Risk-Needs Assessment in Juvenile Justice

• Published date: 01 January 2011
• DOI: 10.1177/0093854810386000
• Author: Nathan E. Cook,John Chapman,Gina M. Vincent
• Date: 01 January 2011
• Subject Matter: Articles

42

CRIMINAL JUSTICE AND BEHAVIOR, Vol. 38 No. 1, January 2011 42-62

DOI: 10.1177/0093854810386000

© 2011 International Association for Correctional and Forensic Psychology

AUTHORS’ NOTE: We wish to thank Randy Borum, PsyD, for his comments on an earlier version of this

article. Please direct correspondence and requests for reprints to Gina M. Vincent, PhD, Center for Mental

Health Services Research, Department of Psychiatry, University of Massachusetts Medical School, 55 Lake

Avenue North, Worcester, MA 01604; e-mail: Gina.Vincent@umassmed.edu.

RISK-NEEDS ASSESSMENT IN

JUVENILE JUSTICE

Predictive Validity of the SAVRY,

Racial Differences, and the

Contribution of Needs Factors

GINA M. VINCENT

University of Massachusetts Medical School

JOHN CHAPMAN

State of Connecticut Judicial Branch, Court Support Services Division

NATHAN E. COOK

University of Massachusetts Medical School

The authors conducted a prospective study of the predictive validity of the Structured Assessment of Violence Risk in Youth

(SAVRY) using a 5-year follow-up period and a sample of 480 male adolescents assessed by juvenile detention personnel.

Analyses were conducted to examine differential validity by race-ethnicity, the relative contribution of structured profes-

sional judgments of risk level, and the incremental validity of dynamic to static risk factors. Overall, the SAVRY total scores

were significantly predictive of any type of reoffending with some variability across racial-ethnic groups. Youths rated as

moderate to high risk by evaluators using structured professional judgment had greater odds of rearrest, but these risk ratings

did not have incremental validity over numeric scores. Static factors were most strongly predictive of nonviolent rearrest, but

dynamic factors (social-contextual) were the most predictive of violent rearrest. Implications for use of risk-needs assessment

tools in juvenile justice programs and areas in need of further investigation are discussed.

Keywords: juvenile justice; SAVRY; youth risk assessment; dynamic risk factors; race

Juvenile justice officials are routinely expected to make decisions involving the handling

of youths who come into contact with the law at various points in the system. When

dealing with youths, who are in a developmental period during which the chance of reha-

bilitation is high, one can argue for “evidence-based” approaches as opposed to dictating

punishments based on the crime alone. Decisions about how to process youths through the

justice system should involve the youths’ threat to public safety, the likelihood of benefit-

ing from interventions, and which interventions are most likely to result in a positive

change (Fagan & Zimring, 2000; Mulvey, 2005). Consistent with best practice and research

on youth development, these decisions can be informed by a proper assessment.

Vincent et al. / PREDICTIVE VALIDITY OF THE SAVRY 43

Researchers and policy makers have suggested for the past couple decades that justice

systems implement structured tools in an effort to increase the consistency and accuracy of

decisions and risk assessments (e.g., Gottfredson & Tonry, 1988; Guarino-Ghezzi &

Byrne, 1989). Structured decision-making models should use standardized risk assessment

tools that incorporate findings from the social sciences about the factors that place youths

at highest risk for reoffending (Hoge, 2002). Arguably, the highest-risk youths should

receive the most intensive interventions. To ensure treatment is done right, offenders must

first be properly assessed to identify their risk level and treatment needs (Austin, 2006;

Grisso, Vincent, & Seagrave, 2005; Skowrya & Cocozza, 2007) to follow what Andrews

and Bonta (2010) refer to as the risk-need-responsivity principles. Such tools were designed

to assist juvenile justice personnel and the court in reaching appropriate dispositions, sanc-

tions, and interventions in a structured, unbiased manner.

On the surface, it appears that implementation of risk for reoffending assessment tools

in juvenile justice programs should increase the consistency and accuracy of decision mak-

ing. However, use of such tools could have deleterious effects on the system if used

improperly or if the tools themselves are invalid (see Grisso, 2005). Juvenile justice sys-

tems and agencies should implement tools that have sufficient evidence, meaning that the

tool has, among other characteristics, evidence of rater agreement and predictive validity

from independent parties (Austin, 2006; Vincent, Terry, & Maney, 2009). Although many

tools designed for use with juveniles are beginning to attain sufficient evidence, others

have yet to be tested in the field. In other words, much research has been conducted in

“contrived” situations where the tool was scored by research assistants on the basis of file

information and then analyzed to determine whether the tool could predict recidivism.

Although that is the first step in analyses, the next step is to establish the generalizability

of predictive accuracy estimates to real-life situations. One cannot assume these estimates

will generalize when they are scored by juvenile justice personnel with multiple competing

demands on their time and added political and bureaucratic pressures. As such, this study

reports the predictive validity of the Structured Assessment of Violence Risk for Youth

(SAVRY; Borum, Bartel, & Forth, 2006) following its implementation in a juvenile deten-

tion facility. Additional analyses were conducted to examine racial-ethnic differences.

Current debates about the optimal structure of risk assessment tools are described below.

FACTORS RELEVANT TO YOUTH RISK ASSESSMENT

Two general types of risk factors enter into these assessments, static risk factors and

dynamic risk factors. Static risk factors are events and characteristics that are significantly

associated with negative outcomes and that are not subject to change. These are generally

historical variables, although other types of variables may also function as static risk fac-

tors. Age at first offense is an example of a factor statistically associated with reoffending that

is not subject to modification. These generally do not serve as causal variables, but they can

serve as markers associated with the probability of future offending or antisocial behavior.

Dynamic risk factors, on the other hand, include individual and circumstantial character-

istics significantly related to future antisocial behavior that are subject to change or reme-

diation. Dynamic risk factors are seen as having a likely causal link to the antisocial behavior

that, if changed, could reduce the likelihood of the future offending and antisocial behavior