You are on page 1of 8

Assessing juvenile offenders: Preliminary data for the Australian

Adaptation of the Youth Level of Service/Case Management Inventory


(Hoge & Andrews, 1995)*
ANTHONY P. THOMPSON & ZOE POPE
Charles Sturt University, Wagga Wagga, New South Wales, Australia
Abstract
The developmental phase and preliminary psychometric data are reported for an Australian adaptation of an assessment
inventory for juvenile offenders. Specically, the Australian Adaptation of the Youth Level of Service/Case Management
Inventory (YLS/CMI-AA, Hoge, & Andrews, 1995) is used to assess risks, needs and strengths to inform decision making
with juvenile offenders. Data from a sample of 290 juvenile offenders were used to analyse item and score characteristics
which, with few exceptions, performed in keeping with traditional psychometric standards. Predictive validity in a subsample
of 174 males followed for recidivism between 6 and 32 months resulted in a correlation of 0.28 and area under the receiver
operating characteristic (ROC) curve of 0.67 for the total score on the inventory. The results and use of the inventory are
placed in the context of related developments in other jurisdictions.
Systematic assessment of the risks and needs of
juvenile offenders is widely accepted as a key
component of informed responses to juvenile crime
(Day, Howells, & Rickwood, 2003, 2004; Howell,
1995). This trend is consistent with a growing
professional psychology emphasis on using forensi-
cally relevant tests to address criminal justice and
psycho-legal issues (Borum, 1996; Lally, 2003,
Martin, Allan, & Allan, 2001; Tolman & Mullendore,
2003). In recent years, the Department of Juvenile
Justice (New South Wales) has adopted such an
approach in a fashion that has progressively linked
research and practice. Central to this process has been
an inventory for assessing risk factors, psychosocial
needs and major strengths. The inventory is referred
to as the Australian Adaptation of the Youth Level of
Service/Case Management Inventory (YLS/CMI-
AA, Hoge, & Andrews, 1995). This report provides
psychometric results from data collected during the
adaptation phase and draws links to related develop-
ments in other jurisdictions.
The philosophical, theoretical, empirical and prac-
tical underpinnings of risk assessment in juvenile
justice have been elaborated by various authors
(e.g. Bonta, 2002; Hoge, 2002; Thompson, 2001;
Thompson &Putnin s, 2003). Essentially, the approach
views adverse developmental outcomes as arising from
the effect over time of antecedent risk factors. Juvenile
delinquency is one type of adverse outcome that is
benecially conceptualised within the risk framework,
but the perspective can be applied to a wide range of
other health and social problems (Goldenring &Rosen,
2004). These propositions are supported by an
extensive empirical and theoretical literature (Cicchetti
& Rogosch, 2002; Durlak, 1998; Farmer & Farmer,
2001; Farrington, 2002; Loeber & Stouthamer-Loe-
ber, 1998; Maluccio, 2002; McLaren, 2000). Such
factors are also prominent in theories of juvenile
delinquency (Agnew, 2001; Andrews & Bonta, 1994;
Henggeler et al., 1998).
A structured approach to assessing such risk factors
is valuable because of its systematic and empirical
features (Bonta, 2002; Hoge, 2002; Quinsey et al.,
1998). It is important to be aware that risk assessment
serves case management and intervention (Andrews,
1991). Assessments can be linked to future likelihood
of offending and high-risk cases should be managed
prudently. However, assessments are also designed to
Correspondence: Dr Anthony P. Thompson, Humanities and Social Sciences, Charles Sturt University, Locked Bag 678, Wagga Wagga, NSW 2678, Australia.
E-mail: athompson@csu.edu.au
*Some of the psychometric results were presented by Thompson, A. P., & Pope, Z. (2003). The conceptual and psychometric basis for risk need assessment in
juvenile justice. In M. Katsikitis (Ed.), Proceedings of the 38th APS Annual Conference (pp. 224 228). Melbourne: The Australian Psychological Society.
Australian Psychologist, November 2005; 40(3): 207 214
ISSN 0005-0067 print/ISSN 1742-9544 online 2005 Australian Psychological Society Ltd
Published by Taylor & Francis
DOI: 10.1080/00050060500243491
identify modiable difculties that become targets for
change, particularly among higher risk offenders.
These ingredients are appealing because they are in
concert with some of the central principles underlying
contemporary models of criminal justice.
In the juvenile justice sector, there has been a
proliferation of risk need assessment inventories.
Various jurisdictions in the United States have
developed instruments (see for example Ashford &
LeCroy, 1990; Gavazzi et al., 2003a; Howell, 1995;
LeCroy, Krysic, & Palumbo, 1998; National Council
on Crime and Delinquency, 2000; Schwalbe et al.,
2004). A risk need assessment tool has also been
recently developed and adopted by the youth justice
system in England and Wales (Baker et al., 2003). In
Australia, the Secure Care Psychosocial Screening
assessment (Putnin s, 1999) has been used for youth in
detention since the beginning of 1994. The Depart-
ment of Human Services, Victoria also developed a
risk need inventory (Day, Howells, & Rickwood,
2003; Greville, 2002). In juvenile justice in NSW, the
inventory that has been used is a version of the Youth
Level of Service/Case Management Inventory (YLS/
CMI, Hoge, & Andrews, 2002) and preliminary data
from the NSW trial adaptation are presented here. The
parent inventory (YLS/CMI) was developed over a
number of years in Ontario, Canada (Hoge, 2002;
Hoge & Andrews, 1996; Jung & Rawana, 1999) and is
itself an adaptation of a prominent inventory used in
adult corrections for over 20 years (The Level of
Service Inventory Revised; Andrews, &Bonta, 2000).
The multiplicity of risk need inventories for use
with juvenile offenders is a situation with both
benets and disadvantages. Widespread efforts at
bringing structure to the assessment process in a way
that distills much of what is known about juvenile
offending is a decided advantage. So, too, is the
development of inventories that cater to the jurisdic-
tional context of offenders and juvenile justice
systems. A drawback, however, is that effort is
distributed and the research needed to support a
given inventory is slow to accumulate. It will also be
a challenge to draw together and integrate ndings
from such a diversity of activity a considerable
proportion of which is yet to appear in peer-reviewed
journals. However, increasingly more research is
appearing, particularly in relation to the psycho-
metric features of such inventories and the current
report is in keeping with that objective.
Method
Australian Adaptation of the Youth Level of Service/
Case Management Inventory (YLS/CMI-AA)
An Australian adaptation of the inventory began
prior to the commercial availability of the parent
inventory. Permission was obtained in 1999 from
Robert D. Hoge, Carleton University to make
changes to the unpublished inventory to suit its
anticipated use in the NSW Department of Juvenile
Justice
1
. In particular, an Australian trial version of
Part I (Assessment of Risks and Needs) and Part II
(Summary of Risk and Needs) was adapted. The
50-item (Part I) trial version corresponded closely
with the unpublished 42-item parent version.
Changes involved: (1) revisions in language to reect
the New South Wales context (e.g. custodial order
replaced detention, supervised order replaced
probation, wags and misses classes replaced
missing school days or skipping classes, (2)
inclusion of several new items that were empirically
or conceptually related to relevant risk domains (e.g.
age at rst court order because of the link between
early onset of offending and recidivism, home-
lessness because of its relevance to family
circumstances and risk/need considerations, occa-
sional alcohol use so that degree of use could be
evaluated consistent with the existing drug use items),
(3) revising the itemrelated to prior probation with an
item concerning the outcome and nature of the rst
court order, (4) tightening the operational denition
of selected items (the meaning and parameters of
various items were specied further and a number of
these clarications have been included in the com-
mercialised parent version), (5) minor reorganisation
of items in some domains to improve the logical ow
(e.g. personality items followed by behaviour items
rather than mixed together), (6) the addition of three
items to identify major strengths that may operate
as protective factors, and (7) printing operational
denitions of all items on the inventory rather than in
a separate manual.
As with the parent version, items in the Australian
trial version were organised into the following
domains: (1) Prior and current offences (eight
items), (2) Family and living circumstances (seven
items), (3) Education/employment (seven items),
(4) Peer relations (four items), (5) Substance abuse
(six items), (6) Leisure/recreation (three items),
(7) Personality/behaviour (seven items), and (8)
Attitudes/orientation (ve items). A new domain,
Assessment of major strengths, was included for
the three items related to protective factors. Items are
scored in a binary fashion to indicate whether the
operationally dened item describes the young
person. One item related to the outcome of the rst
conviction had two parts, both scored in the binary
fashion. One item related to age at rst court order
was scored 0, 1, and 2 with more weight given to
younger offenders. Major strengths are indicated as
present or not at the individual level (social and
personal skills), family level (strong, positive parent-
child relationship) and community level (support
208 A. P. Thompson & Z. Pope
outside the family). Endorsed items in each domain
are tallied to provide a domain total and an overall
risk need score is calculated based on all domains
except Major strengths.
The Australian trial version was used in hard copy
format (46A4-paged booklet) by a sample of Juvenile
Justice Ofcers ( JJOs). The inventory is intended for
use by professionals working in the juvenile justice
sector. The manual for the parent version (Hoge &
Andrews, 2002) recommends training prior to use. A
training workshop was offered in selected community
supervision ofces in Sydney and regional NSW to
familiarise staff with the background, rationale and
use of the inventory. Ethical approval for the research
was granted through the university Ethics Committee
and through the Department of Juvenile Justice. JJOs
were asked to complete the inventories within 1
month of a young person coming on a supervised
order. They were also asked to re-assess clients with
the inventory at the 6-month mark or at le closure
whichever came rst. Staff were encouraged by their
managers to participate in the research project but
involvement and motivation varied both within and
between ofces, as did completion of the inventories
within the suggested time-frames. Data were col-
lected over an 18-month period up to mid-2001 for
young people under community supervision. Ap-
proximately 10% of the inventories were unusable for
reasons such as duplication, incomplete or dubious
information and untraceable identication numbers.
Complete, useable data were obtained for a sample of
290 young offenders (50 female, 240 male). These
inventories were provided by 44 staff from nine
different ofces. The number of inventories com-
pleted by each JJO varied from one to 23, with the
median number being six. The majority (90%) were
from ofces in the metropolitan Sydney area. At the
time of assessment, the mean age of the sample was
16.55 years (SD=1.32 years). Although the age
range was 13 years to 20 years, most youth were
between 14 and 18 years (14 years, 11%; 15 years,
17%; 16 years, 26%; 17 years, 29%; 18 years, 12%).
Follow-up data on re-offending were obtained for a
subsample of 174 males. Useable retest data were
obtained for a subsample of 11 female and 62 male
clients. These data were provided by 19 JJOs, who
completed between one and 12 retest inventories
(median =3).
Preliminary psychometric properties
Male and female data were combined for analyses
unless otherwise indicated. Gender- and ethnic-
based comparisons are important to investigate but
will be considered in subsequent research, as a larger
sample becomes available. One of the items we
added under Prior and current offences was
dropped from the analysis because it related to the
collection of data in relation to sex offending. Data
were checked for logical consistency (i.e. if regular
drug or alcohol use was identied as a risk, so was
occasional use) and for data entry accuracy.
Item and subscore statistics
Endorsement proportions for the dichotomous items
were in the range of 0.08 0.84, with 59% of the
items in the 0.30 0.70 range. Least endorsed items
included those dealing with ever being in custody
(0.08), inated self-esteem (0.10) and regular
alcohol use (0.11). Most endorsed items were those
dealing with the outcome of the rst court order
(0.84), delinquent friends (0.77) and delinquent
acquaintances (0.72). In terms of Major strengths,
endorsement proportions indicated that many of the
juvenile offenders in this sample were judged to have
strengths at the individual (44%), family (51%) or
community (23%) level. Extreme endorsement
proportions have the potential to limit the psycho-
metric properties of a scale, but Nunnally (1970)
emphasises item-total correlations as the more
important standard. Corrected item-total correla-
tions were computed for each item and the total
score in the domain to which it belonged. Most (43
of 50) were between 0.30 and 0.67 and would be
considered good items for scale development (Nunn-
ally, 1970). One questionable item was in the Prior
and current offences domain with an item total
correlation of 70.21. This item asked whether the
young person had received custody or supervision as
a rst court order and was one of the items we added.
The item concerning delinquent friends correlated
only 0.10 with the Peer relations total. It will be
recalled that these two items had the highest
endorsement proportions.
Table 1 provides the intercorrelations, means,
standard deviations and internal consistency esti-
mates for the domain and total scale scores of the
YLS/CMI-AA. The distribution of scores was, to
varying degrees, positively skewed for all domains
except for Leisure/recreation, which was nega-
tively skewed. The internal consistency estimate
(coefcient alpha) was high for the total score
(0.91) and acceptable (0.69 0.79) for most do-
mains. The internal consistency coefcient was poor
for Prior and current offences (0.56), Peer
relations (0.45) and Major strengths (0.50).
Principal component analysis
Reise, Waller, & Comrey (2000) note that principal
components analysis provides a summary of correla-
tions between variables. As such, we used it to
examine whether the empirical coherence of
Assessing juvenile offenders 209
YLS/CMI-AA items resembled the risk domains.
Two items with poor item-total correlation (noted
above) were excluded, as were the three items
dealing with major strengths. This left 45 items for
analysis. All principal components with an eigenva-
lue greater than 1 were extracted and oblique
rotation (direct oblimin) was applied. Oblique
rotation allows extracted factors to be correlated
(Reise, Waller, & Comrey, 2000), which concep-
tually would be expected for risk domains related to
juvenile offending. The analysis yielded 12 compo-
nents that accounted collectively for 61.72% of the
variance. Approximately 60% of the items had
component loadings (0.30 and above) on more than
one component. Interpretation of the structure was
based largely on examining items with the highest
loadings on each component and the results pro-
vided support for some of the YLS/CMI-AA
domains. For example, the rst component ac-
counted for 23.19% of the variance and was
characterised by prominent loadings (0.56 0.79)
on ve items in the Family and living circum-
stances domain, with nine items from several other
domains loading in the 0.30 0.43 range. The
second component accounted for 5.31% of the
variance, with prominent loadings (0.70 0.85) on
four of the Substance abuse items and only four
items from other domains loading in the 0.30 0.45
range. The third component (5.17% of variance)
seemed to represent aggressiveness with three items
(tantrums, verbally aggressive, physically aggressive)
from the Personality/behaviour domain and the
item concerning violent school behaviour loading
between 0.52 and 0.72. The fourth component
(4.72% variance) represented the background
(static) risk dimension with prominent loadings
(0.65 0.76) on four items from the Prior and
current offences domain. The fth component
(3.62% variance) incorporated all items from the
Personality/behaviour and Attitudes/orientation
domains (loadings 0.39 0.77) except one item,
related to physical aggression. The sixth component
(3.58% of variance) supported the Peer relations
domain. The two items related to the absence of
positive friends and acquaintances both loaded highly
(0.86), with many items from other domains loading
in the 0.3 0.4 range. Among the remaining six
components there was some support for the Educa-
tion/employment and Leisure/recreation domains.
Predictive validity
The relationship between YLS/CMI-AA scores and
future offending was investigated for a subsample of
males. Ofcial recidivism was determined by reading
the offence record of each case on the Client
Information Data System of the NSW Department
of Juvenile Justice. The length of the follow-up was
based on the time (rounded to the nearest month)
from risk need assessment until follow-up data
system access or the young persons 18th birthday,
whichever came rst. Only those young peoples with
at least a 6-month follow-up period were considered.
Thus, the follow-up sample consisted of 174 males
who were followed-up between 6 and 32 months with
a median follow-up period of 17 months. In this
group, 70 males, or 40% of the sample, had convic-
tions after the time they were assessed with the risk
need inventory. The median time to conviction for
reoffence was 7.5 months and 79% of recidivists had
registered a new conviction within a year
2
.
Predictive validity for recidivism produced a point
biserial correlation with the YLS/CMI-AA total
score of 0.28 (p <0.001, two-tailed). The highest
Table 1. Intercorrelation, mean, standard deviation and internal consistency for the YLS/CMI-AA subcomponents
YLS/CMI-AA domain
YLS/CMI-AA domain 1 2 3 4 5 6 7 8 Total Strengths
1. Prior and current (0.56)
2. Family and living 0.39 (0.79)
3. Education/employment 0.37 0.63 (0.71)
4. Peer relations 0.27 0.33 0.43 (0.45)
5. Substance abuse 0.22 0.40 0.41 0.24 (0.77)
6. Leisure/recreation 0.26 0.44 0.44 0.38 0.26 (0.74)
7. Personality/behaviour 0.30 0.52 0.65 0.39 0.40 0.37 (0.79)
8. Attitudes/orientation 0.34 0.59 0.53 0.35 0.37 0.46 0.67 (0.69)
Total score 0.58 0.80 0.82 0.56 0.61 0.61 0.79 0.77 (0.91)
Major strengths 70.19 70.43 70.39 70.29 70.20 70.49 70.33 70.46 70.48 (0.50)
Possible score range 0 9 0 7 0 7 0 4 0 6 0 3 0 7 0 5 0 48 0 3
Mean 3.43 2.60 2.62 2.32 1.90 1.77 1.99 1.41 18.06 1.19
Standard deviation 1.86 2.18 1.98 1.14 1.83 1.18 2.05 1.47 9.66 1.02
Note. N=290. All correlations are significant at the 0.01 level (two-tailed). Coefficient alpha on the diagonal.
210 A. P. Thompson & Z. Pope
correlation with recidivism was 0.32 (P<0.001,
two-tailed) for the Prior and current offences
domain. The Personality and behaviour domain
had the next strongest relationship with recidivism
(r =0.25; P<0.01, two-tailed). A receiver operating
characteristic (ROC) curve analysis available in
SPSS was used to further elaborate the predictive
validity of the YLS/CMI-AA. ROC analysis plots the
true positive rate (sensitivity) against false alarm rate
(1 minus specicity) for all possible cut-points and
the resulting area under the curve is a useful index of
overall accuracy (Bennett et al., 1999; Rice & Harris,
1995). The area under the ROC curve for the total
score was 0.67, indicating that the probability that a
randomly selected recidivist will score higher on the
YLS/CMI-AA than a randomly selected non-recidi-
vist is 67%. The values provided by the ROC analysis
indicate the accuracy and the inaccuracy that would
result from using successive scores as cut-points for
risk predictions. For example, if a total score of 24 on
the YLS/CMI-AA were used as a cut-point, 53% of
the recidivists would be among those scoring equal to
or higher than 24, but there would be a 27% false
positive rate. Area under the ROC curve was
essentially the same for the predictive capability of
domain scores for Prior and current offences
(area =0.68) and Personality/behaviour (area =
0.67). However, the range of these subscores is
substantially less than for the YLS/CMI-AA total
score and hence there are fewer cut-point options for
selecting true positive versus false positive ratios.
There was no signicant relationship between the
total strength score and recidivism, but the strength
item dealing with individual social and personal skills
did bear a signicant relationship in the expected
direction (r =70.15, P<0.05, two-tailed).
Test retest stability
The mean test retest period was 5.00 months
(range =1 16 months, SD=2.85 months). Stability
coefcients for the eight risk domains ranged from
0.61 (Peer relations) to 0.85 (Prior and current
offences) and all were signicant (P40.001, two-
tailed). The stability coefcient for the Major
strengths domain was 0.67 (P40.001, two-tailed)
and for the total risk score across the two assessments
it was 0.79 (P40.001, two-tailed). Overall, the total
risk score declined from the rst assessment
(M=18.14) to the second assessment (M=12.95)
and this decline was signicant (t(72) =7.13,
P40.001, two-tailed). The effect size for this
difference was large (d =0.83).
Among the sample assessed twice and followed-up
for at least 6 months, there were 18 recidivists and 41
non-recidivists. On the rst assessment, the mean
total risk score for recidivists was 22.44 compared
with 17.24 for non-recidivists. Although this differ-
ence was not signicant (t(57) =1.89, P=0.06, two-
tailed), the effect size was moderate (d =0.53). On
the second assessment, recidivist scored 19.28 versus
10.63 for non-recidivists. This difference was sig-
nicant (t(24.51, equal variances not assumed) =
2.99, P<0.01) and the effect size was large
(d =0.97). Correlation between total risk score and
the dichotomous criterion of recidivism was not
signicant for rst assessment (r =0.24, P=0.06, two-
tailed), but was for the second assessment (r =0.42,
P=0.001, two-tailed). It is important to note that in
the retest sample conviction was prior to or contem-
poraneous with completion of the second assessment
for eight of the 18 recidivists. Such information may
have inuenced the retest evaluation.
Discussion
We have presented contextual information and the
preliminary psychometric data for the Australian
Adaptation of the Youth Level of Service/Case
Management Inventory (YLS/CMI-AA, Hoge, &
Andrews, 1995). The data were collected in a trial
phase and include follow-up data on recidivism and a
sample of test retest data. The item analysis
revealed that with few exceptions items were
functioning in keeping with accepted psychometric
standards. Item statistics are not included in the
manual for the parent version (Hoge & Andrews,
2002), although some are reported by Flores, Travis, &
Latessa (2003) for a large sample of juvenile
offenders in Ohio. Internal consistency indices in
the current study were generally equivalent or better
than those provided by Flores et al., except for the
Prior and current offences domain. Gavazzi et al.
(2003) provided item analyses for an 86-item global
risk indicators measure with domains that corre-
spond to those on the YLS/CMI-AA. In one study,
Gavazzi et al. found internal consistency estimates
were poorest for domains related to previous
offences, peer relations and attitudes/orientations.
This corresponds with the pattern in the current
study. Gavazzi et al. improved internal consistency
by adding items to the subscales. On the revised
113-item inventory, all alpha coefcients were above
0.87. Item statistics pointed to concerns about two
items on the trail version of the YLS/CMI-AA. Both
items have been retained (one with clarication) and
will be re-examined in the future. Department-wide
staff training, increased familiarity with the inventory
and more systematic data collection may inuence
some of the psychometric properties of the risk
domains. It will also be important to supplement
traditional item analysis with Item Response Theory
procedures (Embretson, 1996). Low internal con-
sistency for Major strengths might be expected.
Assessing juvenile offenders 211
The three items concern protective factors from
different sources (individual, family, community)
that conceptually and empirically might not be highly
inter-related. Nevertheless, it is important to con-
sider potential protective factors (Howell, 1995).
The principle component analysis provided some
empirical support for the domains that are used
to organise risk need items. However, only one
component related to family functioning accounted
for a sizeable proportion of variance. The manual for
the YLS/CMI (Hoge & Andrews, 2002) provides no
data on the factor structure of the parent version and
the factor structure of the adult version has been
described as inconsistent (Andrews & Bonta, 2000).
Using conrmatory factor analysis with the global
risk indicator measure, Gavazzi et al. (2003) found
support for risk domains identical to those on the
YLS/CMI, but the structure was better for the longer
as opposed to the shorter instrument. Clearly, this is
an area for ongoing investigation with the YLS/CMI
and its Australian adaptation. The organisation of
items into risk domains as used on the YLS/CMI
makes sense conceptually and in terms of risk factors
that have been associated with juvenile recidivism
(Cottle, Lee, & Heilbrun, 2001). However, those
factors need to be translated into assessment items
that are empirically homogeneous. Also, domains are
implicitly weighted in the total risk need score by
virtue of their constituent items (3 8) and this also
needs empirical justication.
Predictive validity studies for risk need assess-
ment inventories are beginning to accumulate.
However, results are being reported over varying
time-periods, with various criteria (e.g. arrest,
reconviction, frequency or seriousness of offence on
reconviction, intensity of diversion services), and in
varying ways such as percentage correctly predicted,
accuracy across score bands, correlation, ROC
analysis, survival analysis and mean group differ-
ences (e.g. Baker et al., 2003; Gavazzi et al., 2003b;
Jung & Rawana, 1999; National Council on Crime
and Delinquency, 2000; Schwalbe et al., 2004,
Putnin s, in press). The predictive validity coefcients
from the trial phase of the YLS/CMI-AA (r =0.28
and area under the ROC curve =0.67 for total score
and recidivism, are in keeping with similar indices
reported elsewhere. Flores et al. (2003) found YLS/
CMI total score correlated 0.31 with re-arrest for
males. The risk index score used by Putnin s (in
press) for male youth in secure care correlated 0.32
with recidivism 6 months after release. Validity
correlation coefcients around 0.3 are considered
to be a more than respectable result (Meyer et al.,
2001). Catchpole & Gretton (2003) found area
under the ROC curve of 0.74 for the YLS/CMI total
score and general recidivism in a sample of 74 violent
offenders in British Columbia, Canada. Total score
on the structured assessment used in the United
Kingdom produced an area under the ROC curve
value of 0.72 in a 1-year follow-up of reconviction in
over 1000 cases (Baker et al., 2003).
Predictive validity results for domain scores in the
current study (in particular, r =0.32 between Prior
and current offences and recidivism) are also
consistent with research that shows static risk
factors are especially useful in predicting recidivism.
Flores, Travis, & Latessa (2003) found the Prior
and current offences domain score correlated 0.37
with re-arrest in their large sample. A meta-analysis
of risk factors and recidivism showed certain
offence history variables to be the best (Cottle
et al., 2001). Empirically, there may be grounds
for separating risk assessment, which could be
undertaken with a relatively small set of static
variables, from a more comprehensive consideration
of dynamic needs. This is in keeping with the
assessment framework recommended by the
US Ofce of Juvenile Justice and Delinquency
Prevention (Howell, 1995).
Overall, the retest results in the current data set
were lower at the second assessment than at the rst.
This is encouraging at the system level. Given that
offender needs are not static states, high degrees of
test retest stability are not likely nor, for high-risk
offenders, are they desirable. Interventions should
maintain low-risk offenders at that level but
reduce the risk associated with medium- to high-risk
offenders. Retest scores also provided some support
for discriminating between recidivists and non-
recidivists. It has to be acknowledged, however, that
the retest methodology in this study incorporated
many unknown sources of variation, including
idiosyncratic selection of cases for revaluation,
potential access to scores from the rst assessment
and possible knowledge of reconviction. Rigorous
research on score changes is essential and future
efforts should attempt to differentiate cohorts (e.g.
those with score increases versus score decreases) as
well as explore the expectations and practices of
juvenile justice personnel responsible for assess-
ments. One overseas study in adult corrections
found some manipulation of risk assessment inven-
tory results in response to work demands and lack of
condence in the inventory (Schneider, Ervin, &
Snyder-Joy, 1996).
The experience and data from the trial phase with
this inventory led the NSW Department of Juvenile
Justice to include it in standard assessment proce-
dures for all juvenile offenders. Prior to nalising the
inventory for department wide use, there was further
ne-tuning of operational denitions and layout.
Two items were also changed substantively for
conceptual rather than psychometric reasons. The
item about inconsistent parenting was replaced with
212 A. P. Thompson & Z. Pope
an item related to antisocial/criminal family values.
The reason was that parenting practices seemed well
covered by three other Family and living circum-
stances items and the new item extended the
content domain. A similar rationale applied to the
other change. The item relating to short attention
was expanded to dene the broader concept of
impulsivity. A number of departmental procedures
were adopted to support implementation and quality
control including organisational support, operational
guidelines, a training programme and ongoing
monitoring. The inventory has been used in compu-
terised format since October 2002 under a licensing
agreement with the test publisher, Multi-Health
Systems Incorporated. With the inventory in regular
use, we are able to accumulate a sizeable YLS/
CMI-AA database and to undertake ongoing re-
search. For example, we are investigating the ways
in which Juvenile Justice Ofcers undertake and
understand risk assessment in their work with young
offenders, either with or without the benets of
structured risk assessment. It is important to remem-
ber that the inventory is not only about risk. Its
domains provide a basis for case planning and
intervention. It serves as a basis for helping young
people to lead better lives. Ward (2002) has argued
that we can lose sight of this objective when needs are
viewed only as liabilities to be overcome to prevent
offending. It is important that inventories such as the
YLS/CMI-AA are not simply imported into practice
without a commitment to ongoing evaluation, re-
search and reexivity. Our understanding of juvenile
offending and of best practice in the eld is evolving,
and we should embrace this reciprocal process.
Acknowledgements
The authors would like to thank the Collaborative
Research Unit of the NSW Department of Juvenile
Justice for their assistance in undertaking this
research. The opinions here do not necessarily reect
the views of the NSW Department of Juvenile
Justice, or any of its ofcers. Zoe Pope is now at
Forensic Services, Mental Health ACT, Australia.
Notes
1 Apart from suiting the Australian context, the adaptation was
needed to accommodate, in particular, an older age-range. The
Canadian inventory was developed for use with 12 16-year-old
offenders and norms are for 12 17 years. In NSW, a single
government department deals with 10 18-year-old offenders.
2 There were seven instances in which the time to reconviction
was less than 2.5 months. It is possible that these were
outstanding rather than new charges. Reported predictive
validity analysis included these cases, as the results were
virtually identical to when they were excluded.
References
Agnew, R. (2001). Building on the foundation of general strain
theory: Specifying the types of strain most likely to lead to
crime and delinquency. Journal of Research in Crime and
Delinquency, 38, 319 361.
Andrews, D. A. (1991). Recidivism is predictable and can be
inuenced: Using risk assessments to reduce recidivism. Forum
on Corrections Research, 1(2), 11 17.
Andrews, D. A., & Bonta, J. (1994). The psychology of criminal
conduct. Cincinnati, OH: Anderson Publishing.
Andrews, D. A., & Bonta, J. L. (2000). Level of Service Inventory
Revised. Users manual. North Tonawanda, NY: Multi-Health
Systems Inc.
Ashford, J. B., & LeCroy, C.W. (1990). Juvenile recidivism: A
comparison of three predictor instruments. Adolescence, 25,
441 450.
Baker, K., Jones, S., Roberts, C., & Merrington, S. (2003).
ASSET The evaluation of the validity and reliability of the Youth
Justice Boards assessment for young offenders. Findings from the
rst two years of the use of ASSET. University of Oxford: Centre
for Criminological Research.
Bennett, K. J., Lipman, E. L., Brown, S., Racine, Y., Boyle,
M. H., & Offord, D. R. (1999). Predicting conduct problems:
Can high-risk children be identied in kindergarten and grade
1? Journal of Consulting and Clinical Psychology, 67, 470 480.
Bonta, J. (2002). Offender risk assessment: Guidelines for
selection and use. Criminal Justice and Behavior, 29, 355 379.
Borum, R. (1996). Improving the clinical practice of violence risk
assessment. Technology, guidelines, and training. American
Psychologist, 51, 945 956.
Catchpole, R. E. H., & Gretton, H. M. (2003). The predictive
validity of risk assessment with violent young offenders:
A 1-year examination of criminal outcome. Criminal Justice
and Behavior, 30, 688 708.
Cicchetti, D., & Rogosch, F. A. (2002). A developmental
psychopathology perspective on adolescence. Journal of
Consulting and Clinical Psychology, 70, 6 20.
Cottle, C. C., Lee, R. J., & Heilbrun, K. (2001). The prediction of
criminal recidivism in juveniles: A meta-analysis. Criminal
Justice and Behavior, 28, 367 394.
Day, A., Howells, K., & Rickwood, D. (2003). The Victorian juvenile
justice rehabilitation review. Melbourne, Victoria, Australia:
Department of Human Services, Victorian Government.
Day, A., Howells, K., & Rickwood, D. (2004). Current trends in
the rehabilitation of juvenile offenders. Trends & Issues in Crime
and Criminal Justice No. 284. Canberra, Australia: Australian
Institute of Criminology.
Durlak, J. A. (1998). Common risk and protective factors
in successful prevention programs. American Journal of
Orthopsychiatry, 68, 512 520.
Embretson, S. E. (1996). The new rules of measurement.
Psychological Assessment, 8, 341 349.
Farmer, T. W., & Farmer, E. M. Z. (2001). Developmental
science, systems of care, and prevention of emotional
and behavioral problems in youth. American Journal of
Orthopsychiatry, 71, 171 181.
Farrington, D. P. (2002). Key results from the rst forty years of
the Cambridge study in delinquent development. In T. P.
Thornberry & M. D. Krohn (Eds), Taking stock of delinquency:
An overview of ndings from contemporary longitudinal studies
(pp. 137 183). New York: Kluwer/Plenum.
Flores, A. W., Travis III, L. F., & Latessa, E. J. (2003). Case
classication for juvenile corrections: An assessment of the
Youth Level of Service/Case Management Inventory (YLS/
CMI). Cincinnati, OH: Centre for Criminal Justice Research,
University of Cincinnati.
Assessing juvenile offenders 213
Gavazzi, S. M., Slade, D., Buettner, C. K., Partridge, C.,
Yarcheck, C. M., & Andrews, D. W. (2003a). Toward
conceptual development and empirical measurement of global
risk indicators in the lives of court-involved youth. Psychological
Reports, 92, 599 615.
Gavazzi, S. M., Lim, J. Y., Yarcheck, C. M., & Eyre, E. L.
(2003b). Brief report on predictive validity evidence of global
risk indicators in the lives of court-involved youth. Psychological
Reports, 93, 1239 1242.
Goldenring, J. M., & Rosen, D. S. (2004, January). Getting into
adolescent heads: An essential update. Contemporary Pediatrics,
21, 64 90.
Greville, C. (2002). Predicting the risk of juvenile recidivism using
the new Victorian Juvenile Justice Risk Needs Assessment for young
offenders. Melbourne, Victoria, Australia: Department of
Psychology, Monash University.
Henggeler, S. W., Schoenwald, S. K., Borduin, C. M., Rowland,
M. D., & Cunningham, P. B. (1998). Multisystemic treatment of
antisocial behavior in children and adolescents. New York:
Guilford Press.
Hoge, R. D. (2002). Standardized instruments for assessing risk
and need in youthful offenders. Criminal Justice and Behavior,
29, 380 396.
Hoge, R. D., & Andrews, D. A. (1995). Australian Adaptation of
the Youth Level of Service/Case Management Inventory. North
Tonawanda, NY: Multi-Health Systems Inc.
Hoge, R. D., & Andrews, D. A. (1996). Assessing the youthful
offender: Issues and techniques. New York: Plenum.
Hoge, R. D., & Andrews, D. A. (2002). Youth Level of Service/Case
Management Inventory (YLS/CMI). Users manual. North
Tonawanda, NY: Multi-Health Systems Inc.
Howell, J. C. (Ed.) (1995). Guide for implementing the comprehensive
strategy for serious, violent, and chronic juvenile offenders.
Washington, DC: Department of Justice, Ofce of Justice
Programs, Ofce of Juvenile Justice and Delinquency Preven-
tion.
Jung, S., & Rawana, E. A. (1999). Risk and need assessment of
juvenile offenders. Criminal Justice and Behavior, 26, 69 89.
Lally, S. J. (2003). What tests are acceptable for use in forensic
evaluations? A survey of experts. Professional Psychology:
Research and Practice, 34, 491 498.
LeCroy, C. W., Krysik, J., & Palumbo, D. (1998). Empirical
validation of the Arizona risk/needs instrument and assessment
process. Tucson, Arizona: LeCroy & Milligan Associates Inc.
Loeber, R., & Stouthamer-Loeber, M. (1998). Development of
juvenile aggression and violence: Some common misconcep-
tions and controversies. American Psychologist, 53, 242 259.
Maluccio, A. N. (2002). Resilience: A many-splendored
construct. American Journal of Orthopsychiatry, 72, 596 599.
McLaren, K. L. (2000). Toughis not enoughGettingsmart about youth
crime. Areviewof research on what works to reduce offending by young
people. Wellington, NewZealand: Ministry of Youth Affairs.
Martin, M.-A., Allan, A., & Allan, M. M. (2001). The use of
psychological tests by Australian psychologists who do
assessments for the courts. Australian Journal of Psychology,
53, 77 82.
Meyer, G. J., Finn, S. E., Eyde, L. D., Kay, G. G., Moreland,
K. L., Dies, R. R., et al. (2001). Psychological testing and
psychological assessment: A review of evidence and issues.
American Psychologist, 56, 128 165.
National Council on Crime and Delinquency (2000). Almeda
county placement risk assessment validation. Oakland, CA: Author.
Nunnally, J. C. Jr (1970). Introduction to psychological measurement.
New York: McGraw-Hill.
Putnin s, A. L. (1999). Secure Care Psychosocial Screening: Manual
for SECAPS v.4. Adelaide, Australia: Department of Human
Services.
Putnin s, A. L. (in press). Assessing recidivism risk among young
offenders. Australian and New Zealand Journal of Criminology.
Quinsey, V. L., Harris, G. T., Rice, M. E., & Cormier, C. A.
(1998). Violent offenders: Appraising and managing risk.
Washington, DC: American Psychological Association.
Reise, S. P., Waller, N. G., & Comrey, A. L. (2000).
Factor analysis and scale revision. Psychological Assessment,
12, 287 297.
Rice, M. E., & Harris, G. T. (1995). Violent recidivism: Assessing
predictive validity. Journal of Consulting and Clinical Psychology,
63, 737 748.
Schneider, L. A., Ervin, L., & Snyder-Joy, Z. (1996). Further
exploration of the ight from discretion: The role of risk/need
instruments in probation supervision decisions. Journal of
Criminal Justice, 24, 109 121.
Schwalbe, C. S., Fraser, M. W., Day, S. H., & Arnold, E. M.
(2004). North Carolina Assessment of Risk (NCAR): Relia-
bility and predictive validity with juvenile offenders. Journal of
Offender Rehabilitation, 40, 1 22.
Thompson, A. P. (2001). Juvenile offending: An overview in
support of assessing risk factors, needs, and strengths.
The Australian Educational and Developmental Psychologist, 18,
39 56.
Thompson, A. P., & Putnin s, A. L. (2003). Risk need assessment
inventories for juvenile offenders in Australia. Psychiatry,
Psychology and Law, 10, 324 333.
Tolman, A. O., & Mullendore, K. B. (2003). Risk evaluations for
the courts: Is service quality a function of specialization?
Professional Psychology: Research and Practice, 34, 225 232.
Ward, T. (2002). The management of risk and the design of good
lives. Australian Psychologist, 37, 165 171.
214 A. P. Thompson & Z. Pope