Comparative Analyses of the YLS/CMI, SAVRY, and PCL:YV in Adolescent Offenders: A 10-year Follow-Up Into Adulthood

• Author: Mary Ann Campbell,Carolyn Houlding,Fred Schmidt
• Published date: 01 January 2011
• Date: 01 January 2011
• DOI: http://doi.org/10.1177/1541204010371793
• Subject Matter: Articles

Youth Violence and Juvenile Justice
9(1) 23-42
Comparative Analyses of
ª The Author(s) 2011
Reprints and permission:
the YLS/CMI, SAVRY, and
sagepub.com/journalsPermissions.nav
DOI: 10.1177/1541204010371793
http://yvj.sagepub.com
PCL:YV in Adolescent
Offenders: A 10-year
Follow-Up Into Adulthood
Fred Schmidt1,2, Mary Ann Campbell3, and Carolyn Houlding2
Abstract
A growing body of research has been dedicated to developing adolescent risk assessment
instruments, but much of this research has been limited to short-term tests of predictive validity.
The current study examined the predictive and incremental validity of the Youth Level of Service/
Case Management Inventory (YLS/CMI), Structured Assessment of Violence Risk in Youth (SAVRY),
and Psychopathy Checklist: Youth Version (PCL:YV) in adolescent offenders over a mean 10-year
follow-up period. Each instrument predicted general recidivism with moderate- (YLS/CMI area
under the curve [AUC] ¼ .66) -to-large effect sizes (SAVRY AUC ¼ .74; PCL:YV AUC ¼ .79). How-
ever, there was variation in predictive validity across types of recidivism, and all three instruments
were better at predicting recidivism in males than females. SAVRY total also demonstrated incre-
mental validity over its structured professional judgment of risk. Clinical implications and future
directions for youth risk assessment are discussed.
Keywords
juvenile, risk assessment, longitudinal, recidivism, gender
Assessing the risk of criminal and violent behavior is a common role for professionals who work
with youth in the mental health and youth justice systems. A reliable and useful risk evaluation
requires systematic and comprehensive data collection, as well as expertise in the organization and
interpretation of assessment findings. Such a standardized assessment is foundational to the devel-
opment of an effective intervention plan (Hoge, 2002). Historically, clinicians relied on clinical
judgment to determine risk. However, several important meta-analyses have demonstrated the
superiority of structured risk tools and actuarial assessment procedures over unstructured clinical
1 Children’s Centre Thunder Bay, ON, Canada
2 Lakehead University, ON, Canada
3 University of New Brunswick at Saint John Campus, Canada
Corresponding Author:
Fred Schmidt, Children’s Centre Thunder Bay, ON, Canada, P7B 6G6
Email: fschmidt@childrenscentre.ca
23

---

24
Youth Violence and Juvenile Justice 9(1)
judgment in estimating recidivism risk (Ægisdo´ttit et al., 2006; Grove, Zald, Lebow, Snitz, &
Nelson, 2000). In addition to standardizing the assessment process, structured risk instruments can
guide clinical decision making regarding rehabilitation efforts intended to reduce risk, particularly
when used within a correctional intervention framework such as the Risk-Need-Responsivity model
(RNR; Andrews & Bonta, 2006). Although there is considerably more empirical information avail-
able on risk assessment measures for adults, ethical and clinical concerns preclude their use with
adolescents. For this reason, validation of existing adolescent risk instruments is critical.
Three promising adolescent risk measures are the Youth Level of Service/Case Management
Inventory (YLS/CMI; Hoge & Andrews, 2002), the Structured Assessment of Violence Risk in
Youth (SAVRY; Borum, Bartel, & Forth, 2006), and the Psychopathy Checklist: Youth Version
(PCL:YV; Forth, Kosson, & Hare, 2003). A recent meta-analysis of adolescent risk instruments,
including these three instruments, found that the overall association between risk instruments and
recidivism (r ¼ .25) was lower but not significantly different from the average association found
between adult risk instruments and recidivism (r ¼ .30; Schwalbe, 2007). This is an encouraging
finding that supports the value of developing instruments for adolescent risk assessment.
The YLS/CMI is an actuarial-based professional-completed instrument that provides a cumula-
tive risk score for general recidivism. This risk score can then be used to classify youth into one of
four pre-determined risk categories: low, moderate, high, and very high. The YLS/CMI also has
semistructured guidelines for case planning to guide community supervision and identify targets for
intervention. Of the three identified risk measures noted above, the YLS/CMI has been the most
empirically investigated and widely used. For example, the YLS/CMI has been implemented in parts
of Canada and the United States, with an adapted version being used in Australia (YLS/CMI-AA;
Upperton & Thompson, 2007). A recent meta-analysis has shown that the YLS/CMI consistently
predicted both violent and nonviolent recidivism in high-risk adolescent offender populations
(Olver, Stockdale, & Wormith, 2009).
A more recently developed and professional-completed risk instrument is the SAVRY. It consists of
24 items capturing empirically based predictors of violence in adolescents (Borum et al., 2006). In addi-
tion, six protective factors are rated. The SAVRY differs from the risk classification approach used
within the YLS/CMI. Rather than using an actuarial approach that requires the use of a fixed algorithm
to sum all risk items for a total score, the SAVRY uses a structured professional judgment (SPJ) method
of classification. The SPJ method requires the clinician to make a clinical judgment of risk as low, mod-
erate, or high based on a structured review of all risk and protective factors evident in a youth’s profile.
No cutoff score is used and the SAVRY manual specifically discourages the use of a total score to deter-
mine risk level. Although limited empirical information is available on the SAVRY, the initial results
are promising. The SAVRY can predict violent and nonviolent recidivism over 1- and 3-year follow-up
periods in samples of adolescent offenders (Catchpole & Gretton, 2003; Lodewijks, Doreleijers, & De
Ruiter, 2008; Meyers & Schmidt, 2008). Although a unique strength of the SAVRY is its incorporation
of protective factors into the risk formulation, very little information is available on the validity of these
factors when predicting desistence from criminal behavior. The few studies available suggest that these
factors are important predictive components of adolescent risk assessment over the 1-year follow-up
periods (Dolan & Rennie, 2008; Lodewijks et al., 2008). However, Rennie and Dolan (2010) found
SAVRY protective factors to have low predictive validity over a 3-year follow-up period.
A third instrument used in adolescent risk assessment is the PCL:YV, which was designed to
assess personality traits associated with the construct of psychopathy. The PCL:YV was modeled
after the Psychopathy Checklist–Revised (PCL-R; Hare, 2003), which has a significant body of lit-
erature supporting its predictive validity for multiple recidivism outcomes (Douglas, Vincent, &
Edens, 2006). The PCL:YV generates a total score capturing the severity of psychopathic traits,
as well as 4-facet scores that reflect specific interpersonal, affective, behavioral, and antisocial psy-
chopathic trait domains. With a few exceptions (e.g., Edens & Cahill, 2007), the growing body of
24

---

Schmidt et al.
25
research on the PCL:YV has shown strong predictive validity for violence and general recidivism
with Caucasian adolescent male populations (Olver et al., 2009) but has obtained much weaker
results among female youth (Edens, Campbell, & Weir, 2007; Schmidt, McKinnon, Chattha, &
Brownlee, 2006). Importantly, ethnicity has been identified as a potential moderating factor with
more multiethnic samples resulting in lower PCL:YV prediction values (Edens & Cahill, 2007;
Edens et al., 2007). Gretton, Hare, and Catchpole (2004) conducted one of the few long-term pre-
dictive validity studies of the PCL:YV and found it to be a significant predictor of violent behavior
in adolescent males over a 10-year period. Nonetheless, due to concerns with the implications of
labeling a youth as ‘‘psychopathic,’’ questions about the long-term stability of psychopathy from
childhood to adulthood, and its questionable applicability to female youth (e.g., Seagrave & Grisso,
2002), there is much controversy with the use of this instrument with adolescent offenders.
Given the growing research on adolescent risk assessment, Olver et al., (2009) recently conducted
a meta-analysis of published and unpublished adolescent risk research and identified 22 studies on
the YLS/CMI, 28 on the PCL:YV, and 9 on the SAVRY. Their results indicated that the YLS/CMI,
SAVRY, and PCL:YV were equivalent in their predictive power across both violent and nonviolent
measures of recidivism. Despite their apparent comparability, few studies have compared these
instruments within the same sample to assess their predictive and incremental validity. In their
within-sample comparison, Dolan and Rennie (2008) found the SAVRY outperformed the PCL:YV
over a 12-month follow-up in a sample of adolescent males. Similarly, Welsh, Schmidt, McKinnon,
Chattha, and Meyers (2008) found that the SAVRY possessed incremental predictive validity over
the PCL:YV and YLS/CMI over a mean 3-year follow-up period. Thus, in both studies, the SAVRY
outperformed the other measures in predicting violent and nonviolent outcomes.
An additional gap in adolescent risk assessment is the lack of comparative analyses of long-term
predictive validity of available instruments. This is particularly true for the YLS/CMI and SAVRY.
The longest mean recidivism follow-up period of 5.7 years for the YLS/CMI comes from an unpub-
lished study by...