulties were found in this sense. Data on juvenile and young adult recidivism were collected with mean follow-up periods of 13.74 and 20.19 months, respectively. The youth ages ranged from 18.03 to 23.01 years old during the young adult follow-up period.

Instrument

The YLS/CMI by Hoge and Andrews (2006), translated into Spanish by Garrido et al. (2006), is an instrument used to evaluate the risk of a youth reoffending. The information needed to complete the YLS/CMI is compiled from several sources, including an interview with the youth and their family, their previous charges, social services involvement, educational institutions attended, and so on. This YLS/CMI consists of 42 items grouped into eight risk factors. In each factor, the Youth Offending Team marks the risk items that can be applied to the juvenile (1 = presence; 0 = absence). Each factor has between three and seven items. The factors included in the questionnaire are (a) Prior and Current Offenses and Dispositions; (b) Family Circumstances/Parenting; (c) Education/Employment; (d) Peer Relations; (e) Substance Abuse; (f) Leisure/Recreation; (g) Personality/Behavior; and (h) Attitudes, Values and Beliefs. The total score obtained for the risk of recidivism can be classified in several ranges: low (0–8 points), moderate (9–22), high (23–32), and very high (33–42 points). The Youth Offending Team decides on the kind of educational measure(s) that should be assigned to the young person according to their overall risk level obtained from the YLS/CMI.

According to Hoge and Andrews (2006), the original YLS/CMI shows adequate psychometric properties: α total YLS/CMI = .91 and α area ranging from .56 to .82. The Spanish version of the YLS/CMI obtained similar values, with α values ranging between .62 and .80, with the exception of the prior and current offenses factor, which was .48 (Cuervo & Villanueva, 2015), and .91 (Cuervo et al., 2017) for all the items in the YLS/CMI. The low value of the Prior and current offenses factor may be related to the difference of the original legal context of the Inventory and the legal system of Spain. In our system, the common procedure is to be prosecuted for one conviction at a time and therefore it is unlikely that these items can be marked. As a result, there are very few minors who score on this factor. The total Cronbach’s alpha in this study was .81.

Procedure

Juvenile data were extracted from an analysis of the records of the Juvenile Court of a Spanish province. In Spain, minors must be at least 14 years old to be charged under the juvenile legal system. According to the Organic Law 5/2000, Spain has a specialized system for juveniles. The juvenile’s age and the type of offense determine the Jurisdiction of the Juvenile court. The juvenile’s best interest is the premise guiding the juvenile justice system, addressed by this Law. The educational measure to be chosen has to be guided by this principle and any intervention has to be proportional to the seriousness of the offense that the juvenile has committed.

The Youth Offending Team of the Juvenile Court assesses the minor when she or her is charged with committing a crime. These professionals interview both the minor themselves and his or her legal representatives about the individual, educational, family, and social aspects in the youth’s environment. The legal representatives were parents or close relatives in most cases. However, some juveniles may be under the protection system and may be fostered by social work professionals. The YLS/CMI is scored based on this information, providing a recidivism risk for each subject.

The individual interviews to obtain both a profile of the young person and the information needed to complete the YLS/CMI were carried out by the Justice Department in the offices of the Juvenile Court’s Technical Team. The interviews took place at the Juvenile Court around 3 to 6 months after the charge. Interrater reliability data on the Inventory were not available for the present study, although all the members of the Youth Offending Team must complete a formation process. They were trained for 1 month in the use of the YLS/CMI Inventory, by an expert on the instrument. First assessments were carried out with continuous supervision by this expert who made sure that the criteria were common in practice.

The criterion used to classify a minor as a juvenile reoffender was as follows: Any juvenile who after being sentenced (baseline offense) is charged with another offense within the follow-up period (until the age of 18). This period ranged from 0 to 46.16 months, with a mean of 13.74. The period of 0 days is due to one juvenile who committed two offenses the same day (so the follow-up period according to the criteria is 0). The first reoffending was registered in 2011, and the last in 2015, hence the period for reoffending ranged from 2011 to 2015. However, these new juvenile offenses only included sentences of probation and confinement in custody centers. Other less severe sentences, such as community service and reprimands, could not be taken into account for youth recidivism due to limited access to data.

Nonviolent youth recidivism was defined as any new criminal conviction for a nonviolent offense (e.g., theft, drug offenses). Violent recidivism was defined as any new criminal conviction for a violent offense (e.g., assault, robbery, murder, sexual offenses). Although no classification of violent/nonviolent crime is proposed by the legal system in Spain, the charges were classified based on previous studies (Cuervo et al., 2018; Junger-Tas et al., 2010). The data for the adult criminal proceedings were collected at the Offices of the Coordinator of Administrative Justice in a Spanish province, between the youth’s age of majority (18 years old in Spain) and November 2016, when the data were extracted. Different follow-up periods were therefore obtained for each juvenile, ranging from 0.39 to 62.65 months (M = 0.19). This depended on the youth’s age in the study period, which was the age when the baseline offense was committed. Unfortunately, it was impossible to gather any information about the type of crime (violent/nonviolent) for adult recidivism. The Office of the Coordinator of Administrative Justice provided us with the overall number of adult proceedings, but not the specific details (type of crime, duration, type of penalty, etc.).

The outcome variables for youth and adult recidivism were measured in two different ways: dichotomously (reoffender/non-reoffender) and quantitatively (number of subsequent charges). Juvenile and adult recidivism periods were calculated as mutually exclusive. Only minors who recidivated in the juvenile period were considered for juvenile recidivism and individuals who recidivated in the adult period were counted for the adult period. Hence, offenses were measured as absent or present in every juvenile (0/1; even if more than one offense took place). Total recidivism gathers the presence of any reoffending in the juvenile period, in the adult period, or in both periods.

Data Analyses

The predictive validity of the YLS/CMI was tested for various recidivism outcomes using ROC analyses. The AUC is an index for interpreting the overall predictive validity that can range from 0 (perfect negative prediction) to 1 (perfect prediction) (e.g., Rice & Harris, 1995). Kaplan–Meier survival analysis and Cox regression survival analysis were then conducted to check for predictive validity (Kleinbaum & Klein, 2005). The objective of the Kaplan–Meier survival analysis was to compare the juvenile and adult recidivism rates, scoring on the four YLS/CMI risk levels (i.e., low, moderate, high, and very high) while controlling for individual differences in the follow-up period. Cox regression survival analysis was also performed to examine the relative predictive validity of the eight YLS/CMI subscales. This method calculates survival probabilities (i.e., the likelihood of the juvenile remaining offense-free during the follow-up period). All the analyses were subsequently repeated for both the total score and for the YLS/CMI subscales. A life table analysis was also carried out. A life table is a concise way of displaying the probability of a subject taking part in a particular event. In this study, the life tables examine changes in recidivism during different periods of time.

Results

The mean number of months between the baseline conviction and the first reoffense in the juvenile system was 13.74 months (SD = 11.15; range = 0–46.16 months). The mean time until the first criminal proceedings in the adult system was 20.19 (SD = 15.18; range = 1–62.65 months). The recidivism rates were 27.27% for juvenile recidivism, 25.51% for adult recidivism, and 44.69% for total recidivism (juvenile and adult).

Table 1 shows recidivism among both minors and adults for each risk level, as measured by the YLS/CMI. As there were only two subjects in the very-high-risk group, the high and very high groups have been collapsed (high-risk group, from this point forward). As shown in Table 1, the distribution over the two different follow-up periods (juvenile and adult) is remarkably similar when the subject reoffends: 26.46% of minors recidivated in the juvenile period and 26.01% in the adult period. Only 24 individuals (10.76%) recidivated in both periods. Furthermore, the proportion of minors reoffending increases as the risk level increases.

Table 1:

Distribution in Percentages of Each Risk Level According to the Type of Recidivism

	Juvenile recidivist	Adult recidivist	Total recidivism
Risk level	% (n)	% (n)	% (n)
Low (37)	2.70 (1)	10.81 (4)	13.51 (5)
Moderate (154)	29.49 (46)	26.28 (41)	45.51 (71)
High (30)	40 (12)	43.33 (13)	66.67 (20)
Total (223)	26.46 (59)	26.01 (58)	43.05 (96)
χ²(N = 223)	14.29**	9.12*	20.37**

Note. High-risk-level and very-high-risk-level groups have been collapsed.

p ≤ .05. **p ≤ .001.

The distribution of juveniles and adults in each level of juvenile reoffending is almost the same in the three risk groups. This means that the validity of predictions in the juvenile and adult periods was very similar. Overall, 66.67% of minors classified as high risk recidivated in either the juvenile period or the adult period. It seems that those classified as moderate or high risk who did not reoffend in the juvenile period did so in the adult period.

Predictive Validity of YLS/CMI Total Scores in Youth and Adult Recidivism

ROC Analyses

The predictive validity (AUC) of the YLS/CMI was examined across various recidivism outcomes (Table 2). The YLS/CMI demonstrated moderate-to high predictive validity for all the recidivism outcomes (youth, youth nonviolent, youth violent, adult, and total recidivism), with AUC values ranging from .67 to .72. The YLS/CMI prediction was more accurate for total recidivism (youth and adults combined). There was little difference between the predictions for youth and adult recidivism. The lowest level of validity was observed in the group of minors who recidivated in violent offenses.

Table 2:

Predictive Validity of the YLS/CMI: Overall Sample (n = 223)

Type of recidivism	Base rate (%)	AUC	95% CI
Youth recidivism	26.46	.69**	[0.61, 0.76]
Nonviolent	14.77	.68**	[0.59, 0.76]
Violent	10.98	.67*	[0.57, 0.76]
Adult recidivism	26.01	.68**	[0.60, 0.76]
Total recidivism	43.05	.72**	[0.65, 0.78]

Note. YLS/CMI = Youth Level of Service/Case Management Inventory; AUC = area under a receiver operating characteristic curve; CI = confidence interval.

p ≤ .05. **p ≤ .001.

Survival Time Analyses

The following life table (Table 3) is divided based on the risk levels of the YLS/CMI, according to intervals of juvenile recidivism. “Number entering interval” is the number of surviving cases (nonrecidivist) at the beginning of the interval. “Number withdrawing during interval” is the number of censored cases (i.e., minors who have not reoffended). “Proportion termining” is the number of cases exposed to risk. The minors that scored as high risk in the first interval (12 months) had the highest probability of recidivating in the juvenile period (34%). They are followed by minors in the moderate risk in the third year of the follow-up (36 months; 29%), and minors classified as high risk from 12 to 24 months (27%). The last interval of the moderate risk level and the first two intervals of the high risk level are the periods with the highest probability of recidivism.

Table 3:

Life Table for Risk Level and Months

Risk level	Interval in months	Number entering interval	Number withdrawing during interval	Proportion termining
Low	0	33	12	.00
	12	21	13	.07
	24	7	4	.00
	36	3	3	.00
Moderate	0	146	39	.21
	12	80	41	.15
	24	30	23	.05
	36	6	5	.29
High	0	28	9	.34
	12	11	7	.27
	24	2	2	.00

Kaplan–Meier survival analyses examine the relative rates of the four YLS/CMI risk groups for juvenile and adult recidivism. Survival time was measured in months between offenses. The follow-up period was considered from the first offense to the second, and censored data were considered until the age of 18 (N = 252, M = 13.74 months). The adult follow-up period was considered from the age of 18 to the first offense after that age. Censored data are considered until November 1, 2016 (N = 263, M = 20.19 months).

The juvenile recidivism is shown in Figure 1, where each line represents each YLS/CMI risk level. The Kaplan–Meier survival analyses present higher and faster rates of reoffending as the risk level increases. The same applies to Figure 2 for adult recidivism. Log-rank tests for the juvenile period showed significant pairwise differences between minors in the low and moderate groups according to the time of reoffending—χ²(n = 207) = 7.11, p = .008—and between the low- and high-risk groups—χ²(n = 207) = 12.41, p < .001. Recidivism rates were higher for the groups at higher risk of recidivism (Figure 1).

Figure 1:

Kaplan–Meier Survival Curves: Minor Recidivism Rates as a Function of the YLS/CMI

Figure 2:

Kaplan–Meier Survival Curves: Adult Recidivism Rates as a Function of the YLS/CMI Risk Level

The results of the Kaplan–Meier survival analyses with adult recidivism are presented in Figure 2. Log-rank tests did not show significant differences between the low and moderate groups—χ²(n = 217) = 2.90, p = .089—and did show it between the low- and high-risk groups—χ²(n = 217) = 5.95, p = .015.

There were no significant differences using two-group analysis of variance (ANOVA) analysis between ages of the minors and risk level that could bias the follow-up periods. Mean age of juveniles in each risk level was 16.53 (low risk), 16.38 (moderate risk), and 16.63 (high risk), F(2, 215) = .93, p = .394. Moreover, the differences between the follow-up period were also analyzed within the risk levels. There were significant differences between risk level and months of the follow-up in the juvenile group, but not in the adult group. The mean number of months according to risk level were: 16.94 (low risk), 14.20 (moderate risk), and 9.77 (high risk), F(2, 215) = 3.34, p = .037. Mean number of months according to risk level were: 21.47 (low risk), 17.80 (moderate risk), and 21.77 (high risk), F(2, 219) = 1.50, p = .224. Finally, when checking the differences between follow-ups and recidivism, significant differences were obtained. Juvenile recidivists had shorter follow-up periods (10.45) than nonrecidivists (14.95), F(1, 250) = 8.32, p = .004. Something similar occurred in adults. The follow-up period was shorter for recidivists (16.47) than nonrecidivists (21.59), F(1, 261) = 6.06, p = .014. These results support the validity of prediction analyses, since even when recidivists had shorter follow-up periods, recidivism was significantly predicted.

Finally, the Cox regression survival analyses in Table 4 show that the YLS/CMI is able to predict all types of recidivism (juvenile, violent juvenile, nonviolent juvenile, and adult recidivism). Gender was not related to recidivism, and age was negatively related to total juvenile recidivism and violent recidivism.

Table 4:

Cox Regression Survival Analyses

Type of recidivism	B	SE	Wald	df	Exp(B)	95% confidence interval
Juvenile recidivism
Total risk score	0.08**	0.02	16.34	1	1.08	[1.04, 1.13]
Sex	0.14	0.41	0.12	1	1.15	[0.51, 2.59]
Age	−0.34*	0.15	4.67	1	0.70	[0.51, 0.96]
Violent recidivism
Total risk score	0.09*	0.03	8.75	1	1.10	[1.03, 1.17]
Sex	−0.45	0.52	0.75	1	0.63	[0.23, 1.76]
Age	−0.61**	0.21	8.19	1	0.54	[0.35, 0.82]
Nonviolent recidivism
Total risk score	0.07**	0.02	8.40	1	1.08	[1.02, 1.13]
Sex	1.81	1.02	3.15	1	0.82	[0.82, 45.07]
Age	−0.17	0.21	0.63	1	0.55	[0.55, 1.28]
Adult recidivism
Total risk score	0.07**	0.02	13.736	1	1.08	[1.03, 1.12]
Sex	0.68	0.47	2.045	1	1.97	[0.77, 4.98]
Age	0.00	0.159	0.00	1	0.99	[0.72, 1.35]

p ≤ .05. **p ≤ .001.

Predictive Validity of YLS/CMI Subscales

Descriptive statistics are provided in Table 5 using independent sample t tests with the Bonferroni correction. The t test analyses have been conducted to check the differences between the juveniles who recidivate and the ones who did not in each type of reoffending (violent, nonviolent, youth, adult, and total recidivism). These analyses have been realized for each area of the YLS/CMI. In general, subjects who reoffended had higher mean values than subjects who had not reoffended. However, only particular areas were significant, which are marked on the first column of each type of recidivism. There were significant differences related to violent crime in the area of personality and in the total score of the YLS/CMI. Minors who reoffended for nonviolent crime showed higher mean values for leisure than those who did not reoffend for this type of crime. Significant differences were found in the scores for Education/Employment, Leisure/Recreation, and in the total score between minors who reoffend in the juvenile period and those who did not. There were also differences in both Prior offenses and Leisure/recreation in adults, and total recidivism.

Table 5:

Means and Standard Deviations of the YLS/CMI Subscales for All Types of Recidivism

	Violent recidivists (n = 23)	Violent nonrecidivists (n = 199)	Nonviolent recidivists (n = 32)	Nonviolent nonrecidivists (n = 190)	Youth recidivists (n = 58)	Youth nonrecidivists (n = 164)	Adult recidivists (n = 57)	Adult nonrecidivists (n = 165)	Total recidivists (n = 95)	Total nonrecidivists (n = 127)
Risk subscales	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)
Prior offenses	0.65 (0.83)	0.73 (0.98)	1.22 (1.09)	0.64 (0.92)	0.97 (1.00)	0.64 (0.95)	1.19 (1.23)*	0.56 (0.80)	1.05 (1.12)*	0.48 (0.75)
Family/parenting	3.39 (1.40)	2.62 (1.59)	3.47 (1.48)	2.57 (1.57)	3.38 (1.43)	2.46 (1.57)	3.21 (1.65)	2.53 (1.52)	3.21 (1.53)	2.32 (1.52)
Education/employment	3.57 (0.94)	2.85 (1.42)	3.38 (1.15)	2.85 (1.42)	3.38 (1.10)*	2.76 (1.46)	3.54 (1.35)	2.71 (1.35)	3.45 (1.21)	2.53 (1.4)
Peers	2.74 (1.21)	2.34 (1.35)	3.03 (1.12)	2.27 (1.34)	2.93 (1.15)	2.19 (1.35)	2.81 (1.14)	2.24 (1.38)	2.86 (1.18)	2.02 (1.34)
Drugs/alcohol	1.35 (1.22)	0.98 (1.14)	1.28 (1.22)	0.98 (1.14)	1.29 (1.20)	0.93 (1.13)	1.33 (1.25)	0.92 (1.10)	1.27 (1.22)	0.83 (1.07)
Leisure/recreation	2.65 (0.57)	2.51 (0.84)	2.84 (0.36)*	2.47 (0.85)	2.76 (0.47)**	2.44 (0.89)	2.68 (0.71)*	2.47 (0.84)	2.69 (0.63)**	2.39 (0.91)
Personality/behavior	2.87 (1.18)*	1.91 (1.82)	2.41 (1.93)	1.95 (1.75)	2.47 (1.71)	1.85 (1.78)	2.40 (1.85)	1.88 (1.74)	2.36 (1.78)	1.76 (1.75)
Attitudes/orientation	1.61 (1.27)	1.13 (1.27)	1.69 (1.35)	1.09 (1.25)	1.62 (1.29)	1.02 (1.24)	1.51 (1.26)	1.07 (1.26)	1.55 (1.31)	0.91 (1.18)
Total score	18.83 (4.77)*	15.10 (7.03)	19.27 (5.99)	14.82 (6.87)	18.78 (5.56)*	14.29 (6.98)	18.67 (6.66)	14.36 (6.67)	18.45 (6.23)	13.24 (6.57)

Note. YLS/CMI = Youth Level of Service/Case Management Inventory.

p ≤ .05. **p ≤ .001.

ROC Analyses

The predictive validity of the YLS/CMI subscales was examined according to different recidivism outcomes (juvenile, violent juvenile, nonviolent juvenile, adult, and total recidivism) using ROC analyses (Table 6). The YLS/CMI subscales demonstrated moderate-to-high predictive validity for all recidivism outcomes, with AUC values ranging from .59 to .69. All the subscales were significant predictors of total recidivism. All areas except for Leisure/Recreation were significant predictors of juvenile recidivism. Furthermore, all the scales except for Leisure/Recreation and Personality/Behavior again appeared as significant predictors for adult recidivism; Family/Parenting, Education/Employment, and Personality/Behavior predicted violent recidivism, whereas Prior Offenses, Family/Parenting, Companions, and Attitudes/Orientation predicted nonviolent offenses.

Table 6:

Predictive Validity of YLS/CMI Subscales for Different Recidivism Out

	Juvenile recidivism		Violent recidivism		Nonviolent recidivism		Adult recidivism		Total recidivism
Risk subscales	AUC	95% CI	AUC	95% CI	AUC	95% CI	AUC	95% CI	AUC	95% CI
Prior offenses	.60*	[0.51, 0.68]	.49	[0.37, 0.61]	.66**	[0.56, 0.76]	.65***	[0.57, 0.74]	.65***	[0.58, 0.73]
Family/parenting	.67***	[0.59, 0.75]	.65*	[0.54, 0.76]	.66**	[0.56, 0.75]	.61*	[0.52, 0.70]	.66***	[0.58, 0.73]
Education/employment	.62**	[0.54, 0.69]	.64*	[0.54, 0.74]	.60	[0.50, 0.69]	.68***	[0.59, 0.76]	.69***	[0.62, 0.76]
Companions	.65***	[0.57, 0.73]	.58	[0.46, 0.69]	.66**	[0.56, 0.75]	.62**	[0.54, 0.70]	.66***	[0.59, 0.74]
Drugs/alcohol	.59*	[0.50, 0.67]	.59	[0.47, 0.71]	.57	[0.46, 0.68]	.59*	[0.50, 0.67]	.61***	[0.53, 0.68]
Leisure/recreation	.57	[0.49, 0.65]	.51	[0.40, 0.63]	.60	[0.51, 0.69]	.56	[0.48, 0.65]	.58*	[0.51, 0.66]
Personality/behavior	.60*	[0.52, 0.68]	.68**	[0.59, 0.76]	.56	[0.45, 0.67]	.57	[0.49, 0.66]	.59*	[0.52, 0.67]
Attitudes/orientation	.63**	[0.55, 0.71]	.61	[0.50, 0.73]	.62**	[0.52, 0.73]	.61*	[0.53, 0.69]	.64***	[0.56, 0.71]

Note. YLS/CMI = Youth Level of Service/Case Management Inventory; AUC = area under a receiver operating characteristic curve; CI = confidence interval.

p ≤ .05. **p ≤ .001.

Survival Time Analyses

Cox regression survival analyses were run on the eight YLS/CMI subscales for juvenile and adult recidivism. The eight factors were entered simultaneously to examine their unique relationships to recidivism. No subscale was significant in predicting youth recidivism. We found differences when the type of recidivism was taken into account: Prior offenses predicted nonviolent recidivism—Wald = 7.14; Exp(B) = 1.63; p = .008. However, no subscale significantly predicted violent crimes. Two subscales significantly predicted adult recidivism: Prior offenses—Wald = 4.87; Exp(B) = 1.37; p = .027—and Education/Employment—Wald = 6.45; Exp(B) = 1.41; p = .011.

Discussion

The main aim of this study was to analyze the predictive validity of the YLS/CMI for youth and adult recidivism in a sample of Spanish juveniles committing serious crimes who had served in probation and custody centers. The first hypothesis proposed that the YLS/CMI would demonstrate adequate predictive validity for both Spanish juveniles committing serious crimes and various recidivism outcomes (youth/adult/total). This hypothesis was broadly supported by the results obtained. The AUC values were nearly identical for the prediction of recidivism outcomes (youth/adult/total). When compared with previous studies, these AUC values are in the upper range of predictive validity (Schwalbe, 2007; Thompson & Pope, 2005). Particularly, these findings support the first hypothesis regarding the cultural context. That is, the YLS/CMI seems to be an adequate risk instrument in predicting recidivism among Spanish youth involved in the Juvenile Justice system. The AUC values, the total risk means, and the predictive models obtained in this study were in the adequate range, in comparison to previous studies carried out with the original YLS/CMI (Olver et al., 2014; Thompson & Pope, 2005) and with the general population of Spanish youth involved in the Justice system (Cuervo et al., 2017; Cuervo & Villanueva, 2015; López et al., 2016).

The YLS/CMI was similarly effective at predicting either youth or young adult recidivism; in fact, the highest predictive validity was for total recidivism, that is, juvenile and adult combined. These results are consistent with previous studies showing the predictive validity of the YLS/CMI, not only for youth recidivism but also for adult recidivism (Olver et al., 2012; Schmidt et al., 2011; Schwalbe, 2007).

The distribution of participants according to risk levels and reoffending (cross charts) was also similar in youth and adults. The YLS/CMI that captures the dynamic nature of juvenile risk factors also had adequate predictive validity for adult recidivism, showing that this kind of risk assessment can be a valid instrument for emerging young adults, and increasing in predictive value when follow-up periods into adulthood are taken into account. The classifications suggested that the YLS/CMI was accurate when classifying a juvenile as high risk. In other words, when this reoffending does not take place in the juvenile period, it tends to occur in the immediate adult period. This shows the crucial role of updated and frequent assessments in some decision-making processes, given the high predictive validity of this juvenile risk assessment into emerging adulthood.

The most critical period for recidivism was the first year after being classified as having a high risk of youth reoffending, according to the YLS/CMI (Capdevila et al., 2008; Schmidt et al., 2005). The analyses yielded higher and faster rates of reoffending as the risk level increased, the same pattern appeared in adult recidivism. For example, minors with a high risk presented the highest probability of reoffending in the first interval (12 months). They were followed by minors with the same risk level (12–24 months), and minors who scored a moderate risk (from 24 to 36 months). This validity was adequate even when recidivists had significantly shorter follow-up periods than nonrecidivists. This fact highlights the predictive power in high-risk youth, which are detected even with shorter follow-up periods. That is, the YLS/CMI showed a coherent pattern of recidivism by risk level. Similar results were found by Schmidt et al. (2005), in a follow-up study of youth lasting nearly 3 years. As the risk level increased, the time until reoffending decreased.

The second hypothesis focused on the prediction of violent/nonviolent recidivism. As predicted, there was little difference in the YLS/CMI’s predictive validity for violent /nonviolent recidivism, as reported in previous studies (Catchpole & Gretton, 2003; Schmidt et al., 2011). This is not surprising, given that the YLS/CMI is an instrument for general offending. After taking into account the type of crime in the regression models, only Prior offenses (the only static risk factor included in the YLS/CMI) predicted nonviolent reoffending in youth.

In addition, gender was not related to any recidivism outcome, and age was only related to total juvenile recidivism and violent recidivism (the younger the youth when they committed the first offense, the higher the rate of both total and violent reoffending), thus confirming the results of previous studies (Moffitt, 2006). The criminal persistency found during young adulthood increases according to the number of crimes committed previously. Furthermore, it has been found that accumulated charges are related to early recidivism, criminal propensity, and an increase in the youth’s risk level (Farrington et al., 2016; Piquero et al., 2010). Some studies have examined criminal trajectories, studying long-term circumstances and risk factors associated with life environments. They found that the risk of recidivism also tends to decrease as time passes, meaning that as more time passes without the juvenile reoffending, they are less likely to reoffend (Bergman & Andershed, 2009; Farrington et al., 2009; Sarnecki, 2009). Thus, a “spider’s web” effect may appear: coming into contact with the criminal system and having committed these offenses at an early age lead the juvenile to a more difficult desistance. The more contact with the system, the more difficult desistance becomes.

The third hypothesis, namely, youth risk factors will present similar values when predicting youth recidivism than when predicting young adult recidivism, was supported by data. As mentioned above, there were no substantial differences between juvenile and young adult prediction; as such, the risk subscales were able to predict youth recidivism as accurately as adult recidivism. All subscales (with the exception of Leisure/free time) indicate a good predictive validity for youth and young adult recidivism. Previous literature has also observed the weak links between the Leisure/free time subscale and the various types of reoffending outcomes (it is not a Big Four factor; Hilterman et al., 2014; Olver et al., 2012). In fact, there are small differences between the means scores of recidivist and nonrecidivists on the leisure scale. This may be due to the general poor use of time for organized activities in youth. Hence, this factor is not strong enough to predict a type of recidivism separately, but still it does predict the total recidivism, presenting an overall impact on recidivism.

The YLS/CMI total score was able to predict all types of recidivism (violent, nonviolent, juvenile, and adult recidivism), taking into account the time to reoffend. Surprisingly, no youth risk subscale was individually able to predict the total or violent youth reoffending (Cox regression analyses). One possible variable that may interfere in these results is the length of the follow-up period; the period is longer in adults (20.19 months) than in juveniles (13.74 months). As Cox regression analyses takes into account the time for recidivism and censored data, the type of analysis may differ in its prediction. In some cases, the follow-up period lasted only a couple of months, and as such not all the real youth reoffending could have been collected. In fact, most studies agree that the majority of youth reoffending takes place within 2 years of the first offense (Mulder et al., 2011).

The Prior Offenses and Education/Employment subscales were especially significant predictors for both youth and adult recidivism. Although the factor related to Prior offenses is a static risk indicator, Education is a dynamic aspect that may be critical for planning interventions to prevent future offending. Previous evidence suggests that programs and services targeting specific reoffending risk factors are five times more effective than those without an adequate conceptual model (Izzo & Ross, 1990). Both factors—Prior Offenses and Education/Employment—appeared as predictors of adult recidivism and related to the chronicity of offending, starting with school maladjustment (Olver et al., 2012). Hence, education is a crucial dynamic factor to focus interventions to reduce recidivism.

For the fourth hypothesis posited (i.e., the prediction of risk factors for violent and nonviolent recidivism), significant validity in violent crime was obtained for the Personality, Family, and Education factors, whereas in nonviolent offenses, the predictors were Prior Offenses, Family, Peers, and Antisocial Attitudes. As observed, Family/Parenting is the only subscale present in both outcomes (violent/nonviolent). Moreover, positive family relationships seem to reduce juveniles’ risk of recidivism, as poor attachment and social bonding within families is associated with reoffending. Classic previous studies have extensively shown that criminal behavior occurs because of weak social bonds with prosocial subjects and adjusted society (Chan et al., 2015; Hirschi, 1969). Our findings suggest that the social context plays an important role in predicting nonviolent recidivism among Spanish youth involved in the Justice system. Aside from inconsistent parenting in relation to the factor of family, other detrimental environmental contexts such as delinquent peers, previous contact with crime and antisocial attitudes are of crucial importance in shaping nonviolent crime. The association between peer deviancy and nonviolent crime was also observed in the study of high-risk minors by Kalvin and Bierman (2017).

On the contrary, when analyzed in detail, violent crime seems to be related to more individual factors such as personality and behavior (guilt, aggressiveness, self-esteem) and education (items related to disruptive behavior and problems with peers and teachers); as such, there seems to be an association between the two. Nevertheless, future research must deeply study these possible associations between individual/contextual factors and types of crimes.

Finally, two important limitations should be addressed. First, the sample came from a specific region in Spain. Future studies should replicate this research with a nationally representative sample. Second, some short follow-up periods may produce data that were “missing by design.” Therefore, additional research, using longer follow-up periods for different types of recidivism, is needed to provide stronger evidence about YLS/CMI predictive validity among young adults. Accordingly, the age of the first offense, which was not available in this study may be also considered in future research as a predictor of juvenile and adult recidivism, to analyze the whole trajectories of the juveniles in detail. Despite these limitations, this study suggests that the YLS/CMI is a tool capable of predicting recidivism in the Spanish context, in both the juvenile and the young adult periods. Violent crimes seem to be related to more individual factors, whereas antisocial environmental contexts seem to predict nonviolent crimes. The identification and detection of a particular need related to a specific type of recidivism could be useful for intervention, increasing the suitability of the programs (Olver et al., 2012). To conclude, the YLS/CMI is a proficient tool for risk classification, the allocation of services, and to regulate the intensity and urgency of the supervision, in particular, by combining the subscales of Prior Offenses and Education/Employment.

Footnotes

Authors’ Note:

The authors thank the Juvenile Justice Court and the Juvenile Court’s Technical Team for their support in this research. This study was funded by Generalitat Valenciana, Conselleria d’Innovació, Universitats, Ciència i Societat Digital (reference number: GV/2019/089), and Jaume I University (reference number: UJI-A2019-09).

ORCID iD

Keren Cuervo

Keren Cuervo, PhD, is senior researcher at the Department of Developmental, Educational and Social Psychology and Methodology, Jaume I University, Spain. Her main research interests are forensic psychology, criminology, and juvenile offenders, focusing on juvenile corrections and risk assessment.

Lidón Villanueva, PhD, is an associate professor in Developmental Psychology at Jaume I University, Castellón, Spain, developing her research in collaboration with the Juvenile Court and probation and custody center entities. Her main topics include the analysis of risk and protective factors in youth offending, to prevent recidivism.

Miguel Basto-Pereira, PhD, is a forensic psychologist, codirector of the Master’s in Forensic Psychology, and a lecturer at the ISPA—Instituto Universitário, Lisbon. His research interests include child abuse, factors that promote the abandonment of criminal behavior and analysis of trajectories of juvenile delinquency.

References

Arnett

J. J.

(2000). Emerging adulthood: A theory of development from the late teens through the twenties. American Psychologist, 55(5), 469–480. https://doi.org/10.1037/0003-066X.55.5.469

Basto-Pereira

Começanha

Ribeiro

, & Maia, Â. (2015). Long-term predictors of crime desistance in juvenile delinquents: A systematic review of longitudinal studies. Aggression and Violent Behavior, 25, 332–342. https://doi.org/10.1016/j.avb.2015.09.012

Benuto

L. T.

Leany

B. D.

Thaler

N. S.

Yonesawa

(2014). Assessing risk, recidivism, and dangerousness in Asians. In Benuto

L. T.

Thaler

N. S.

Leany

B. D.

(Eds.), Guide to psychological assessment with Asians (pp. 443–451). Springer.

Bergman

L. R.

Andershed

(2009). Predictors and outcomes of persistent or age limited registered criminal behavior: A 30-year longitudinal study of a Swedish urban population. Aggressive Behavior, 35(2), 164–178. https://doi.org/10.1002/ab.20298

Bonta

Andrews

D. A.

(2017). The psychology of criminal conduct. Routledge.

Brame

Mulvey

E. P.

Schubert

C. A.

Piquero

A. R.

(2018). Recidivism in a sample of serious adolescent offenders. Journal of Quantitative Criminology, 34(1), 167–187. https://doi.org/10.1007/s10940-016-9329-2

Campbell

C. A.

Miller

Papp

Barnes

A. R.

Onifade

Anderson

V. R.

(2019). Assessing intervention needs of juvenile probationers: An application of latent profile analysis to a risk–need–responsivity assessment model. Criminal Justice and Behavior, 46(1), 82–100. https://doi.org/10.1177/0093854818796869

Capdevila

Ferrer

Luque

(2008). Evolució del perfil dels joves infractors ingressats en centres educatius i taxa de reincidencia [Profile evolution of young offenders in custody centers and recidivism rate]. JUSTIDATA. Estadística Básica de la Justicia a Catalunya, 48, 1–16.

Catchpole

R. E.

Gretton

H. M.

(2003). The predictive validity of risk assessment with violent young offenders: A 1-year examination of criminal outcome. Criminal Justice and Behavior, 30(6), 688–708. https://doi.org/10.1177/0093854803256455

10.

Chan

H. C.

T. W.

Zhong

L. Y.

Chui

W. H.

(2015). Criminal recidivism of incarcerated male nonviolent offenders in Hong Kong. International Journal of Offender Therapy and Comparative Criminology, 59(2), 121–142. https://doi.org/10.1177/0306624X13502965

11.

Chu

C. M.

Lee

Zeng

Yim

Tan

C. Y.

Ang

Chin

Ruby

(2015). Assessing youth offenders in a non-western context: The predictive validity of the YLS/CMI ratings. Psychological Assessment, 27(3), 1013–1021. https://doi.org/10.1037/a0038670

12.

Cox

S. M.

Kochol

Hedlund

(2016). The exploration of risk and protective score differences across juvenile offending career types and their effects on recidivism. Youth Violence and Juvenile Justice, 16(1), 77–96. https://doi.org/10.1177/1541204016678439

13.

Cuervo

Villanueva

(2015). Analysis of risk and protective factors for recidivism in Spanish youth offenders. International Journal of Offender Therapy and Comparative Criminology, 59, 1149–1165. https://doi.org/10.1177/0306624X14557917

14.

Cuervo

Villanueva

(2018). Prediction of recidivism with the Youth Level of Service/Case Management Inventory (reduced version) in a sample of young Spanish offenders. International Journal of Offender Therapy and Comparative Criminology, 62(11), 3562–3580. https://doi.org/10.1177/0306624X17741250

15.

Cuervo

Villanueva

Born

Gavray

(2018). Analysis of violent and non-violent versatility in self-reported juvenile delinquency. Psychiatry, Psychology and Law, 25(1), 72–85. https://doi.org/10.1080/13218719.2017.1347935

16.

Cuervo

Villanueva

Prado-Gascó

(2017). Predicción de la reincidencia juvenil mediante el Inventario YLS/CMI y baremos para su valoración [Youth recidivism prediction with the YLS/CMI Inventory and norms for assessment]. Revista Mexicana de Psicología, 34(1), 24–36.

17.

De Ruiter

Hildebrand

. (2009). Youth Level of Service/Case Management Inventory (YLS/CMI): Handleiding. Pearson.

18.

Farrington

D. P.

Loeber

Howell

J. C.

(2012). Young adult offenders: The need for more effective legislative options and justice processing. Criminology & Public Policy, 11(4), 729–750. https://doi.org/10.1111/j.1745-9133.2012.00838.x

19.

Farrington

D. P.

Ttofi

M. M.

Coid

J. W.

(2009). Development of adolescence-limited, late-onset, and persistent offenders from age 8 to age 48. Aggressive Behavior, 35, 150–163. https://doi.org/10.1002/ab.20296

20.

Farrington

D. P.

Ttofi

M. M.

Piquero

A. R.

(2016). Risk, promotive, and protective factors in youth offending: Results from the Cambridge study in delinquent development. Journal of Criminal Justice, 45, 63–70. https://doi.org/10.1016/j.jcrimjus.2016.02.014

21.

Flores

A. W.

Travis

L. F.

Latessa

E. J.

(2004). Case classification for juvenile corrections: An assessment of the Youth Level of Service/Case Management Inventory (YLS/CMI). U.S. Department of Justice.

22.

Garrido

Lopez

Silva

López

M. J.

Molina

(2006). El modelo de la competencia social de la ley de menores [Social competence model in minors criminal law]. Tirant lo Blanch.

23.

Gendreau

Little

Goggin

(1996). A meta-analysis of the predictors of adult offender recidivism: What works! Criminology, 34(4), 575–608. https://doi.org/10.1111/j.1745-9125.1996.tb01220.x

24.

Guy

L. S.

Packer

I. K.

Warnken

(2012). Assessing risk of violence using structured professional judgment guidelines. Journal of Forensic Psychology Practice, 12(3), 270–283. https://doi.org/10.1080/15228932.2012.674471

25.

Henggeler

S. W.

(2004). Decreasing effect sizes for effectiveness studies-implications for the transport of evidence-based treatments: Comment on Curtis, Ronan, and Borduin (2004). Journal of Family Psychology, 18(3), 420–423. https://doi.org/10.1037/0893-3200.18.3.420

26.

Hilterman

E. L.

Nicholls

T. L.

van Nieuwenhuizen

(2014). Predictive validity of risk assessments in juvenile offenders: Comparing the SAVRY, PCL: YV, and YLS/CMI with unstructured clinical assessments. Assessment, 21(3), 324–339. https://doi.org/10.1177/1073191113498113

27.

Hirschi

(1969). Causes of delinquency. University of California Press.

28.

Hofstede

(1980). Culture’s consequences: International differences in work-related values. SAGE.

29.

Hofstede

G. J.

Minkov

(2010). Cultures and organizations (3rd ed.). McGraw-Hill.

30.

Hoge

R. D.

Andrews

D. A.

(2006). Youth Level of Service/Case Management Inventory (YLS/CMI). Multi-Heath Systems.

31.

Izzo

R. L.

Ross

R. R.

(1990). Meta-analysis of rehabilitation programs for juvenile delinquents: A brief report. Criminal Justice and Behavior, 17(1), 134–142. https://doi.org/10.1177/0093854890017001008

32.

Junger-Tas

Marshall

I. H.

Enzmann

Killias

Steketee

Gruszczynska

(2010). Juvenile delinquency in Europe and beyond: Results of the second international self-report delinquency study. Springer.

33.

Kalvin

C. B.

Bierman

K. L.

(2017). Child and adolescent risk factors that differentially predict violent versus nonviolent crime. Aggressive Behavior, 43(6), 568–577. https://doi.org/10.1002/ab.21715

34.

Kleinbaum

D. G.

Klein

(2005). Competing risks survival analysis. In Kleinbaum

D. G.

(Ed.), Survival analysis: Statistics for biology and health (pp. 391–461). Springer.

35.

López

Garrido

V. J.

López

J. J.

López

M. J.

Galvis

M. J.

(2016). Predicción de la reincidencia con delincuentes juveniles: Un estudio longitudinal [Recidivism prediction in juvenile offenders: A longitudinal study]. Revista Española de Investigación Criminológica: REIC, 14(6), 1–22.

36.

Markus

H. R.

Kitayama

(1991). Culture and the self: Implications for cognition, emotion, and motivation. Psychological Review, 98, 224–253. https://doi.org/10.1037/0033-295x.98.2.224

37.

Moffitt

T. E.

(1993). Adolescent-limited and life-course persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100, 674–701.

38.

Moffitt

T. E.

(2006). Life-course persistent versus adolescence-limited antisocial behaviour. In Cicchetti

Cohen

D. J.

(Eds.), Developmental psychopathology (pp. 570–598). John Wiley.

39.

Moffitt

T. E.

(2007). A review of research on the taxonomy of life-course persistent versus adolescence-limited antisocial behavior. In Flannery

D. J.

Vazdonyi

Waldman

I. D.

(Eds.), The Cambridge handbook of violent behavior and aggression (pp. 49–74). Cambridge University Press.

40.

Mori

Takahashi

Kroner

D. G.

(2017). Can unstructured clinical risk judgment have incremental validity in the prediction of recidivism in a non-Western juvenile context? Psychological Services, 14(1), 77–86. https://doi.org/10.1037/ser0000107

41.

Mowen

T. J.

Boman

J. H.

(2018). A developmental perspective on reentry: Understanding the causes and consequences of family conflict and peer delinquency during adolescence and emerging adulthood. Journal of Youth and Adolescence, 47(2), 275–289. https://doi.org/10.1007/s10964-017-0794-1

42.

Mulder

Brand

Bullens

van Marle

(2011). Risk factors for overall recidivism and severity of recidivism in serious juvenile offenders. International Journal of Offender Therapy and Comparative Criminology, 55(1), 118–135. https://doi.org/10.1177/0306624X09356683

43.

Olver

Stockdale

Wong

(2012). Short and long-term prediction of recidivism using the Youth Level of Service/Case Management Inventory in a sample of serious young offenders. Law and Human Behavior, 36(4), 331–344. https://doi.org/10.1037/h0093927

44.

Olver

Stockdale

Wormith

(2009). Risk assessment with young offenders. Criminal Justice and Behavior, 36(4), 329–353. https://doi.org/10.1177/0093854809331457

45.

Olver

Stockdale

Wormith

(2014). Thirty years of research on the level of service scales: A meta-analytic examination of predictive accuracy and sources of variability. Psychological Assessment, 26, 156–176. https://doi.org/10.1037/a0035080

46.

Onifade

Davidson

Campbell

Turke

Malinowski

Turner

(2008). Predicting the who, what and when of delinquency with the Youth Level of Service Case Management Inventory. Criminal Justice and Behavior, 35, 474–483. https://doi.org/10.1177/0093854807313427

47.

Ortega-Campos

García-García

Zaldívar-Basurto

(2017). The predictive validity of the structured assessment of violence risk in youth for young Spanish offenders. Frontiers in Psychology, 8, 577–585. https://doi.org/10.3389/fpsyg.2017.00577

48.

Papp

Campbell

C. A.

Anderson

V. R.

(2019). Assessing the incremental validity of Andrews and Bonta’s “moderate four” predictors of recidivism using a diverse sample of offending and truant youth. International Journal of Offender Therapy and Comparative Criminology, 63(6), 854–873. https://doi.org/10.1177/0306624X18814185

49.

Pimentel

Quintas

Fonseca

Serra

(2015). Estudo normativo da versão portuguesa do YLS/CMI: Inventário de avaliação do risco de reincidência e de gestão de caso para jovens [Normative study of the Portuguese version of the YLS/CMI: Youth Level of Service/Case Management Inventory]. Análise Psicológica, 33(1), 55–71. https://doi.org/10.14417/ap.883

50.

Piquero

Farrington

Nagin

Moffitt

(2010). Trajectories of offending and their relation to life failure in late middle age: Findings from the Cambridge study in delinquent development. Journal of Research in Crime & Delinquency, 47(2), 151–173. https://doi.org/10.1177/0022427809357713

51.

Pusch

Holtfreter

(2017). Gender and risk assessment in juvenile offenders: A meta-analysis. Criminal Justice and Behavior, 45(1), 56–81. https://doi.org/10.1177/0093854817721720

52.

Rennie

Dolan

(2010). Predictive validity of the Youth Level of Service/Case Management in custody sample in England. The Journal of Forensic Psychiatry & Psychology, 21(3), 407–425. https://doi.org/10.1080/14789940903452311

53.

Rice

M. E.

Harris

G. T.

(1995). Violent recidivism: Assessing predictive validity. Journal of Consulting and Clinical Psychology, 63(5), 737–748. https://doi.org/10.1037//0022-006x.63.5.737

54.

Rice

M. E.

Harris

G. T.

(2005). Comparing effect sizes in follow-up studies: ROC area, Cohen’s d, and r. Law and Human Behavior, 29(5), 615–620. https://doi.org/10.1007/s10979-005-6832-7

55.

Sarnecki

(2009). Delinquent networks: Youth co-offending. In Schneider

H. J.

(Ed.), Internationales handbuch der kriminologie band [International criminology handbook] (Vol. 2, pp. 995–1023). De Gruyter Recht.

56.

Schmidt

Campbell

M. A.

Houlding

(2011). Comparative analyses of the YLS/CMI, SAVRY and PCL: YV in adolescent offenders: A 10-year follow-up into adulthood. Youth Violence and Juvenile Justice, 9(1), 23–42. https://doi.org/10.1177/1541204010371793

57.

Schmidt

Hoge

R. D.

Gomes

(2005). Reliability and validity analyses of the Youth Level of Service/Case Management Inventory. Criminal Justice and Behavior, 32(3), 329–344. https://doi.org/10.1177/0093854804274373

58.

Schwalbe

C. S.

(2007). Risk assessment for juvenile justice: A meta-analysis. Law and Human Behavior, 31(5), 449–462. https://doi.org/10.1007/s10979-006-9071-7

59.

Singh

J. P.

(2013). Predictive validity performance indicators in violence risk assessment: A methodological primer. Behavioral Sciences & the Law, 31(1), 8–22. https://doi.org/10.1002/bsl.2052

60.

Spanish Statistics Institute. (2018). Estadística de condenados adultos/Estadística de condenados menores [Charged adult statistics/charged youth statistics 2017].

61.

Spanish Statistics Institute. (2019). Cifras de población (CP) a 1 de enero de 2019: Estadística de Migraciones (EM). Año 2018 [Population data 1st January 2019: Migration statistics].

62.

Takahashi

Mori

Kroner

D. G.

(2013). A cross-validation of the Youth Level of Service/Case Management Inventory (YLS/CMI) among Japanese juvenile offenders. Law and Human Behavior, 37(6), 389–400. https://doi.org/10.1037/lhb0000029

63.

Taxman

F. S.

(2018). Risk assessment: Where do we go from here? In Singh

J. P.

Kroner

D. G.

Wormith

J. S.

Desmarais

S. L.

Hamilton

(Eds.), Handbook of recidivism risk/needs assessment tools (pp. 271–284). John Wiley.

64.

Thompson

A. P.

McGrath

(2012). Subgroup differences and implications for contemporary risk-need assessment with juvenile offenders. Law and Human Behavior, 36(4), 345–355. https://doi.org/10.1037/h0093930

65.

Thompson

A. P.

Pope

(2005). Assessing juvenile offenders: Preliminary data for the Australian adaptation of the Youth Level of Service/Case Management Inventory (Hoge & Andrews, 1995). Australian Psychologist, 40(3), 207–214. https://doi.org/10.1080/00050060500243491

66.

Vincent

G. M.

Drawbridge

Davis

(2019). The validity of risk assessment instruments for transition-age youth. Journal of Consulting and Clinical Psychology, 87(2), 171–183. https://doi.org/10.1037/ccp0000366