Predictive Validity of the DYRA and SAFVR: New Zealand Police’s Family Violence Risk Assessment Instruments

• Published date: 01 October 2021
• DOI: 10.1177/0093854821997525
• Author: Apriel D. Jolliffe Simpson,Devon L. L. Polaschek,Chaitanya Joshi
• Date: 01 October 2021
• Subject Matter: Articles

CRIMINAL JUSTICE AND BEHAVIOR, 2021, Vol. 48, No. 10, October 2021, 1487 –1508.

DOI: https://doi.org/10.1177/0093854821997525

Article reuse guidelines: sagepub.com/journals-permissions

© 2021 International Association for Correctional and Forensic Psychology

1487

PREDICTIVE VALIDITY OF THE DYRA

AND SAFVR

New Zealand Police’s Family Violence Risk

Assessment Instruments

APRIEL D. JOLLIFFE SIMPSON

CHAITANYA JOSHI

DEVON L. L. POLASCHEK

The University of Waikato

Rapid access to accurate risk assessment information is essential for effective police responses to family violence (FV) calls

for service. This study describes the predictive validity of the Dynamic Risk Assessment (DYRA) for family violence and

Static Assessment of Family Violence Recidivism (SAFVR), currently in use by the New Zealand Police. We used 1,817

police reports of FV episodes to predict recurrence (i.e., repeat call for police service) over three follow-up periods.

Regardless of follow-up, the DYRA and SAFVR each displayed poor ability to discriminate between episodes with and

without a recurrence. Both instruments substantially over-predicted recurrence and performed relatively consistently across

subsamples (e.g., intimate partners vs. other family relationship; aggressor gender, ethnicity, age). The especially poor per-

formance of the DYRA suggests further research on dynamic risk factors and their contribution to police responses for FV is

needed to make these instruments more useful for agencies working with families.

Keywords: family violence; police risk assessment; actuarial risk assessment

Family violence (FV) causes billions of dollars of societal harm per year, even in a small

country such as New Zealand (Kahui & Snively, 2014). But this harm is distributed

very unevenly; more than half is experienced by 1% of adults (Adams, 2016). In other

words, exposure to FV varies widely, not only in the severity of individual episodes but

also in the likelihood of repeated episodes, and in the extent to which further exposure may

be mitigated by legal and social responses. Only some FV episodes result in calls for ser-

vice to the police; about one in three victims had an episode reported to police in the latest

nationwide crime survey (New Zealand Crime and Victims Survey, 2018). And even

though most episodes are not reported, police still spend an estimated 41% of frontline

time responding to those incidents that are reported (New Zealand Family Violence

AUTHORS’ NOTE: We would like to thank Josephine Ryan, the Integrated Safety Response, and the New

Zealand Police for supplying the data for this manuscript. We also acknowledge The University of Waikato and

the William Georgetti Scholarship for the financial support provided to the first author. Correspondence con-

cerning this article should be addressed to Apriel D. Jolliffe Simpson, School of Psychology, The University of

Waikato, Private Bag 3105, Hamilton 3240, New Zealand; e-mail: adj10@students.waikato.ac.nz.

997525CJBXXX10.1177/0093854821997525Criminal Justice and BehaviorJolliffe Simpson et al. / Short Title

research-article2021

1488 CRIMINAL JUSTICE AND BEHAVIOR

Clearinghouse, 2017). The volume of calls and the demand they place on services make it

an imperative for law enforcement to be able to assess risk rapidly and accurately prioritize

directing limited resources to where they can be most effective: where the level of risk is

the highest. In this article, we report on one aspect of predictive validity for repeat FV

perpetration—discriminative ability—for two risk assessment instruments currently used

on the frontline by New Zealand police (NZP).

DEFINITION OF FV

The definition of FV in the New Zealand Family Violence Act 2018 is broad, including

physical, sexual, or psychological abuse inflicted against a person by any other person with

whom that person is, or has been, in a family relationship (Family Violence Act 2018, § 9).

The legislation further specifies that a family relationship is one between partners, family

members, or people who ordinarily share a household. But family relationships can also

include close personal relationships, based on factors that include the amount of time spent

together, how shared time is spent, and the duration of the relationship. Such relationships do

not have to include shared sexual activity and may be between carers and care recipients.

Applying this definition of FV leads to the inclusion of a wider range of episodes involving a

wider range of people than those in many previous research populations (Williams, 2012),

which were often limited to men prosecuted for intimate partner violence (IPV; defined as the

physical, sexual, or psychological abuse of a current or former intimate partner). In New

Zealand, most FV-related calls for service do not result in charges being laid; approximately

two thirds are “argument-only” calls for service, where police do not report detecting the use

of physical violence. Moreover, approximately one third of episodes are not any type of IPV.

They involve conflict between siblings, parents and children, or people in other familial rela-

tionships. These differences in context give us an opportunity to use New Zealand data to

extend earlier research, by shedding light on the use of FV risk assessment instruments with

the broad range of episodes police attend, some of which may be relatively minor in nature.

RESEARCH ON RISK ASSESSMENT INSTRUMENTS FOR FV

Risk assessment instruments vary along a number of dimensions, including the types of

FV they are designed to predict, the types of items used, the outcome criteria they are vali-

dated against, and the practical purpose for which they were designed. Most risk assessment

instruments are created for predicting and preventing repeat IPV (e.g., Spousal Assault Risk

Assessment; Kropp & Hart, 2000). Meta-analyses show that using these instruments actu-

arially, most have small to moderate Area Under the Curves (AUCs) in predicting IPV

recidivism (.54–.67; Messing & Thaller, 2013), although AUCs can vary widely (e.g., .46–

.87; van der Put et al., 2019). Recidivism in these studies typically refers to the recurrence

of criminal behaviors, and can be measured in different ways, including victim reports of

repeated violence, arrests, convictions, or even death (Graham et al., 2019).

Besides IPV, the main other type of FV for which specific instruments have been con-

structed is child maltreatment (e.g., Child Abuse Risk Assessment Scale; Chan, 2012). A

meta-analysis of 30 studies found that on average actuarial instruments created for child

maltreatment had moderate AUCs (.70; van der Put et al., 2017). Other specific types of FV

are sufficiently uncommon in comparison to child maltreatment and IPV that specialized

risk assessment instruments have been rarely developed for them.