Predictive Validity of the SAVRY Differs Across Ethnic Subgroups

This research identified the presence and severity of salient risk factors for violence and assessed the predictive validity of the Structured Assessment of Violent Risk in Youth (SAVRY) for an Australian young male offender cohort held in detention. As the bulk of previous research has focused on European and North American Caucasian youth, comparisons were made between participants from Australian ethnic subgroups: English-Speaking Background (ESB), Culturally and Linguistically Diverse (CALD), and Indigenous and Torres Strait Islanders (IND). The study found the instrument to moderately predict general and violent recidivism across the larger cohort and the SAVRY Risk Rating was able to differentiate between times to reoffense. However, the predictive validity differed significantly across ethnic subgroups with moderate to strong predictive accuracy for the ESB group, poor predictive accuracy for the CALD group, and only particular SAVRY scores attained significant accuracy for the IND group. Findings on subgroup risk factors were considered in light of contemporary understandings of the unique experiences and trajectories of minority youth. Future investigation is necessary to differentiate and characterize the risk factors and offending patterns of the ethnicities within the CALD classification.

youth violence, risk assessment, recidivism, ethnicity, juvenile offending

This research examined the presence and severity of salient risk factors as well as the predictive validity of the Structured Assessment of Violence Risk in Youth (SAVRY) for a group of 175 detained Australian young male offenders. Offenders were divided into 3 cultural categories: those who self-identified as English Speaking Background (ESB; n = 84), Culturally and Linguistically Diverse (CALD; n = 59), or Indigenous groups and Torres Strait Islanders (IND; n = 32). SAVRY total scores, domain scores, and item scores were examined for the total group as well as for the cultural subgroups.  In addition, follow-up data were collected for 139 of these offenders for at least 6 months after release from detention.

“Results showed that both the Indigenous and ESB groups had significantly higher Total and Historical Domain scores compared with the CALD group. Overall, the Indigenous group had the highest mean scores for the SAVRY total score, Historical Domain, and Socio/Contextual Domain, whereas the ESB group had the highest mean Individual/Clinical Domain score of the three groups. There were no significant differences for protective factor scores across groups” (p. 34).

“During the follow-up period, 104 (74.8%) of the total sample was charged by police for any new offense [General Recidivism (GR)] and 82 (59.0%) for a new violent offense [Violent Recidivism (VR)]. Indigenous participants reoffended (GR: 86.2%, VR: 69.0%) at higher rates than ESB (GR: 72.5%, VR: 58.0%) and CALD (GR: 70.7%, VR: 53.7%) participants. The leading re-offense categories for the total sample included: Theft offenses 43.6%, Assault 27.9%, Property Damage 11.5%, and Drug Offenses 6.7%. There were no significant differences in the proportion of general or violent reoffending across the ethnic groupings” (p. 35).

“For participants rated as high risk, 85.1% generally reoffended, with 70.3% reoffending violently. In contrast, of participants given a low risk rating, 47.6% reoffended generally and 38.1% reoffended violently” (p. 35).

With respect to the predictive validity of the SAVRY for the overall sample, results indicated that both the Total Score and the Summary Risk Rating significantly predicted future general recidivism and violent recidivism at a moderate level.

With respect to the predictive validity of the SAVRY for each of the three cultural subgroups, results indicates that the Total Score, Summary Risk Rating, and Domain scores were predictive of both general and violent recidivism in the ESB group whereas none of these scores were predictive of general or violent recidivism in the CALD group. Results for the IND group were mixed, with scores on the Socio/Contextual domain, Individual domain, and Protective factor domain showing some predictive validity for general recidivism in this group and the Total Score, Summary Risk rating, and Historical domain scores showing some predictive validity for violent recidivism in this group.

“The study provided evidence for the utility of the SAVRY in predicting general and violent recidivism. The overall Total Score was able to moderately forecast both forms of re-offense reflecting previous validation literature. The SAVRY Summary Risk Rating was found to be associated with both forms of recidivism and time at risk. Participants who received a low Risk Rating were less likely to reoffend and exhibited the highest mean survival time. Conversely High-Risk offenders had higher re-offense rates and their times at risk were significantly lower than Low-Risk offenders for both general and violent recidivism” (p. 39).

When the predictive validity of the SAVRY was examined across different ethnic subgroups, SAVRY scores were unable to predict any category of recidivism for the culturally and ethnically diverse group. The authors speculate that these results may be explained by the extreme heterogeneity of the grouping, which comprised participants from a number of diverse ethnicities. Caution is advised when using the SAVRY to predict violent recidivism in culturally and linguistically diverse groups until the characteristics of the group membership are further explored.

These results serve to underscore the general point that risk assessment instruments may not demonstrate equivalent predictive utility for general or violent recidivism across different groups. Clinicians who use these instruments to evaluate individuals should be mindful of the degree to which the personal characteristics of the evaluee reflect those of the validation samples for the risk assessment instrument.

The results section also presents data on the SAVRY item scores and SAVRY domain scores for each of the three ethnic groups, which may provide useful information for determining appropriate risk management strategies and interventions.

As always, please join the discussion below if you have thoughts or comments to add!