Addictive Behaviors 39 (2014) 1773–1778

Contents lists available at ScienceDirect

Addictive Behaviors

Violent offending severity among injecting drug users: Examining riskfactors and issues around classification

Michelle Torok a,⁎, Shane Darke a, Fiona Shand b, Sharlene Kaye a

a National Drug and Alcohol Research Centre, University of New South Wales, Sydney, Australiab Black Dog Institute, University of New South Wales, Sydney, Australia

H I G H L I G H T S

• There is currently no way to uniformly classify the severity of violent behaviour.• Four severity classification schemes were tested among injecting drug users (IDU).• Severely violent IDU differed significantly in risk profile from lower level IDU.• Higher cumulative risk exposure was associated with more severe violent offending.• There was considerable lack of uniformity in correlates of severe violent offending.

⁎ Corresponding author at: National Drug and AlcoholNew South Wales, Sydney, NSW 2052, Australia. Tel.: +9385 0222.

E-mail address: m.torok@med.unsw.edu.au (M. Torok

http://dx.doi.org/10.1016/j.addbeh.2014.07.0020306-4603/© 2014 Elsevier Ltd. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Available online 10 July 2014

Keywords:Violent offendingSeverityInjectingDrugs

Objective: There is a paucity of research as to how injecting drug users (IDU)might be differentiated in the sever-ity of their violent offending. This paper reported on the risks associated with severity, as well as issues aroundseverity classification and the impact on observed relationships with known major risk factors.Method: A cross-sectional survey administered to 300 IDU, who had injected drugs weekly or more in the past12 months. A structured questionnaire addresses potential substance use and early-life risk factors for violentoffending.

Results: Four severity groupswere identified: non-violent (24%), low (34%),moderate (22%) and high (20%) leveloffenders. Higher severity groups hadmoreprevalent andmore severe histories of childhoodmaltreatment, childpsychopathology and dysfunctional trait personalities, as well as more severe substance use problems than low-level and non-violent IDU. Regression analyses found that only two of 15 risk factors remained uniformly asso-ciated with violent offending across the four classification schemes tested: (1) having committed violenceunder the influence and (2) having more impulsive trait personalities.Conclusions: Disaggregating IDU into distinct subgroups showed that the extent and severity of predispositionaland substance use risk exposure corresponded to the severity of violent offending. There is a need to establish asystematic method for classifying severity given that there were clinically meaningful differences betweengroups which require further exploration and replication, and because there was extensive variation in therisks associated with severity across schemes.

© 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Violent offending is a major problem among injecting drug users(IDU), with up to 90% having ever committed a violent offence (Darke,Torok, Kaye, Ross, &McKetin, 2010; Neale, Bloor, &Weir, 2005). Despitethe high prevalence of violent offending, surprisingly little is knownabout the differences in the seriousness of violent offending among

Research Centre, University of61 02 9385 0104; fax: +61 02

).

IDU, and whether specific risks are associated with differences in theseverity of violent behaviour. Understanding whether higher- andlower-level violent IDU are uniquely characterised by specific risks hasimplications for the targeted management of violent behaviour. Deter-mining risks associated with severity, however, relies on having a con-sistent, systematic method of classifying violent offending. Currently,no such classification system exists.

Difficulty establishing a systematic severity classification schemecan be attributed to two key issues: (1) difficulties reaching consensusin definitions of violent offending, and (2) differences in severitycriteria. In respect to the former, whilst legal definitions of violence in-clude common assault, aggravated assault and robbery, sexual assault,

1774 M. Torok et al. / Addictive Behaviors 39 (2014) 1773–1778

manslaughter, attempted homicide and homicide (Pink, 2011), these def-initions are not readily accepted in research. Research studies often ex-clude common assault, considering it to be over inclusive, encompassingbehaviourswhichwould not necessarily be considered violent (e.g. push-ing, shoving) (Howard & Dixon, 2011; Kenny & Press, 2006). Omittingcommon assault from research is, however, problematic. As the least seri-ous violent offence, itwould be feasible to assume that thosewho commitonly common assault are likely to be characteristically different fromthose who are committing serious, injurious forms of violence. Indeed,where common assault been omitted in the context of drug and alcoholresearch, it has been found that the prevalence of violent offending isabout half that of studies which have included it (Barratt, Mills, &Teesson, 2011). A substantial proportion of violent drug users appear tobe only committing common assault, and itmay not be possible to gener-alise the risks associatedwith serious violent offending to this lower levelgroup. Moreover, by omitting low level violent drug users from research,there is the implicit assumption that they are no different fromdrug userswho are not violent.Whilst thismay be the case, we currently lack the ev-idence to support such a hypothesis. The general lack of evidence regard-ing severity only serves to highlight the importance of differentiatingdrugusers into hierarchically ordered subgroups to properly understand howtheymight differ in their risk profiles, andwhere to focus intervention ef-forts. Efforts to establish classification systems are further problematisedby extensive variation in severity criteria, which may include statutorymaxima, index offence (most recent offence for which the individualwas taken into custody), weighted indexing systems, or self-report/idiosyncratic indictors (Kenny & Press, 2006). Limitations and differentconceptual definitions attached to each of these criteria arguably affectthe capacity to develop a cohesive body of knowledge about risks associ-ated with severity.

Of the few studies which have examined differences in violentoffending severity among drug users, there are significant methodologi-cal limitations. For instance, measures have been used that do not reflectlegal definitions of violence (e.g. slapped/beat up, weapon involvement,frequency) (Chermack&Blow, 2002; Torok, Darke, &Kaye, 2012),mean-ing that severity categorisations may be arbitrary and difficult to repli-cate. The lack of consistency in severity measures confers to difficultiesidentifying what are reliable correlates of severity, a problem furthercompounded by the inconsistencies in the risk factors controlled for inthese studies. For example, Chermack and Blow (2002) only controlledfor demographic and substance use risks, thus finding that substanceuse was related to severity of violent offending, whilst Torok et al.(2012) controlled for predisposing risks (e.g. psychopathology, childabuse) and substance use histories, finding that only predispositionalfactors were related to a more severe course of violent offending. Thereis no consistency in the extant research findings, and as such, it is notpossible to identify which IDU are at greater risk of committingmore se-rious, costly forms of violence than others.

Based on prior research, it appears to be important to control forboth substance use and predisposing risks for violent offending. Thecurrent study aims to address gaps in our knowledge on the relationshipbetween these risks and violent offending severity by: (1) determiningwhat types of violent offences are being committed by IDU; (2) deter-miningwhether substance use and predispositional riskswere uniquelycorrelated severity subgroups of violent IDU; and (3) examining howdifferent methods of classifying violent offending severity impact onrisk factor identification.

2. Methods

2.1. Procedure

A targeted sample of 300 regular (i.e. weekly or more) IDU wererecruited from needle and syringe programmes (NSPs) locatedthroughout the greater Sydney metropolitan area, as well as by wordof mouth. Recruitment took place from August 2011 until August

2012. Flyers were placed in NSPs and interested persons were requiredto contact the interviewer. All interested participants were screened foreligibility either in person, or by phone, prior to being given an inter-view appointment. In total, 313 respondents were screened, of whom13 (4.2%) did not meet study criteria. All participants fully completedthe questionnaire, and there were no instances of discontinued partici-pation. To be eligible, participants had to be aged 18 years or older; haveinjected illicit opiates and/or psychostimulants weekly or more in the12 months preceding interview; and, to not be intoxicated at the timeof interview. Additionally, respondents were asked ‘dummy’ questions(e.g. current treatment status) to disguise the study criteria and mini-mise the risk of false responses. Eligible participants were administereda face-to-face structured questionnaire, which took an average of30 min to complete. Participation was voluntary. Interviews were com-pleted by the first author, who has completed Composite InternationalDiagnostic Interview training. During the consenting procedure, partic-ipants were assured that any information given was both confidentialand anonymous. Upon completion of the interview, participants werereimbursed AU$30 for out-of-pocket expenses, and provided with con-tact numbers for mental health and social support networks. Ethical ap-proval was obtained from University of New South Wales and SydneySouth West Area Health Service Human Research Ethics Committees.

2.2. Measures

Questions were asked about lifetime and past six month use ofalcohol, tobacco, opioids, methamphetamine, cocaine, ecstasy, benzodi-azepines, hallucinogens, antidepressants, inhalants, and cannabis. Sub-stance use questions were adapted from the Australian TreatmentOutcome Study (Darke et al., 2005). Ages of onset of alcohol intoxica-tion, illicit drug use (non-injecting), injecting drug use, and regularinjecting were obtained. Hazardous and harmful alcohol use wasscreened using the Alcohol Use Disorders Identification Test (AUDIT)(Babor, Higgins-Biddle, Saunders, & Monteiro, 2001).

Participantswere asked about lifetime andpast 12 month prevalenceof violent offending in respect to specific offence types (i.e. common as-sault, aggravated assault, aggravated robbery, aggravated sexual assault,manslaughter, attempted murder, murder). Violent offence coding wasbased on the 2011 Australian and New Zealand Standard Offence Classi-fication (Pink, 2011), consistent with legal definitions of violence. Ifparticipants had committed a violent offence, questions were askedabout age of onset, recency, and number of incidents committed.

Questions regarding childhood maltreatment were adapted fromthe Christchurch Trauma Assessment (Fergusson, Horwood, Shannon,& Lawton, 1989), which has been used in previous research on illicitdrug users (Conroy et al., 2009). Participants were asked whether theyhad been physically abused in childhood (e.g. severely beaten, kicked,burnt with hot objects), had experienced emotional abuse or neglect(e.g. verbally abused by parents, lack of emotional support/care, poorparental supervision), sustained injury from childhood abuse, and thenumber of times they had been assaulted. A Diagnostic and StatisticalManual (4th edition) (DSM-IV) diagnosis of conduct disorder (CD)was obtained using a modified version of the Diagnostic InterviewSchedule (DIS) (Robins, Helzer, Croughan, & Ratcliff, 1981). Participantshave to endorse three or more of 15 symptoms, which must have onsetbefore age 15 years, tomeet criteria for CD. Attention deficit hyperactiv-ity disorder (ADHD) was assessed using a screener adapted from theWorld Health Organisation Composite International Diagnostic Inter-view (American Psychiatric Association, 2000). For diagnosis, six ormore symptoms must have presented before age seven. Symptomsmust have persisted for at least six months and have caused significantimpairment across two or more settings (e.g. social, academic or occu-pational domains). Trait impulsivity was screened for using the BarrattImpulsivity Scale — Short Form (BIS-15) (Spinella, 2007). A normativescore in a non-clinical community sample is 32.8 (Spinella, 2007).Trait aggression was measured using the 12-item short-form of the

1775M. Torok et al. / Addictive Behaviors 39 (2014) 1773–1778

Buss–Perry Aggression Questionnaire (BPAQ-SF) (Bryant & Smith,2001). Normative scores are 30.7 for males and 26.5 for females(Wolff, Morgan, Shi, Huening, & Fisher, 2011).

2.3. Participants

The sample consisted of 300 IDU. Themajorityweremale (70%), andthe mean age was 37.1 years (SD 7.0, range 21–62 years). Two-thirds(66%) were not in a relationship. Mean years of secondary schoolingwere 9.8 (SD 1.4, range 6–12), 85% were currently unemployed. Meanage of first illicit drug use was 15.5 years (SD 2.8, range 6–26), and19.8 years (SD 4.7, range 10–43) at first injection. These werepolysubstance users, whose main drugs of choice were heroin (96%)and methamphetamine (97%). The lifetime mean number of drugclasses used was 9.1 (SD 1.8, range 4–12), with 5.9 (SD 1.7, range2–9) having been used in the preceding six months. Just under half(43%) met criteria for risky alcohol use according to the AUDIT(Babor et al., 2001).

2.4. Data analysis

Severity of violent offending was used as the outcome variable ineach classification. Those who had ever committed a violent offencewere classified into discrete subgroups based on the most seriousoffence they had committed, where seriousness was decided accordingto the maximum sentence receivable under the New South Wales(NSW) criminal legislation (NSW Crimes Act 1900). This approach pro-vides an objective, replicable method of classification.

The sample was divided into four categorical groups; (1) ‘Non’(never committed a violent crime; 24%, n=72); (2) ‘Low’ (most seriousoffence: common assault; 34%, n = 101); (3) ‘Mod’ (most seriousoffence: aggravated assault; 22%, n = 66) and (4) ‘High’ (most seriousoffence: sexual assault, aggravated robbery, manslaughter, attemptedmurder OR murder; 20%, n = 61) (Table 1). Though participants wereallocated to a specific group based on the most severe offence commit-ted, higher severity groups (i.e. Mod, High) could also have committedlower level violent offences (e.g. common assault) in addition to theirmost serious offence. Indeed, 99% of the Mod group had committedcommon assault, whilst 98% of the High group had.

2.4.1. Classification 1: non v. low v. moderate v. high (ordinal measure)This hierarchical classification was guided by the maximum

sentence a person could receive under NSW legislation (Crimes Act1900). Participants were classified as Low, Mod, or High based on themaximum sentence that could legally be applied to the most seriouscrime they had ever committed.

2.4.2. Classification 2: severe v. non-severe (dichotomous measure)This classification compared those in the High group to the rest of

the sample.

2.4.3. Classification 3: high/moderate v. low/non (dichotomous measure)This dichotomous classification code compared those in the

High/Mod groups to those in the Low/Non groups.

Table 1Maximum sentence for NSW and group status based on most serious offence.

Violent offence types Maximum sentence Grouping status allocation N (%)

Non-violent – Non 72 (24)Common assault 2 years Low 101 (34)Aggravated assault 7 years Mod 66 (22)Sexual assault 20 years High 61 (20)Aggravated robbery 25 years HighManslaughter 25 years HighAttempted murder 25 years HighMurder Life High

2.4.4. Classification 4: weighted sentence (continuous measure)Severity was coded as a continuous measure. Each violent offence

typewas given aweight according to themaximumsentence receivableunder NSW legislation. No violent offending= 0, common assault = 1,aggravated assault = 2, sexual assault = 3, aggravated robbery, man-slaughter and attempted murder all = 4 (assigned the same weight asall have the same maximum sentence), and murder = 5. The weightof each offence committed was added up to give a total score (e.g.1 + 2 + 3 = 6). Higher scores represent greater severity.

The violent and non-violent offending, predispositional, andsubstance use risk profiles were compared across the four groups.Chi-square analyses were used to determine if there are differences be-tween groups on categorical variables, with the χ2 statistic reported,along with Cramer's V (φ) effect size (small effect: 0.10; medium:0.30; large: 0.50). Where there were significant χ2 values, pairwiseexaminations were performed to determine where significant groupdifferences existed, and odds ratios (ORs) and 95% confidence intervals(CIs) reported. Where variables were continuous and distributions nor-mative, one-way analysis of variance was used with means, standarddeviations (SD), and eta-squared effect (η2) sizes reported (small effect:0.01; medium: 0.06; large: 0.14). Where the F-value was significant,Scheffe post-hoc comparisons (which have the least chance of makingtype-1 errors) were performed to examine where means were signifi-cantly different from each other. Multivariate regressions were per-formed to determine the factors which remained independentlyassociated with each severity classification. For classification 1, an ordi-nal regression was performed using the most severe offenders (‘High’)as the reference group. Where outcome variables were categorical(classifications 2 and 3), binary logistic regressions were performed.Where the outcome variable was continuous (classification 4) a linearregression was conducted and the β used to determine the directionof the relationship. For each regression the same set of independent var-iables were included (see Table 3). The B (unstandardised regressioncoefficient) and SE (standard error) indicate the contribution of eachvariable to explaining severity of violent offending. Model fit wasassessed using Nagelkerke's R2, with high values indicating bettermodel fit. All analyses were conducted using SPSS Statistics version20.0 (SPSS, 2011).

3. Results

3.1. Violent offending profile

The majority of participants (76%) had ever committed a violentoffence, with no significant differences reported between males andfemales (76% v. 77%; p = 0.81). The High group had initiated violentoffending at a much younger age than the Low group (High: 14.8 yearsv. Low: 19.6 years; p b 0.001), were more likely to have committed vio-lence in the past 12 months compared to the Low group (High: 54% v.Low: 25%; OR 2.22, CI 1.16–4.25, p = 0.02), and were more likely tohave committed violence before using drugs (High: 59% v. Low: 25%;OR 2.71, CI 1.41–5.22, p= 0.002). The Mod group had also initiated vio-lence much younger than the Low group (Mod: 16.3 years v. Low:19.6 years, p b 0.001), were more likely to have committed violence inthe past 12 months (Mod: 53% v. Low: 25%: OR 2.13, CI 1.13–4.02,p = 0.02), and committed violence before using drugs (Mod: 52% v.Low: 25%: OR 1.43, CI 1.01–2.06, p = 0.01). No significant differenceswere reported between High andMod groups on any of these indicators.

3.2. Psychosocial profiles

Predispositional and substance use risks were compared across thefour groups (Table 2). Significant group differences were noted on al-most all variables, with the exception of beingmale, age, and having ex-perienced childhood neglect. Group comparisons showed that the Highand Mod groups were relatively similar in their risk exposure and

1776 M. Torok et al. / Addictive Behaviors 39 (2014) 1773–1778

severity, differing on only two variables: (1) severity of alcohol use, and(2) screening positive for childhood ADHD. The High group was moresevere on both these factors. The Non group had the most favourablerisk profile, as indicated by the lowest prevalence and severity of allrisk factors. The Low group were quite different from the Non groupin their overall risk exposure, with significant differences noted be-tween the two groups on the majority of variables (Table 2).

3.3. Regressions to determine correlates of violent offending severity

Table 3 shows the covariates which remained independently associ-ated with severity of violent offending in each classification scheme.Only two of 15 variables were uniformly correlated with severity ofviolent offending in every classification scheme. These variables were:committing violence whilst intoxicated and scoring higher in the traitimpulsivity measure. Demographic and child maltreatment variableswere not associated with violent offending severity in any classification(Table 3).

4. Discussion

Injecting drug users were disaggregated into distinct subgroups ac-cording to the severity of their lifetime violent offending to determineif there were differences in their risk profiles, and whether the way inwhich severity was classified influenced the relationships with ob-served variables. The findings showed significant differences betweenhigher level, lower level, and non-violent IDU in their risk exposure.The way in which severity was classified had a significant bearing onthe risks identified, with extensive variation noted across schemes.

Lifetime violent offending was prevalent among IDU, at levels com-parable to those noted in previous studies (Darke et al., 2010; Nealeet al., 2005). Disaggregating IDU according to their most serious violentoffence showed that the majority were ‘low-level’ offenders (34%) (e.g.only ever committed common assault). In the extant literature,

Table 2Psychosocial profiles by severity of offending group status.

N (%) or mean ± SD NonN = 72

LowN = 101

ModN = 66

HighN = 61

Male 71 65 70 75Age (years) 39.0 ± 8.1 36.0 ± 7.5 36.5 ± 7.4 37.3 ± 8.6

Substance useAge of onset (years) 16.3 ± 3.1 15.9 ± 2.5 15.1 ± 2.7 14.3 ± 3.0Polysubstance use 8.5 ± 1.9 9.2 ± 1.6 9.3 ± 1.6 9.2 ± 1.9AUDIT score 5.3 ± 5.7 7.6 ± 7.9 8.8 ± 7.9 17.0 ± 12.1

Predispositional risksCD +ve 20 (28) 64 (63) 56 (85) 51 (84)

CD symptom count 1.9 ± 1.9 3.5 ± 2.4 5.6 ± 3.0 6.8 ± 3.6

ADHD +ve 11 (15) 26 (26) 20 (30) 27 (44)Global impulsivity score 36.5 ± 9.3 40.0 ± 9.7 43.3 ± 9.2 45.0 ± 9.6Global aggression score 24.7 ± 10.2 35.6 ± 10.7 45.2 ± 10.1 46.4 ± 10.8

No. times abused 7.2 ± 8.4 11.8 ± 8.7 16.0 ± 8.8 18.8 ± 7.2

Child physical abuse 17 (24) 36 (36) 38 (58) 46 (75)

Childhood neglect 55 (76) 89 (88) 60 (91) 53 (87)Parental substance use 22 (31) 37 (37) 46 (70) 43 (71)

ViolenceViolent whilst intoxicated – 48 (48) 40 (61) 48 (79)Violent for systemic reason – 23 (23) 32 (49) 43 (71)

AUDIT = Alcohol Use Disorder Identification Test; +ve = screened positive; CD = conduct d⁎ p b 0.05.⁎⁎ p b 0.01.⁎⁎⁎ p b 0.001.

common assault is often excluded for being over-inclusive (Howard &Dixon, 2011; Kenny & Press, 2006), however, the current findingsshow that by excluding low severity perpetrators, we are foregoingknowledge of risks associated with violent behaviour in a substantialminority of IDU. The twomost severe groups (High,Mod) reported sim-ilarly stable violent careers (i.e. earlier age of onset, continuing to com-mit violence in the past 12 months), which were significantly differentfrom the Low group. There appears to be a strong relationship betweenseverity and temporal stability. Also, the majority of those in the HighandMod groups had committed violence before using drugs, suggestingthat the liability to severe violent offending is established quite early onand for reasons other than drug use.

A central question of this study was whether subgroups of violentIDU were characterised by unique risk factors. It was found that theprevalence and severity of risk exposure correspondedwith the severityof violent offending across groups. That is, the High and Mod groups,who had highly similar risk profiles, were more likely to screen positivefor childhood psychopathology, havemore dysfunctional trait personal-ities, more extensive histories of child maltreatment, and moreproblematic substance use than other groups. The lack of significant dif-ferences between High and Mod groups in their risk profiles suggeststhat aggravated assault may be a useful cut-off point for defining ‘highseverity’. At the other end of the spectrum, non-violent IDU were de-fined by a relative lack of exposure to these same risks, particularlytrait aggression which occurred at lower than normative levels (Wolffet al., 2011). The Low group lay between these extremes. Of note, theLow group were significantly differentiated from non-violent IDU onthemajority of risk indicators, highlighting that thosewho only commitcommon assault are uniquely differentiated from non-violent andmoresevere groups, thus merit inclusion as a distinct subgroup in future em-pirical research. Evidence of a positive correlation between extent/severity of risk exposure and severity of violent offending in this studysuggests that differences in the severity of violent offending amongIDU are likely to be best explained by a coherent underlying risk

Group comparisons

χ2 = 1.89, p = 0.60; φ = 0.79F = 2.25; p = 0.08; η2 = 0.02

F = 6.99, p b 0.001; η2 = 0.06; High b Low⁎⁎; High b Non⁎⁎⁎F = 2.05, p = 0.01; η2 = 0.04; High N Non⁎; Mod N Non⁎; Low N Non⁎F = 23.14, p b 0.001; η2 = 0.19; High N Mod⁎⁎⁎; High N Low⁎⁎⁎; High N Non⁎⁎⁎

χ2 = 21.33, p b 0.001; φ = 0.46; High N Low⁎⁎; High N Non⁎⁎⁎; Mod N Low⁎⁎;Mod N Non⁎⁎⁎; Low N Non⁎⁎⁎F = 44.20, p b 0.001; η2 = 0.31; High N Low⁎⁎⁎; High N Non⁎⁎⁎; Mod N Low⁎⁎;Mod N Non⁎⁎⁎; Low N Non⁎⁎χ2 = 16.94, p b 0.01; φ = 0.24; High N Mod⁎⁎; High N Low⁎; High N Non⁎⁎⁎F = 10.45, p b 0.001; η2 = 0.10; High N Low⁎; High N Non⁎⁎⁎; Mod N Non⁎⁎F = 63.93, p b 0.001; η2 = 0.39; High N Low⁎⁎⁎; High N Non⁎⁎⁎; Mod N Low⁎⁎⁎;Mod N Non⁎⁎⁎; Low N Non⁎⁎⁎F = 2.98, p b 0.001; η2 = 0.20; High N Low⁎⁎⁎; High N Non⁎⁎⁎; Mod N Low⁎;Mod N Non⁎⁎⁎; Low N Non⁎⁎χ2 = 45.90, p b 0.001; φ = 0.39; High N Low⁎⁎⁎; High N Non⁎⁎⁎; Mod N Low⁎⁎⁎;Mod N Non⁎⁎⁎; Low N Non⁎χ2 = 7.13, p = 0.07; φ = 0.15χ2 = 38.55, p b 0.001; φ = 0.36; High N Low⁎⁎⁎; High N Non⁎⁎⁎; Mod N Low⁎⁎⁎;Mod N Non⁎⁎⁎

χ2 = 93.5, p b 0.001; φ = 0.56; High N Mod⁎; High N Low⁎⁎⁎

χ2 = 86.6, p b 0.001; φ = 0.54; High N Mod⁎; High N Low⁎⁎⁎; Mod N Low⁎⁎

isorder; ADHD = attention deficit hyperactivity disorder.

Table 3Regressions determining factors associated with severity of violent offending.

Classification 1a

High v. Mod v. Low v. non

Classification 2

High v. the rest

Classification 3

High/Mod v. Low/Non

Classification 4

Weighted sentencing

B SE p-value B SE p-value B SE p-value β SE p-value

Sex −.02 .34 .90 .31 .57 .75 −.23 .53 .39 .27 .35 .44

Age −.006 .016 .73 .017 .025 .50 .024 .023 .30 .03 .02 .81

Age onset drug use −.14 .08 .47 −.22 .13 .53 −.23 .11 .54 −.12 .07 .26

AUDIT score .06 .01 <.001 .10 .02 <.001 .03 .03 .22 .06 .02 <.001

No. polydrug classes used −.07 .10 .48 −.18 .17 .21 −.13 .16 .61 −.25 .10 .02

Violent for pharmacologic −1.42 .39 <.001−1.40 .71 .04 .83 .40 .001 .87 .37 .003

Violent for systemic reason −1.60 .37 <.001−1.78 .54 .002 1.57 .39 .61 1.50 .35 <.001

CD symptom count .26 .09 .13 .33 .13 .005 .28 .14 .052 .30 .09 <.001

ADHD +ve .53 .38 .17 −.02 .57 .98 −.87 .47 .06 .07 .38 .86

Aggression score .06 .02 <.001 .01 .04 .90 .13 .04 <.001 .02 .02 .005

Impulsivity score .09 .02 <.001 −.03 .04 .03 −.12 .04 .004 .14 .06 <.001

Child physical abuse 1.60 .52 .056 −1.39 1.10 .21 .85 .44 .06 .51 .26 .08

Child neglect −.78 .62 .64 −1.67 1.10 .13 −2.21 1.11 .26 −.13 .30 .70

No. times abused .04 .02 .22 .08 .04 .27 .07 .03 .42 .05 .02 .14

Parental substance use −.011 .274 .97 −.33 .48 .49 .67 .37 .07 .15 .26 .55

Nagelkerke's R2 61.1 51.0 60.4 54.3

aReferent group is the ‘High’ group; shaded cells indicate significant covariates; +ve = screened positive; ADHD = attention deficit hyperactivity disorder.

1777M. Torok et al. / Addictive Behaviors 39 (2014) 1773–1778

continuum, rather than any specific risk factor. That is, greater cumula-tive risk exposure is linked to more severe violent behaviour.

The secondmajor focus of this study was whether the way in whichseverity is classified, matters. It does. Across the four classificationschemes, only two variables were uniformly associated with severity,despite themajority of these schemes being structured on the same un-derlying subgroups (i.e. different combinations of High/Mod/Low/Non).These two variables were (1) having committed violence whilst intoxi-cated and (2) having a higher impulsive trait personality, suggestingthat these are reliable indicators of those who are at risk of committingserious violence. The extensive variation across remaining risk factors,however, highlights the need to establish greater consensus in how se-verity is classified in order to be able to develop a reliable knowledgebase.Whilst evidence linking chemical effects of drug use to the generallikelihood of committing violence amongdrug users is generally consid-ered to be weak (Boles & Miotto, 2003; Erickson, 2001), it appears thatpharmacologic risk may be an important marker for understanding nu-anced differences in the severity of violence. Also, higher trait impulsiv-ity may lead to more serious violence as such individuals are moreprone to poorly thought out responses in challenging or provocative sit-uations (Kumari et al., 2009;Moore & Foreman-Peck, 2009). As trait im-pulsivity is thought to be present from infancy, and remain stable acrossthe life course (Nestor, 2002), early intervention in antisocial childrendisplaying atypically high levels of impulsivity is likely to be critical toreducing the liability to both drug use and violent offending. Additional-ly, standard drug and alcohol treatment also appears to be an importantpoint of intervention, as beingmore likely to commit violence whilst in-toxicated was a consistent indicator of severe violence. Abstinencewould be needed, however, to sustain reductions in violent behaviour.

All research bear caveats. The main limitation of this study was thereliance on retrospective self-report. Studies have shown, however,that when confidentiality and anonymity are guaranteed self-reportdata can be a reliable and valid source of information for both illicit

drug use (Napper, Fisher, Johnson, & Wood, 2010) and violence(Cartier, Farabee, & Prendergast, 2006). Self-report data was not corrob-oratedwith official reports. Additionally, whilst caremust be taken in ex-trapolating to other regular substance users, the demographic and druguse characteristics of these IDU were typical of those reported interna-tionally (Bargagli et al., 2006; Gossop,Marsden, Stewart, & Treacy, 2002).

To summarise, disaggregating IDU into distinct subgroups showedthat the extent and severity of predispositional and substance use riskexposure corresponded to the severity of violent offending, indicatingrisk commonality across subgroups. There is a clear need to establish asystematic method for classifying severity given that there were clini-cally meaningful differences between groups which require further ex-ploration and replication.

Role of funding sourceFunding for this study was provided by the Australian Government under the

Substance Misuse Prevention and Service Improvements Grants Fund. The funding bodyhad no further role in study design, in the collection, analysis and interpretation of data,in the writing of the report, or in the decision to submit the paper for publication.

ContributorsMichelle Torok and Shane Darke designed the study and wrote the questionnaire.

Fiona Shand and Sharlene Kaye contributed to the statistical analysis, and MichelleTorokwrote thefirst draft of themanuscript. All authors contributed to andhave approvedthe final manuscript.

Conflict of interestAll authors declare that they have no conflicts of interest.

AcknowledgementsThe National Drug and Alcohol Research Centre at the University of NSW is supported

by funding from the Australian Government under the Substance Misuse Prevention andService Improvements Grants Fund. The authors would also like to thank staff and clientsat participating agencies for their time and support.

1778 M. Torok et al. / Addictive Behaviors 39 (2014) 1773–1778

References

American Psychiatric Association (2000). Diagnostic and statistical manual of mental healthdisorders: Text revision (4th ed.). Washington, USA: American Psychiatric Association.

Babor, T. F., Higgins-Biddle, J. C., Saunders, J. B., & Monteiro, M. G. (2001). The Alcohol UseDisorders Identification Test: Guidelines for use in primary care (2nd ed.). Geneva:World Health Organization.

Bargagli, A.M., Faggiano, F., Amato, L., Salamina, G., Davoli, M., Mathis, F., et al. (2006).VEdeTTE, a longitudinal study on effectiveness of treatments for heroin addiction inItaly: Study protocol and characteristics of study population. Substance Use andMisuse, 41, 1861–1879.

Barratt, E., Mills, K., & Teesson, M. (2011). Hurt people who hurt people: Violence amongindividual with comorbid substance use disorders and post-traumatic stress disorder.Addictive Behaviors, 36, 721–728.

Boles, S. M., & Miotto, K. (2003). Substance abuse and violence: A review of the literature.Aggression and Violent Behavior, 8, 155–174.

Bryant, F. B., & Smith, B.D. (2001). Refining the architecture of aggression: Ameasurementmodel for the Buss–Perry Aggression Questionnaire. Journal of Research in Personality,35, 138–167.

Cartier, J., Farabee, D., & Prendergast, M. L. (2006). Methamphetamine use, self-reportedviolent crime, and recidivism among offenders in California who abuse substances.Journal of Interpersonal Violence, 21, 435–444.

Chermack, S. T., & Blow, F. C. (2002). Violence among individuals in substance abusetreatment: The role of alcohol and cocaine consumption. Drug and Alcohol Dependence,66, 29–37.

Conroy, E., Degenhardt, L., Mattick, R. P., & Nelson, E. C. (2009). Child maltreatment as arisk factor for opioid dependence: comparison of family characteristics and typeand severity of child maltreatment with a matched control group. Child Abuse &Neglect, 33, 343–352.

Darke, S., Ross, J., Teesson, M., Ali, R., Cooke, R., Ritter, A., et al. (2005). Factors associatedwith 12 months continuous heroin abstinence: Findings from the Australian Treat-ment Outcome Study (ATOS). Journal of Substance Abuse Treatment, 28, 255–263.

Darke, S., Torok, M., Kaye, S., Ross, J., & McKetin, R. (2010). Comparative rates of violentcrime amongst regular methamphetamine and opioid users: Offending andvictimisation. Addiction, 105, 916–919.

Erickson, P. G. (2001). Drugs, violence and public health: What does the harm reductionapproach have to offer? Vancouver Fraser Institute.

Fergusson, D., Horwood, L., Shannon, F., & Lawton, J. (1989). The Christchurch ChildDevelopment Study: A review of epidemiological findings. Paediatric and PerinatalEpidemiology, 3, 302–325.

Gossop, M., Marsden, J., Stewart, D., & Treacy, S. (2002). Change and stability of changeafter treatment of drug misuse. 2 year outcomes from the National TreatmentOutcome Research Study (UK). Addictive Behaviors, 27, 155–166.

Howard, P., & Dixon, L. (2011). Developing an empirical classification of violent offencesfor use in the prediction of recidivism in England and Wales. Journal of Aggression,Conflict and Peace Research, 3, 141–154.

Kenny, D. T., & Press, A. L. (2006). Violence classifications and their impact on observedrelationships with key factors in young offenders. Psychology, Public Policy, and Law,12, 86–105.

Kumari, V., Barkataki, I., Goswami, S., Flora, S., Das, M., & Taylor, P. (2009). Dysfunctional,but not functional, impulsivity is associated with a history of seriously violent behav-iour and reduced orbitofrontal and hippocampal volumes in schizophrenia.Psychiatry Research: Neuroimaging, 173, 39–44.

Moore, S.C., & Foreman-Peck, J. (2009). Alcohol consumption predicts violent victimiza-tion, impulsive decision making predicts violence. The Open Behavioral ScienceJournal, 3, 28–33.

Napper, L. E., Fisher, D.G., Johnson, M. E., & Wood, M. M. (2010). The reliability andvalidity of drug users' self reports of amphetamine use among primarily heroin andcocaine users. Addictive Behaviors, 35, 350–354.

Neale, J., Bloor, M., &Weir, C. (2005). Problem drug users and assault. International Journalof Drug Policy, 16, 393–402.

Nestor (2002).Pink, B. (2011). Australian and New Zealand Standard Offence Classification (ANZSOC).

(1234.0). Australia: Australian Bureau of Statistics.Robins, L. N., Helzer, J. E., Croughan, J., & Ratcliff, K. S. (1981). National Institute of Mental

Health Diagnostic Interview Schedule: Its history, characteristics, and validity.Archives of General Psychiatry, 38, 381–389.

Spinella, M. (2007). Normative data and a short form of the Barratt Impulsiveness Scale.International Journal of Neuroscience, 117, 359–368.

SPSS Inc. (2011). IBM SPSS statistics 20.0. Chicago, IL: SPSS, Inc.Torok, M., Darke, S., & Kaye, S. (2012). Predisposed violent drug users versus drug users

who commit violence: Does the order of onset translate to differences in the severityof violent offending? Drug and Alcohol Review, 31, 558–565.

Wolff, N., Morgan, R. D., Shi, J., Huening, J., & Fisher, W. H. (2011). Thinking styles andemotional states of male and female prison inmates by mental disorder status.Psychiatric Services, 62, 149–1485.