Discussion and Recommendations

The key findings from this study so far are:

Sample/offence related variables

  • The ethnicity of the 69 participants in the study was evenly split between European and Maori although this differed between the three YOU’s in the study with the Waikeria YOU having the largest percentage of Maori and the Christchurch unit, the largest percentage of European inmates;
  • The average age of participants at first conviction was 15.62 years of age with the mean current age when assessed in this study 17.5. The Waikeria unit was found to have a lower current mean age than the other two units;
  • Participants had a mean sentence length of 2 years 8 months with the most frequent index offences for Aggravated Robbery followed by Dishonesty. However, when all violent and sexual crimes were added, 70% had index offences for serious violence/sexual offending. No significant differences were found between the units for index offences;
  • They had a mean of 17 total convictions (both past and current), with a mean number of four different offences types (criminal versatility). Mean total violent convictions were 2.20, and sexual 0.43. Those in the Hawkes Bay and Christchurch units were found to have significantly higher means for total convictions and criminal versatility, variables related to higher risk of recidivism than the Waikeria unit.

Risk measures

  • All four measures were found to have a normal score distribution, although with a positive skew reflecting the high mean risk scores;
    • RoC*RoI (M = 0.58, SD = 0.17)
    • PCL: YV (M = 25.09, SD = 7.29)
    • RSYO (M = 38.48, SD = 12.97)
    • YLS/CMI (M = 23.49, SD 6.38)
  • All risk measures identified a proportion of the sample as high risk with both the RoC*RoI and PCL: YV identifying approximately 30%, the RSYO 10%, and the YSL/CMI 62% in this risk category;
  • The PCL: YV was the only risk measure that differed in mean score between units. The Hawkes Bay and Christchurch units were found to have significantly higher mean scores for this instrument than the Waikeria unit.

Distribution of dynamic risk factors

  • The YLS/CMI’s inclusion of dynamic risk factors identified the following percentage of the sample as having high rated dynamic needs:
    • Family 20%
    • Education/employment 78.5%
    • Peer relations 58%
    • Substance abuse 74%
    • Leisure/recreation 33%
    • Personality/behavior 42%
    • Attitudes/orientation 32%
  • Seventy-one percent of the sample had high scores for PCL: YV Factor 2: Affective deficits (i.e., lack of remorse, shallow affect, callous/lack of empathy, fails to accept responsibility);
  • Eighty percent of the sample had high scores for PCL: YV Factor 3: Behavioral deficits (i.e., Stimulation seeking, impulsivity, irresponsible, parasitic orientation, lacks goals).

Validity of risk measures

  • Moderate correlations were found for all risk measures and the offence related variables, total convictions and criminal versatility. The strongest for both being the RoC*RoI and PCL: YV scores;
  • Only the PCL: YV (r = .35) and RoC*RoI (r = .25) scores were found to have significant correlations with total violent convictions;
  • Moderate correlations were found between all measures, with the RSYO having a moderate/high relationship with all other measures. This was encouraging in view of the RoC*RoI, PCL: YV, and YLS/CMI all being proven predictors of recidivism and violence.

Descriptive information YOU group

While not all current YOU inmates were included in the sample, the notable exclusion due to funding constraints being the Wellington YOU, the sample is believed to be representative. Few of the study variables revealed differences between the three units included in the study. One variable that did differ was ethnicity. The ethnicity of the 69 participants in the study was evenly split between European and Maori although this differed significantly with the Waikeria YOU having the largest percentage of Maori and the Christchurch unit the largest percentage of European inmates.

There was no difference among the sample between the three units for age of first conviction. The average age of participant’s first conviction was 15.62 years of age. However, this only reflects their first court appearance not first contact with the Police. While details on early antisocial behaviour were not included in this report, information on antisocial behaviour before 10 years of age was part of both the RSYO (scale 2) and the PCL: YV (item 12). It is noted that the RSYO scale 2 scores had a mean that was the same as the scale cut-off detailed in the description of the instrument in Appendix A. The cut-off indicating early onset offending and potential membership of the ‘chronic’ offender developmental pathway. One question in the structured interview related to numbers of previous Family Group Conferences attended. It is estimated that 90% of the sample had attended at least one conference, with one offender reporting attending fourteen.

The mean current age when assessed in this study was 17.5 with the vast majority of the youth offenders aged 16-17. The Waikeria unit was found to have a significantly lower current mean age than the other two units. There was no clear reason for this lower current age for the Waikeria unit, however it is noted that this unit also had a number of offenders convicted of murder/manslaughter usually at an early age.

Participants had a mean sentence length of 2 years 8 months with the most frequent index offences for Aggravated Robbery followed by Dishonesty. However, when all violent and sexual crimes were added, 70% had index offences for serious violence/sexual offending. No significant differences were found between the units for index offences, although the Christchurch unit had the most youth offenders imprisoned for sexual offences. All had approximately 30% with an index offence for Aggravated Robbery.

In keeping with the serious nature of the offenders in the study they had a mean of 17 total convictions (both past and current), with a mean number of four different offences types (criminal versatility). Mean total violent convictions were 2.20, and sexual 0.43. Those in the Hawkes Bay and Christchurch units were found to have significantly higher means for total convictions and criminal versatility, variables related to higher risk of recidivism than the Waikeria unit.

In summary, the sample was found to be characterised by serious violent offending, with large numbers of offences, usually in a number of offence categories. There were no ethnic or age differences in relation to their offending. The implications of their offence profile is that they require specialist violence prevention programmes and for at least 14.5% of the sample, specialist sex offender treatment. The current group programme, EQUIP is not designed for intervention with such offenders. The other concern in relation to treatment is the relative short sentences of imprisonment (M =2 years 8 months) especially with most eligible for parole after a third of their sentence. This means that any intervention to address their offending needs to be flexible and targeted at the highest priority criminogenic need. While it is preferable for as many dynamic risk variables as possible to be addressed, treatment needs such as inappropriate beliefs about violence should be prioritised before release, rather than variables that may not be directly related to future violent behaviour, such as substance abuse, which may not be a necessary part of violent behaviour.

While group treatment is the normal intervention to ensure the greatest number receives treatment, a number of variables relating to responsivity such as the short periods of imprisonment, high risk, and a range of index offence (including 14.5% for sexual offences) mean individual specialist psychological treatment services are required. It was noted during this study in which all institutional files were reviewed that only 7 of the 69 had Psychological Service assessment reports and few, less than 5 were currently receiving treatment from a Departmental psychologist for their offending.

Finally, in relation to addressing offending while some of the dynamic needs could be addressed through targeted psychological interventions others will require inclusion in a comprehensive reintegration plan. For example employment or association with antisocial peers. Further discussion of reintegrative needs will be included in a second study on this issue currently being completed by the principal researcher in this study and Dr Llew Richards-Ward.

Risk measures

This study provided support for the effectiveness of all four measures. All four measures were found to have a normal score distribution, although with a positive skew reflecting the high mean risk scores. The RoC*RoI mean was above the intervention cut-off with approximately 30% .70 or greater, the cut-off used by the NZ Parole Board and Psychological Service to identify high risk offenders. It is noted that approximately 28% of all male inmates in prison currently have RoC*RoI scores of .70 or greater. While RoC*RoI scores for youth offenders have been treated with caution in the past it should be noted that the original development sample for the instrument contained large number of youth offenders. In addition, this study has shown a normal distribution for RoC*RoI, with moderate/high correlations with offence related variables and the other risk measures. All providing support for the use of this measure in combination with other measures in decision making in regard to youth offenders.

The PCL: YV also performed well with a mean score (25.09) that was similar to male adolescents imprisoned for violence in the validation studies from the manual (Forth, Kosson, & Hare, 2003). A recent study into the predictive validity of the PCL: YV by Catchpole and Gretton (2003) for violent young offenders (N = 74) with 54% of European descent, 29.7% of Aboriginal, 8% Asian, and 5.45 other, had a mean of 23.8 (SD = 6.9). Of this sample 26.9% were classified as high on psychopathy (high risk) based on scores of 30-40. The current study had 27% in this score range. Catchpole and Gretton found that the PCL: YV high risk group (scores 30-40) had a 40% violent recidivism rate within 12 months of release. Odds ratio analysis identifying this high risk group as four times more likely to commit a violent offence than others in the sample. Using Receiver Operating Characteristic (ROC) analysis, the PCL: YV was found to have an Area Under the Curve (AUC) of .78 for general offending and .73 for violence. This indicates a moderate level of predictive ability. While no predictive validity analysis was able to be under taken in the current study due to the prospective design, the PCL: YV score had the highest significant correlation with previous total violent convictions (r = .35). A similar correlation was found in a large validation study for the instrument by Forth (2002) (cited in Forth et al., 2003).

In the current study the PCL: YV was also the only risk instrument that picked up on apparent differences in risk across the YOU’s. Finding a significantly lower mean score for offenders from the Waikeria Unit compared to the Hawkes Bay unit. It is noted that two other variables associated with higher risk of recidivism, namely, total convictions and criminal versatility also found the same between unit differences. While the PCL: YV performed well in the current study and has a growing body of overseas research on its predictive validity it is a specialist instrument that takes a day to administer and score and can only be administered by trained Psychological Service Psychologists. It is recommended as a valuable support for cases in which the other risk measures in this study have indicated high risk to assist in Parole Board assessments, it will also be a valuable tool in assessing potential Section 107 applications (Parole Act 2002).

The RSYO also had a normal score distribution. This instrument, which is fully described in Appendix A, relies on a developmental pathways model to identify chronic offenders from adolescent limited offenders. The instrument is easily administered, however, considerable difficulties were experienced in applying the reliability scale (scale 1) with significant others. Many could not be located and for those interviewed a high degree of impression management was apparent. Reliability was able to be effectively established through collateral sources. It is recommended that this approach receive further validation to assist the application of this instrument in YOU’S.

The RSYO also had high/moderate correlations with the other instruments, especially the PCL: YV and YLS/CMI. This convergent validity provides further support for this instruments ability to identify recidivism risk in youth offenders. However, the RSYO still requires predictive validity to be established for a serious youth offender sample. It is hoped that the recommended follow up of the study sample at a later date will provide this essential validity information. It was of concern in this study that the RSYO identified the smallest percentage of the sample (10%) as high risk and did not correlate with previous violent convictions, with the lowest correlations in relation to the variables total convictions, and criminal versatility of all risk measures. While it is recommended that the RSYO is used in combination with other risk measures (such as RoC*RoI and YLS/CMI) it is not established in this study that it is a reliable measure for predicting risk for imprisoned youth offenders.

The YLS/CMI had a normal distribution and a mean that reflected a positively skewed distribution (M = 23.49, SD 6.38). However, the score distribution tended to be narrow when the total score was divided into four risk categories based on manual based recommended categories (Low = 0-8; Moderate = 9-22; High = 23-34; Very High = 35 -42). Only one Low risk and one Very High risk score was found for participants, with 62% of the sample classified as high risk in comparison to the PCL: YV and RoC*RoI assigning only 30% to a high risk category. Other research using this instrument with a high risk youth offender population has found a similar narrow distribution (Catchpole & Gretton, 2003). The YLS/CMI did correlate significantly with total convictions and criminal versatility but not with total violent and sexual convictions. It did have high/moderate correlations with the other risk measures, and in addition to this convergent validity provided information on established dynamic risk variables that could be used to treat or manage offenders. The correlation between the YLS/CMI and the PCL: YV total scores was r = .62. A study looking at risk in a larger sample of 179 male adolescent offenders using the

PCL: YV and YLS/CMI found a similar correlation between scores for the instruments (r = .61) (Forth et al., 2003). Catchpole and Gretton found in their study comparing the PCL: YV and YLS/CMI that the YLS had an AUC = .74 for general recidivism over 12 months and .73 for violent reoffending in the same time period. This indicates a moderate level of predictive ability. Therefore, it is recommended that the YLS/CMI is used along with RoC*RoI and the RSYO in assessing risk and dynamic risk needs for youth offenders. The instrument is simple to use, requires only a 30-40 minute interview, and is designed for use by a variety of corrections employees.

Distribution of dynamic risk factors

The YLS/CMI’s inclusion of dynamic risk factors identified a range of criminogenic needs in the sample. No difference between YOU’s was found for these needs. When the YLS/CMI manual guidelines for classifying degree of need were used to establish high need, Education/Employment was the variable with the highest percentage (78.5%), followed by Substance abuse (74%), and Peer relations (58%). The lowest percentage of high need was in the area of Family relations. It was of interest that the two scales related to difficulties in engagement in treatment, the scale associated with antisocial personality (Personality/behavior 42%) and hostile/antisocial beliefs (Attitudes/orientation 32%) accounted for approximately a third of the sample.

Further support for the needs and responsivity barriers identified in the YLS/CMI assessment came from the PCL: YV assessments. Seventy-one percent of the sample had high scores for PCL: YV Factor 2: Affective deficits (i.e., lack of remorse, shallow affect, callous/lack of empathy, fails to accept responsibility). In addition, 88 % of the sample had high scores for PCL: YV Factor 3: Behavioral deficits (i.e., Stimulation seeking, impulsivity, irresponsible, parasitic orientation, lacks goals). These beliefs and behaviours provide further support for the recommendation that specialist psychological service assessment and treatment is required for a significant proportion of youth offenders in the YOU’s.

Future research suggestions

The study sample has been limited by the funding available to gather data and the short time frame (four months), preventing any follow up after release to establish predictive validity for the measures. It is recommended that a research proposal is considered for the next REASC funding round to extend the sample size, both by including the Wellington YOU and a number of youth offenders on community based sentences. This will ensure a sample of sufficient size to provide robust analysis, as well as widening the risk profile to establish the predictive ability and reliability of the measures for a wider range of youth offenders. With the protocol established and the principal researcher well versed in these assessments costs in gathering another 80 assessments (for a total sample size of 150) are not expected to be more than $40,000, including the predictive validity follow-up analysis in approximately 12-16 months time.

Conclusions/Recommendations

  • The study has established a risk-needs profile for a representative sample of imprisoned youth offenders from YOU’s. The four risk measures, RoC*RoI, PCL: YV, RSYO, and YLS/CMI were found to distinguish a range of risk and to have significant correlations with offence variables and each other. Besides identifying that the majority of offenders were imprisoned for serious violent/sexual offences a number of criminogenic needs were identified in the sample.
  • While youth offenders were imprisoned for serious or repeated crimes their sentence lengths in the main were relatively short, especially with most eligible for parole after a third of their sentence.
    • Therefore, any intervention to address their offending needs to be flexible and targeted at the highest priority criminogenic need.
    • While group treatment is the norm to ensure the greatest number receive treatment, short periods of imprisonment, high risk, and a range of index offence (including 14.5% for sexual offences) mean individual specialist psychological treatment services are required.
    • While some of the dynamic needs can be addressed through targeted psychological interventions others will require inclusion in a comprehensive reintegration plan. For example employment or association with antisocial peers.
  • That in view of the high number of the sample with violent index offences a targeted and intensive youth violence prevention programme is developed for use in the YOU’s;
  • The four risk measures used in this study were all found to identify a range of risk in the sample, with three, the PCL: YV, RoC*RoI and YLS/CMI classifying at least 30-60% as at high risk of further serious recidivism. The use of multiple measures of risk is recommended to reduce decision error when treatment is being targeted at those in greatest need.
  • It is recommended based on the results that RoC*RoI, YLS/CMI, and the RSYO are used with all youth offenders to identify those at high risk. The YLS/CMI and RSYO can be scored from a short 30-40 minute interview by trained PPS staff.
  • If further evidence of high risk is required by parole authorities or to support a Section 107 application (Parole Act 2002), especially for risk of violence, then the PCL: YV should be administered by trained departmental psychologists.
  • Finally, convergent validity has been found for the four measures based on their relationship with offence related variables and the strong correlations between the established specialist youth risk instruments (PCL; YV and YLS/CMI) and the New Zealand instruments (RoC*RoI; RSYO). However, validity will only be confirmed when predictive validity is established for the risk measures and the study sample (i.e., how good were the risk measure scores at predicting reoffending outcomes);
  • It is recommended that the sample is followed up again in approximately a years time when it is estimated that at least 70-80% of will have been release for a minimum of six months to provide some preliminary data on predictive validity.