An Empirical Test of the Factor Structure of the Violence Risk Scale and Its Measurement Invariance Across Time and Cultural Groups

• Author: Yilma Woldgabreal,Andrew Day,Michael Daffern,Caleb Lloyd,Joe Graffam
• DOI: http://doi.org/10.1177/00938548221084984
• Published date: 01 September 2022
• Date: 01 September 2022
• Subject Matter: Articles

CRIMINAL JUSTICE AND BEHAVIOR, 2022, Vol. 49, No. 9, September 2022, 1255 –1275.

DOI: https://doi.org/10.1177/00938548221084984

Article reuse guidelines: sagepub.com/journals-permissions

© 2022 International Association for Correctional and Forensic Psychology

1255

AN EMPIRICAL TEST OF THE FACTOR

STRUCTURE OF THE VIOLENCE RISK SCALE

AND ITS MEASUREMENT INVARIANCE ACROSS

TIME AND CULTURAL GROUPS

YILMA WOLDGABREAL

Deakin University

Central Queensland University

ANDREW DAY

University of Melbourne

MICHAEL DAFFERN

CALEB LLOYD

Swinburne University of Technology

JOE GRAFFAM

Deakin University

Violence risk assessment instruments are used to inform key decisions about treatment planning and delivery, release on

parole, and intensity of supervision in the community. Yet, limited published information is available about psychometric

properties other than predictive validity. The purpose of this study was to examine the factor structure and measurement

invariance across pretreatment to posttreatment and cultural groups of one of the most widely used violence risk assessment

instruments, the Violence Risk Scale (VRS). Data from 366 completed assessments at preintervention and postintervention

phases for adults serving custodial sentences for violent offenses in an Australian jurisdiction were subject to confirmatory

factor analysis. The results indicated four intercorrelated but conceptually independent dimensions. Furthermore, measure-

ment invariance was established across pretreatement and posttreatment occasions and different cultural groups. However,

latent means comparison revealed significant difference across cultural groups, raising questions about sensitivity and gen-

eralizability of the VRS when used with diverse cultural groups.

Keywords: violence risk assessment; dynamic risk; risk factor; measurement; treatment

INTRODUCTION

Violence, whether it is physical or psychological, will often have serious consequences.

It can not only result in traumatic injuries, even death, and high levels of psychological

harm but also lead to the breakdown of significant relationships, problems with

AUTHORS’ NOTE: The authors are grateful to Gene Mercer for the initial support in facilitating access to

the archival data. The authors further extend their special thanks and appreciation to Rachel Howe for her help

with data extraction from a departmental archive. Correspondence concerning this article should be addressed

to Yilma Woldgabreal, P.O. Box 2278, Port Augusta, South Australia 5700, Australia; e-mail: y.woldgabreal@

cqu.edu.au.

1084984CJBXXX10.1177/00938548221084984Criminal Justice and BehaviorWoldgabreal et al. / FACTOR STRUCTURE OF THE VIOLENCE RISK SCALE

research-article2022

1256 CRIMINAL JUSTICE AND BEHAVIOR

occupational and educational functioning, and more broadly, weaken the social, economic,

and political fabric of communities and society at large (Fuller, 2015). As a consequence,

there is considerable interest in the development and delivery of violence prevention pro-

grams (VPPs) for those who have been convicted of violent offenses (Polaschek, 2019).

Violence risk assessment instruments play a crucial role in this respect; they are not only

used to inform decisions about treatment planning and delivery, but also inform judgments

about classification, suitability for release on parole, and the intensity of supervision in the

community (Olver & Wong, 2019). It is thus particularly important to ensure that instru-

ments that are used are psychometrically sound.

The primary aim of this study is to examine the factor structure and establish measure-

ment invariance of one of the most established instruments for use with adults in prison, the

Violence Risk Scale (VRS; Wong & Gordon, 2006). The VRS consists of 26 actuarial risk

indicators (we prefer this term to “dynamic risk factor” as a factor commonly connotes the

presence of multi-items, but each of the VRS item is scored singularly). The VRS indicators

are grouped under two risk domains: (a) six static (unchangeable) risk indicators (current

age, age at first violent conviction, number of juvenile convictions, violence through lifes-

pan, prior release failures or escapes, and stability of family upbringing) and (b) 20 dynamic

(changeable) risk indicators (violent lifestyles, criminal personality, criminal attitudes,

work ethic, criminal peers, interpersonal aggression, emotional control, violence during

institutionalization, weapon use, insight into violence, mental disorder, substance abuse,

stability of relationships, community supports, release to high-risk situations, violence

cycle, impulsivity, cognitive distortions, compliance with supervision, and security level at

release institution). A modified stages of change rating scale is also incorporated which is

designed to provide an indication of overall motivation and treatment readiness and when

rated at posttreatment, purports to quantify treatment change (Wong & Gordon, 2003).

Although it has been reported that the VRS indicators were originally selected for inclu-

sion based on a review of what have been called the “Central Eight” covariates of criminal

conduct (see Olver & Wong, 2019), many of the final risk indicators do not cleanly align to

these domains. Wong and Gordon (2006) did, however, report that seven of the 20 dynamic

risk indicators (criminal attitude, interpersonal aggression, weapon use, insight into vio-

lence, violence cycle, impulsivity, and cognitive distortion) are associated with the

“Antisocial Attitudes” domain described in the Psychology of Criminal Conduct (PCC;

Andrews & Bonta, 2003). Another, the “Criminal Personality” indicator, is used to reflect

the PCC conceptualization of antisocial personality. The authors also describe the other

VRS indicators (violent lifestyle, work ethic, substance abuse, stability of relationships

with significant others, community support, released to high-risk situations, and compli-

ance with community supervision) as reflections of problematic behavior in the domains of

home, work, school, and leisure. However, seven of the VRS dynamic risk indicators

(impulsivity, emotional control, mental disorder, criminal attitudes, substance abuse, inter-

personal aggression, and insight into violence) were identified from the work of Douglas

and Skeem (2005) and were included by Wong and Gordon (2006) on the basis that they

represented “promising” factors. Two other dynamic risk indicators (violence during insti-

tutionalization and security level at release institution) identified in the work of Duncan

et al. (1995) were also considered relevant and included.

Although Wong and Gordon (2006) originally conceptualized the VRS with reference to

three broad frameworks—the Risk-Need-Responsivity principles of effective correctional