Sage Journals: Discover world-class research

Abstract

Clinical discretion by youth probation officers to override the actuarial risk assessment classification on the Youth Level of Service/Case Management Inventory (YLS/CMI) is an important feature of case management practices. In a sample of 1,259 youth, the impact of clinical override on the predictive validity of the YLS/CMI and factors associated with override use were studied. Using receiver operating characteristic (ROC) analyses, the YLS/CMI demonstrated large effect size predictive validity values for general recidivism. Override occurred with 10.8% of youth and was more likely to be used with younger youth who had a sexual or violent index offense or those at the upper end of the moderate risk level. For youth where override was used, the YLS/CMI predictive validity became non-significant and fell to chance levels. The implications of these findings for field case management use, training, and probation officer decision-making practices are highlighted.

Keywords

risk–need–responsivity juvenile justice predictive validity risk assessment recidivism

The Risk–Need–Responsivity (RNR) model is a well-established and heavily researched intervention strategy, which guides effective rehabilitation and case management of offenders in the corrections system (Andrews & Bonta, 2010; Bonta & Andrews, 2017). The RNR model, first developed by Andrews et al. (1990), empirically identified RNR factors as being helpful in reducing recidivism. The first principle of “risk” refers to the probability that an individual will commit another offense. According to this principle, reducing recidivism requires the appropriate matching of intervention intensity with risk level. That is, higher risk individuals should receive more intensive intervention than those identified at a lower risk level (Andrews & Bonta, 2010; Bonta & Andrews, 2017). The second principle of “need” guides intervention by focusing rehabilitation efforts on criminogenic factors, which are distinct from non-criminogenic intervention targets. An extensive body of research has identified eight broad criminogenic need domains (i.e., prior offenses, family circumstances/parenting, education, peer relations, substance use, leisure/recreation, personality/behavior, attitudes/orientation) empirically linked to criminal behavior (Bonta & Andrews, 2017). Interventions focused on addressing these criminogenic factors effectively reduce recidivism, while interventions addressing non-criminogenic variables (e.g., self-esteem, anxiety, or depression) are important and needed for the well-being of an offender but do not address the source of re-offending behavior. The third component of the RNR model, “responsivity,” refers to the benefit of considering each offender’s unique characteristics (specific responsivity), such as motivation level, cognitive skills, learning style, or culture, and intentionally matching these individualized strategies to general responsivity factors based on social cognitive learning methods (Bonta & Andrews, 2017). Although the RNR model was initially designed for adult populations, it has been effectively used with youth (e.g., Vieira et al., 2009).

Multiple psychometric tools have been developed to support the implementation of the RNR model in both the adult and youth correction systems. Within the youth justice system, the most established and well-researched RNR risk tool is the Youth Level of Service/Case Management Inventory (YLS/CMI 2.0; Hoge & Andrews, 2011). The YLS/CMI consists of 42 individual items, which identify criminogenic needs organized into eight broad domains empirically related to criminal behavior and recidivism. Summing these 42 individual items provides a YLS/CMI total score (from 0 to 42) which can be categorized into one of four risk level classifications: low, moderate, high, and very high. Responsivity factors are included in the case management plan, allowing a full RNR intervention plan to reduce recidivism (Hoge & Andrews, 2011). Several meta-analyses have demonstrated sound psychometric properties of this instrument (e.g., Olver et al., 2009, 2014).

The YLS/CMI manual allows for the upward or downward adjustment of the Total Score risk classification level through a clinical override feature. The rationale for this provision is that an actuarial or mechanical approach to prediction does not account for unusual or rare factors that can alter risk level such as a youth with psychotic symptoms characterized by violence or aggression. It is suggested that allowing this flexibility will enhance the clinical use and predictive validity of the YLS/CMI. Despite the common clinical practice in the corrections field to include clinical override with risk instruments, there is limited research outlining the strengths and weaknesses of this approach or specific practice guidelines of when to use. The YLS/CMI manual states that the override principle “should only be used in rare circumstances and should be supported by logical arguments and reasonable evidence” and that “final decisions about the client must rest with the professional responsible for the youth” (pp. 9; Hoge & Andrews, 2011). In reference to the adult version of the YLS/CMI, the Level of Service/Case Management Inventory (LS/CMI), Guay and Parent (2018) report that probation officers in their sample were trained to limit clinical override use to no more than 5% of offenders. This type of guideline limiting override use has not been reported in other similar studies. It is important to note that instructions found in the LS/CMI manual restricts the lowering of risk by no more than one category but places no limit on the number of categories risk can be adjusted upward. While this latter information provides more detail or guidance on clinician-based adjustments to risk, the YLS/CMI 2.0 manual does not have similar guidelines for clinical override use (Hoge & Andrews, 2011).

Research examining the use and efficacy of clinical override in the adult corrections field has been reported over the past 20 years. Much of this adult research investigated override use with the LS/CMI (Frechette & Lussier, 2021; Guay & Parent, 2018; Wormith et al., 2012, 2015), or an alternate version called the Level of Service Inventory–Ontario Version (LSI-OR; Girard & Wormith, 2004; Orton et al., 2021). There have also been multiple studies investigating the use of clinical override with adults convicted of a sexual offense (Cohen et al., 2016; Duwe & Rocque, 2018; Orton et al., 2021; Storey et al., 2012; Wormith et al., 2012). Across these studies, override use has been found to vary greatly ranging from a low of 3% (Girard & Wormith, 2004) to a high of roughly 50% in a sample of offenders who committed a sexual offense (Duwe & Rocque, 2018). In most circumstances, clinical override was used to increase the risk classification of the offender, was used more often with offenders who had a sexual rather than non-sexual offense, and resulted in a detrimental impact on the predictive accuracy of the risk instrument used (Duwe & Rocque, 2018; Guay & Parent, 2018; Storey et al., 2012; Wormith et al., 2012, 2015). One exception to this pattern was the finding of Orton et al. (2021) who reported that the predictive validity of the LSI-OR was diminished for general, violent, and nonviolent recidivism, but resulted in improved predictive validity for adults with a sexual offense.

Many of the adult studies reporting on the prevalence of clinical override use and impact on predictive validity also examined possible factors associated with its use. Wormith et al. (2012) found pro-criminal attitudes, anti-social patterns, personal problems, mental health issues, and responsivity factors to be associated with the increased use of clinical override. In a later study, the same authors found Indigenous adults to be less likely to have override used (Wormith et al., 2015), while Orton et al. (2021) found that Indigenous offenders were more likely to receive an adjustment. Frechette and Lussier (2021) found clinical override use occurred more frequently in low-risk community offenders with a violent offense or intimating/controlling personality and with nonviolent offenders who minimized their offense. Guay and Parent (2018) found offenders who were male, detained, had pro-criminal attitudes, and experienced family/marital difficulties were more likely to have clinical override adjusted upward. These authors also found that greater education/employment difficulties were negatively related to upward adjustments, while the total strengths score on the LS/CMI was associated with downward adjustments (Guay & Parent, 2018). Finally, a common finding across all studies was the disproportionately higher rate of clinical override use with adults who committed a sexual offense (Cohen et al., 2016; Guay & Parent, 2018; Orton et al., 2021).

Investigation of clinical override use in the youth justice system has a smaller research base. One of the first studies by Chappell et al. (2013) examined clinical override use when making detention decisions for youth who were considered serious chronic offenders, a community threat, or unlikely to appear at court. In their sample of 4,683 detention youth, clinical override was used in 13.9% of cases. The authors found several factors associated with the use of override (e.g., offense seriousness, offense type, prior intake referrals), but did not report on the impact use had on the predictive accuracy of the risk assessment tool. In the Scottish youth justice system, Vaswani and Merone (2014) studied the use of clinical override with the YLS/CMI in a sample of 1,138 youth. With 14.1% of cases receiving a clinical override, predictive accuracy deteriorated for general recidivism and dropped to non-significance for serious violent re-offending. In a Canadian sample, Schmidt et al. (2016) examined the use of clinical override in a group of youth who committed sexual offenses and a randomly selected matched group of non-sexual offending youth. Clinical override was used with 74% of the sexually offending youth and 41.6% with the non-sexual offending youth. Predictive validity of the YLS/CMI with youth receiving an override adjustment fell to non-significant levels. The authors found that mental health and substance use difficulties were associated with clinical override use in the non-sexually offending group. In a large sample of 11,008 youth across multiple counties in Ohio, clinical override was used in 7% of cases, with 98% of adjustments used to increase risk (McCafferty, 2017). While predictive accuracy was slightly lower after the use of clinical override, it was not significantly different. McCafferty (2017) suggested that, although override use did not improve accuracy, it was not significantly detrimental.

Most recently, Parent et al. (2022) reported on the use of clinical override with the YLS/CMI in a sample of 597 justice-involved youth in Quebec, Canada. The authors found clinical override to be used in 33.5% of cases, with predominantly upward adjustments used (i.e., 30.3% vs. 3.2%). In this sample, the YLS/CMI predictive validity was found to be in the small effect size range and was not negatively impacted by the use of clinical override. Being male and Black were associated with a higher likelihood of upward clinical override adjustment, while older White youth were more likely to have their risk level lowered. All the YLS/CMI domain scores, except for substance use, were positively correlated with increased clinical override, while a lower score on Family Circumstances/Parenting, Personality/Behavior, and Attitudes/Orientation domains were associated with downward override use.

Parallel to adult research on the use of clinical override, the smaller body of youth research highlights considerable variability in use ranging from 7% to 33.5% with one outlier study reporting up to 74% in a sample of youth who committed a sexual offense. In both adult and youth research, use of clinical override generally leads to a decrease in predictive validity of the respective risk tool, with a small number of studies showing no detrimental effect. Moreover, it is noted that adult studies have investigated potential factors associated with the use of clinical override with less information available on youth. The goal of the present study is to build on the existing youth research. First, this study examined probation officers’ utilization of the clinical override feature on the YLS/CMI in a sample of justice involved youth. Second, the impact of clinical override use on the YLS/CMI’s ability to predict general recidivism was examined. Finally, the current study explored possible youth demographic factors, offense characteristics, and criminogenic need profiles associated with the use of clinical override by probation officers.

Method

Participants

Participants for this study included 1,300 randomly selected and de-identified youth who were active in the province of Ontario, Canada, youth justice system in the year 2015. Information on each youth’s risk/need assessments (YLS/CMI) as completed by trained probation officers, convictions while in the youth justice system, clinical override use, and demographic data (i.e., age, sex, Indigenous status) were available and complete for the period of 2009 to 2018. This allowed the ability to track each youth’s general recidivism over an extended time frame following the completion of each youth’s initial YLS/CMI. Within this sample, 27 youth (2.08%) were 19 years of age or older at the time of their first YLS/CMI completion and were excluded from analysis. Another 14 youth (1.08%) were excluded as they did not have a full 12-month follow-up period from the time of their first YLS/CMI to the end of recidivism data collection. The final sample of 1,259 individuals included 1,000 male (79.4%) and 259 female (20.6%) youth. There were 157 Indigenous youth (12.5%). No information was available on the culture or racial background of the other youth who were then lumped together into a non-Indigenous category. The average age of youth at the time of their first YLS/CMI was 16.30 years, with a range of 12.46 to 18.96 years of age. The first YLS/CMI in the youth’s file was used for predictive validity and clinical override analyses. This represented each youth’s first risk assessment, which was available for all youth and was the most uniform and fair point upon which to classify each youth and examine the use of clinical override by probation officers.

Measures

Youth Level of Service/Case Management Inventory (YLS/CMI)

The YLS/CMI is a standardized risk assessment tool, supporting the RNR model of intervention, by identifying risk for recidivism (Hoge & Andrews, 2011). This risk tool consists of 42 items scored as either absent or present resulting in a score range of 0 to 42. The YLS/CMI Total Score is broken into risk classifications of low (0–8), moderate (9–22), high (23–34), and very high (35–42). Each risk item identifies a factor associated with recidivism, grouped into eight domains. The first domain, called Prior and Current Offenses/Dispositions, consists of five static historical factors reflecting past criminal behavior. The remaining items are dynamic and changeable, amenable to intervention, and grouped into the following seven domains: Family Circumstances/Parenting, Education/Employment, Peer Relations, Substance Abuse, Leisure/Recreation, Personality/Behavior, and Attitudes/Orientation. As a risk measure, predictive validity is a key psychometric construct. Considerable research has been done in this latter area with several meta-analyses demonstrating sound predictive validity across different client groups and countries (Olver et al., 2009, 2014). The YLS/CMI has also demonstrated sound internal consistency (Catchpole & Gretton, 2003; Thompson & Putnins, 2003), inter-rater reliability (Schmidt et al., 2005), and test–retest reliability (Thompson & Putnins, 2003). In addition to the structured assessment of risk and criminogenic needs, the YLS/CMI allows for practitioner discretion to override the actuarial risk score if the clinician believes the risk score does not accurately reflect the youth’s actual risk level (Hoge & Andrews, 2011).

Recidivism Data

The youth justice history of each randomly selected youth, including their criminal history of offense convictions, was pulled on December 31, 2018. The 2009–2018 time frame resulted in a lengthy follow-up period, with a mean time frame of 1,471.47 days (range of 374–3,467 days). Given the large variability across youth in this sample, a fixed follow-up period to track recidivism was used. This method has been recommended by Harris et al. (2003) as a means of improving the measurement of instrument accuracy. General recidivism was defined as any conviction which occurred within the first 24 months (730 days) following completion of the first YLS/CMI, while ensuring a minimum 12-month availability to re-offend. Any convictions which occurred prior to the first YLS/CMI were used to determine index offense classification types as described below. Recidivism data were Ontario-specific and available for any offense convictions while under the age of 18 or on a youth justice order after the age of 18. It did not include adult convictions following release from the youth justice system or offenses outside of the province of Ontario.

Index Offense Classification

Included in each youth’s history of conviction data were an existing provincial 26-offense type coding scheme, which ranked offenses from most to least serious. This existing Most Serious Offense (MSO) data were used to automatically re-code offenses into violent, nonviolent, sexual, and technical offense categories. The violent category included homicide and related offenses, serious violent offenses, weapons offenses, assault and related offenses, and arson. Nonviolent offenses included break-and-enter, drug trafficking, fraud, theft/possession, moral offenses, drug possession, traffic offenses, impaired driving, public order offenses, other federal offices, highway traffic act (provincial), liquor control act (provincial), other provincial offenses, municipal bylaws, and unknown. Sexual offenses included violent and nonviolent sexual offenses. Technical offenses consisted of obstruction of justice, administration of justice, and parole violations. Each youth was classified into these four index offense classification types so that analyses could be completed on the possible role that initial criminal offense characteristics (i.e., violent and/or sexual index offense) may play on the use of clinical override. Research approval for this study was obtained from the Youth Justice Services division of the Ontario Ministry of Children and Youth Services and ethics approval was obtained from the local university.

Analytical Strategy

All analyses were conducted using SPSS v. 28. The frequency of clinical override use and percent of youth in each risk classification level was done using descriptive statistics. A key question of this study was the impact of clinical override use on the ability of the YLS/CMI to predict general recidivism. To that end, both receiver operating characteristic (ROC) and correlational analyses were completed for youth who did and did not receive a clinical override. ROC analysis, using area under the curve (AUC), is one of the most frequently used methods to determine a risk tools predictive accuracy (Viljoen et al., 2021). AUC values of 1.00 represent perfect prediction, whereas values around .50 represent poor or chance prediction (Rice & Harris, 1995). Rice and Harris (2005) described AUC values of .556 as small, .639 as medium, and .714 as a large predictive validity effect size. Similarly, interpretation of continuous correlational analyses was done according to guidelines recommended by Cohen (1988) for small, medium, and large correlation effect sizes represented by .10, .30, and .50, respectively. For point biserial correlations, Rice and Harris (2005) report slightly adjusted correlation values for small (.10), medium (.24), and large (.37) effect sizes. To determine whether the predictive validity of the YLS/CMI differed across youth who did and did not have a clinical override, an ROC comparative analysis was done as recommended by Hanley and McNeil (1983). Comparison of group differences across independent samples can be completed by using a critical z-score which must reach ±1.96 to achieve statistical significance at the .05 level.

A final objective of this study was to explore possible factors associated with the use of clinical override. To answer this question, one set of analyses involved the completion of partial point biserial correlations with available youth information (demographics, index offense type, and YLS/CMI domain scores) and clinical override use, while controlling for overall risk level using the YLS/CMI Total Score. A second component of analyses involved a decision tree analysis (DTA) using available youth offender characteristics. DTA is a non-parametric method used to classify and examine the relationship between variables and a specific outcome (Song & Lu, 2015).

In the current study, DTA was completed using the chi-square automatic interaction detector (CHAID) method. This is an exploratory data-mining technique whereby the data set are systematically searched to find the predictor that shows the greatest degree of differentiation in the outcome variable (i.e., clinical override). Differentiation of associated predictors is represented by a split in the data and the formation of a decision tree with specific nodes identifying the significant variables associated with the outcome. This process is repeated to form a multi-tiered or hierarchical decision tree model which is repeated until all possible nodes are exhausted. This process is built through chi-square analysis and use of p values with Bonferroni corrections to determine each node. DTA has a number of unique advantages over other types of analyses and has been effectively used in past research on clinical override (Frechette & Lussier, 2021). One strength of DTA is the ability to handle a large number of independent variables that may be in mixed formats (i.e., nominal, ordinal, or interval). It also has advantages over traditional statistical approaches, such as regression models, because of its iterative process and ability to identify factors and place them in a hierarchical nested framework. This provides the capacity to capture unique combinations of factors in a way not possible with regression models and place factors in an order of importance or association with the outcome variable. Perhaps one of the most significant benefits of a DTA approach is the ability to look at a combination of factors which is not in a simple additive relationship. DTA looks at the interaction and combination of variables in a more complex manner that allows the opportunity to look at predictor variables which are nested within other predictor variables.

Results

Sample Description

All continuous data variables were screened for outliers and standard error of kurtosis was within ±2, and skewness within ±3. The YLS/CMI profiles for the full sample and youth who did and did not receive a clinical override are shown in Table 1. The overall mean YLS/CMI Total Score was 14.4 (SD = 7.7). The YLS/CMI Total Score of 14.2 for youth who did not receive a clinical override was significantly lower than the 16.1 obtained by youth who did receive a clinical override, t (1257) = −2.76, p = .005, representing a small effect size difference (d = .25). Additional analyses on YLS/CMI Total Scores on important subgroups revealed that girls (15.1) were not different from boys (14.2), t (1257) = −1.59, p = .11, d = .11, but that Indigenous youth (M = 16.1; SD = 6.9) had significantly higher Total Score values than non-Indigenous youth (M = 14.2; SD = 7.8), t (1257) = −3.04, p = .002, d = .26. Youth with an index sexual offense (M = 10.1, SD = 7.7) were significantly lower than youth with a non-sexual index offense (M = 14.7, SD = 7.6), t (1257) = 5.01, p < .001, falling at a moderate effect size level (d = .60).

Table 1:

Initial YLS/CMI Assessment Scores for Total Sample and by Risk Override Use

YLS/CMI domains	Total sample(N = 1,259)	No risk override use(n = 1,123)	Risk override only(n = 136)
YLS/CMI domains	M (SD)	M (SD)	M (SD)
Total score	14.4 (7.7)	14.2 (7.6)	16.1 (8.4)
Prior history	0.2 (0.6)	0.2 (0.6)	0.3 (0.6)
Family circumstances	2.5 (1.7)	2.5 (1.7)	2.8 (1.7)
Education/employment	2.9 (1.9)	2.9 (1.9)	3.1 (2.2)
Peer relationships	1.9 (1.3)	1.9 (1.3)	2.0 (1.4)
Substance use	1.7 (1.5)	1.7 (1.4)	1.5 (1.6)
Leisure/recreation	1.4 (1.0)	1.4 (1.0)	1.4 (1.0)
Personality/behavior	2.6 (2.0)	2.5 (2.0)	3.3 (2.1)
Attitudes/orientation	1.2 (1.3)	1.1 (1.3)	1.6 (1.5)

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

The average overall Total Score of 14.4 in this sample was slightly higher than that seen by Parent et al. (2022) (M = 13.4; SD = 6.8) and slightly lower than Vaswani and Merone (2014) (M =15.2; SD = 9.4). However, the Total Score in this sample was lower than that found in youth referred for court-ordered assessments (M = 16.9; SD = 9.3; Schmidt et al., 2005), and much lower than a sample of high-risk youth (M = 25.1; SD = 7.7; Olver et al., 2012), or another community sample of youth (M = 20.34; SD = 9.28; Luong & Wormith, 2011).

The rate of general recidivism of 65.4% (n = 823) in this sample was similar to other youth studies on clinical override use. McCafferty (2017) reported a 1-year general recidivism rate of 44%, while Vaswani and Merone (2014) reported a higher level of 73.6% over the same time frame. Over an average follow-up period of 957 days, Schmidt et al. (2016) reported a 63.2% general recidivism rate in a subsample of non-sexual offending youth. Finally, Parent et al. (2022) obtained a 51.3% general recidivism rate in youth followed over a 2-year time period. Comparison of important subgroups in this sample did not reveal a difference in recidivism between male (66.6%) and female (60.6%) youth, χ² (1, n = 1,259) = 3.25, p = .07, or youth with an index sexual (65.3%) and non-sexual (65.4%) offense, χ² (1, n = 1,259) = .00, p > .05. Differences in general recidivism were found between Indigenous (73.2%) and non-Indigenous (64.2%) youth, χ² (1, n = 1,259) = 4.92, p = .027. For those youth who re-offended, the average time to first general recidivism event was 156.39 days (SD = 166.0 days), with a range of 2 to 726 days.

Patterns of Clinical Override Use

Examination of clinical override use in the full sample revealed that 10.8% of youth had their risk level adjusted (n = 136). In all cases, clinical override was used to increase risk-level classification. An increase of one risk level was used most frequently (8.6%; n = 108), followed by an increase of two risk levels (1.9%; n = 24), and an increase of three risk levels (0.3%; n = 4). For the 136 youth who received a clinical override, their original risk classification, based on the YLS/CMI Total Score, resulted in 27.9% falling into the low category, 67.6% in moderate, and 4.4% in high. Following clinical override use, 0.0% were low, 12.5% were moderate, 75.0% were high, and 12.5% were very high.

YLS/CMI Predictive Validity

Predictive validity results using ROC analyses and point biserial correlations for YLS/CMI Total Score, original YLS/CMI risk level, and override YLS/CMI risk level classifications are displayed in Table 2. The YLS/CMI Total Score was found to have sound predictive validity for general recidivism across the full sample with an AUC value of .71, which falls within the large effect size range. This large effect size AUC was also observed in the initial predictive validity of the YLS/CMI for both youth who received (AUC = .73) and did not receive (AUC = .71) a clinical override. No statistically significant difference in AUC values was found between groups where clinical override was applied and not applied, with a z-score of .95 obtained. In addition to ROC and point biserial correlation analyses, predictive validity of the YLS/CMI was examined through the number of days to first re-offense and the number of recidivism events following the first YLS/CMI. The correlation between the YLS/CMI Total Score and days to first general recidivism event for the 823 youth who recidivated was a small effect size of −.12 (p < .001) and for number of recidivism events was .24 (p < .001), falling between a small and medium effect size.

Table 2:

Point Biserial Correlations and AUC Values for YLS/CMI Total Score, Original Risk Level, and Risk Override Level With General Recidivism

	Total score		Original risk level		Override risk level
YLS/CMI total score	r	AUC [95% CI]	r	AUC [95% CI]	r	AUC [95% CI]
Total sample^a	.35**	.71 [.68, .74]	.30**	.65 [.62, 68]	.28**	.65 [.61, .68]
Risk override only^b	.34**	.73 [.63, .83]	.33**	.68 [.57, .79]	.07	.53 [.42, .65]
No risk override^c	.34**	.71 [.68, .74]	.30**	.65 [.62, .68]

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

N = 1,259. ^b n = 136. ^c n = 1,123.

p < .05. **p < .01.

Table 2 also displays the AUC values comparing the YLS/CMI predictive validity values based on risk-level classifications. Before and after the use of clinical override, the YLS/CMI predictive validity for the total sample was unaffected. However, the AUC value for the clinical override only group fell sharply going from a moderate effect size value of .68 to a non-significant value of .53 after clinical override. Examination of point biserial correlation coefficients closely followed the results of the ROC analyses.

Factors Associated With Clinical Override Use

Two levels of analyses were completed to examine possible factors associated with clinical override use. First, point biserial correlations were completed between clinical override use and available demographic and index offense information available in the data set, while controlling for the YLS/CMI Total Score. These variables included gender, Indigenous status, age, and the occurrence of four index offense types including violent, nonviolent, sexual, and technical offenses. Clinical override use was also correlated with the eight YLS/CMI domain scores. The results of these analyses are shown in Table 3. Small effect size associations were found for age and violent index offense, while a sexual index offense was found to have an effect size between the medium to large range. Younger age youth and those with a violent or sexual index offense were more likely to receive a clinical override adjustment. There were also positive associations between three of the eight YLS/CMI domains and clinical override use, all at a small effect size level. Higher scores on the Personality and Attitudes/Orientation domains were associated with a higher likelihood of clinical override use. However, youth with a higher substance use score were less likely to receive a clinical override adjustment.

Table 3:

Partial Point Biserial Correlation Matrix With Demographic Characteristics, Index Offense Types, and YLS/CMI Domains With the Use of Clinical Override Controlling for YLS/CMI Total Score (N = 1,259)

Demographic and index offense types	Correlation	YLS/CMI domains	Correlation
Male	−.048	Prior/current offense	.004
Indigenous	−.030	Family	.004
Age	−.055*	Education	−.037
Violent offense	.075**	Peer relations	−.032
Nonviolent offense	−.039	Substance use	−.101**
Sexual offense	.308**	Leisure/recreation	−.045
Technical offense	−.035	Personality	.108**
		Attitudes/orientation	.094**

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

p < .05. **p < .01.

Second, to more fully understand the offender characteristics related to clinical override use, a DTA was done. Based on the available demographic variables, index offense types, and YLS/CMI variables, there were five factors found to be associated with clinical override use. These results are shown in Figure 1. The first-level factor involved the presence of an index sexual offense. Youth who had an index sexual offense were roughly 5 times more likely to have override used (54%) when compared with the sample average of 10.8%. At the second level of analysis, specific to youth with a non-sexual offense, a YLS/CMI Total Score of 22 was related to a greater likelihood of override use. Importantly, these are youth who fall at the very top end of the moderate risk level classification range (i.e., 9–22). This pattern can be seen through node 11 in Figure 1, where youth who had a YLS/CMI Total Score of 22 were twice as likely to be given a clinical override adjustment (i.e., 23.6%). In addition, for other youth in the moderate risk level range who had a Total Score between 14 and 21 and an elevation on the YLS/CMI Personality domain with a score of either 5 or 6 (node 10) were also 2 times more likely to have clinical override used (i.e., 22.6%). Finally, youth who fell within the low-risk classification level and were verbally aggressive (node 8) were given a clinical override 14.5% of the time.

Figure 1:

Decision Tree Analysis of Demographic, Index Offense Types, and Youth Level of Service/Case Management Inventory Domain and Item Scores Associated With Clinical Override Use

Discussion

The current study examined the frequency of clinical override use and the resulting impact on predictive validity of the YLS/CMI in a large representative sample of youth in the corrections system. The rate of clinical override use across the full sample was 10.8% and was always used to increase a youth’s risk-level classification. The YLS/CMI predictive validity fell to chance levels and was non-significant for youth who were given a clinical override. Several youth variables were associated with the use of clinical override by probation officers including younger youth and those who had a violent or sexual index offense. In addition, override was more likely to be used if, based on the YLS/CMI risk profile, they were verbally aggressive, had higher levels of Personality and Attitude/Orientation domain scores, or lower levels of Substance Use risk. In addition, youth within the upper end of the moderate risk-level classification range were 2 times more likely to be given a clinical override.

Clinical Override Use and YLS/CMI Predictive Validity

The use of clinical override with 10.8% of youth in this sample appears to be slightly higher than the 5% benchmark reported in the study by Guay and Parent (2018) or the “rare circumstance” described in the YLS/CMI manual (Hoge & Andrews, 2011). This finding, along with the existing literature on clinical override use with risk tools, reflects a larger practice issue in the risk assessment field regarding the use of professional judgment. There is an established history suggesting that actuarial and empirically based risk assessments, on average, are superior to clinical judgment alone approaches (Ægisdóttit et al., 2006; Grove et al., 2000). While some have argued that clinical judgment does not increase incremental validity when used in combination with actuarial risk tools for dangerous sex offender predictions (Abbott, 2011), the risk assessment field has generally accepted the use of professional judgment in this manner. For example, there is a strong perspective in the violence risk assessment field to use a structured professional judgment (SPJ) approach over a pure actuarial method (Hart et al., 2017), placing a premium on clinical judgment to integrate relevant factors into a risk-level determination.

A preference for clinical override flexibility was reinforced by the results of Guy et al. (2014) who reported probation officers’ preference to have professional discretion when using the YLS/CMI rather than strictly adhering to the actuarial score. Moreover, in a recent in-depth and thorough review of the risk prediction literature, Viljoen et al. (2021) suggested that the strength of the research base to support the conclusion that actuarial approaches are superior to clinical judgment may have some important gaps which require further investigation. When taken together, these differing perspectives of how to integrate clinical judgment into risk assessments leaves some uncertainty about how and when to use the clinical override feature. The value placed on clinical judgment in everyday field practice is clearly seen in the override literature where considerable variability and high rates of override use, ranging from 3% to 50%, exist in both the adult and youth justice systems. The high and variable rate of clinical override use across studies suggests that there may be uncertainty about best practices and/or a lack of confidence in strictly relying on actuarially based risk tools.

Looking specifically at the 136 youth in this study where clinical override was used, the predictive validity of the YLS/CMI, using the risk-level classification, dropped from an AUC of .68 (confidence interval [CI] = [.57, .79]) to a non-significant value of .53 (CI = [.42, .65]). This represents a drop from a moderate effect size to chance levels of prediction, significantly compromising the predictive validity of the YLS/CMI in those youth. This pattern is not unique to this study and has been found in similar research. The use of professional judgment through the clinical override feature frequently appears to compromise the predictive validity of a risk tool (Duwe & Rocque, 2018; Orton et al., 2021; Schmidt et al., 2016; Storey et al., 2012; Vaswani & Merone, 2014; Wormith et al., 2012, 2015) or, in some instances, will show no or minimal detrimental effects (McCafferty, 2017; Parent et al., 2022). It does not appear, however, to improve or enhance predictive validity as might be suggested by the SPJ perspective. Even within two studies where specific probation officer training on the LS/CMI was explicitly directed to not use clinical override adjustments beyond 5% of clients, the resulting rates of 6.5% and 4.1% of clinical override use still appeared to compromise predictive validity (Frechette & Lussier, 2021; Guay & Parent, 2018).

These current results, along with the existing small body of research in this area, have important implications for training of probation officers and how best to effectively implement an RNR model of intervention. For example, in the YLS/CMI manual, Hoge and Andrews (2011) suggest that the clinical override feature should be used in “rare” circumstances. In the training that probation officers received in the Guay and Parent (2018) study, this was interpreted to mean no more than 5% of cases. However, it is currently not clear from other researchers and practitioners in the field what is or is not an acceptable level of clinical override use. The current results suggest that greater attention must be given to this question in both academic research and clinical practice. Given the current knowledge base, it would appear prudent to use clinical override less often and closely follow the guidance given in the YLS/CMI manual to use in “rare” circumstances only. To that end, the establishment of clearer and monitored evidence-based practice parameters for probation officers is needed and would likely enhance the application of clinical override in the RNR model of intervention.

An important finding in this study was the practice of probation officers who only increased and never decreased clinical override adjustments. A bias toward upward risk adjustments is consistently found within the existing literature (McCafferty, 2017; Orton et al., 2021; Schmidt et al., 2016; Wormith et al., 2012, 2015) and suggests a possible risk aversion bias in probation officer decision-making. Theoretically, one might expect that risk adjustment could be lowered equally often as it would be raised. However, this is not seen in practice and was not seen in this sample.

The bias toward increasing risk ratings may be related to the potential consequences of a recidivism outcome. Officers may view increasing risk scores as a cautious and less problematic approach when compared with lowering risk levels. That is, the occurrence of a false positive is less concerning than underestimating the risk of a youth who re-offends. This pattern was clearly seen in youth who had an index sexual offense. These sexually offending youth had a very high rate of clinical override use (i.e., 54.7% compared with non-sexual offending youth rate of 8.0%). Probation officers may have been concerned about the implications of a youth sexually re-offending or that the YLS/CMI was not adequate to identity sexual recidivism risk in this group and decided to use the clinical override feature to compensate. A similar tendency to increase risk classifications was found by Childs et al. (2014) when comparing probation officers’ determination of risk to a statistical method of risk determination. Moreover, in a qualitative youth justice study by Ballucci (2012), risk inflation practices by probation officers were found. Officers consistently raised, not lowered, risk classification as a means of managing community safety and personal liability issues. This potential risk adjustment bias, reinforced by the results of this study, has important practice implications. Specific attention may need to be given on how to address risk issues with youth who commit sexual offenses in addition to addressing possible inherent decision-making biases to increase risk-level classifications. Beyond risk aversion, Schaefer and Williamson (2018) identified other factors that may affect experienced probation officers’ decision-making. These authors described the role that emotional exhaustion, depersonalization, and years of experience had on officers’ non-compliance with expected RNR case management practices (Schaefer & Williamson, 2018). When taken together, there are multiple factors to consider when trying to understand the use of clinical override adjustments with risk/case management tools such as the YLS/CMI.

Factors Associated With Clinical Override Use

A gap in the literature on clinical override use has been the identification of factors associated with clinical override use. In this study, multiple youth variables were identified. Most salient was the strong effect for an index sexual offense. This factor fell between a medium and large effect size correlation with override use (r = .308) and was the first-level factor identified on the DTA. The presence of a sexual offense has been consistently found in both the adult (Frechette & Lussier, 2021; Guay & Parent, 2018; Wormith et al., 2012) and youth (Parent et al., 2022; Schmidt et al., 2016) research on clinical override use. This may not be surprising given that general recidivism risk tools such as the LS/CMI and YLS/CMI are not designed to predict sexual re-offending. Probation officers may have been overcompensating for this limitation and used the override feature more frequently because of the serious implications of a youth or adult who sexually re-offended. Considering this finding, it may be important to integrate the use of a sexual offending–specific risk tool for these youth rather than relying on a general recidivism measure such as the YLS/CMI. This may give probation officers more confidence in their risk assessment and decrease the need to do a clinical override, which dropped the YLS/CMI predictive accuracy to chance levels.

The demographic variables of age, gender, and Indigenous status were also examined for a possible relationship with clinical override use. Gender and Indigenous status did not appear to be related to override use in this study. Being male has been associated with greater clinical override use in past research with adult offenders (Guay & Parent, 2018; Orton et al., 2021) and in youth (Parent et al., 2022), but not related in other studies (Chappell et al., 2013; Frechette & Lussier, 2021; Wormith et al., 2012). It may be that the relationship of gender to clinical override use is dependent with or affected by other proxy factors such as nature of index offense or anti-social personality characteristics and behavior. This may account for some of the inconsistent results in gender and clinical override use. Indigenous status was also not related to clinical override use in this study, although Orton et al. (2021) did find a positive association in adults. The current result for Indigenous youth does parallel previous research (Wormith et al., 2012) where Indigenous adults were less likely to have override used, which the authors suggested may be due to a ceiling effect of the initially higher risk/need scores of Indigenous offenders in their sample. The Indigenous youth in this sample also had the highest YLS/CMI Total Scores when compared with other demographic groups and this may be a factor in the current finding. A final demographic factor examined was age. In this study, there was a small negative correlation suggesting that younger youth were slightly more likely to have override used. This differs from the study of youth by Chappell et al. (2013) where a positive association was found. This latter study, however, involved decision-making regarding detention practices and did not use the YLS/CMI risk tool. It is difficult to reach any conclusions regarding age given the limited studies that examined this variable and the differing purposes for clinical override use.

The eight YLS/CMI domains were examined as possible factors associated with clinical override use. These factors have not been previously studied in youth, although Parent et al. (2022) recently looked at these factors with respect to both upward and downward clinical override adjustments. Parent et al. (2022) found that the YLS/CMI Total Score and domain scores, except for substance use, were positively correlated with increased clinical override adjustments. In this sample, after controlling for the YLS/CMI total score, the YLS/CMI Personality and Attitudes/Orientation domains were positively related to override use. There was also a small positive correlation between clinical override use and having a violent, but not nonviolent, index offense and that, as a group, youth who received a clinical override had a higher YLS/CMI Total Score when compared with youth who did not get an override. This is consistent with past adult (Frechette & Lussier, 2021; Guay & Parent, 2018; Wormith et al., 2012) and youth research (Parent et al., 2022) and suggests that probation officers may be using salient criminogenic needs to identify youth which require more intensive case management practices.

The pattern of flagging youth with more anti-social characteristics was highlighted on the DTA where the combination of a moderate risk level classification and elevated Personality domain scores as well as low-risk youth who were verbally aggressive were most likely to have clinical override used. These factors, related to anti-social or aggressive tendencies, may be used by officers to increase risk classification only if the youth is within one of the lower risk level classifications and may parallel the association found with a violent index offense and override use. The DTA also identified a high prevalence of clinical override use for youth who fell within the upper end of the moderate risk level classification and at the cut-off between moderate and high-risk classification. Officers were much more likely to increase risk classification for youth at the upper end of the moderate risk classification range (YLS/CMI Total Score of 22) when compared with other youth in the moderate range.

The use of clinical override is a complex decision-making process. Thus, rather than risk override being a function of an officer’s specific risk prediction or effort to mitigate potential liability for future recidivism, it may also reflect a risk management intervention strategy as suggested by Frechette and Lussier (2021). That is, raising the risk classification of a youth may be performed to direct and apply more resources and a higher level of monitoring to a specific youth. Youth at the upper end of the moderate risk range or youth with anti-social personality characteristics may be viewed by officers as needing more resources or a higher level of monitoring than that identified by the YLS/CMI Total Score. Future research may require a more complex and interactive analysis of such multiple variable combinations to better understand the application of clinical override and probation officers’ reasoning for adjustment. The use of the DTA analysis, as done in this study and by Frechette and Lussier (2021), is one approach which may enhance the ability to look at the combination of multiple factors at one time. Deeper investigation of these associated factors, individually and in combination, may help establish practice guidelines and more effective training practices for the use of case management risk tools such as the YLS/CMI. For example, youth with individual risk items (i.e., verbally aggressive) or one higher domain score (i.e., Personality) but falling in the low or moderate risk classification may require particular attention in training practices. Given that the use of clinical override in these circumstances compromised accuracy, it will be important for future training practices to emphasize that clinical override decisions are best based on an overall actuarial score rather than focusing on any one individual item or one domain. Training should incorporate clearer guidelines about how to manage anecdotal clinical information or isolated items from the YLS/CMI in the overall risk classification. Together with practitioners’ general bias to inflate or increase risk assessment determinations, it will be important to highlight possible practitioner biases and the drop in risk accuracy when too much weight is given to an individual risk item(s).

Study Limitations

The current study has several strengths, including the large and representative sample of youth in the Ontario correctional system. However, it is not without some important limitations that need to be recognized and considered when interpreting the results. The current study only examined offender characteristics when looking at factors potentially associated with the use of clinical override. There was no information available regarding offender strengths or protective factors, probation officer characteristics, the training conducted with probation officers, or any information on institutional practices. These latter factors may have a significant impact on clinical override use and may provide a richer understanding of clinical override decision-making if considered in future research. The available recidivism data were collected while the youth were involved with the youth justice system and did not include adult recidivism data or offenses which may have occurred outside the province of Ontario. Finally, the data used for this study were based on information from a provincial dataset completed by probation officers as part of everyday clinical practice. There was no opportunity to check data accuracy or corroborate the data entered with information directly from clients or to assess inter-rater reliability among groups of probation officers. It may also be possible that there was a bias by officers who used clinical override adjustments to have poorer fidelity to YLS/CMI assessment and interpretation guidelines.

Conclusion and Clinical Applications

The results of the current study reinforce and complement many of the findings found in the growing body of research on clinical override use in the correctional system. First and foremost, the YLS/CMI was found to have sound predictive validity in youth, with AUC values in the large effect size range. However, for youth where the clinical override adjustment was applied, the predictive validity was compromised and decreased to non-significant levels. Practically, these results have significant implications on the use of clinical override when evaluating youth within an RNR case management plan. It may be beneficial to have clearer practice guidelines for risk tools such as the YLS/CMI. This has been previously recommended by Guy et al. (2014) and still appears to be a valid and important consideration today. As one part of more comprehensive practice guidelines, the existing research on clinical override use could be proactively integrated into YLS/CMI training along with enhanced supervision of probation officers in this area of practice. Key in this training and implementation process is strong leadership support and clear institutional practice guidelines on override use. Vincent et al. (2012) have written an implementation manual for a similar and related risk assessment tool called the Structured Assessment of Violence Risk in Youth (SAVRY). The use of a similar guiding document may be helpful for other risk tools such as the YLS/CMI. Finally, there are important clinical practice questions which require further research. Which youth require a clinical override adjustment? When should risk adjustments be raised? When should risk adjustments be lowered? Additional information on these questions and others could improve standard case management practices and the effective application of risk tools such as the YLS/CMI in the youth justice system.

Footnotes

AUTHORS’ NOTE:

The authors wish to express their appreciation to the Ontario Ministry of Children and Youth Services (MCYS) for permission to use the data described in this study. A special thank you is extended to Dr. Laura Freedman for her considerable assistance. The views expressed in this paper do not necessarily reflect the views of the MCYS.

ORCID iD

Fred Schmidt

Fred Schmidt is a clinical and forensic psychologist working at the Children’s Centre Thunder Bay who has worked across the children’s mental health,child welfare,and youth justice systems. He is also an adjunct professor at Lakehead University.

Amy Killen is a quality improvement data analyst with the Children’s Centre Thunder Bay. She is heavily involved in analyzing and producing reports which support the Centre’s ongoing evaluation of services and enhancing organizational continuous quality improvement practices.

Dilys Haner is a clinical and forensic psychologist working with children,adolescents,adults,and families involved in,or at risk of involvement with,the justice system in southwestern Ontario. Their work at Navigating Onward and London Family Court Clinic includes providing adolescent psychological assessments under sec. 34 of the Criminal Youth Justice Act (CYJA) and other Court-ordered assessments.

Elaine Toombs is an adjunct professor in the Department of Psychology at Lakehead University. Her research primarily focuses on First Nations mental health using community-based methods,including how eHealth interventions can be used in a culturally relevant manner with Indigenous people seeking treatment for substance use. In addition to her academic work,she is a registered clinical psychologist and provides assessment,intervention,and consultation services for First Nation children,adolescents,and adults.

References

Abbott

B. R.

(2011). Throwing the baby out with the bath water: Is it time for clinical judgment to supplement actuarial risk assessment? Journal of the American Academy of Psychiatry and the Law, 39(2), 222–230. https://jaapl.org/content/jaapl/39/2/222.full.pdf

Ægisdóttir

White

M. J.

Spengler

P. M.

Maugherman

A. S.

Anderson

L. A.

Cook

R. S.

Nichols

C. N.

Lampropoulos

G. K.

Walker

B. S.

Cohen

Rush

J. D.

(2006). The meta-analysis of clinical judgment project: Fifty-six years of accumulated research on clinical versus statistical prediction. The Counseling Psychologist, 34(3), 341–382. https://doi.org/10.1177/0011000005285875

Andrews

D. A.

Bonta

(2010). Rehabilitating criminal justice policy and practice. Psychology, Public Policy, and Law, 16(1), 39–55. https://doi.org/10.1037/a0018362

Andrews

D. A.

Bonta

Hoge

R. D.

(1990). Classification for effective rehabilitation: Rediscovering psychology. Criminal Justice and Behavior, 17(1), 19–52. https://doi.org/10.1177/0093854890017001004

Ballucci

(2012). Subverting and negotiating risk assessment: A case study of the LSI in a Canadian youth custody facility. Canadian Journal of Criminology and Criminal Justice, 54(2), 203–228. https://doi.org/10.3138/cjccj.2010.E.27

Bonta

Andrews

D. A.

(2017). The psychology of criminal conduct (6th ed.). Routledge.

Catchpole

R. E.

Gretton

H. M.

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

Chappell

A. T.

Maggard

S. R.

Higgins

J. L.

(2013). Exceptions to the rule? Exploring the use of overrides in detention risk assessment. Youth Violence and Juvenile Justice, 11(4), 332–348. https://doi.org/10.1177/1541204012470849

Childs

Frick

P. J.

Ryals

J. S.

Jr. Lingonblad

Villo

M. J.

(2014). A comparison of empirically based and structured professional judgment estimation of risk using the Structured Assessment of Violence Risk in Youth. Youth Violence and Juvenile Justice, 12(1), 40–57. https://doi.org/10.1177/1541204013480368

10.

Cohen

(1988). Statistical power analysis for the behavioral sciences (2nd ed.). Lawrence Erlbaum.

11.

Cohen

T. H.

Pendergast

VanBenshoten

S. W.

(2016). Examining overrides of risk classifications for offenders on federal supervision. Federal Probation, 80, 12–21. https://www.uscourts.gov/sites/default/files/80_1_2_0.pdf

12.

Duwe

Rocque

(2018). The home-field advantage and the perils of professional judgment: Evaluating the performance of the Static-99R and the MnSOST-3 in predicting sexual recidivism. Law and Human Behavior, 42(3), 269–279. https://doi.org/10.1037/lhb0000277

13.

Frechette

Lussier

(2021). Betting against the odds: The mysterious case of the clinical override in risk assessment of adult convicted offenders. International Journal of Offender Therapy and Comparative Criminology, 67(9), 887–909. https://doi.org/10.1177/0306624X211049181

14.

Girard

Wormith

(2004). The predictive validity of the Level of Service Inventory-Ontario Revision on general and violent recidivism among various offender groups. Criminal Justice and Behavior, 31(2), 150–181. https://doi.org/10.1177/0093854803261335

15.

Grove

W. M.

Zald

D. H.

Lebow

B. S.

Snitz

B. E.

Nelson

(2000). Clinical versus mechanical prediction: A meta-analysis. Psychological Assessment, 12(1), 19–30. https://doi.org/10.1037/1040-3590.12.1.19

16.

Guay

J. P.

Parent

(2018). Broken legs, clinical overrides, and recidivism risk: An analysis of decisions to adjust risk levels with the LS/CMI. Criminal Justice and Behavior, 45, 82–100. https://doi.org/10.1177/0093854817719482

17.

Guy

L. S.

Nelson

R. J.

Fusco-Morin

S. L.

Vincent

G. M.

(2014). What do juvenile probation officers think of using the SAVRY and YLS/CMI for case management, and do they use the instruments properly? International Journal of Forensic Mental Health, 13, 227–241. https://doi.org/10.1080/14999013.2014.939789

18.

Hanley

J. A.

McNeil

B. J.

(1983). A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology, 148(3), 839–843. https://doi.org/10.1148/radiology.148.3.6878708

19.

Harris

G. T.

Rice

M. E.

Quinsey

V. L.

LaLumiere

M. L.

Boer

Lang

(2003). A multisite comparison of actuarial risk instruments for sex offenders. Psychological Assessment, 15(3), 413–425. https://doi.org/10.1037/1040-3590.15.3.413

20.

Hart

S. D.

Douglas

K. S.

Guy

L. S.

(2017). The structured professional judgement approach to violence risk assessment: Origins, nature, and advances. In Boer

D. P.

Beech

A. R.

Ward

Craig

L. A.

Rettenberger

Marshall

L. E.

Marshall

W. L.

(Eds.), The Wiley handbook on the theories, assessment, and treatment of sexual offending (pp. 643–666). Wiley Blackwell.

21.

Hoge

R. D.

Andrews

D. A.

(2011). Youth Level of Service/Case Management Inventory 2.0 (YLS/CMI 2.0) user’s manual. Multi-Health Systems.

22.

Luong

Wormith

J. S.

(2011). Applying risk/need assessment to probation practice and its impact on the recidivism of young offenders. Criminal Justice and Behavior, 38(12), 1177–1199. https://doi.org/10.1177/0093854811421596

23.

McCafferty

J. T.

(2017). Professional discretion and the predictive validity of juvenile risk assessment instrument exploring the overlooked principle of effective correctional classification. Youth Violence and Juvenile Justice, 15(2), 103–118. https://doi.org/10.1177/154/204015622255

24.

Olver

M. E.

Stockdale

K. C.

Wong

S. C. P.

(2012). Short and long-term prediction of recidivism using the youth level of service/case management inventory in a sample of serious young offenders. Law and Human Behavior, 36(4), 331–344. https://doi.org/10.1037/h0093927

25.

Olver

M. E.

Stockdale

K. C.

Wormith

J. S.

(2009). Risk assessment with young offenders: A meta-analysis of three assessment measures. Criminal Justice and Behavior, 36(4), 329–353. https://doi.org/10.1177/0093854809331457

26.

Olver

M. E.

Stockdale

K. C.

Wormith

J. S.

(2014). Thirty years of research on the Level of Service Scales: A meta-analytic examination of predictive accuracy and sources of variability. Psychological Assessment, 26(1), 156–176. https://doi.org/10.1037/a0035080

27.

Orton

L. C.

Hogan

N. R.

Wormith

J. S.

(2021). An examination of the professional override of the Level of Service Inventory—Ontario Revision. Criminal Justice and Behavior, 48(4), 421–441. https://doi.org/10.1177/0093854820942270

28.

Parent

Bilodeau

M. P.

Laurier

Guay

J. P.

(2022). Clinical overrides with the YLS/CMI: Predictive validity and associated factors. Criminal Justice and Behavior, 50(1), 101–117. https://doi.org/:10.1177/00938548221131958

29.

Rice

M. E.

Harris

G. T.

(1995). Violent recidivism: Assessing predictive validity. Journal of Consulting and Clinical Psychology, 63(5), 737–748. https://doi.org/10.1037/0022006X.63.5.737

30.

Rice

M. E.

Harris

G. T.

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

31.

Schaefer

Williamson

(2018). Probation and parole officers’ compliance with case management tools: Professional discretion and override. International Journal of Offender Therapy and Comparative Criminology, 62(14), 4565–4584. https://doi.org/10.1177/0306624X18764851

32.

Schmidt

Hoge

R. D.

Gomes

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

33.

Schmidt

Sinclair

S. M.

Thomasdottir

(2016). Predictive validity of the Youth Level of Service/Case Management Inventory with youth who have committed sexual and non-sexual offenses: The utility of professional override. Criminal Justice and Behavior, 43(3), 413–430. https://doi.org/10.1177/0093854815603389

34.

Song

(2015). Decision tree methods: Applications for classification and prediction. Shanghai Archives of Psychiatry, 27, 130–135. https://doi.org/10.11919/j.issn.1002-0829.215044

35.

Storey

J. E.

Watt

K. A.

Jackson

K. J.

Hart

S. D.

(2012). Utilization and implications of the STATIC-99 in practice. Sexual Abuse: A Journal of Research and Treatment, 24(3), 289–302. https://doi.org/10.1177/1079063211423943

36.

Thompson

A. P.

Putnins

A. L.

(2003). Risk-need assessment inventories for juvenile offenders. Psychiatry, Psychology, and Law, 10(2), 324–333. https://doi.org/10.1375/132187103322742158

37.

Vaswani

Merone

(2014). Are there risks with risk assessment? A study of the predictive accuracy of the Youth Level of Service–Case Management Inventory with young offenders in Scotland. British Journal of Social Work, 44, 2163–2181. https://doi.org/10.1093/bjsw/bct059

38.

Vieira

T. A.

Skilling

T. A.

Peterson-Badali

(2009). Matching court-ordered services with treatment needs: Predicting treatment success with young offenders. Criminal Justice and Behavior, 36(4), 385–401. https://doi.org/10.1177/0093854808331249

39.

Viljoen

J. L.

Vargen

L. M.

Cochrane

D. M.

Jonnson

M. R.

Goosens

Monjazeb

(2021). Do structured risk assessment predict violent, any, and sexual offending better than unstructured judgment? An umbrella review. Psychology, Public Policy, and Law, 27(1), 79–97. https://doi.org/10.1036/law0000299

40.

Vincent

G. M.

Guy

L. S.

Grisso

(2012). Risk assessment in juvenile justice: A guidebook for implementation. John D. and Catherine T. MacArthur Foundation. https://njjn.org/uploads/digital-library/Risk_Assessment_in_Juvenile_Justice_A_Guidebook_for_Implementation.pdf

41.

Wormith

J. S.

Hogg

S. M.

Guzzo

(2012). The predictive validity of a general risk/needs assessment inventory on sexual offender recidivism and an exploration of the override. Criminal Justice and Behavior, 39(12), 1511–1538. https://doi.org/10.1177/0093854812455741

42.

Wormith

J. S.

Hogg

S. M.

Guzzo

(2015). The predictive validity of the LS/CMI with aboriginal offenders in Canada. Criminal Justice and Behavior, 42(5), 481–508. https://doi.org/10.1177/0093854814552843

Clinical Override Use With the Youth Level of Service/Case Management Inventory: Predictive Validity and Factors Associated With Case Management Use

Abstract

Keywords

Method

Participants

Measures

Youth Level of Service/Case Management Inventory (YLS/CMI)

Recidivism Data

Index Offense Classification

Analytical Strategy

Results

Sample Description

Patterns of Clinical Override Use

YLS/CMI Predictive Validity

Factors Associated With Clinical Override Use

Discussion

Clinical Override Use and YLS/CMI Predictive Validity

Factors Associated With Clinical Override Use

Study Limitations

Conclusion and Clinical Applications

Footnotes

AUTHORS’ NOTE:

ORCID iD

References