A New Perspective on SAVRY in Youth Violence Assessments

Psychological Assessment | 2023, Vol. 35, No. 10, 856–867.

A Proof of Concept Study of Promotive, Mixed, and Risk Effects Using the SAVRY Assessment Tool Items With Youth With Sexual Offenses

Calvin M. Langton- Department of Psychology, University of Windsor

Jennifer A. Ranjit- Department of Psychology, University of Windsor

James R. Worling - Toronto, Ontario, Canada

There is currently a lack of consensus about the nature of strengths in forensic assessments. With 273 justice-involved male youth and a fixed 3-year follow-up, this study adopted the approach of Farrington and colleagues to investigating the nature of associations between trichotomized variables, representing risks and strengths, and outcomes using pairs of odds ratios (ORs) and percentage point changes from base rates. Items from the Structured Assessment of Violence Risk in Youth (SAVRY), a structured professional judgment tool used to assess risk and protective factors in justice-involved youth, were employed for this purpose. In the literature, the accuracy of SAVRY summed totals for its Risk Factor item sets (each item rated using a trichotomy) has been generally in the moderate range in predicting future violence. But the total for its summed Protective Factor items (each rated using a dichotomy) has been less consistently encouraging. In this study, contrary to their labels, the majority of SAVRY Risk and Protective Factors (rated using trichotomies) exerted a risk effect at one end of their trichotomy (risk item ratings of 2, protective item ratings of 0) and a promotive effect at the other end (risk item ratings of 0, protective factor ratings of 2) for a new violent (including sexual) offense and any new offense. Subsets of items conservatively weighted using ORs (capturing risk and strength) were statistically significantly more accurate in predicting outcomes than their originally rated counterpart subsets. Implications for understanding the nature of strengths and for applied assessment practices are discussed.

juvenile offenders, risk assessment, protective factors, desistance, Structured Assessment of Violence Risk in Youth (SAVRY)

"... [In] this proof of concept study, we focused on investigating three types of direct effects: risk, promotive, and mixed effects, per Farrington et al.'s (2016) corresponding "factors."... analyses with the individual items comprising the Structured Assessment of Violence Risk in Youth (SAVRY), a structured professional judgment (SPJ) tool widely used in applied youth justice settings... the SAVRY was of particular interest because it includes both risk and protective factors in distinct sections, the items for which can be rated trichotomously and so investigated using Farrington et al.'s approach" (p. 857).

"SPJ tools are manualized sets of assessment guidelines intended to structure the assessor's consideration of a set of relevant constructs, operationalized as items, for which ratings are given anchored by explicit criteria ...The items labeled risk factors or protective factors can be summed to create totals as well as used to inform a summary judgment rating (also typically using a three-point ordinal scale of low, moderate, or high) to indicate overall risk for the adverse outcome or, in the case of protective factors, the degree to which risk is thought to be reduced by their presence or the extent to which desistance (as an outcome rather than a process) is deemed likely….because an SPJ tool's individual items represent clearly operationalized constructs, rated using a trichotomy, they also invite investigation of the nature of their associations with outcomes using the approach of Farrington and his colleagues, which would have implications for informing evidence-based applied assessment practices. Indeed, the findings from a study by Li et al. (2019) suggest that understanding of the nature of purported risks and strengths in applied assessment practices may be enhanced with this approach" (p. 858).

"...The SAVRY includes distinct sets of items operationalized as risk or protective factors, which lend themselves to a systematic investigation of the types of effects items in each set can be demonstrated to exert in relation to recidivism when examined using Farrington et al.'s (2016) approach….SAVRY indices have been shown to predict multiple types of recidivism outcomes with various populations of justice-involved adolescent males, although the evidence is more equivocal both with justice-involved adolescent females and for the summed total of the Protective Factor items…

…The sample [consisted] of male adolescents from a major urban area in Southern Ontario referred for specialized assessment services for youth who had sexually abused others and who were at risk to reoffend starting from 2003 up to 2014 (n = 323). Of these 323 adolescents, only those for whom a fixed follow-up of 3 years was available were used, resulting in an n ≤ 273… Of these 273, 8% had one or more prior convictions for a violent (nonsexual) offense. Two percent had two or more prior convictions for a sexual offense, 14% had one prior sexual offense conviction, and 84% had no prior conviction for a sexual offense… for 131 of these 273 participants; using Statistics Canada categories, 47% of these were of Europeanorigin, 16% were of Caribbean origin, 13% were of African origin (Central, North, South, East, West), 10% were of Asian origin (South, East and South East, West Central, and Middle Eastern), 8% were of Latin, Central and South American origin, and 6% were of First Nations, Indigenous, Inuit, or Métis origin…" (p. 859).

"Support for the hypotheses was found for this outcome…The AUC for the OR-recoded sum of 25 SAVRY items was statistically significantly higher than those for the same subset of 25 SAVRY items (i.e., the same 21 Risk Factor items summed minus the same four Protective Factor items summed, using ratings of 0, 1, or2), and the full set of 30 SAVRY items (i.e., 24 Risk Factor items summed minus the six Protective Factor items summed, using ratings of 0, 1, or 2), with ds ≥ 0.38… The only exception was the comparison with the sum of the Individual/Clinical Risk Factor items: z = 1.63, p = .052 (1-tailed), d = 0.20. The AUC for the OR-recoded sum of 25 SAVRY items was statistically significantly higher than those for the same subset of 25 SAVRY items (i.e., the same 21 Risk Factor items summed minus the same four Protective Factor items summed, using ratings of 0, 1, or 2), and the full set of 30 SAVRY items (i.e., 24 Risk Factor items summed minus the six Protective Factor items summed, using ratings of 0, 1, or 2), with ds ≥ 0.38.

…Analyses revealed a more complex picture of the types of effects exerted by items constituting the SPJ tool, the SAVRY, than might be assumed on the basis of its general category labels of Risk Factors and Protective Factors… As well, using the pairs of ORs for each item calculated to select items and inform conservative recoding to reflect risk and strengths effects, when present, it was possible to generate new item sets of both the SAVRY's Risk Factors and Protective Factors. The sums for these were statistically significantly more accurate in predicting both outcomes than a simple summing of various item sets using their 0, 1, or 2 ratings…Of course, these findings do not represent a sufficient evidence base for departure from the developers' instructions for the SAVRY in applied settings" (p. 862).

"...Replication work with the SAVRY's unipolar trichotomous Risk Factors and investigations of potential risk, promotive, and mixed effects using Farrington et al.'s (2016) approach with bipolar operationalizations of constructs, with a 0-point representing the absence of risk and strength, would be illuminating. If an effect at only one pole was demonstrated (whether a promotive or risk effect) for such an operationalization, it would mean the view of a Protective Factor as only the obverse of Risk Factor, per Harris and Rice (2015), (and vice versa) would have to be expanded. Ultimately, revised versions of tools such as the SAVRY may be shown to be more accurate in predicting outcomes (both adverse and adaptive/positive) and their utility for intervention planning enhanced if their items are operationalized as unipolar or bipolar based on empirical work demonstrating their range of effects.

…Research of this kind has the potential to elucidate the complex nature of distinct operationalizations of constructs in relation to specific outcomes that have been theoretically or empirically implicated in the persistence in or desistance from criminal behaviors" (p. 865- 866).

"As it is, in applied practice, a rating of 0 for any of the SAVRY's six Protective Factor items may result in that construct being identified as an area or domain in which the strength should be developed or introduced with an intervention in order to facilitate desistance from future involvement in crime. With the possible exception of (the operationalization of) Strong Social Support (#2), this would be supported by the findings of the present study. But for three of the six Protective Factor items, 0 was associated with an increased likelihood of future involvement in crime, on the basis of the risk ORs. This has implications for intervention planning because there may be important practical and theoretical differences between targeting a construct as a strength that should be introduced or fostered to increase the likelihood of desistance from crime and targeting a construct as a risk to be managed or removed to decrease the likelihood of recidivism" (p. 865).

As always, please feel free to check out the resources below and comment to add more to the discussion!

• Gender Differences Apparent in the Use and Interpretation of the SAVRY for Juvenile Offenders
• YLS/CMI and SAVRY helpful for Juvenile Probation Officers’ case planning decisions
• In the Grey Areas: Violence Across the Spectrum of Adolescence