Evaluation of the Predictive Validity of a Risk-Need-Responsivity Assessment Tool (RNR-A) in the Swedish Prison and Probation Service

• Author: Maria Danielsson,Dan Andersson,Peter Johansson Bäckström,Louise C. Starfelt Sutton
• DOI: 10.1177/00328855221095561
• Published date: 01 June 2022
• Date: 01 June 2022
• Subject Matter: Articles

Evaluation of the

Predictive Validity of a

Risk-Need-Responsivity

Assessment Tool

(RNR-A) in the Swedish

Prison and Probation

Service

Peter Johansson Bäckström,

Maria Danielsson, Louise

C. Starfelt Sutton,

and Dan Andersson

Abstract

To appraise the “real-world”implementation of the risk principle, this study

examined the predictive validity of a Risk-Need-Responsivity assessment in

the Swedish Prison and Probation Service. Reconviction rates at 24 months

follow-up in a cohort of 2,442 offenders were used to assess calibration and

discrimination indices. Results indicated acceptable predictive accuracy

(AUC =.68–.74), with scope for improvement among young adult offenders.

The tool’s utility was supported foremost by its ability to screen out low-risk

offenders, while over-prediction of recidivism among medium- and high-risk

offenders calls for more comprehensive assessment to inform the effective

planning of rehabilitative service intensity.

Keywords

risk assessment, Risk-Need-Responsivity, predictive validity, Sweden, recidivism

The Swedish Prison and Probation Service, Norrkoping, Sweden

Corresponding Author:

Louise C. Starfelt Sutton, The Swedish Prison and Probation Service, Slottsgatan 78, 601 80,

Norrkoping, Sweden.

Email: louise.starfeltsutton@kriminalvarden.se

Article

The Prison Journal

2022, Vol. 102(3) 367–390

© 2022 SAGE Publications

Article reuse guidelines:

sagepub.com/journals-permissions

DOI: 10.1177/00328855221095561

journals.sagepub.com/home/tpj

Introduction

Since the introduction of assessment tools for predicting recidivism, they have

become an integrated part of prison and probation services as well as youth

care (e.g., Andrews et al., 1990, 2006). Historically, the assessment of recid-

ivism risk has relied solely on clinical judgment. The f‌irst generation of risk

assessments thus typically involved correctional staff (i.e., probation off‌icers

or prison staff) or clinical professionals (i.e., psychologists, psychiatrists, or

social workers) conducting unstructured assessments. One of the main advan-

tages of this type of assessment is f‌lexibility, which is benef‌icial in unique

cases. Given a lack of transparency, however, it is diff‌icult, if not impossible,

to describe the decision-making process to measure the agreement between

different assessments or assessors (Brown & Singh, 2014). Unstructured

risk assessments are therefore hard to replicate and have shown evidence of

personal biases (e.g., Bonta & Andrews, 2007; Van Voorhis & Brown,

1996) and limited predictive accuracy (Mills, 2017).

In the 1970’s, there was a growing recognition of the benef‌its of evidence-

based actuarial risk assessment tools in place of professional judgment. As a

consequence, a second generation of assessment tools was developed. These

tools provided quantitative methods for assessing future risk, using static risk

factors that were linked to recidivism. Such factors were identif‌ied through

research on criminal populations –mostly white males (Hannah-Moffat,

2013). The utility of these tools has been questioned due to a lack of f‌lexibil-

ity, despite increased accuracy in prediction when compared to unstructured

assessments (e.g., Ægisdóttir et al., 2006). Further, the exclusive reliance on

static factors obscures the complexity and variability of individual function-

ing and undercuts any possibility to identify intervention targets (Andrews

et al., 1990; Campbell et al., 2009).

1

With a growing interest in treatment and risk-reducing interventions and as

a reaction against a widespread treatment pessimism (“nothing works”;

Martinson, 1974), the research focus shifted in the late 1970’s and early

1980’s. This shift involved a change from a one-sided emphasis on static

risk factors to extensive empirical enquiry into risk factors that were

dynamic in nature to measure changes in the likelihood of violent or criminal

behavior. There is nothing inherently different with actuarial tools that justif‌iy

the exclusion of dynamic factors (Skeem & Monahan, 2011). Criminal history

has thus continued to be an important factor in third-generation risk assess-

ment tools, combined with dynamic factors that are sensitive to changes in

an offender’s circumstances (Bonta & Andrews, 2007).

Afourthgenerationofriskassessmentinstruments has also been described in

the literature. These instruments are designed as an integrated part of risk

368 The Prison Journal 102(3)