Holsinger, Lowenkamp, Latessa, 2006

\\server05\productn\J\JCJ\30-2\JCJ207.txt unknown Seq: 1 13-MAR-06 15:50

PREDICTING INSTITUTIONAL MISCONDUCTUSING THE YOUTH LEVEL OF SERVICE/

CASE MANAGEMENT INVENTORY

ALEXANDER M. HOLSINGER, Ph.D.†University of Missouri at Kansas City

CHRISTOPHER T. LOWENKAMP, Ph.D.University of Cincinnati

EDWARD J. LATESSA, Ph.D.University of Cincinnati

ABSTRACT: Offender assessment in corrections has advanced considerably overthe last several decades. Currently, it is not uncommon to find correctional pro-fessionals using any number of objective standardized assessment instruments.While many of these instruments possess face validity as well as statistical predic-tive validity, more work is needed to test classification protocol on new popula-tions, and in various correctional environments. The current paper investigatesthe predictive validity of the Youth Level of Service/Case Management Inventory(YLS/CMI) within an institutional setting. Specifically, the composite score ren-dered from the YLS/CMI is used to predict institutional misconduct. The YLS/CMI was found to effectively differentiate between two levels of offender risk/need, and was significantly related to outcome using several different statisticaltechniques.

INTRODUCTION

Offender classification has emerged nearly unequivocally as part ofthe ‘best practices’ within correctional intervention (National Instituteof Corrections, 2000). The importance of general offender classificationmakes intuitive sense, but has been demonstrated in the correctionalliterature base as well. Offender classification can take on many forms,and may utilize any number of procedures. For example, the impor-tance of merely grouping juvenile offenders with similar problems to-gether has been demonstrated, as has the validity of somewhat complex

† Direct all correspondence to: Alexander M. Holsinger, University of Missouri,Kansas City, Department Sociology/Criminal Justice and Crimingology, 208 Haag Hall,Kansas City, MO 64110. Email: [email protected].

AMERICAN JOURNAL OF CRIMINAL JUSTICE, Vol. 30 No. 2, 2006 2006 Southern Criminal Justice Association


268 PREDICTING INSTITUTIONAL MISCONDUCT

and comprehensive actuarial assessments (Dembo & Schmeidler, 2003;Hoge, 2002).

Specifically dynamic offender assessment and classification hasemerged as one of the most important initial functions of any correc-tional agency. Various forms of actuarial offender classification havebeen employed over the last several decades, with differing levels ofeffectiveness. The development of offender classification has beencharacterized as falling into three ‘generations’ with unstructuredclinical interviews comprising the first of these generations (Bonta,Law, & Hanson, 1998), whereby the correctional practitioner reliesheavily on qualitative impressions and correctional experience to makedecisions about a supervision and treatment strategy .

The second generation is typically characterized by the develop-ment and use of actuarial tools, inventories, or checklists to help informcorrectional decision making. While the use of assessment tools devel-oped from quantitative research represents a marked improvement inthe predictive validity of various classifications systems, second-genera-tion tools rely heavily on static, or unchangeable variables typicallygleaned from the offender’s criminal history (Bonta et al., 1998; Grove,Zald, Lebow, Snitz & Nelson, 1995; Hanson & Bussiere, 1998). Thethird generation of offender assessment methods employs actuarialtools as well, but these tools may include dynamic as well as staticpredictors, in an effort to better meet several goals of classification.These goals include improved predictive validity, and the ability to im-plement case planning that utilizes rehabilitative programming. Whendone well, dynamic offender assessment not only builds the initial prac-titioner’s knowledge base about the offender, but may ultimately in-form other agencies (such as treatment programs) that work with theoffender after the intake process.

There is a growing body of literature demonstrating the utility ofcomprehensive actuarial assessment for juvenile offenders. In one com-parative analysis three well-known actuarial assessments (generally cov-ering different criminogenic needs with some overlap) were shown topossess good predictive validity, and to facilitate decision-making aswell as treatment recommendations (Hoge, 2002). Other studies havedemonstrated the benefits of intensively examining one particular do-main and its predictive validity (e.g., personality characteristics usingthe MMPI, or assessing likelihood of violence using neuropsychologicalmeasures) (Caggiano, 2000; Calhoun, Glaser, & Petrocelli, 2002).Overall, what may be of most use to correctional agencies servingyouthful offenders with multiple risk and need factors is an actuarialassessment that incorporates both static and dynamic factors. Whilemeta-analyses have revealed the validity of both types of predictors in a


HOLSINGER, LOWENKAMP, AND LATESSA 269

global sense (Cottle, Lee, & Heilbrun, 2001; Gendreau, Little, & Gog-gin, 1996), there have also been several validation studies of specificinstruments that cover both static and dynamic predictors (Ayers et al.,1999; Barnoski, 1998; Flores, Travis & Latessa, 2003; Hoge, 2002; Jones& Harris, 1999).

DEVELOPING AND USINGACTUARIAL ASSESSMENTS

Most actuarial assessment processes result in the assignment of anumber, or score, that then gets translated into a ‘level’ (e.g., a risk/need level). The offender’s assigned level then in turn may impact thelevel of supervision they receive as well as motivate the use of otherrestrictive options such as house arrest, or curfews. The number orscore is gleaned from the investigation and weighting of various crimi-nogenic items. The literature base investigating the most salient crimi-nogenic risk/need factors has greatly informed the development ofseveral offender assessment tools. Most notably, meta-analysis hasbeen used to synthesize large volumes of correctional literature, and toa substantial degree has identified several criminogenic domains thattend to be included in many current risk/need assessment tools (An-drews & Bonta, 1994; Gendreau, Little, & Goggin, 1996; Simourd, 1997;Simourd & Andrews, 1994). These criminogenic domains typically in-clude (but are not necessarily limited to) criminal history, educationaland/or employment experiences, the significant relationships in the of-fender’s life, substance use/abuse, mental health issues, and antisocialcognitions. Some combination of these criminogenic domains is typi-cally measured in some form, on most third-generation assessmentinstruments.

The use of dynamic actuarial assessments does contain some pit-falls. For example, the process typically involves and relies heavily on astructured interview between the offender and a trained correctionalprofessional, necessitating the consideration of the reliability of mea-surement (Lowenkamp, Holsinger, Brusman-Lovins, & Latessa, 2004).In addition, agencies wishing to utilize any new assessment process mayhave to confront issues surrounding the applicability of the tool to of-fenders from differing ethnicities or genders (Holsinger, Lowenkamp,& Latessa, 2003; Lowenkamp, Holsinger, & Latessa, 2001). Conductingresearch on any new assessment process is necessary to investigate thepredictive validity of any actuarial assessment tool particularly when aprocess is utilized on new offender populations (Andrews & Bonta,1998).



IMPLEMENTING THE RISK PRINCIPLEUltimately, effective offender assessment and classification makes

the incorporation of the “risk principle” possible (Andrews, Bonta, &Hoge, 1990). In very broad terms, the risk principle states that offend-ers who need the most services (or supervision, as the case may be)should receive the most attention and services. Almost without ques-tion, those offenders who ‘need the most’ will be those that posses thegreatest number or amount of criminogenic risk/need factors. At thesame time, offenders who indeed may have committed some criminalact that brought them to the attention of the criminal justice system, butwho by-and-large do not possess many criminogenic risk/need factorsshould not receive much correctional intervention. In fact, research hasdemonstrated that when low-risk offenders are placed in an environ-ment that was designed for high-risk offenders, their risk levels actuallyincrease (Lowenkamp & Latessa, 2005). This may represent what Clearrefers to as “penal harm,” or incorporating correctional strategies thatare too harsh given the extenuating circumstances. This in effect mayactually make low-risk offenders worse off (Clear, 1994). The “riskprinciple” of classification applies to both community correctionalagencies as well as institutional facilities.

Correctional agencies of any type are virtually unable to incorpo-rate the “risk principle” of correctional intervention without the utiliza-tion of a validated assessment process that identifies relevantcriminogenic risk factors. This is not to say that risk/need assessment isthe only form of assessment necessary for successful intervention—farfrom it, in fact—but effective risk/need assessment is an extremely im-portant and substantial step toward building a successful interventionprotocol.

THE YOUTH LEVEL OF SERVICE/CASEMANAGEMENT INVENTORY

The Youth Level of Service/Case Management Inventory (YLS/CMI) is one example of a ‘third-generation’ offender assessment toolthat is gaining widespread use in the United States (Hoge & Andrews,1996). As the title suggests, the YLS/CMI was designed for use withthe juvenile offender population (those typically between the ages of 12and 16). The instrument measures 42 items that cover eight crimi-nogenic domains including criminal history, education, family circum-stances/parenting, peer relationships, substance use, leisure/recreation,personality characteristics, and attitudes/orientations (the 42 items arenot spread evenly across all eight domains, however). Criminal historyincludes historical/static items pertaining to the youth’s prior adjudica-



tions; education examines the youth’s current experiences in school(disciplinary infractions, achievement level); family circumstances/parenting assesses the parenting received by the youth (e.g., consis-tency, appropriateness of discipline); ‘peer relationships’ measures theyouth’s social network and whether or not they are involved in antiso-cial behavior; substance abuse assesses the youth’s current substanceuse (if any); leisure and recreation measures how involved the youth isin pro-social activity; personality characteristics measures constructssuch as impulsivity and the expression of anger; finally, attitudes andorientations assesses some measures of antisocial cognitions.

The YLS/CMI is currently used in a variety of different correc-tional settings including various community correctional options as wellas residential and treatment facilities. The assessment process is alsoutilized at various points throughout the correctional system (e.g. at in-take and/or discharge, as well as at various points throughout a youth’ssupervision). Due to the dynamic nature of several of the items on theYLS/CMI, the instrument has the potential for measuring offenderchange over time, within the domains that are assessed. This in turnmay assist agencies with the option of recommending early dischargefrom programming and/or supervision provided the youth has demon-strated pro-social change.

Utilizing the YLS/CMI requires a structured interview between atrained correctional practitioner and the offender that typically lasts be-tween 35 and 45 minutes. It is also generally recommended that theprofessional conducting the assessment contact the parent(s) or guard-ian(s) of the youth in order to gather additional information and/or ver-ify what was learned from the initial interview. Collateral informationsuch as previous case files, or interviews with other professionals whohave had recent contact with the youth are also beneficial. After gath-ering adequate information the professional then scores each item, andrenders a total composite score (potentially ranging between 0 and 42)that will place the youth into a specific category of criminogenic risk/need. Higher scores on the assessment in theory indicate a higher riskfor recidivism and a greater presence of criminogenic factors in need ofintervention.

To date, the research specifically investigating the predictive valid-ity of the YLS/CMI is limited (particularly regarding explicit tests ofpredictive validity). As a whole, many of the studies testing the predic-tive validity of the YLS/CMI find that the composite score does possessmoderate predictive validity with correlations of .15 to .35, dependingon sub-population and correctional setting (Catchpole & Gretton, 2003;Flores, Travis, & Latessa, 2003; Jung & Rawana, 1999; Marczyk, Heil-brun, Lander, & DeMatteo, 2005; Schmidt, Hoge, & Gomes, 2005;



Shields & Simourd, 1991). In addition, other research has examined theconcurrent validity that the YLS/CMI may have with other assessments.For example, the YLS/CMI was found to have concurrent validity withthe J-SOAP (a sex-offender risk assessment) and correlated highly withthe summary score gleaned from that instrument (Righthand et al.,2005). In addition, the YLS/CMI has been found to perform as well orbetter than other commonly used risk/need assessments such as the Psy-chopathy Checklist: Youth Version, and the Structured Assessment ofViolence Risk in Youth (Catchpole & Gretton, 2003). More research isneeded where the predictive validity of the YLS/CMI is tested, usingnew offender populations, and, utilizing additional measures of out-come. The current study utilizes institutional misconduct as the pri-mary measure of outcome, a measure of recidivism that is currentlylacking in the literature that examines the YLS/CMI.

CURRENT STUDY

The purpose of the current study is to investigate the YLS/CMIusing a sample of incarcerated young offenders, and as such utilizes thecomposite score to predict institutional misconduct. Utilizing the YLS/CMI in this manner may illuminate the potential use for this type ofclassification system in a residential setting, and may hold implicationsfor the administration of such facilities and the services containedtherein. It is common for residential agencies serving youth to utilize aclassification methodology of some sort, in order to facilitate placementwithin the facility and assess overall risk for misconduct. Interview-based actuarial assessment (that typically utilize dynamic factors) havebeen shown to offer the best predictive validity (Gendreau, Goggin, &Law, 1997; Hoge, 2002). In addition, dynamic factors in and of them-selves have been demonstrated to possess better predictive validity thanstatic, unchanging variables (Andrews & Bonta, 1998; Barnoski, 1998;Motiuk, 1995). While instruments utilizing static predictors do possessgood predictive validity in a variety of correctional environments(Jones, Harris, Fader, Burrell, & Fadeyi, 1999), they are limited in scopeand utility. A residential facility utilizing a comprehensive dynamicrisk/need assessment protocol at the beginning of a youthful offender’sstay will be in a much better position to make a variety of decisions(security classification; roommate assignment; programming recom-mendations), and perhaps more importantly, measure progress of theyouth over time through re-assessment, using the same tool. The cur-rent study is restricted in scope to the prediction of institutional miscon-duct as an outcome; however, if an agency via better placement andcase management decisions is able to anticipate and circumvent miscon-



duct, this has potential implications for virtually all other parts of ayouth’s journey through the justice system.

METHODS

ParticipantsThe participants in this study are youthful offenders sentenced to a

correctional facility for youth, in an unnamed state (the agency thatprovided the data for this research requested that the state not benamed explicitly in this manuscript). The facility is secure, and offerssome treatment services (but not a comprehensive battery of services,nor are the services implemented explicitly according to need). Whilecorrectional officers may have had access to the assessment data, at thetime there was no classification strategy in place that would have read-ily identified any particular youth as “high” or “moderate” risk to thecorrectional staff working with them. Similarly, the youth themselveswere not privy to their assessment data, nor were they aware of theirscore and/or their overall risk/need assessment. As such, it is unlikelythat at any point the youth would be responding to a label of being‘high’ risk for example. The youth in this study were randomly selectedby correctional staff for inclusion in the study. It should be noted, how-ever, that this random selection was not scientific and no representa-tions are made about the generalizability of this sample to the rest ofthe incarcerated youthful population. A total of 80 youth were assessedfrom August 2001 through March 2002. The age range of the partici-pants is from 14.6 to 18.9 years with an average age of 17.0 years.

ProceduresThe youth in this study were assessed with the YLS/CMI by agency

staff that had previously been trained in the administration of the YLS/CMI. Staff underwent three days of initial training and a follow-uptraining that included a review of a video taped assessment submittedby each staff member. The YLS/CMI was scored based on interviewswith the youth and a review of collateral information regarding criminalhistory, family, and education. Misconducts committed by youth whileincarcerated were gathered from the records of the correctionalinstitution.

MeasuresThe measures for the independent variables in this study included

the age of the youth, the subcomponents of the YLS/CMI, the totalYLS/CMI score, and days spent in the institution. Age was utilized as a



control variable due to the potential impact it has on rates and preva-lence of antisocial behavior. While the intention of the current researchdoes not include the specific effects of the individual subcomponentscores, correlations between subcomponent scores and measures of out-come are presented in order to display the disaggregated parts of theinstrument. Future research may warrant an investigation of the pre-dictive validity of the individual subcomponents, although the develop-ers of the instrument maintain that the predictive validity of the totalcomposite score is the most appropriate test. In addition, the YLS/CMIscore is dichotomized for some of the analyses in order to compare thetwo risk groups that are represented in the sample (moderate and high)across the outcome measures. When used as a continuous measure, theYLS/CMI score possesses a slightly irregular distribution, due to a lim-ited number of cases, coupled with the potential variation in the scoresthemselves. Days spent in the institution is included as a predictor vari-able due to the effect that time under correctional control can have onthe accrual of misconducts.

Globally the dependent variable included several measures of insti-tutional misconduct. A youth was considered to have received a mis-conduct if the staff filed a misconduct report with the facility detailingan incident. In other words, it is unknown whether or not the miscon-duct resulted in any disciplinary action, nor is it known what the ulti-mate outcome of the misconduct hearing was. Misconducts areclassified according to severity (low, moderate, high, and greatest). Assuch, conceptualizations of the dependent variable include percentageof the sample that received any type of misconduct (any of the fourlevels), as well as percent that received low, percent that received mod-erate, percent that received high, and the percent that received greatest.In addition, the total number of misconducts received was calculated,which allowed for difference of means testing between the two risk clas-sification levels present in the sample. Finally, each measure of institu-tional misconduct was dichotomized (e.g., if the offender received anymisconducts yes/no; if the offender received any low-level misconductsyes/no, and so on) in order to allow logistic regression analyses predict-ing a dichotomous dependent variable.

RESULTS

Table 1 presents descriptive information (mean, standard devia-tion, and range) for each of the eight YLS/CMI subcomponents as wellas the total YLS/CMI composite score. Also presented are descriptivestatistics for each of the four measures of institutional misconduct, thetotal number of misconducts committed by the sample, and the amount



of time the youth spent in an institution. While the eight domains havebeen disaggregated for the purpose of this table (and subsequent analy-ses), it should be noted that none of the individual subcomponents weredesigned to serve as a comprehensive assessment on their own.

TABLE 1Descriptive Statistics of Continuous Measures

Measure Mean Min Max SD

Prior and current offenses and dispositions 3.04 0.00 5.00 1.12Family circumstances/parenting 2.76 0.00 6.00 1.62Education/employment 3.34 0.00 6.00 1.97Peer relations 3.11 0.00 4.00 0.98Substance abuse 2.88 0.00 5.00 1.21Leisure/recreation 1.93 0.00 3.00 0.90Personality/behavior 3.20 0.00 7.00 1.77Attitudes/orientation 1.59 0.00 5.00 1.59Total YLSI 21.84 3.00 34.00 7.02Greatest Misconducts 1.10 0.00 8.00 1.69High Misconducts 2.28 0.00 15.00 2.86Moderate Misconducts 3.99 0.00 21.00 4.94Low Misconducts 1.65 0.00 16.00 2.58Total misconducts 9.01 0.00 56.00 10.76Time in institution 209.36 2.00 816.00 143.36

The mean YLS/CMI score for the sample was 21.84, a score whichfalls in the “moderate risk” category according to the classificationguidelines issued by the assessment’s publishers (Multi-Health Sys-tems). The YLS/CMI composite scores for the sample range from a lowof 3 to a high of 34 (although the range of possible scores is from 0 to42). Summary statistics for each of the four types of institutional mis-conduct are presented in Table 1 as well. On average the sample gener-ated 1.10 misconduct classified as Type A (hereafter greatestmisconduct), 2.28 misconduct classified as Type B (hereafter high mis-conduct), 3.99 misconduct classified as Type C (hereafter moderate mis-conduct), and 1.65 misconduct classified as Type D (hereafter lowmisconduct.

Greatest misconducts include behaviors such as sexual assault, kill-ing, use of force or threats against a correctional worker, escape from alocked institution, possession of weapons or firearms, and rioting. Highmisconducts include fighting, extortion, assault without a weapon, es-cape from an open institution, tampering with a lock, giving or offering



a bribe, use of drugs or paraphernalia, and threatening another personother than a correctional officer with bodily harm. Moderate miscon-ducts include sexual acts, indecent exposure, wearing a disguise, theft,misuse of medication, and refusing to obey an order. Low misconductsinclude possessing property belonging to someone else, using abusive orobscene language to staff, smoking, tattooing, unauthorized use of mailor telephone, or feigning an illness. On average the sample generated amean of 9.01 institutional misconduct. The mean amount of time spentin the institutional setting was just over 209 days, with a range of 2 to816 days.

All but one of the offenders in the sample fell into one of two riskclassifications (moderate, or high). Table 2 presents the results whenexamining the differences between these two offender groupings, re-garding rates of misconduct (the one offender that fell into the “low”risk classification level was eliminated for these analyses). The recom-mended ‘cut-off’ scores from Multihealth systems (the proprietors ofthe YLS/CMI) are (a) low risk: 0 to 8, (b) moderate risk: 9 to 22, (c)high risk: 23 to 34, and (d) very high risk: 35 to 42. Of those offenderswho were classified as being of moderate risk, 62% had at least onemisconduct of some kind, while 95% of the high risk offenders receivedat least one misconduct, representing a substantial and statistically sig-nificant difference (p<.01). The high risk offenders received greatestmisconduct at twice the rate of moderate risk offenders (67% and 31%respectively), which was also a statistically significant difference. Whilethere was a substantial difference between high and moderate risk of-fenders regarding rates of high misconduct (87% and 52% respec-tively), the two groups were not significantly different from oneanother. For moderate misconduct, analyses revealed a significant dif-ference between moderate risk offenders (50% receiving a misconduct)and high risk offenders (87% receiving a misconduct of this type). Sim-ilar results were also found when comparing moderate to high risk of-fenders regarding low misconduct, with 41% of moderate offendersreceiving at least one misconduct for this behavior, and 81% of the highrisk offenders. Overall, the moderate risk offenders received a mean of6.11 misconduct, while the high risk offenders received a mean of 12.54misconduct, which also generated a statistically significant differencebetween the groups. With only one exception, the moderate and highrisk offenders differed significantly from one another across each mea-sure outcome.

Table 3 presents the zero-order correlations between each measureof institutional misconduct, and each subcomponent of the YLS/CMI,the composite YLS/CMI score, and days incarcerated. Each correlationcoefficient measuring the relationship between the composite YLS/CMI



TABLE 2Misconduct Rates and Average Number ofMisconducts by Risk Level

Measure Moderate High c2

Risk Risk

Percent with Any Misconduct*** 62% 95% 11.948Percent with Greatest Misconducts*** 31% 67% 10.564Percent with High Misconducts 52% 87% 2.640Percent with Moderate Misconducts*** 50% 87% 11.860Percent with Low Misconducts*** 41% 81% 13.458Average Number of Total Misconducts** 6.11 12.54** p=.01, *** p=.001

score and institutional misconduct was statistically significant (p < .05).In addition, each of the correlation coefficients between YLS/CMI totalscore and each measure of misconduct was positive (in the anticipateddirection), and of low/moderate to moderate strength, ranging from alow of .267 (total YLS/CMI x high misconduct), to a high of .434 (totalYLS/CMI x low misconduct). The total YLS/CMI score was correlatedwith total misconduct at r=.397. Table 3 also reveals the fact that thenumber of days incarcerated was significantly related to receiving a mis-conduct for each measure of institutional misconduct, except for great-est misconduct. These findings indicated the need to statisticallycontrol for the amount of time a youth spent in the institution, whiledetermining the predictive ability of the YLS/CMI composite score.

Table 4 presents several logistic regression models, each utilizingthree variables (days spent in the institution, age of the offender at in-take, and total YLS/CMI score) to predict outcome (the four types ofinstitutional misconduct were dichotomized so 0=did not receive a mis-conduct, and 1=received a misconduct), as well as receiving any institu-tional misconduct (where 0=no, 1=yes). For each model, days spent inthe institution as well as total YLS/CMI score were revealed as signifi-cant predictors. For each model, the coefficient representing YLS/CMIscore was in the anticipated direction (positive) indicating that as theYLS/CMI score increased, so did the likelihood of receiving a miscon-duct. These results were gleaned while statistically controlling for timespent in the institution which was also revealed as a significant predictorin each model.



TA

BL

E 3

Co

rrel

atio

ns

bet

wee

n Y

LS

/CM

I an

d M

isco

nd

uct

s

Scor

eG

reat

est

Hig

hM

oder

ate

Low

Tota

lM

isco

nduc

tM

isco

nduc

tM

isco

nduc

tM

isco

nduc

tM

isco

nduc

ts

r(S

ig.)

r(S

ig.)

r(S

ig.)

r(S

ig.)

r(S

ig.)

Cri

min

al H

isto

ry.3

20(.

004)

.103

(.36

1).1

74(.

122)

.250

(.02

5).2

18(.

052)

Fam

ily.0

64(.

572)

−.05

9(.

602)

.075

(.50

7).1

70(.

131)

.070

(.53

9)

Edu

cati

on.2

75(.

014)

.192

(.08

9).3

05(.

006)

.287

(.01

0).3

03(.

006)

Pee

rs.1

38(.

221)

.043

(.70

5)−.

005

(.96

5).2

26(.

044)

.085

(.45

3)

Subs

tanc

e A

buse

.118

(.29

6).1

05(.

352)

.081

(.47

8).1

32(.

242)

.115

(.30

8)

Lei

sure

and

Rec

reat

ion

.155

(.16

8).1

76(.

119)

.214

(.05

7).1

91(.

090)

.215

(.05

5)

Per

sona

lity

.466

(.00

0).4

32(.

000)

.503

(.00

0).4

94(.

000)

.538

(.00

0)

Att

itud

es a

nd O

rien

tati

on.3

32(.

003)

.242

(.03

0).2

12(.

059)

.313

(.00

5).2

89(.

009)

Tota

l.3

96(.

000)

.267

(.01

6).3

47(.

002)

.434

(.00

0).3

97(.

000)

Day

s In

carc

erat

ed.2

17(.

054)

.326

(.00

3).3

69(.

001)

.357

(.00

1).3

76(.

001)



TABLE 4Logistic Regression Analyses Predicting InstitutionalMisconducts

Variable Any Greatest High Moderate Low

Days In Institution 0.011** 0.006** 0.005* 0.009** 0.010***Age 0.185 0.209 −0.100 0.293 0.279Total YLS/CMI 0.193** 0.139** 0.086* 0.135** 0.187**Constant −7.572 −8.030 −0.659 −8.623 −10.305c2 32.635 24.778 13.682 26.805 38.577Nagelkerke R2 .503 .355 .213 .395 .515*p=05, **p=.01, ***p=.001

DISCUSSIONThe analyses presented above reveal support for the YLS/CMI

composite score as a predictor of institutional misconduct. These re-sults are similar from previous validity studies using the YLS/CMI, butdiffer regarding the use of institutional misconduct as the primary mea-sures of outcome.

The offenders included in the sample were classified into one oftwo categories for most of the analyses (moderate or high risk) basedon the composite YLS/CMI score they were assigned. For each mea-sure of outcome, the high risk offenders committed a misconduct atsubstantially higher rates than the offenders who were classified asmoderate risk. Similarly, zero-order correlation analyses revealed sev-eral significant relationships of moderate strength between the YLS/CMI and outcome. Not surprisingly, correlation analyses also revealeddays spent in the institution as being significantly related to institutionalmisconduct as well. As a result, both these variables (in addition to ageof the offender at the time of intake) were entered into a series of logis-tic regression models. The multivariate analyses ultimately reinforcedwhat was revealed previously in the bivariate analyses by displaying thecomposite YLS/CMI score as a significant predictor of outcome, alongwith time spent in the institution.

Residential facilities designed to house and serve youthful offend-ers stand to benefit from the implementation of a valid risk/need classi-fication system. Several decision points can be informed using a soundassessment model. While the facilities from which these data weregathered do not pick and choose which offenders they receive, it wouldbe feasible to use the YLS/CMI composite score (as well as other as-



sessment protocol) to decide who is most appropriate for the residentialcorrectional environment. As such, an initial recommendation for thefacilities from which these data were gathered would be to determinefirst and foremost who is (and who is not) most appropriate for theenvironment and services the institution offers (although the ability toimplement this recommendation is admittedly limited). For those of-fenders who are admitted into the facilities, it is recommended that thecomposite score be used to make security-level designations (e.g., a‘high-risk’ pod or dorm, versus a ‘moderate-risk’ pod, and so on).Overall, regarding intake classification, implementation of the YLS/CMI and the appropriate use of the information gleaned from it couldassist with the identification of youth who are most likely to targetother youth within a facility, as well as staff. Basing tracking and hous-ing decisions on valid criteria would assist this (or any) facility by deter-mining appropriate staffing and programming levels and thereby avertpotential victimization and other behavioral problems within theinstitution.

Regarding treatment, due to the domains that are assessed via theYLS/CMI, a treatment-oriented case plan could be initiated that wouldplace the offender into needed treatment services. As such, yet anotherrecommendation for the agencies involved in the current study wouldbe to insure that a comprehensive case plan is built directly from theYLS/CMI – one that targets the criminogenic domains that are preva-lent in the offender’s life. Similarly, the overall progress that an of-fender may make while participating in residential treatment servicescould be gauged via reassessment using the same dynamic measures.As such, it is recommended that of the YLS/CMI be conducted at in-take, and that re-assessment using the YLS/CMI become a priorityupon successful participation in treatment services.

More often than not, youthful offenders are assigned to variouscorrectional environments by the juvenile court system, with the ulti-mate decision typically wresting with a juvenile judge or magistrate.Ideally, the decision to place a youth in an institutional setting would beinformed at least in part by validated risk/need protocol in order toinsure that those youth who pose the greatest risk for reoffending re-ceive the level of correctional intervention that they need. Doing thiswould initiate the risk principle and serve to reduce the over-incarcera-tion of low and moderate-risk offenders avoiding the damage exacer-bated by punishment that is too great (Clear, 1994). Despite the factthat residential facilities do not pick and choose offenders, a valid as-sessment protocol would still be beneficial in order to place offenderson various ‘tracks’ and avoid mixing populations of offenders of differ-ent risk levels.



Irregardless of the ‘placement’ or ‘tracking’ issues referred toabove, treatment recommendations can also be informed greatly via anassessment protocol. Insuring offenders who need specific services doindeed receive them (and in turn insuring that offenders who do notneed certain services do not receive them) further reinforces the ‘riskprinciple’ by avoiding the over-programming of certain (lower risk) of-fenders. This in turn may streamline services potentially making reha-bilitative efforts more efficient and effective. While the issue ofmeasuring offender progress and change is beyond the scope of the cur-rent article, these are important issues that relate directly to the efficacyof rehabilitative programming and can be informed via the assessmentof objective dynamic criteria.

There are several methodological and statistical limitations to thecurrent study. The limited size of the current sample provides somelimitations to the power of the analyses presented above. Although theanalyses conducted were not inappropriate in light of the low samplesize, more definitive conclusions may have been reached if a highernumber of cases had been available.

As noted above, the selection of youth for inclusion in the currentstudy was ultimately conducted by the agency from which the data weregleaned. Efforts were undertaken to implement a completely random-ized selection protocol, however, due to due process constraints andlimited logistic flexibility, truly random selection could not be achieved.As such, the generalizability of the sample to even the state from whichthe data came may be limited, not to mention generalizability to thepopulation of youthful offenders as a whole.

Another limitation involves the length of stay of the offenders inthe facilities. As a variable, length of stay was a significant predictor ofinstitutional misconduct across all models. Unfortunately the data werenot available in order to further investigate the extent to which lengthof stay was a driving force in the youths’ commission of institutionalmisconducts. Similarly, it remains possible that the results were evenfurther driven by a tandem effect of length of stay combined with risklevel. For example, it is feasible that a higher risk youth spent moretime in the institution, thereby making it possible for them to accumu-late more institutional misconducts.

Even if complete randomization were achieved, the results wouldnot necessarily be generalizable to the population of youthful offenders.These data came from a far western state of the United States that isarguably very different demographically from the rest of the UnitedStates. As such, even a totally randomized sample would have limitedgeneralizability to the United States as a whole.



While the YLS/CMI covers eight conceptually important crimi-nogenic domains, the analyses presented above were limited to theseeight domains, in addition to days spent in the institution, and age ofthe offender at intake. Arguably, other factors that were left unmea-sured could be conceptually and statistically related to institutional mis-conduct. For example, any number of responsivity factors such asmotivation or intelligence could be related to the various types of insti-tutional misconduct that were used as outcome variables in these analy-ses. While the YLS/CMI does measure some factors that may beindicative of antisocial personality in the domain labeled “personality,”a more in-depth analysis of the offenders’ personality could have poten-tially informed the results. Some clues to this possibility may be re-vealed in Table 3 where the “personality” sub-component of the YLS/CMI showed the strongest zero-order correlations with the variousmeasures of outcome. This is indeed merely a clue as the individualsub-domains should not be treated as full assessments in and of them-selves. In sum, even if additional measures and variables that couldimpact institutional misconduct had been available, the limited numberof cases would still have inhibited the incorporation of further statisticalcontrol. Future research utilizing the YLS/CMI and/or other assess-ment protocol will ideally include additional cases (larger sample sizes)and will statistically control for potentially extraneous influences on in-stitutional misconduct.

Finally, the results presented above are limited regarding the out-come measures that were predicted. While the current research maycontribute to an under-researched area (utilizing a dynamic risk/needprotocol to predict institutional misconduct) future research will ideallyincorporate other outcome measures that may be considered more im-portant to long-term correctional goals. Because institutional miscon-duct is by definition specific to the institutional environment,conclusions about behavior in the community cannot be made. Recidi-vism while in the community is perhaps of most interest to policy mak-ers and the public, as opposed to what occurs behind closed doorswithin an institutional setting.

REFERENCES

Andrews, D.A., & Bonta, J. (1994). The psychology of criminal conduct. Cincinnati, OH:Anderson.

Andrews, D.A., & Bonta, J. (1998). The psychology of criminal conduct (2nd ed.). Cin-cinnati, OH: Anderson.

Andrews, D.A., Bonta, J., & Hoge, R. (1990). Classification for effective rehabilitation:Rediscovering psychology. Criminal Justice & Behavior, 17(1), 19-52.



Ayers, C.D., Williams, J.H., Hawkins, J.D., Peterson, P.L., Catalano, R.F., & Abbott,R.D. (1999). Assessing correlates of onset, escalation, de-escalation, and desistanceof delinquent behavior. Journal of Quantitative Criminology, 15(3), 277-306.

Barnoski, R. (1998). Juvenile rehabilitation administration assessments: Validity reviewand recommendations. Washington State Institute for Public Policy. Olympia, WA:U.S. Government Printing Office.

Bonta, J., Law, M., & Hanson, R.K. (1998). The prediction of criminal and violent recidi-vism among mentally disordered offenders: A meta-analysis. Psychology Bulletin,39, 127-144.

Caggiano, A. D. (2000). Identifying violent-toward staff juvenile delinquents via the mil-lon adolescent clinical inventory and neuropsychological measures. Journal of Of-fender Rehabilitation, 32(1/2), 147-165.

Calhoun, G.B., Glaser, B.A., & Petrocelli, J.V. (2002). Personality characteristics of malejuvenile offenders by adjudicated offenses as indicated by the MMPI-A. CriminalJustice and Behavior, 29(2), 183-201.

Catchpole, R.E., & Gretton, H.M. (2003). Predictive validity of risk assessment with vio-lent young offenders: A 1-year examination of criminal outcome. Criminal Justiceand Behavior, 30(6), 688-708.

Clear, T. (1994). Harm in American penology: Offenders, victims, and their communities.Albany, NY: SUNY Press.

Cottle, C.C., Lee, R.J., & Heilbrun, K. (2001). Prediction of criminal recidivism injuveniles: A meta-analysis. Criminal Justice and Behavior, 28(3), 367-394.

Dembo, R., & Schmeidler, J. (2003). Classification of high-risk youths. Crime and Delin-quency, 49(2), 201-230.

Flores, A.W., Travis, L.F., & Latessa, E.L. (2003). Case classification for juvenile correc-tions: An assessment of the youth level of service/case management inventory (YLS/CMI). National Institute of Justice. Rockville, MD: U.S. Government PrintingOffice.

Gendreau, P., Goggin, C., & Law, M. (1997). Predicting prison misconducts. CriminalJustice and Behavior, 24(4), 414-431.

Gendreau, P., Little, T., & Goggin, C. (1996). A meta-analysis of the predictors of adultoffender recidivism: What works! Criminology, 34, 575-607.

Grove, W.M., Zald, D.H., Lebow, B.S., Snitz, B.E., & Nelson, C. (1995). Clinical vs.mechanical prediction: A meta-analysis. Unpublished manuscript.

Hanson, R.K., & Bussiere, M.T. (1998). Predicting relapse: A meta-analysis of sexualoffender recidivism studies. Journal of Consulting and Clinical Psychology, 66, 348-362.

Hoge, R. (2002). Standardized instruments for assessing risk and need in youthful offend-ers. Criminal Justice and Behavior, 29(4), 380-396.

Hoge, R., & Andrews, D.A. (1996). The youth level of service/case management Inven-tory: description and evaluation. Manuscript presented at the American Psychologi-cal Association, Toronto, Canada.

Holsinger, A.M., Lowenkamp, C.T., & Latessa, E.J. (2003). Ethnicity, gender, and thelevel of service inventory—Revised. Journal of Criminal Justice, 31(4), 309-320.

Jones, P. R., & Harris, P. W. (1999). Developing an empirically based typology of delin-quent youths. Journal of Quantitative Criminology, 15(3), 251-276.

Jones, P.R, Harris, P.W., Fader, J., Burrell, J., & Fadeyi, A. (1999). Identifying chronicjuvenile offenders. Corrections Compendium, 24(8), 1-23.



Jung, S., & Rawana, E.P. (1999). Risk and need assessment of juvenile offenders. Crimi-nal Justice & Behavior, 26(1), 69-89.

Lowenakmp, C.T., Holsinger, A.M., Brusman-Lovins, L., & Latessa, E.J. (2004). Assess-ing the inter-rater agreement of the level of service inventory—Revised. FederalProbation, 68(3), 34-38.

Lowenkamp, C.T., & Latessa, E.J. (2005). Increasing the effectiveness of correctionalprogramming through the risk principle: Identifying offenders for residential place-ment. Criminology & Public Policy, 4(2), 263-290.

Lowenkamp, C.T., Holsinger, A.M. & Latessa, E.J. (2001). Risk/need assessment, of-fender classification, and the role of childhood abuse. Criminal Justice & Behavior,28(5), 543-563.

Marczyk, G.R., Heilbrun, K., Lander, T., & DeMatteo, D. (2005). Juvenile decertifica-tion: Developing a model for classification and prediction. Criminal Justice & Behav-ior, 32(3), 278-301.

Motiuk, L. L. (1995). Using familial factors to assess offender risk and need. Forum onCorrections Research, 7(2), 19-22.

National Institute of Corrections (2000). Promoting public safety using effective interven-tions with offenders. Washington, DC.

Righthand, S., Prentky, R., Knight, R., Carpenter, E., Kecker, J.E., & Nangle, D. (2005).Factor structure and validation of the juvenile sex offender assessment protocol (J-SOAP). Sexual Abuse: A Journal of Research and Treatment, 17(1), 13-30.

Schmidt, F., Hoge, R., & Gomes, L. (2005). Reliability and validity analyses of the YouthLevel of Service/Case Management Inventory. Criminal Justice & Behavior, 32(3),329-344.

Shields, I.W, & Simourd, D.J. (1991). Predicting predatory behavior in a population ofincarcerated young offenders. Criminal Justice & Behavior, 18(2), 180-194.

Simourd, D.J. (1997). The criminal sentiments scale-modified and pride in delinquencyscale. Criminal Justice and Behavior, 24, 52-70.

Simourd, D.J., & Andrews, D.A. (1994). Correlates of delinquency: A look at gender-differences. Forum on Corrections Research, 6(1), 26-31.

Documents

Holsinger, Lowenkamp, Latessa, 2006