Jump to ContentJump to Main Navigation
From Juvenile Delinquency to Adult CrimeCriminal Careers, Justice Policy and Prevention$

Rolf Loeber and David P. Farrington

Print publication date: 2012

Print ISBN-13: 9780199828166

Published to Oxford Scholarship Online: September 2012

DOI: 10.1093/acprof:oso/9780199828166.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2019. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use.  Subscriber: null; date: 23 October 2019

Prediction and Risk/Needs Assessments

Prediction and Risk/Needs Assessments

Chapter:
(p.150) 6 Prediction and Risk/Needs Assessments
Source:
From Juvenile Delinquency to Adult Crime
Author(s):

Robert D. Hoge

Gina Vincent

Laura Guy

Publisher:
Oxford University Press
DOI:10.1093/acprof:oso/9780199828166.003.0006

Abstract and Keywords

The focus of the chapter is on available knowledge regarding the prediction of early adult offending from information available during the early juvenile years. Standardized screening and assessment tools for formulating these predictions are also reviewed. The chapter begins with a discussion of general issues regarding risk prediction and assessment, including dynamic risk factors (need factors) and protective or strength factors. This is followed by a summary of risk factors identified in the theoretical and empirical literatures as known to be associated with criminal activity. Technical issues in the conduct of risk assessments are discussed, followed by reviews of the major established juvenile and adult assessment tools. Two major conclusions emerge from the chapter. First, while we have considerable knowledge of the risk and need factors associated with adolescent offending and offending during the adult years, information about the prediction of the initiation and desistance from criminal activities specifically for the early adult years (roughly 18 to 25) is lacking. Second, while a number of standardized screening and assessment tools demonstrate sound psychometric properties in the prediction of adolescent offending and adult offending, relatively less information is available regarding the ability of the adolescent measures to predict early adult offending. The chapter concludes with a set of recommendations for researchers and practitioners.

Keywords:   criminogenic risk factors, criminogenic need factors, protective factors, risk assessment, risk management, developmental criminology, assessment, ethics

This chapter provides a review of available knowledge regarding the prediction of early adult offending (focusing on the age range of 18 to 29) on the basis of information available during the juvenile years and of assessment tools for formulating these predictions (see Vincent, Terry, & Maney [2009] for a review of juvenile instruments and meta-analyses cited in this chapter for reviews of adult instruments). The chapter begins with a discussion of general issues regarding prediction and assessment, including the parameters of risk prediction and legal and ethical issues associated with risk assessment. We then provide a brief summary of risk factors associated with criminal activity. Technical issues in the conduct of risk assessments are discussed, followed by reviews of the major established juvenile and adult risk assessment tools. The chapter concludes with research and clinical recommendations relating to risk assessment and prediction.

Parameters of Risk Prediction and Assessment with Youth

The prediction of the onset and persistence of criminal activity depends on early identification of serious and violent individuals and circumstantial factors that facilitate such identification. Three concepts are relevant to our analysis of risk assessment and prediction. Risk factors refer to characteristics of youths or their circumstances that increase the likelihood that they will engage in delinquency (e.g., a history of conduct disorder). Criminogenic need factors (also known as dynamic risk factors) are risk factors that can be changed, and, if changed, could reduce the likelihood of engagement in antisocial behaviors (e.g., antisocial peer associations). Strength or protective factors are features of the youth or his or her situation that can buffer the effects of risk factors (e.g., a positive bond between youth and parent can reduce the impact of negative peer associations).

The identification of the risk, need, and protective factors is both a theoretical and empirical issue (see Thornberry et al., Chapter 3, this volume; also, Farrington, (p.151) 2005; Guerra, Williams, Tolan, & Modecki, 2008; Rutter, Giller, & Hagell, 1998) and will be explored in more detail later in this chapter. However, several points need to be stressed here. First, different risk and protective factors may predict the onset of and desistance from delinquency. For example, early drug abuse may be associated with the onset of criminal activities, but the establishment of a positive social bond may be associated with desistance from such activities. Second, the relative importance of risk, need, and protective factors may vary with developmental age. For example, some drug and alcohol use assumes decreasing influence through adolescence, whereas peer group influences assume increasing importance. This presents a challenge to assessing risk for delinquency during late adolescence because we are dealing with a period of transition between older adolescence and early adulthood. However, much of the available research focuses either on adolescents or adults, and relatively little information is available for the transition years.

The Contexts of Risk Assessments

Risk assessments are relevant in a range of criminal and civil legal decisions, such as pre-charge diversion, pre-trial detention, eligibility for alternative measures programs, and waivers to adult and mental health proceedings, sentencing, and dispositions (see Howell et al., Chapter 8, this volume). While some of these decision contexts may call for a narrow focus on risk for future delinquency because a quick decision needs to be made, most decisions require an assessment of criminological needs as well. As we noted earlier, criminogenic needs are essentially dynamic risk factors that are driving individuals’ offending. Assessing the criminogenic need factors underlying risk is important in any decision where a disposition or intervention is to be provided to address the risk factor. If, for example, negative peer associations and substance abuse are major risk factors for delinquency, then these factors can be identified as criminogenic need factors to be addressed in any intervention effort. The term risk management will be used when referring to the identification of criminogenic needs for purposes of reducing risk.

The selection of the risk assessment tool, and consequently the amount of information needed from the tool, depends on the nature of the decision in question. Pre-trial detention or classification decisions will often require an estimate of the likelihood of committing a violent offense over some short period of time. Similarly, decisions such as those relating to eligibility for pre-charge diversion or pre-trial alternative measures may call for general predictions of the likelihood of re-offending. However, longer-term decisions, such as disposition, case planning, or management may call for an assessment of the likelihood of re-offending, in addition to the criminogenic needs underlying the risk of offending. Also, longer range planning like this can benefit from risk assessment tools that contain strength or protective factors.

(p.152) Another aspect of this issue concerns the relation between the assessment and the ultimate legal question. As Heilbrun (2010) stresses, a direct link does not always exist between the two. A prediction of the likelihood of engaging in a violent sexual assault is directly linked with a designation of a violent sexual predator. However, an estimate of the likelihood of engaging in a criminal act might be only one consideration in a decision about placement within an institutional or community program. Care must always be taken in evaluating the relevance of the assessment to the judicial decision to be made.

Considerable variability exists in the way in which risk/needs assessments are formulated and reported. Variations occur, first, in the source of the assessment. In many cases a mental health professional will conduct the assessment. This is particularly true for waiver or civil commitment proceedings requiring a professional psychological assessment. Personnel other than mental health professionals, such as probation officers, intake officers, correctional officers, or youth workers, also conduct risk/needs assessments; these individuals should have received specialized training.

Variability also exists in the procedures employed to conduct the assessments. Some judicial and correctional systems employ no formal or structured assessment procedures. For example, a correctional officer may interview the youth and on the basis of his or her experience form a judgment about the risk level. In other cases more or less structured procedures might be employed. These would normally be based on clinical interviews and possibly the use of standardized tests. Tests are common in the case of psychologists or psychiatrists assigning mental health diagnoses. However, and as we will see below, a number of standardized and validated procedures have been developed to guide the collection and synthesis of information to yield estimates of risk and needs. These are often suitable for use by both mental health professionals and trained judicial and correctional personnel.

Relatively little information is available regarding the frequency of use of standardized risk/needs assessment procedures in correctional or forensic decision contexts. Informal observations would suggest that many systems employ only unstructured assessments or locally developed and nonvalidated instruments. Mulvey and Iselin (2008) have suggested that many of the reports provided to judges to aid in disposition decisions are based on informal assessment procedures. This is unfortunate because, as we will see below, a number of validated, standardized instruments are available to guide such decisions.

Legal and Ethical Considerations

Legal and ethical issues arise in connection with many forensic risk assessments (Grisso & Applebaum, 1992; Heilbrun, 2001, 2010; Heilbrun, Grisso, & Goldstein, 2008; Melton, Petrila, Poythress, & Slobogin, 2007). For example, in the United (p.153) States the Federal Rules of Evidence specify criteria relevant to the admissibility of expert testimony, and this includes testimony based on risk assessments. Other sources of guidance are from professional practice and ethical guidelines (Heilbrun, 1992, 2010; Melton et al., 2007). The American Psychological Association’s (APA) Ethical Principles of Psychologists and Code of Conduct (APA, 2002) provides a broad range of guidelines regarding psychological practice. A group within a division of APA, the Committee on Ethical Guidelines for Forensic Psychologists (1991), provides standards specifically applicable to forensic practices, including the conduct and reporting of risk assessments.

Practice guidelines regarding the use of tools for conducting risk assessments also are available. Heilbrun (1992) offered the following recommendations regarding the selection and use of such instruments: they should be commercially available, have a manual, and be critically reviewed in the literature; have adequate reliability and validity demonstrated; be relevant to the forensic decision; be administered using standard procedures; scores should be interpreted with reference to populations and contexts that are similar to that of the evaluee; when possible, objective tests and actuarial procedures should be used; and response style should be considered, as distorted responding may negate the results of the assessment. Importantly, it must be kept in mind that practice and ethics guidelines only govern the actions of the relevant mental health providers. Other professionals such as probation or correctional officers may not be affected by any professional guidelines relevant to conducting risk assessments.

A second source of guidance in the conduct of risk or risk/needs assessments may be found in policy statements and practice guidelines developed in specific jurisdictions. Many agencies in the United States have adopted standardized tools to guide the conduct of these assessments and an increasing number have developed case planning and management procedures directly tied to the assessments.

Current Knowledge Regarding the Prediction of Offending in Early Adulthood

Three related areas of theory and research relevant to the prediction of late or early adult-onset offending have emerged: (a) analyses of the trajectories of criminal offending, (b) identification of risk and protective factors, and (c) a developmental life-course perspective integrating these two areas. We will provide only brief summaries of these literatures in this chapter but will refer the reader to more comprehensive reviews as well as to Piquero et al. (Chapter 2, in this volume).

Efforts to identify stable trajectories of criminal careers have focused on two patterns labeled life-course persistent delinquency and adolescent-limited delinquency (Loeber, Slot, & Stouthamer-Loeber, 2006; Loeber & Stouthamer-Loeber, 1998; Moffitt, 2003; Thornberry, 2005). The life-course persistent pattern describes cases where evidence of conduct problems appears during the preschool years, (p.154) escalates through early childhood and adolescence, and persists into adulthood. Adult offenders convicted of very serious crimes often exhibit this pattern. The adolescent-limited pattern on the other hand describes cases where antisocial acts first appear in adolescence. The criminal activity in this case is generally not of a serious nature and the individual usually desists by later adolescence.

The issue of trajectories is of particular concern in the present case because of our focus on ages 18 to 29. This is the period when youth with early signs of a life-course persistent pattern either persist or desist in these activities. The first appearance of criminal activities after age 18 has generally been considered relatively rare, but it does occur and can be considered an extension of the adolescent-onset pattern. Although the available evidence is equivocal regarding the number or nature of the trajectories (Piquero, Farrington, & Blumstein, 2007; Piquero et al., Chapter 2, this volume; Sampson & Laub, 2003), the information provides some guidance regarding the prediction of criminal activity during the early adult years.

The second area of relevance to prediction concerns the identification of risk and protective factors associated with the onset, persistence, and desistance from criminal activity (see Hoge & Andrews, 2010; Lipsey & Derzon, 1998; Loeber & Stouthamer-Loeber, 1987; Thornberry & Krohn, 2003). A large and growing literature on these factors derives from both cross-sectional and longitudinal studies. Of particular value are the long-term longitudinal studies such as the Cambridge Study in Delinquent Development (Farrington, 2004; Farrington et al., 2006), the Pittsburgh Youth Study (Loeber, Stouthamer-Loeber, & Farrington, 2008) and the Rochester Youth Development Study (Thornberry, 2005; Thornberry, Ireland, & Smith, 2001). A wide range of contextual (e.g., criminality in family of origin, high crime neighborhood) and individual (e.g., antisocial attitudes, negative peer associations) factors have been identified in this research.

Several cautions should be noted. First, conceptual and methodological issues persist in this research (O’Connor & Rutter, 1996; Rutter et al., 1998). Conceptual issues relate to the difficulty of identifying causal factors among the correlates identified. Methodological problems involve subject loss and the comparability of measures at different points in data collection. Second, the applicability of the risk factors across gender, cultural group, and age has not always been well established. Another issue is that few efforts have been developed to identify strength or protective factors. These are important because they represent potential moderators of the effects of risk and are associated with desistance from criminal activity. Two categories of potential risk factors are worthy of further discussion and will be described later: mental health and personality variables.

The third relevant development is represented in efforts to formulate a developmental life-course perspective on the prediction of criminal behavior (Catalano & Hawkins, 1996; Farrington, 2004, 2005; Guerra et al., 2008; Thornberry, 2005). Evaluating risk requires consideration of the developmental stage and social context (Mulvey, 2005). Even the relevance of risk factors can change across time (for a review see Odgers, Vincent, & Corrado, 2002). For example, smoking prior to (p.155) age 12 is a significant risk factor, but smoking at age 15 when experimentation is a normal part of development, or in early adulthood when smoking is legal, would not be risk factors for offending.

Another key developmental concept for assessments of risk for violence and serious offending is the impact of maturation on the time frame for which predictions remain accurate. A significant limitation on attempts to identify youth who will become chronic and violent offenders is the potential for a high false positive rate. A significant number of youth who engage in violent behavior at one stage of development do not continue to do so as their development proceeds. Indeed, at least 50% of children who initiate pervasive and serious antisocial behavior between ages 6 and 12 do not develop into seriously antisocial adults (Patterson et al., 1998; Robins, 1974), and an even greater portion of serious offending adolescents do not develop into antisocial adults (Moffitt & Caspi, 2001; see also Piquero et al., Chapter 2, this volume). The National Youth Survey (Elliott, Huizinga, & Menard, 1989) found that, for about 50% of youths, violent behavior persisted into adulthood if their first violent acts occurred before age 11, about 30% persisted if their violence started between ages 11 and 13, and about 10% persisted if their first violent acts occurred in adolescence. Taken as a whole, the research suggests that even youth who engage in the most serious violence or antisocial behavior at a young age have only a 50–50 chance of persisting.

Mental Health Variables in Risk Assessment

The relation of mental health problems to offending and violence is complicated. On the one hand, many individuals with mental disorders do not have a violent or delinquent history. On the other hand, many individuals with a mental disorder are involved in the justice system. Single-site studies using forms of the Diagnostic Interview for Children (DISC) suggest that about two-thirds of youths involved with the juvenile justice system meet criteria for one or more psychiatric disorders, even after excluding conduct disorder (Atkins, Pumariega, & Rogers, 2003; Teplin, Abram, McClelland, Dulcan, & Mericle, 2002; Wasserman, McReynolds, Lucas, Fisher, & Santos, 2002). This is substantially higher than prevalence rates found in youths in the general population (around 14 to 22%; Kazdin, 2000). More recently, a nationwide study of self-reported mental health symptoms among justice-involved youth found that 72% of girls and 63% of boys had at least one clinically significant mental health problem (Vincent, Grisso, Terry, & Banks, 2008). Further, in a complex analysis of longitudinal data, researchers reported that 19.5% of adult crime among women and 28.7% of adult crime among men was attributable to individuals with childhood psychiatric disorders, and those diagnosed with multiple disorders were at even greater risk for later criminal behavior (Copeland, Miller-Johnson, Keeler, Angold, & Costello, 2007). The percentages remained significant even after adjustments removing conduct disorder and juvenile offense history (20.6% for females and 15.3% for males, respectively).

(p.156) There are several potential explanations for this relation between mental disorder and offending, and possibly even persistent offending, into early adulthood. First, some child and adolescent mental disorders, or symptoms of mental disorder, may be causally connected to violence and antisocial behavior, namely, disruptive behavior disorders and attention deficit–hyperactivity disorder (ADHD). This potential causal connection is evidenced in a concept known as a hyperactive-impulsive-attention deficit (HIA) syndrome (Loeber, 1990), which when combined with conduct problems, seems to be found in children with early initiation of antisocial behavior that is frequent and severe (Lynam, 1996).

With respect to other types of disorders, mood disorders may relate to violent offending in cases where the mood disorders manifest in anger or hostility (Vincent & Grisso, 2005). Many youths with conduct problems have some form of anxiety (Frick, Lilienfeld, Ellis, Loney, & Silverthorn, 1999), and post-traumatic stress disorder in particular can underlie tendencies to react aggressively (Charney, Deutch, Krystal, Southwick, & Davis, 1993). Less is known about the connection between psychosis and offending or violence among youth.

In some cases, mental disorder may not be connected to actual offending, but youths and adults with mental health problems wind up in the justice system as a consequence of the lack of community mental health services. Until further research can disentangle the relationships, the evidence implies that most risk assessment schemes should contain an item or items related to attention problems and impulsive behavior. Further, until we know more about the connection between other symptoms of mental illness and offending among youth, consideration should be given to risk assessment tools, including those that contain some form of override feature for use with cases for which examiners have reason to believe that symptoms of mental illness are connected to violent or antisocial outcomes for a particular youth.

Personality and Callous-Unemotional Traits in Risk Assessment

The features of some personality disorders (e.g., Borderline, Narcissistic, Antisocial) are likely to get some individuals in trouble with the law. Given that these disorders tend to have their roots in childhood and adolescence, these personality traits may be related to early offending and may be a good predictor of offending into early adulthood. Psychopathic personality disorder probably has been the most widely studied with respect to its relation to future offending, particularly violent offending (for a review see Hare, 2003). The association between psychopathic personality and later offending and violence among adults has been documented in a number of meta-analyses of prospective studies (e.g., Hemphill, Hare, & Wong, 1998; Leistco, Salekin, DeCoster & Rogers, 2008). Many psychopathic features get people in trouble with the law, including impulsive and sensation-seeking behavior, callous and guiltless emotions, and an arrogance characterized by a desire to exert power over others.

(p.157) The relation between psychopathy and early adult offending is more complicated when assessing young people. Many scholars acknowledge that diagnosing or labeling a youth as a “psychopath” is inappropriate given developmental changes that affect personality through adolescence. There is a youth psychopathy assessment tool designed to assess psychopathic traits among adolescents (reviewed later), which has demonstrated a small to moderate effect for prediction of violence and re-offending, but most research has not examined its prediction of offending into early adulthood.

Instead, the focus for youth has been on a “syndrome” that involves the combination of callous-unemotional traits AND serious conduct problems, referred to as CU-CD. Callous-unemotional traits distinguish sub-groups of seriously conduct-disordered children and adolescents who experience minimal distress when engaging in criminal behaviors (Frick, O’Brien, Wootton, & McBurnett, 1994) and are more severe and stable in their offending patterns. In a prospective study, Frick, Kimonis, Dandreaux, and Farell (2003) discovered that callous-unemotional features were relatively stable during childhood for children who had CU scores falling in the lower or upper quartiles at age 6 if they also had serious conduct problems. Parent and teacher ratings of Interpersonal Callousness (IC) in children ages 7 to 12 appear to predict adult psychopathy ratings in the same youths at ages 18 to 19 (Burke, Loeber, & Lahey, 2007).

The stability of these traits from childhood or adolescence into adulthood has proven still to be only modest. Longitudinal studies examining measurement invariance in measures of IC or psychopathic features indicate that there is measurement invariance from late adolescence to young adulthood (6 years, Loney, Taylor, Butler, & Iacono, 2007). However, one longitudinal study found that, of the juveniles who scored in the top 20% of psychopathic traits at age 13, the vast majority (86%) did not score above a diagnostic threshold on a measure of psychopathic traits as adults 11 years later (Lynam, Caspi, Moffitt, Loeber, & Stouthamer-Loeber, 2007). Although the fact that 14% continued to score high as adults makes this construct relatively stable as far as youth disorders are concerned, practically speaking, use of assessments of these traits in childhood and adolescence to predict who would be psychopathic as adults would lead to a large number of false predictions. This body of literature indicates that it is crucial for risk assessment schemes to contain an item or items related to impulsivity, attention problems, sensation-seeking, and low empathy and remorse or callousness, all of which are strongly related to chronic offending.

Risk/Needs Assessment Formats

The four types of assessment formats are unstructured clinical, static actuarial, static/dynamic actuarial, and structured professional judgment, although this categorization is something of an oversimplification (Borum & Verhaagen, 2006; (p.158) Hoge, 2008). Unstructured clinical assessments depend on the unguided collection of information and the formulation of a judgment about level of risk through subjective interpretation of the information. For example, a psychologist may interview the client in an unstructured manner and, on the basis of his or her education and experience, will formulate a judgment about the likelihood that the individual will engage in violence or an antisocial act. Research has shown that unstructured clinical assessments are associated with poor levels of reliability and validity. Indeed, when mental health professionals make specific, unstructured predictions that a person “will” or “will not” be violent, they will likely be accurate in no more than one-third of cases (Grisso & Tomkins, 1996; Monahan, 1996; Rubin, 1972).

The different approaches to assessing risk can vary in terms of the amount of structure imposed on the three central decisions that arise in the assessment process: 1) which risk factors to consider and how to measure them; 2) how to combine the risk factors; and 3) how to generate the final risk estimate (Monahan, 2008). Standardized risk assessments are based on structured procedures for the collection and synthesis of information. Actuarial measures constitute a specific type of standardized measure. Structure is imposed on each of the three major decisions in the actuarial assessment approach, there is no discretion in terms of selecting, measuring, or combining risk factors, and the final risk estimate is determined by a priori, fixed rules. Items on these measures are often empirically derived using a construction sample with known outcomes (the developers know who recidivated and who did not) to identify the factors that predicted re-offending in that group. An algorithm is created to categorize people according to the likelihood of reoffending.

Actuarial decision-making means that specific risk predictions are formulated based on a statistical formula. Static actuarial measures include only historical and invariant items. Bonta (1996) described these as second-generation risk assessment instruments (unstructured clinical procedures constitute the first generation). Although scholars asserted that actuarial tools are superior to clinical judgment in the prediction of violence and re-offending (Grove & Meehl, 1996; Quinsey, Harris, Rice, & Cormier, 1998), the incremental gain in predictive validity is minimal (Litwack, 2001). Moreover, Bonta and several other critics have cautioned that static actuarial instruments are for the most part atheoretical, cover only a limited range of predictor variables, and are not useful for intervention planning or reassessments to measure individual progress (Borum, 1996; Dvoskin & Heilbrun, 2001; Hart, 2003; Hoge & Andrews, 2010).

Static/dynamic actuarial measures, termed third-generation measures by Bonta (1996), incorporate both static and dynamic risk factors. For the most part, these tools include static and dynamic risk factors that were selected due to a known empirical association with later offending, as opposed to selecting items that were predictive in a particular construction sample. As such, these tools are generally theoretically and empirically grounded. In the juvenile offender area, (p.159) these instruments are often referred to as risk/need actuarial tools. These tools can be used for reassessment and for intervention planning. They often include an override procedure to the final risk level to account for idiosyncratic factors that may affect an individual’s risk level but would not be reflected in the overall score.

The fourth category of risk assessment tools is structured professional judgment (SPJ). In this approach to assessing risk for violence, structure is imposed on which risk factors should be considered and how they should be measured, but the way in which factors are combined is left to the discretion of the evaluator. The evaluator’s discretion similarly is valued in terms of generating the final estimate of risk. Like the static/dynamic actuarial tools, SPJ tools are informed by the state of the discipline in clinical theory and empirical research on static and dynamic factors to include factors that guide decisions about risk and treatment planning. The intent was to improve human judgment by adding structure, and to improve actuarial decision-making by adding more rater discretion (Borum & Douglas, 2003). These instruments emphasize “prevention” as opposed to “prediction.” They contain static and dynamic risk factors and protective factors, assuming that risk can change as a result of treatment quality and quantity, developmental factors, protective factors, context, and the passage of time. The difference between SPJ tools and the static/dynamic actuarial tools is that SPJ tools lead to a final judgment by the rater regarding the overall level of risk (typically communicated as low, moderate, or high) based on a combination of risk factors, protective factors, and idiosyncratic factors present. No algorithm is used to produce a quantitative index of risk level.

Summarizing Debates

Considerable research has been conducted on the relative efficacy of unstructured clinical judgment versus standardized assessments (Baird & Wagner, 2000; Bonta, Law, & Hanson, 1998; Grove & Meehl, 1996; Grove, et al., 2000). The research consistently demonstrates that standardized assessments, whether based on actuarial or structured judgment procedures, yield better predictions of future behavior than unstructured clinical assessments. This evidence is strong and at this point is rarely disputed. However, some disagreements in the field remain.

The first disagreement is the issue of actuarial versus SPJ tools for estimates of an individual’s risk level. There is an accumulating body of literature regarding adult populations that suggests that SPJ-based decisions about risk may have incremental predictive validity over simple score-based decisions (Douglas, Yeoman, & Boer, 2005). The validity or reliability of actuarial overrides is still largely unknown. More recently, a very comprehensive meta-analysis of adult risk assessment tools indicated that, on average, when compared directly within the same sample, SPJ and actuarial tools have equivalent predictive validity for re-offending (Guy, 2008).

Another related debate is the value of including criminogenic need factors in risk tools. One argument is that the inclusion of these dynamic variables in (p.160) determinations of risk level (e.g., low, moderate, or high risk) can diminish the predictive accuracy of risk tools and, therefore, should be in a separate tool. The premise of this argument is that the goal of a risk assessment should be prediction. Of course, research has demonstrated that certain dynamic risk factors or criminogenic needs can elevate risk for delinquency (e.g., Farrington & West, 1993; Moffitt & Caspi, 2001), which suggests that they should be included in any risk tools (Austin, 2006). Further, the inclusion of these criminogenic need factors is essential for measuring changes or progress in risk level and for case management.

Statistically speaking, the debate revolves around whether dynamic risk or criminogenic need factors add incremental predictive value to quantitative risk levels. Practically speaking, however, this debate should consider that once an offender is labeled as high-risk by a tool with unchangeable factors, there is no room for documenting developmental changes.

One issue that continues to complicate this discussion is the definition of “needs.” Dynamic risk factors would be considered criminogenic needs when they have a causal connection with the individual’s risk for re-offending. Criminogenic needs theoretically are changeable. This is in contrast to other types of needs an offender may have that have not shown a significant relation to later offending, such as depression. One might refer to these as well-being needs. In our view, criminogenic needs are an essential feature of any comprehensive risk assessment tool; however, well-being needs belong in a separate assessment tool or, at the least, may be included within a risk assessment tool but should not be included in quantitative estimates of risk.

Review of Risk Assessment Instruments

Tools were selected for inclusion in this chapter if they: (1) were developed to assess risk for general antisocial behavior or violence in the community; (2) were designed to be generalizable rather than jurisdiction- or sample-specific (this excludes tools that are modified for each site, such as the Wisconsin classification system, Baird, 1981); (3) are administered by a trained rater/examiner or professional (i.e., not self-report inventories); and (4) have enough research evidence (including peer-reviewed research) to be considered evidence-based or promising at the time of this review. We defined evidence-based and promising as having some evidence of inter-rater reliability (IRR) and predictive validity for re-offending and enough information for replicable assessments, such as having a test manual. Categorization of a tool as evidence-based (versus promising) requires reliability evidence and validation from an independent party that receives no economic gain from the tool (Austin, 2006). We review tools used with youth (ages 12–17) and adults (aged 18 and older); no instrument has been developed specifically for the transition-ages (18 to 29). Table 6.1 provides a summary of the measures reviewed and identifies the variables included.

Table 6.1 Empirically-based risk factors contained in each risk assessment tool reviewed.

Youth Tools

Adult Tools

Risk Factors

WSJCA

YASI

YLS/

CMI

SAVRY

NCAR

LSI/

CMI

SIR

VRAG

HCR-20

COVR

Personal Risk Factors - Static

History of joblessness

X

X

X

X

X

X

History of school suspensions/expulsions

X

X

X

X

X

X

Early school leaving

X

X

X

X

History of homelessness

X

X

X

Early conduct problems

X

Early age of onset for offending

X

X

X

X

X

X

X

X

Early violence/aggression

X

X

Prior probation/custody

X

X

X

X

X

X

X

X

X

X

Exposure to violence (early)

X

X

X

X

X

X

Early caregiver disruption

X

X

X

History of violence

X

X

X

X

X

X

X

X

X

History of nonviolent delinquency/offending

X

X

X

X

X

X

X

X

X

History of abuse/maltreatment

X

X

X

X

X

X

X

Contextual Risk Factors - Static

Antisocial attitudes, values, beliefs, family

X

X

X

X

Criminality in family

X

X

X

X

X

X

X

Mental illness in family

X

X

X

Marital conflicts in family

X

X

X

X

X

Joblessness in family

X

X

Low family income

X

X

X

Inadequate family housing

X

X

Dysfunctional neighborhood/community disorganization

X

X

High crime neighborhood

X

X

X

Personal Risk Factors - Dynamic

Antisocial attitudes, values, beliefs

X

X

X

X

X

X

Attentional disorder

X

X

X

Psychopathy (adult tools)

X

X

X

Low empathy/remorse

X

X

Inflated self-worth

X

X

X

Sensation seeking

X

X

Impulsivity

X

X

X

X

X

Chronic lying

Hostile, aggressive, violent

X

X

X

X

X

X

X

X

Anger management issues

X

X

X

X

Lack of motivation

X

X

Drug abuse

X

X

X

X

X

X

X

X

Alcohol abuse

X

X

X

X

X

X

X

X

X

Poor school adjustment

X

X

X

X

X

X

X

X

Poor school performance

X

X

X

X

X

X

X

Poor coping ability

X

X

X

X

X

Mental illness/health

X

X

X

X

X

X

X

Contextual Risk Factors - Dynamic

Dysfunctional family environment

X

X

X

X

X

X

Ineffective parenting

X

X

X

X

X

Parent-youth conflicts/attachment problem

X

X

X

X

X

X

Lack of family supports

X

X

X

X

Maternal/paternal poor coping

Negative peer associations

X

X

X

X

X

X

X

X

Personal Factors (Positive/Strengths)

Emotional maturity

X

Positive, prosocial attitude

X

X

X

X

X

Good problem-solving skills

X

X

X

Good social skills

X

X

X

X

Motivation to address personal issues/Readiness

X

X

Job stability

X

X

X

X

X

X

Educational commitment/achievement

X

X

X

X

X

Positive hobby/sport involvement

X

X

X

X

X

Intelligence

X

X

Positive attitude toward authority/treatment

X

X

X

X

Resilient personality traits

X

Contextual Factors (Positive/Strengths)

Stable & cohesive family unit

X

X

X

X

X

Supportive parent

X

X

X

X

X

X

Supportive other adult

X

X

X

X

Stable marital/partner relationship

X

X

X

Positive peer associations

X

X

X

X

X

X

Strong attachment and bonds

X

X

X

(p.161) (p.162) (p.163)

(p.164) Risk Assessment Instruments Created for Adolescents

Each of the tools reviewed below is completed based on an interview with the youth and collateral contacts, and a review of file information. While reviewing the evidence, it is important to note that most studies of adolescent tools have used short follow-up periods of approximately one year. A few have tracked arrest rates after three to three-and-a-half years. This means the research, as it stands, offers little in terms of prediction of young adulthood/late-onset offending.

Youth Level of Service/Case Management Inventory (YLS/CMI). The YLS/CMI (Hoge & Andrews, 2006) is a static/dynamic actuarial instrument for assessing risk for recidivism for male and female juvenile offenders aged 12 to 17. The instrument uses an “adjusted actuarial approach,” meaning that it has a total score used to determine risk level, but the evaluator can override. IRR between professionals and probation officers report good ICCs (Intraclass Correlation Coefficients) on most subscales (.71 to .85; Schmidt, Hoge, & Robertson, 2005) and .75 for the Total Risk score (Poluchowiz, Jung, & Rawana, 2000).

Predictive validity has been studied prospectively using both file-based and probation-officer completed assessments. A recent meta-analysis of multiple studies, many from independent researchers, confirmed that there was a substantial amount of variability in effect sizes across predictive validity studies for the YLS/CMI. On average, the effect size (weighted r) was .29 for nonviolent re-offending (based on three samples) and .26 for violent re-offending (based on nine samples; Olver, Stockdale, & Wormith, 2009). Areas under the ROC curve (AUCs) have ranged around .57 to .75 (Schwalbe, 2007). Based on a limited number of studies, Olver et al. (2009) noted that the YLS/CMI seems to predict re-offending equally well across genders and for minority youth in U.S. and Canadian settings.

The Structured Assessment of Violence Risk in Youth (SAVRY). The SAVRY (Borum, Bartel, & Forth, 2006) is an SPJ tool comprising static and dynamic risk factors and protective factors for male and female young offenders aged 12 to 18. The ultimate determination of an examinee’s overall level of violence risk is communicated via the Summary Risk Rating (SRR), which is rated as low, moderate, or high, based on the examiner’s judgment as informed by a systematic appraisal of relevant factors. The IRR is good (e.g., ICCs for total scores around .81 and .77 for SRR, Catchpole & Gretton, 2003) as demonstrated by many studies. Predictive validity for both violent and nonviolent offending has been demonstrated in forensic and young offender populations across multiple studies (for a review, see Olver et al., 2009), including circumstances where the assessment was completed by masters-level clinicians or juvenile justice personnel (Vincent, Chapman, & Cook, 2005).

A recent meta-analysis reported weighted rs of .38 for nonviolent re-offending (based on two samples) and .30 for violent re-offending (based on nine samples) for total scores (Olver et al., 2009) with no statistically significant variability in effect sizes across studies. It should be noted that the meta-analysis summarized the predictive accuracy of SAVRY total scores rather than the SRR, which is how the tool is used in (p.165) practice. Studies examining the prediction of the SRR also have reported high accuracy (AUC = .71 for violent re-offending; Lodewijks, Doreleijers, & de Ruiter, 2008). AUCs for the total score have ranged from .65 to .89 among studies with a follow-up period of one year or more (for a review, see Borum, Lodewijks, Bartel, & Forth, 2010).

The Washington State Juvenile Court Assessment (WSJCA) and Youth Assessment and Screening Instrument (YASI). The WSJCA (Barnoski, 2004) is a risk for re-offending assessment tool for use with offenders aged 12 to 18. Most users utilize a computer-based program referred to as the Back on Track, or, more recently, the Positive Achievement Change Tool (PACT). Another adaptation of this tool is the YASI, which was created by Orbis Partners by expanding the WSJCA to include mental health items. These tools come in three parts: a pre-screen, full assessment, and reassessment. The pre-screen is an actuarial tool comprising mostly static items whereas the other tools contain static and dynamic items. Using a sample of 20,339 pre-screen assessments and 12,187 full assessments for youths on probation, Barnoski (2004) found that the felony recidivism rate and violent recidivism rate of the high-risk group were about three times that of the low-risk group after an 18-month follow-up. The AUC was .64 for both violent and nonviolent recidivism. The predictive validity is good for minority groups and girls, but weaker than the predictive validity for whites and boys (Washington State Institute for Public Policy, 2004) and, therefore, a different item-weighting system is recommended for these groups. One study of the PACT, conducted by an independent examiner using a one-year follow-up, indicated that as the overall level of risk for youth arrestees increased, the odds of rearrest increased by 1.5 times (Baglivio, 2009). The AUC, however, was only .59. The IRR has not been reported for any of the measures, so it is unknown if the lower AUC in this study was due to diminished IRR.

North Carolina Assessment of Risk (NCAR). The NCAR is a brief, actuarial tool designed to evaluate risk for re-offending among juvenile offenders. It consists of nine risk items that together comprise a total risk score. The tool, which can be completed easily by juvenile justice personnel, comprises primarily static risk factors, but also contains a few dynamic risk items. Although the name implies that the NCAR is jurisdiction-specific, it is likely to generalize because the items were selected according to general delinquency research and theory rather than statistical associations between indicators and recidivism in this jurisdiction. To our knowledge, there are no rigorous studies of the tool’s inter-rater agreement, but estimates suggest it is adequate (Schwalbe, Fraser, Day, & Arnold, 2004). Schwalbe, Fraser, Day, and Cooley (2006) examined the predictive validity of the NCAR using a sizeable sample of 9,534 delinquent offenders (M = 13.7 years) over an average one-year follow-up. Cox regressions resulted in a hazard ratio of 1.08, meaning that a one-point increase in the NCAR score would be associated with an additional 8% chance of recidivism. The NCAR was predictive among both white and black boys and black girls, but did not differentiate recidivists among white girls. A meta-analysis with three studies indicated AUCs ranged from .60–.68 (Schwalbe, 2007).

Psychopathy Checklist: Youth Version (PCL:YV). The PCL:YV (Forth, Kosson, & Hare, 2003) is a downward extension of the Psychopathy Checklist-Revised (p.166) (PCL-R) for adults that is used to provide a dimensional assessment of the “prototypical” psychopath among adolescents aged 12 to 18. The PCL:YV is an expert symptom-rating scale that defines psychopathy along interpersonal, affective, and behavioral symptom clusters. Scoring of the tool’s 20 items is based on each symptom’s pervasiveness, severity, and chronicity. Total and factor (i.e., symptom cluster) scores are dimensional, meaning that they represent the level or severity of psychopathic traits. The PCL:YV is not a risk assessment tool per se but is used in risk assessments in light of research that indicates that youth who score high on the PCL:YV have more serious and persistent criminal histories with an earlier onset than those who score low (e.g., Brandt, Wallace, Patrick, & Curtin, 1997). A meta-analysis (Edens, Campbell, & Weir, 2007) of PCL:YV total scores yielded medium effect sizes for general (r = .26) and violent recidivism (r = .23). PCL:YV total scores have high IRR (r = .90–.96) across the various populations (i.e., institutional, probation, or community) tested by many independent researchers.

Generally, high PCL:YV scores in youth may compel a conclusion of high risk over short periods of time. However, given that adolescence is a time of extreme developmental change and most studies employing PCL:YV have not used follow-up periods into adulthood, risk and psychopathic characteristics should be reassessed routinely in high-risk youth to determine if maturation attenuates risk. Since the PCL:YV is not a diagnostic tool, there are no established valid cutoffs for making categorical decisions of whether someone is, or is not, a psychopath (Forth et al., 2003). Other cautions about use of the PCL:YV is that in girl samples there is not yet good evidence that the tool predicts rearrest well (Odgers, Moretti, & Reppucci, 2005; Vincent, Odgers, McCormick, & Corrado, 2008).

Risk Assessment Instruments Created for Adults

Classification of Violence Risk (COVR). The COVR (Monahan et al., 2005) is an interactive software program developed to estimate the risk that a psychiatric in-patient will be violent towards others in the community. In contrast to most actuarial tools that are created using a regression approach, the COVR is based on a multiple iterative classification tree (ICT) method. As such, individuals can be classified at the same level of risk based on different combinations of risk factors. The particular questions administered to an individual depend on the answers given to prior questions. The software can assess 40 risk factors, but in practice only enough risk factors necessary to classify the patient’s violence risk are considered. In the development sample, psychiatric subjects were placed into one of five risk groups whose likelihood of violence to others over the next 20 weeks ranged from 1% (the ‘low-risk’ group) to 76% (the ‘high-risk’ group), with a resulting AUC of .88. The recidivism rates for the low- and high-risk groups in a cross-validation analysis were respectively 9% and 35% and the AUC was .63. An AUC of .70 was obtained when a modified follow-up procedure was used. Despite the relatively large drop in predictive (p.167) accuracy (which is to be expected to some extent when any actuarial method is cross-validated), results indicated that the COVR’s use of multiple ICT models was a useful aid in decision-making.

Historical-Clinical-Risk Management-20 (HCR-20). The HCR-20 (Webster, Douglas, Eaves, & Hart, 1997) adheres to the SPJ model of risk assessment and was developed to facilitate assessments of risk for violence. The tool comprises 20 items across three scales, including an historical scale, which is static, and clinical and risk scales, which are dynamic. Like the SAVRY, evaluators provide SRRs after considering all the factors. Research from independent parties provides support for the use of the HCR-20 across both genders with forensic and civil psychiatric patients, mentally disordered offenders, and general correctional offenders. The median IRR coefficient for the HCR-20 total score across 25 studies was .85 (Douglas & Reeves, 2009). Fewer studies have reported IRR for SRRs; across nine values (from five studies), the median ICC was .65. This is expected to be lower relative to the multi-item index because the SRR is a single item. There have been over 50 evaluations of the tool’s predictive validity across approximately a dozen countries (Guy, 2008). Overall, findings indicate that HCR-20 numeric total scores, numeric scale scores, and SRRs are associated with violence with average effect sizes of moderate to large magnitude, and that the tool performs comparably across genders and countries. In a recent meta-analysis (Guy, 2008), the weighted mean AUC for the numeric total score for violence was .73 (based on 14 effect sizes). Across the six studies that have investigated SRRs, results indicate the categorical estimates are as or more strongly related to violence than are the numeric total, with four studies demonstrating that the SRR added incrementally to the numeric total.

Violence Risk Appraisal Guide (VRAG). The VRAG (Harris, Rice, & Quinsey, 1993; Quinsey et al., 1998) is an actuarial tool developed to assess the risk of violence among male forensic populations. There have been few studies done with women. Most of the tool’s 12 items are static risk factors that in multiple regression analyses made independent and incremental contributions to predictive accuracy among approximately 50 variables that were available in institutional records. Several reliability and validity studies have been conducted by independent examiners. Harris et al. (1993) reported a high IRR (r = .90). The AUC for violent recidivism in the development sample was .76. Among the over 30 independent replications of the VRAG (see www.mhcp-research.com), typically moderate to high predictive accuracy for violent recidivism has been obtained, although the average of cross-validated effect sizes is lower (as expected based on the actuarial derivation approach). Campbell, French, and Gendreau (2009) reported a weighted mean correlation of .32 for violent recidivism (based on 14 effect sizes). Predictive accuracy is not moderated by nationality (there are replications in seven countries), length of follow-up (mean durations have ranged from 12 weeks to 10 years), or how violent recidivism has been operationalized (Quinsey et al., 1998).

(p.168) Level of Service Inventory-Revised (LSI-R). The LSI-R (Andrews & Bonta, 1995) is an actuarial tool developed according to the principles advanced by the Risk/Needs/Responsivity framework (Andrews & Bonta, 2006). It comprises 54 items primarily representing dynamic risk factors (i.e., criminogenic needs), across 10 domains. It is distinct from other actuarial tools in that its items were selected rationally, rather than statistically, on the basis of prior evidence of their association with recidivism. Another unique feature is that it was developed using a gender-balanced sample (40% male; Andrews et al., in press). Although the LSI was developed for use with general offender populations, research also supports its use with mentally disordered offenders and forensic psychiatric patients.

A large body of research supports the reliability and predictive utility of the LSI-R for general recidivism among men and women. For example, when the reassessment interval was less than a month, the mean inter-rater reliability across six samples was .92 (Andrews, Bonta & Wormith, in press). Based on 19 prospective studies, Campbell et al. (2009) reported a weighted mean correlation of .28 for violent recidivism. Gendreau, Goggin, and Smith (2002) meta-analyzed 33 effect sizes and reported the aggregate correlation between the LSI-R and general recidivism to be .37. Meta-analytic evidence (Smith, Cullen, & Latessa, 2009) from 27 effect sizes yielded an association of a similar magnitude (r = .35), and gender did not moderate predictive accuracy in this research.

A revised version of the tool, the Level of Service/Case Management Inventory LS/CMI; Andrews, Bonta, & Wormith, 2004), has been developed in which the 54 LSI-R items were refined and combined into 43 items. The LS/CMI is unique in that it is a fully functioning case management tool.

Statistical Information for Recidivism Scale (SIR). The SIR (Nuffield, 1982) is an actuarial tool developed by the National Parole Board of Canada in 1975 (originally referred to as the General Information for Recidivism Scale, or GSIR) to assess risk for recidivism among offenders released from Canadian penitentiaries. The tool comprises 15 primarily static risk variables covering demographic characteristics and criminal history. Satisfactory levels of inter-rater agreement have been reported. For example, Wormith and Goldstone (1984) reported an 85% agreement rate when comparing parole officer and researcher scores. Adequate levels of predictive validity have also been reported for men, although validity coefficients for female samples tend to be lower than for male samples. Campbell et al. (2009) reported a weighted correlation of .22 based on 17 effect sizes for the prospective prediction of violent recidivism in male offenders from SIR scores. At present, research supports the use of the SIR only with men from nonaboriginal ethnic backgrounds.

Psychopathy Checklist—Revised (PCL-R). The PCL-R (Hare, 1991, 2003) is a structured measure of psychopathic traits that often is used as a risk prediction tool in light of its strong association with future offending (especially violent offending). Its items are organized into two correlated factors representing Affective/Interpersonal (Factor 1) and Socially Deviant/Impulsive Lifestyle (Factor 2) characteristics.

(p.169) The PCL-R manual reports adequate IRR ratings for pooled samples of male offenders (ICC1 = .86, ICC2 = .92), male forensic patients (ICC1 = .88, ICC2 = .93), and female offenders (ICC1 = .94, ICC2 = .97). As noted earlier, several meta-analyses reported strong associations between PCL-R scores and violence in institutions (e.g., Guy, Edens, Anthony, & Douglas, 2005) and the community (e.g., Leistico et al., 2008) with Factor 2 being a stronger predictor of recidivism than Factor 1. In the most recent and comprehensive quantitative synthesis, Leistico and colleagues (2008) aggregated effect sizes from 95 nonoverlapping studies (representing 15,826 individuals). The median weighted Cohen’s d values for the total, Factor 1,and Factor 2 scores were .57, .38, and .58, respectively. Importantly, in this and previous meta-analyses, substantial heterogeneity in effect sizes was observed, which suggests that it is likely that the association between PCL-R scores and recidivism is moderated by several factors (e.g., race, gender, institutional setting, and country in which data were collected).

Review of Meta-Analyses Comparing Risk Assessment Instruments

Several meta-analyses have compared one or more risk assessment instruments used with juveniles (Olver, Stockdale, & Wormith, 2009; Schwalbe, 2007, 2008) or adults (Campbell et al., 2009; Gendreau, Goggin, & Little, 1996; Gendreau, Goggin, & Smith, 2002; Guy, 2008; Hanson & Morton-Bourgon, 2009; Walters, 2006). Taken together, the studies comparing the aggregate predictive validity of various actuarial and SPJ tools indicate that there often is no definitive advantage to either approach with respect to predicting who will re-offend. When differences are observed, they most often are in the direction of the actuarial approach; the magnitudes of the differences, however, typically are small. For example, examining all available studies in which an SPJ and an actuarial measure were studied in the same sample, Guy (2008) found that, regardless of whether the predictor for the SPJ tool was the SRR or the numeric rating, the mean-weighted effect sizes for the SPJ and actuarial approaches were moderate in size and virtually identical for all comparisons. For violent (including sexual) recidivism, the mean-weighted AUC values in the SRR versus actuarial comparison were both .61, and for the SPJ numeric total score versus actuarial comparison, the corresponding mean-weighted AUC values were .71 and .68, respectively.

Examining several tools for use with adults with respect to their predictive validity for general recidivism, Gendreau, Goggin, and Little (1996) reported that the LSI-R had a slightly larger mean effect size (.33, k = 28) compared to the Wisconsin Classification System (Baird, 1981 .32, k = 14) or PCL-R (.29, k = 9). Gendreau and colleagues later compared the predictive validity of the LSI/LSI-R relative to the PCL/PCL-R for violent recidivism (Gendreau, Goggin, & Smith, 2002) and found that the effect sizes were highly similar in studies in which both tools were used in the same sample. Mean correlations for the LSI-R and the (p.170) PCL-R for general recidivism were .37 and .26, respectively, and for violent recidivism were .24 and .22, respectively. A similar pattern was observed for between-study comparisons of the tools.

Campbell et al. (2009) examined prospective evaluations of forensic psychiatric patients and general offenders in which the HCR-20, LSI/LSI-R, PCL/PCL-R, SIR, or VRAG was used. The authors concluded that each tool predicted violent recidivism with at least a moderate magnitude of success, and that although the LSI-R, PCL-R, and SIR yielded the most precise point estimates, no one measure should be singled out as being most effective in predicting violent recidivism. Importantly, the authors did not evaluate the performance of the only SPJ tool included (the HCR-20) as it was intended to be used in clinical practice (i.e., SRRs were not included, which, as noted above, consistently yield larger effect sizes as compared to the numeric total scores for the HCR-20).

With respect to youth instruments, examining the predictive validity of 42 effect sizes coded from 33 samples that reported on 28 risk assessment instruments (primarily actuarial tools), Schwalbe (2007) reported that the mean-weighted AUC for all tools versus recidivism was .64; substantial variability was observed (range of AUCs: .55–.78). Relative to “first” or “second” generation tools, higher levels of accuracy were observed among “third generation” tools, such as the YLS/CMI. Similar to findings for the overall sample, the weighted mean AUC value (derived from 11 effect sizes) for the YLS/CMI was .64 (95% C.I. = .51–.78).

In a more recent meta-analysis of the YLS/CMI, PCL:YV, and SAVRY from 44 samples that represented 8,746 youth, Olver et al. (2009) found that all three measures were significantly associated with general, nonviolent, and violent recidivism with comparable, moderate degrees of accuracy. In head-to-head comparisons of tools used in the same sample, mean effect sizes were comparable for the YLS/CMI and PCL:YV/SAVRY comparisons. Of the two studies in which the YLS and SAVRY were compared directly, predictive accuracy was comparable in one study (Catchpole & Gretton, 2003), and the SAVRY was more predictive in the other (Welsh et al., 2008). However, it is important to note that in the latter study the SAVRY was completed by trained masters level research assistants retrospectively based on file reviews whereas the YLS/CMI had been completed by probation officers in the field. Thus, comparisons should be made with caution. The weighted mean correlation from these two studies was .43 for the SAVRY and .29 for the YLS.

Secondary Analyses

Our review indicates that there has been considerable progress in the area of risk assessment to predict and prevent later offending among offending populations. However, whether these tools maintain their relevance (or predictive accuracy) during different developmental periods is unknown. In other words, the question (p.171) is whether tools are as adept at predicting re-offending behavior occurring during early adulthood as tools are at predicting re-offending behavior during adolescence (for youth tools, which are generally valid for ages 12 to 17) or later adulthood (for adult tools, valid for ages 18 and older).

The co-authors and colleagues conducted a very preliminary investigation into the validity of assessment tools for assessing risk at different developmental periods (Vincent, Fusco, Gershenson, & Guy, 2010, unpublished data). This involved a search of all publications and unpublished datasets known to the authors which, (a) included one of the assessment tools reviewed in this chapter, (b) studied an offender (corrections, pre-trial detention, or probation) or forensic psychiatric population (civil psychiatric samples were not included), and (c) had a measure of re-offending in the community (self-report or official records). The researchers requested datasets from the authors to be used in a secondary data analysis. The datasets obtained and included in these analyses are described in Table 6.2. Follow-up periods for these datasets spanned one to seven years. Data sets were then combined by risk tool to examine the following recidivism outcomes: (1) any re-offending, which includes violent re-offending and all other crime types (except status offenses), and (2) violent re-offending only. Studies varied in the manner that they operationalized recidivism (see Table 6.2)

Table 6.2 Descriptions of studies included in the secondary data analysis.

n

% Female

% Minority

Instrument

Recidivism:

Charges or Convictions

North Carolina Department of Juvenile Justice (unpublished data)

de Vogel, de Ruiter, Hildebrand, Bos, & van de Ven (2004)

Douglas (1999)

Harris, Rice, & Cormier (2002)

Lodewijks, de Ruiter, & Doreleijers (2008)

Loza, Villeneuve, & Loza-Fanous (2002)

Rowe (2002)

Vincent, Quinlan, Nitschelm, & Ogloff (2003)

Vincent, Chapman, & Cook (under review)

Welsh, Schmidt, McKinnon, Chattha, & Meyers (2008)

6,751

119

279

1,190

82

278

408

248

757

133

27.0

10.9

40.1

NR

42.7

0.0

19.9

12.0

30.7

37.6

58.6

18.5

21.5

NR

37.8

NR

NR

26.6

64.1

31.6

NCAR

HCR-20

HCR-20

VRAG

SAVRY

VRAG

YLS

HCR-20; VRAG

SAVRY

SAVRY; YLS

Charges

Convictions

Self-report

Charges

Charges

Charges

Charges

Convictions

Charges

Charges

Note: NCAR = North Carolina Assessment of Risk; HCR-20 = Historical Clinical Risk Management- 20; VRAG = Violence Risk Appraisal Guide; YLS/CMI = Youth Level of Service/Case Management Inventory; SAVRY = Structured Assessment of Violence Risk for Youth; NR = Not Recorded.

.

The first question was whether the tools were accurate at predicting re-offending during early adulthood (defined as ages 18 to 25), relative to adolescent re-offending (for youth tools) or later adult re-offending (for adult tools). This required generating sub-samples that included only cases that had a follow-up period spanning early adulthood. Thus, the researchers included only cases that had been assessed on an instrument (or released from an institution where applicable) prior or equal to age 23, which would provide at least one or two years follow-up into early adulthood. Cases that were not tracked until age 19 or older also were excluded from the analyses. As such, a considerable number of cases were dropped from many of the datasets. The final samples used in subsequent analyses are provided in Table 6.3 for the youth tools and Table 6.4 for the adult tools.

Table 6.3 AUCs and Cox regressions for risk assessment tools for adolescent populations by developmental period of first re-offense.

Whole Usable Sample

Recidivists During Early Adulthood

Recidivists During Adolescence

n

Base rate Re-offense

Base rate Violence

Base rate Re-offense

AUC (CI)

Base rate Violence

AUC (CI)

Base rate Re-offense

AUC (CI)

Base rate Violence

AUC (CI)

NCAR

2129

17%

MIS

NA

NA

17%

.61*** (.58-.64)

MIS

SAVRY

786

84%

41%

7%

.59 (.50-.69)

7%

.63**(.56-.71)

77%

.53 (.48-.58)

34%

.56* (.52-.60)

YLS

425

61%

26%

4%

.67* (.52-.82)

3%

NV

57%

.74**

(.69-.80)

23%

.67*** (.60-.73)

Cox Regressions

NCAR

Re-offense- χ2(1) = 40.31; Exp(β) = 1.07***

SAVRY

Re-offense- χ2(1) = 3.80; Exp(β) = 1.04*

Violence - χ2(1) = 11.89; Exp(β) = 1.07***

Re-offense- χ2(1) = 4.20; Exp(β) = 1.01*

Violence - χ2(1) = 18.09; Exp(β) = 1.03*

YLS

Re-offense- χ2(1) = 5.82; Exp(β) = 1.06*

Re-offense- χ2(1) = 69.17; Exp(β) = 1.06***

Violence- χ2(1) = 23.82; Exp(β) = 1.06***

NOTE: The recidivist columns include only samples of cases that committed their first re-offense during that time period. No early adulthood analyses are provided for the NCAR sample because all recidivists committed their first offense during adolescence. AUC = Area Under the Curve. CI = 95% Confidence Interval. NA = not applicable. NV = cannot provide a valid calculation due to low sample sizes. MIS = cannot provide a calculation due to considerable missing data or information unavailable for all data sets. * = p 〈 .05, ** = p 〈 .01, *** = p 〈 .001.

Table 6.4 AUCs and Cox regressions for risk assessment tools for early to older adult populations by developmental period of first re-offense.

Whole Usable Sample

Recidivists During Early Adulthood

Recidivists During Adulthood

n

Base rate Re-offense

Base rate Violence

Base rate Re-offense

AUC (CI)

Base rate violence

AUC (CI)

Base rate Re-offense

AUC (CI)

Base rate violence

AUC (CI)

VRAG

589

MIS

44%

MIS

15%

.68*(.58-.78)

MIS

30%

.71**(.65-.77)

HCR-20

127

49%

39%

43%

.73**

(.65-.89)

34%

.82*** (.72-.93)

6%

NV

5%

NV

Cox Regressions

VRAG

Violence - χ2(1) = 7.6; Exp(β) = 1.04**

Violence - χ2(1) = 30.71; Exp(β) = 1.05***

HCR-20

Re-offense - χ2(1) = 19.31; Exp(β) = 1.15***

Violence - χ2(1) = 15.87; Exp(β) = 1.14***

NV

Note: The recidivist columns include only samples of cases that committed their first re-offense during that time period. AUC = Area Under the Curve. CI = 95% Confidence Interval. NV = cannot provide a valid calculation due to low sample sizes. MIS = cannot provide a calculation due to considerable missing data or information unavailable for all data sets. * = p 〈 .05, ** = p 〈 .01, *** = p 〈 .001.

The researchers used Receiver Operator Characteristic (ROC) curves, calculated in SPSS, to examine the predictive accuracy of each instrument for any re-offending (includes all types of offenses) and for violent re-offending specifically. Tables 6.3 and 6.4 report the AUC and confidence intervals for each instrument as a function of the developmental period during which cases committed their first re-offense. The tables provide AUCs for both any re-offending and violent re-offending for each tool as the data permitted. AUCs ranged from non-significant at .53 (for the SAVRY’s prediction of any re-offending) to .82 (for the HCR-20’s prediction of violent offending during early adulthood). The MedCalc software was used to test whether AUCs differed significantly between the early adult and other developmental periods for each risk assessment tool. None of the AUCs were significantly different.

What can be said from the findings is that tools with statistically significant accuracy maintain that accuracy regardless of the developmental period in which (p.172) (p.173) re-offending occurred. Although we can compare AUCs (a measure of effect size) within tools, one should not try to interpret differences between tools. The analyses reported here did not account for confounding variables that would affect outcomes across studies, such as length of follow-up, operational definition of recidivism, and sample demographics (e.g., gender, race).

A significant limitation of ROC analyses is that they do not account for the length of time in which individuals had an opportunity to re-offend (time at risk). Therefore, the researchers conducted a second set of analyses using Cox Proportional-Hazards Regressions for each tool in each time period in which samples permitted a valid analysis in order to account for the potential confounding effects of time at risk. Cox regression is a semi-parametric test that models the relation between predictor variables and an event (i.e., any re-offense or violent re-offense) while accounting for time to the occurrence of the event. The dependent variable, time at risk to re-offend, is based on the cumulative survival function; that is, the proportion of cases “surviving” (i.e., not charged with or convicted of a new offense) at a particular point in time. The Cox regressions included all cases regardless of whether they had re-offended by estimating time to a hypothetical event for these censored cases. Inclusion of censored cases is essential because each released individual who has not recidivated theoretically could still be arrested in the future. The preferred index for interpretation is the Hazard function (Exp[β]), a measure of the likelihood (odds) of a case experiencing an event, given it has survived that long. For example, an Exp[β] of 1.40 indicates that a one-unit increase in the risk assessment total score would result, on average, in a 40% increase in the odds of a rearrest, given a case has survived this long.

(p.174) Time at risk was calculated separately for any re-offense and violent re-offense based on the days between the adjudication for the index offense (the offense that resulted in the subject being included in the sample and when the risk assessment was completed) or release from an institution (for samples from correctional or forensic hospital settings) and the first rearrest. For nonrecidivists, time at risk was calculated according to the final follow-up date for the whole sample. Results of the Cox regressions are provided in Tables 6.3 and 6.4. The tables show that the risk tools were significant predictors of both violent and any re-offending after taking time at risk into account, regardless of the developmental period in which the offense occurred. The odds were similar for early adult recidivists and adolescent recidivists for most adolescent tools, with the exception of the SAVRY. For the adult tools, odds were also similar for early adult and adult recidivists.

In sum, risk assessment tools predicted re-offending during early adulthood as well as or better than re-offending that occurred during adolescent years for individuals who were first assessed and first offended during adolescence. Adult risk assessment tools conducted during early adulthood also were adept at predicting re-offending during this developmental period. This was true even after controlling for the amount of time an individual had to re-offend.

Unfortunately, there are several constraints on these analyses that limit the conclusions. Because of the selection of participants, the datasets cannot be used to determine what variables predict early adult-onset offending. Risk assessment tool validation studies tend to include only cases that committed an offense in a set time period onward. For many of the offenders in these studies, their index offense in the study would not be the first offense they ever committed in their life. It would be necessary to use data from longitudinal studies of young community samples in order to determine which variables predicted early adult-onset offending. Risk assessment validation studies tend not to gather data from people before they have ever offended. A second limitation is that the data did not permit examination of continued offending into early adulthood for adolescent samples. The datasets contained each case’s first re-offense (any type), first violent re-offense, and in some instances the first nonviolent re-offense specifically. Therefore, we do not have information about continued offending. The only exception was the NCAR data set. Unfortunately, analyses for early adulthood could not be conducted for the NCAR dataset because a very small number of youth re-offended in early adulthood during the two-year follow-up period (approximately 20).

Conclusions and Recommendations

Recommendations for research. The most important conclusion from our review of the literature concerns the need for more information about the risk, need, and protective factors associated with criminal activity occurring during ages 18 to 29. (p.175) We already have considerable information about factors influencing antisocial behaviors during childhood and adolescence and for ages 18 and older. However, information about the specific periods of later adolescence and early adulthood is sparse. Likewise, information about factors that can identify those who will initiate offending during this age period is very sparse. In addition, there is no evidence yet to show that risk screening instruments can differentiate between different developmental trajectories of offending during the transition between adolescence and adulthood (for trajectories see Piquero et al., Chapter 2, this volume). Researchers are encouraged to focus more carefully on that period and investigate screening instruments that can distinguish between different developmental trajectories of offending.

Of particular concern in assessments is the influence of normative life events for the early adult period, including school and work experiences, changes in bonds with parents, and the establishment of new relationships (see Horney et al., Chapter 4, this volume; Thornberry et al., Chapter 3, this volume). Contacts with the juvenile and adult justice and correctional systems are also of relevance for many youth. These variables are often not represented in the research, or, if they are, interactions among them are often not explored. For example, it is possible that forming a positive romantic relationship during early adulthood can mediate the effects of early school drop-out and criminal convictions on later offending. It is this type of interaction that must be explored before definitive answers can be found to the question of variables influencing early adult persistence and desistance (Thornberry, 2005).

The validity of generalizing current knowledge about the risk, need, and strength factors associated with early adult offending is also of concern. Much of the research is based on samples of American, Canadian, and British males from the majority culture. Our knowledge of the dynamics of early adult crime among females and those from minority ethnic and cultural groups is limited.

Limitations to our knowledge of the factors affecting early adult offending impact, of course, the validity of our assessment tools. We have seen that considerable progress has been made in developing and evaluating instruments for assessing risk, need, and strength factors in youth and in adults. Further, most of these measures display construct validity to the extent that they include the major risk factors associated with criminal activity. Indeed, the meta-analyses reviewed indicate that all the tools selected for review in this chapter are comparable in terms of their predictive accuracy.

The secondary analyses provide some important additional insights into the ability of risk assessment tools to predict re-offending occurring at different developmental periods, especially early adulthood. The tentative conclusion is that these tools predict re-offending during early adulthood as well as other time periods. Future analyses should use longitudinal studies of community samples that would permit post hoc scoring of risk assessment tools using variables available in the dataset (although a limitation of this approach would be that the investigation would be retrospective in nature). Additionally, risk assessment researchers could (p.176) start recording data for every re-offense occurring during their follow-up periods so trajectory analyses can be conducted in the future, permitting examination of the prediction of continued offending versus desistance in early adulthood. Future efforts also could determine whether specific risk and criminogenic need factors within these risk assessment tools differentially predict continued offending.

Implications for practitioners. The conclusions of this review provide some important guidelines for practitioners and policy-makers. First, based on a wealth of findings, there are specific risk factors that arguably should be contained in any risk assessment tool for youth in order to maximize its effectiveness. These factors include impulsivity, remorselessness, callousness-unemotional traits, inconsistent or lax parental discipline, and early onset violence. However, the relevance of these risk factors to the onset of offending or continued offending in early adulthood (ages 18 to 29) is not well established. Further research on the issue of early adult-onset offending may lead to the development of a new instrument or the modification of an existing measure.

Second, it should not be assumed that the factors associated with the initiation in or desistance from criminal activity are the same for early adulthood as for earlier or later developmental stages. In particular, normative transitions relating to school, work, parental bonds, and the establishment of new relationships often assume unique importance during this period. For example, risk assessment tools for adults should contain a measure of psychopathic personality or Cluster B-type personality traits, given the moderate association with violence, yet it is unlikely that psychopathy would be a good predictor of early adult-onset offending since these individuals tend to initiate offending behavior much earlier in life.

We have seen that existing youth and adult risk/needs prediction instruments are of roughly equal value in terms of construct and predictive validity. However, instruments do vary somewhat in the variables represented, and the instrument selected should include variables identified as having particular relevance for the early adult period.

There also may be implications for interventions geared towards risk management or prevention of continued offending. Treatment strategies that lead to desistance among adolescents may not work for those initiating offending later in life. For example, standard adolescent treatment programs for attitude change, peer group affiliation, or family therapies may not address the criminogenic needs of individuals initiating offending in early adulthood.

Finally, lessons for broad systemic changes exist. The period of early adulthood has been traditionally neglected when it comes to educational, vocational, mental health, and social services. Within most systems, individuals aged 17 to 21 are shifted out of the adolescent services systems, and there is often little to replace those services. Counseling and other treatment/support services, to assist individuals to cope with substance abuse, employment, and relationship issues arising during this period, could ease the transition and help individuals avoid the problems that often characterize these years.

(p.177) References

Bibliography references:

American Psychological Association (2002). Ethical principles of psychologists and code of conduct. American Psychologist, 57, 1060–1073.

Andrews, D.A. & Bonta, J. (1995). Level of Service Inventory-Revised. Toronto, ON: Multi-Health Systems.

Andrews, D.A., Bonta, J., & Wormith, S. (2004). Manual for the Level of Service/Case Management Inventory (LS/CMI). North Tonawanda, NY: Multi-Health Systems.

Andrews, D.A., Bonta, J., & Wormith, S.J. (in press). The Level of Service/RNR. Toronto, ON: Multi-Health Systems.

Atkins, D., Pumariega, W., & Rogers K. (1999). Mental health and incarcerated youth. In: Prevalence and nature of psychopathology. Journal of Child and Family Studies, 8, 193–204.

Austin, J. (September, 2006). How much risk can we take? The misuse of risk assessment in corrections. Federal Probation, 70, 58–63.

Baglivio, M. (2009). The assessment of risk to recidivate among a juvenile offending population. Journal of Criminal Justice, 37, 596–607.

Baird, C.S. (1981). Probation and parole classification: The Wisconsin model. Corrections Today, 43, 36–41.

Baird, C., & Wagner, D. (2000). The relative validity of actuarial- and consensus-based risk assessment systems. Children and Youth Services Review, 22, 839–871.

Barnoski, R. (2004). Washington State Juvenile Court Assessment manual, Version 2.1 (Report No. 04-03-1203). Olympia, WA: Washington State Institute for Public Policy.

Boccaccini, M.T., Turner, D.B., & Murrie, D.C. (2008). Do some evaluators report consistently higher or lower PCL-R scores than others? Findings from a statewide sample of sexually violent predator evaluations. Psychology, Public Policy, and Law, 14, 262–283.

Bonta, J. (1996). Risk-needs assessment and treatment. In A.T. Harland (Ed.), Choosing correctional options that work: Defining the demand and evaluating the supply (pp.18–32). Thousand Oaks, CA: Sage.

Bonta, J.S., Law, M, & Hanson, R.K. (1998). The prediction of criminal and violent recidivism among mentally disordered offenders: A meta-analysis. Psychological Bulletin, 122, 123–142.

Borum, R. & Douglas K. (2003). New directions in violence risk assessment. Psychiatric Times, 20, 102–103.

Borum, R. (1996). Improving the clinical practice of violence risk assessment: Technology, guidelines, and training. American Psychologist, 51, 945–956.

Borum, R., Bartel, P., & Forth, A. (2006). Structured Assessment of Violence Risk in Youth (SAVRY). Lutz, FL: Psychological Assessment Resources.

Borum, R., Lodewijks, H., Bartel, P., & Forth, A. (2010). Structured Assessment of Violence Risk in Youth (SAVRY). In R.K. Otto & K.S. Douglas (Eds.), Handbook of violence risk assessment (pp. 63–80). New York: Routledge.

Borum, R., & Verhaagen, D. (2006). Assessing and managing violence risk in juveniles. New York: Guilford Press.

Brandt, J.R., Wallace, A.K., Patrick, C.J., & Curtin, J.J. (1997). Assessment of psychopathy in a population of incarcerated adolescent offenders. Psychological Assessment, 9, 429–435.

Burke, J., Loeber, R., & Lahey, B. (2007). Adolescent conduct disorder and interpersonal callousness as predictors of psychopathy in young adults. Journal of Clinical Child and Adolescent Psychology, 36, 334–346.

(p.178) Campbell, M.A., French, S., & Gendreau, P. (2009). The prediction of violence in adult offenders. Criminal Justice and Behavior, 36, 567–590.

Catalano, R.F., & Hawkins, 1996. The social development model: A theory of antisocial behaviour. In J.D. Hawkins (Ed.), Delinquency and crime: Current theories (pp. 149–197). Cambridge, UK: Cambridge University Press.

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

Charney, D., Deutch, A., Krystal, J., Southwick, S., & Davis, M. (1993). Psychological mechanisms of posttraumatic stress disorder. Archives of General Psychiatry, 50, 294–305.

Committee on Ethical Guidelines for Forensic Psychologists (1991). Specialty guidelines for forensic psychologists. Law and Human Behavior, 15, 655–665.

Copeland, W.E., Miller-Johnson, S., Keeler, G., Angold, A., & Costello, E.J. (2007). Childhood psychiatric disorders and young adult crime: A prospective, population-based study. American Journal of Psychiatry, 164, 1668–1675.

de Vogel, V., de Ruiter, C., Hildebrand, M., Bos, B., & van de Ven, P. (2004). Type of discharge and risk of recidivism measured by the HCR-20: A retrospective study in a Dutch sample of treated forensic psychiatric patients. International Journal of Forensic Mental Health, 3, 149–165.

Douglas, K.S., Ogloff, J.R.P., Nicholls, T.L., & Grant, I. (1999). Assessing risk for violence among psychiatric patients: The HCR-20 Violence Risk Assessment Scheme and the Psychopathy Checklist: Screening Version. Journal of Consulting and Clinical Psychology, 67, 917–930.

Douglas, K.S., & Reeves, K. (2009). The HCR-20 violence risk assessment scheme: Overview and review of the research. In R.K. Otto & K.S. Douglas (Eds.), Handbook of violence risk assessment (pp. 147–186). Oxford UK: Routledge.

Douglas, K.S., Yeoman, M., & Boer, D.P. (2005). Comparative validity analysis of multiple measures of violence risk in a sample of criminal offenders. Criminal Justice and Behavior, 32, 479–510.

Dvoskin, J.A., & Heilbrun, K. (2001). Risk assessment: Release decision-making toward resolving the great debate. Journal of the American Academy of Psychiatry and the Law, 29, 6–10.

Edens, J., Campbell, J. & Weir, J. (2007). Youth psychopathy and criminal recidivism: A meta-analysis of the Psychopathy Checklist measures. Law and Human Behavior, 31, 53–75.

Elliott, D.S., Huizinga, D., & Menard, S. (1989). Multiple problem youth: Delinquency, substance use and mental health problems. New York: Springer-Verlag.

Farrington, D.P. (2004). Conduct disorder, aggression, and delinquency. In R.M. Lerner, & L. Steinberg (Eds.), Handbook of adolescent psychology (2nd ed.; pp. 627–624). New York: Wiley.

Farrington, D.P., & West, D.J. (1993). Criminal, penal, and life histories of chronic offenders: Risk and protective factors and early identification. Criminal Behaviour and Mental Health, 3, 492–523.

Farrington, D.P. (2005). (Ed.) Integrated developmental and life-course theories of offending. New Brunswick, NJ: Transaction.

Frick, P., J. Kimonis, E., Dandreaux, D., & Farell, J. (2003). The 4-year stability of psychopathic traits in non-referred youth. Behavioral Sciences and the Law, 21, 713–736.

(p.179) Frick, P.J., Lilienfeld, S.O., Ellis, M., Loney, B., & Silverthorn, P. (1999). The association between anxiety and psychopathy dimensions in children. Journal of Abnormal Child Psychology, 27, 383–392.

Frick, P., O’Brien, B., Wotton, J., & McBurnett, K. (1994). Psychopathy and conduct problems in children. Journal of Abnormal Psychology, 103, 700–707.

Forth, A.E., Kosson, D.S., & Hare, R.D. (2003). Hare Psychopathy Checklist: Youth Version. Toronto, ON: Multi-Health Systems.

Gendreau, P., Goggin, C., & Smith, P. (2002). Is the PCL-R really the “unparalleled” measure of offender risk? A lesson in knowledge cumulation. Criminal Justice and Behavior, 29, 397–426.

Gendreau, P., Little, T., & Goggin, C.E. (1996). A meta-analysis of the predictors of adult offender recidivism: What works! Criminology, 34, 401–433.

Grisso, T., & Appelbaum, P. (1992). Is it unethical to offer predictions of future violent behavior? Law and Human Behavior, 16, 621–633.

Grisso, T., & Tomkins, A.J. (1996). Communicating violence risk assessments. American Psychologist, 51, 928–930.

Grove, W.M., & Meehl, P.E. (1996). Comparative efficiency of informal (subjective, impressionistic) and formal (mechanical, algorithmic) prediction procedures: The clinical-statistical controversy. Psychology, Public Policy, and Law, 2, 293–323.

Guerra, N.G., Williams, K.R., Tolan, P.H., & Modecki, K.L. (2008). Theoretical and research advances in understanding the causes of juvenile offending. In R.D. Hoge, N. Guerra, & P. Boxer (Eds.), Treating the juvenile offender (pp. 33–53). New York: Guilford Press.

Guy, L.S. (2008). Performance indicators of the structured professional judgment approach for assessing risk for violence to others: A meta-analytic survey. Doctoral dissertation, Simon Fraser University, Burnaby, BC.

Guy, L.S., Edens, J.F., Anthony, C., & Douglas, K.S. (2005). Does psychopathy predict institutional misconduct among adults? A meta-analytic investigation. Journal of Consulting and Clinical Psychology, 73, 1056–1064.

Hanson, R.K., & Morton-Bourgon, K.E. (2009). The accuracy of recidivism risk assessments for sexual offenders: A meta-analysis. Psychological Assessment, 21, 1–21.

Hare, R.D. (1991). Manual for the Hare Psychopathy Checklist—Revised. Toronto, ON: Multi-Health Systems.

Hare, R.D. (2003). Manual for the Hare Psychopathy Checklist—Revised (2nd ed.). Toronto, ON: Multi-Health Systems.

Harris, G., Rice, M., & Cormier, C. (2002). Prospective replication of the Violence Risk Appraisal Guide in predicting violent recidivism among forensic patients. Law and Human Behavior, 26, 377–394.

Harris, G., Rice, M., & Quinsey, V. (1993). Violent recidivism of mentally disordered offenders: The development of a statistical prediction instrument. Criminal Justice and Behavior, 20, 315–335.

Hart, S.D. (2003). Actuarial risk assessment: commentary on Berlin, et al. Sexual Abuse, 15, 377–382.

Heilbrun, K. (1992). The role of psychological testing in forensic assessment. Law and Human Behavior, 16, 257–272.

Heilbrun, K. (2001). Principles of forensic mental health assessment. New York: Kluwer Academic/Plenum.

Heilbrun, K. (2010). Violence risk assessment in adults. Oxford, UK: Oxford University Press.

(p.180) Heilbrun, K., Grisso, T., & Goldstein, A. (2008). Foundations of forensic mental health assessment. New York: Oxford University Press.

Hemphill, J.F., Hare, R.D., & Wong, S. (1998). Psychopathy and recidivism: A review. Legal and Criminological Psychology, 3, 139–170.

Hoge, R.D. (2008). Assessment in juvenile justice systems. In R.D. Hoge, N. Guerra, & P. Boxer (Eds.), Treating the juvenile offender (pp. 54–75). New York: Guilford Press.

Hoge, R.D., & Andrews, D.A. (2006). Youth Level of Service/Case Management Inventory: User’s manual. North Tonawanda, NY: Multi-Health Systems.

Hoge, R.D., & Andrews, D.A. (2010). Evaluation of risk for violence in juveniles. New York: Oxford University Press.

Kazdin A. (2000). Psychotherapy for children and adolescents: Directions for research and practice. New York: Oxford University Press.

Leistico, A.R., Salekin, R.T., DeCoster, J., & Rogers, R. (2008). A large-scale meta-analysis relating the Hare measures of psychopathy to antisocial conduct. Law and Human Behavior, 32, 28–45.

Lipsey, M.W., & Derzon, J.H. (1998). Predictors of violent or serious delinquency in adolescence and early adulthood: A synthesis of longitudinal research. In R. Loeber & D.P. Farrington (Eds.), Serious and violent juvenile offenders: Risk factors and successful interventions (pp. 86–105). Thousand Oaks, CA: Sage.

Lipsey, M.W., & Wilson, D.B. (2001). Practical meta-analysis. Thousand Oaks, CA: Sage.

Litwack, T.R. (2001). Actuarial versus clinical assessments of dangerousness. Psychology, Public Policy, and Law, 7, 409–443.

Lodewijks, H., de Ruiter, C., & Doreleijers, T. (2008). Gender differences in violent outcome and risk assessment in adolescent offenders after residential treatment. International Journal of Forensic Mental Health, 7, 133–146.

Lodewijks, H.P.B., Doreleijers, T.A.H., & de Ruiter, C. (2008). SAVRY risk assessment in violent Dutch adolescents: Relation to sentencing and recidivism. Criminal Justice and Behavior, 35, 696–709.

Loeber, R., & Stouthamer-Loeber, M. (1998). Development of juvenile aggression and behavior: Some common misconceptions and controversies. American Psychologist, 53, 242–259.

Loeber, R. (1990). Development and risk factors of juvenile antisocial behavior and delinquency. Clinical Psychology Review, 10, 1–41.

Loeber, R., & Stouthamer-Loeber, M. (1987). Prediction. In H.C. Quay (Ed.), Handbook of juvenile delinquency (pp. 325–382). New York: Wiley.

Loeber, R., Slot, N.W., & Stouthamer-Loeber, M. (2006). A three-dimensional, cumulative developmental model of serious delinquency. In P.O.H. Wikström & R. Sampson (Eds.), The explanation of crime: Contexts and mechanisms (pp. 153–194). Cambridge, UK: Cambridge University Press.

Loney, B., Taylor, J., Butler, M., & Iacono, W. (2007). Adolescent psychopathy features: 6-year temporal stability and the prediction of externalizing symptoms during the transition to adulthood. Aggressive Behavior, 33, 242–252.

Loza, W., Villeneuve, D., & Loza-Fanous, A. (2002). Predictive validity of the Violence Risk Appraisal Guide: A tool for assessing violent offender’s recidivism. International Journal of Law and Psychiatry, 25, 85–92.

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

(p.181) Lynam, D. R., Caspi, A., Moffitt, T.E., Loeber, R., & Stouthamer-Loeber, M. (2007). Longitudinal evidence that psychopathy scores in early adolescence predict adult psychopathy. Journal of Abnormal Psychology, 116, 155–165.

Melton, G., Petrila, J., Poythress, N., & Slobogin, C. (2007). Psychological evaluations for the courts: A handbook for mental health professionals and lawyers (3rd ed.). New York: Guilford Press.

Moffitt, T.E. (2003). Life-course-persistent and adolescence-limited antisocial behavior: A 10-year research review and research agenda. In B.B. Lahey, T.E. Moffitt, & A. Caspi (Eds.), Causes of conduct disorder and juvenile delinquency (pp. 49–75). New York: Guilford Press.

Moffitt, T.E., & Caspi, A. (2001). Childhood predictors differentiate life-course persistent and adolescence-limited antisocial pathways among males and females. Development and Psychopathology, 13, 355–375.

Monahan, J. (1996). Violence prediction: The last 20 years and the next 20 years. Criminal Justice and Behavior, 23, 107–120.

Monahan, J. (2008). Structured risk assessment of violence. In R.I. Simon, & K. Tardiff (Eds.), Textbook of violence assessment and management (pp. 17–33). Arlington, VA: American Psychiatric Publishing.

Monahan, J., Steadman, H.J., Robbins, P.C., Appelbaum, P., Banks, S., & Grisso, T., et al. (2005). An actuarial model of violence risk assessment for persons with mental disorders. Psychiatric Services, 56, 810–815.

Mulvey, E.P. (2005). Risk assessment in juvenile justice policy and practice. In K. Heilbrun, N.E.S. Goldstein, & R.E. Redding (Eds.), Juvenile delinquency: Prevention, assessment, and intervention (pp. 209–231). New York: Oxford University Press.

Mulvey, E.P., & Iselin, A.R. (2008). Improving professional judgments of risk and amenability in juvenile justice. The Future of Children, 18, 35–57.

Murrie, D.C., Boccaccini, M.T., Johnson, J.T., & Janke, C. (2008). Does interrater (dis)agreement on psychopathy checklist scores in sexually violent predator trials suggest partisan allegiance in forensic evaluations? Law and Human Behavior, 32, 352–362.

Nicholls, T.L., Ogloff, J.R.P., & Douglas, K.S. (2004). Assessing risk for violence among male and female civil psychiatric patients: The HCR-20, PCL:SV, and VSC. Behavioral Sciences and the Law, 22, 127–158.

Nuffield, J. (1982). Parole decision-making in Canada: Research towards decision guidelines. Ottawa, ON: Supply and Services Canada.

O’Connor, T.G., & Rutter, M. (1996). Risk mechanisms in development: Some conceptual and methodological considerations. Developmental Psychology, 32, 787–795.

Odgers, C.L., Moretti, M., & Reppucci, N.D. (2005). Examining the science and practice of violence risk assessment with female adolescents. Law and Human Behavior, 29, 7–27.

Odgers, C., Vincent, G.M., & Corrado, R.R. (2002). A preliminary conceptual framework for the prevention and management of multi-problem youth. In R.R. Corrado, R. Roesch, S.D. Hart, & Gierowski, J.K. (Eds.), Multi-problem violent youth: A foundation for comparative research on needs, interventions and outcomes (pp. 302–329). Amsterdam NL: IOS Press.

Olver, M.E., Stockdale, K.C., & Wormith, J.S. (2009). Risk assessment with young offenders: A meta-analysis of three assessment measures. Criminal Justice and Behavior, 36, 329–353.

Patterson, G.R., Forgatch, M.S., Yoerger, K.L., & Stoolmiller, M. (1998). Variables that initiate and maintain an early-onset trajectory of offending. Developmental Psychopathology, 10, 531–547.

(p.182) Piquero, A.R., Farrington, D.P., & Blumstein, A. (2007). Key issues in criminal career research: New analyses of the Cambridge Study in Delinquent Development. Cambridge, UK: Cambridge University Press.

Poluchowicz, S., Jung, S., & Rawana, E.P. (2000, June). The interrater reliability of the Ministry Risk/Needs Assessment Form for juvenile offenders. Paper presented at the Annual Conference of the Canadian Psychological Association, Montreal, QC.

Quinsey, V.L., Harris, G.T., Rice, M.E., & Cormier, C.A. (1998). Violent offenders: Appraising and managing risk. Washington, DC: American Psychological Association.

Robins L.N. (1974). Antisocial behaviour disturbances of childhood: Prevalence, prognosis, and prospects. In E.J. Anthony & C. Koupernik (Eds.), The child in his family: Children at psychiatric risk. Oxford, UK: Wiley.

Rowe, R.C. (2002). Predictors of criminal offending: Evaluating measures of risk/needs, psychopathy, and disruptive behaviour disorders. Unpublished doctoral thesis. Carleton University, Ottawa, ON.

Rubin, B. (1972). Prediction of dangerousness in mentally ill criminals. Archives of General Psychiatry, 27, 397–407.

Rutter, M., Giller, H., & Hagell, A. (1998). Antisocial behaviour by young people. Cambridge, UK: Cambridge University Press.

Sampson, R.J., & Laub, J.H. (2003). A life-course view of the development of crime. Annals of the American Academy of Political and Social Science, 602, 12–45.

Schmidt, F., Hoge, R., & Robertson, L. (2005). Reliability and validity analyses of the Youth Level of Services/Case Management Inventory. Criminal Justice and Behavior, 32, 329–344.

Schwalbe, C.S. (2007a). A meta-analysis of juvenile justice risk assessment instruments: Predictive validity by gender. Criminal Justice and Behavior, 35, 1367–1381.

Schwalbe, C.S. (2007b). Risk assessment for juvenile justice: A meta-analysis. Law and Human Behavior, 31, 449–462.

Schwalbe, C.S., Fraser, M.W., Day, S.H., & Arnold, E. (2004). North Carolina Assessment of Risk (NCAR): Reliability and predictive validity with juvenile offenders. Journal of Offender Rehabilitation, 40, 1–22.

Schwalbe, C.S., Fraser, M.W., Day, S.H., & Cooley, V. (2006). Classifying juvenile offenders according to risk of recidivism: Predictive validity, race/ethnicity, and gender. Criminal Justice and Behavior, 33, 305–324.

Smith, P., Cullen, F.T., & Latessa, E.J. (2009). Can 14,737 women be wrong? A meta-analysis of the LSI-R and recidivism for female offenders. Criminology and Public Policy, 8, 183–208.

Teplin, L., Abram K., McClelland, G., Dulcan, M., & Mericle, A. (2002). Psychiatric disorders in youth and juvenile detention. Archives of General Psychiatry, 59, 1133–1143.

Thornberry, T.P. (2005). Explaining multiple patterns of offending across the life course and across generations. Annals of the American Academy of Political and Social Science, 602, 156–195.

Thornberry, T.P., & Krohn, M.D. (2003). (Eds.). Taking stock of delinquency: An overview of findings from contemporary longitudinal studies. New York: Kluwer/Plenum.

Upperton, R, A., & Thompson, A.P. (2007). Predicting juvenile offender recidivism: Risk-need assessment and juvenile justice officers. Psychiatry, Psychology and Law, 14, 138–146.

Vincent, G.M., Chapman, J., & Cook, N.E. (2005). Risk/needs assessment in juvenile justice: Predictive validity of the SAVRY, racial differences, and contributions of needs factors. Criminal Justice and Behavior, 32, 42–62

(p.183) Vincent, G.M., Fusco, S.L., Gershenson, B.G. & Guy, L.S., (2010). Secondary analyses for the prediction and assessment of early adult criminal activity. Unpublished report to the National Institute of Justice.

Vincent, G.M., Grisso, T., Terry, A., & Banks, S. (2008). Gender and race differences in mental health symptoms in juvenile justice. Journal of the American Academy of Child and Adolescent Psychiatry, 47, 282–290.

Vincent, G.M., Odgers, C.L., McCormick, A., & Corrado, R.R. (2008). The PCL:YV and recidivism in male and female juveniles: A follow-up into young adulthood. International Journal of Law and Psychiatry, 31, 287–296.

Vincent, G.M., Quinlan, J. C, Nitschelm, S., & Ogloff, J.R.P. (2003, April). Differential recidivism prediction by mental disorder. Paper presented at the 2003 annual conference of the International Association of Forensic Mental Health Services, Miami, FL.

Vincent, G.M., Terry, A., & Maney, S. (2009). Risk/needs tools for antisocial behavior and violence among youthful populations. In J. Andrade (Ed.), Handbook of violence risk assessment and treatment for forensic mental health practitioners (pp. 337–424). New York: Springer.

Walters, G. (2006). Risk-appraisal versus self-report in the prediction of criminal justice outcomes: A meta-analysis. Criminal Justice and Behavior, 33, 279–304.

Washington State Institute for Public Policy. (March, 2004). Assessing risk for re-offense: Validating the Washington State Juvenile Court Assessment (Document No. 04-03-1201). Olympia, WA: WSIPP.

Wasserman, G., McReynolds, L., Lucas, C., Fisher, P., & Santos L. (2002). The Voice DISC-IV with incarcerated male youths: Prevalence of disorder. Journal of the American Academy of Child & Adolescent Psychiatry, 41, 314–321.

Webster, C.D., Douglas, K., Eaves, D. & Hart, S. (1997). HCR-20: Assessing risk for violence, Version 2. Burnaby, BC: Simon Fraser University.

Welsh, J., Schmidt, F., McKinnon, L., Chattha, H., & Meyers, J. (2008). A comparative study of adolescent risk assessment instruments: Predictive and incremental validity. Assessment, 15, 104–115.

Wormith, J.S., & Goldstone, C.S. (1984). The clinical and statistical prediction of recidivism. Criminal Justice and Behavior, 11, 3–34.