bs_bs_banner

International Journal of Mental Health Nursing (2019) 28, 1248–1267 doi: 10.1111/inm.12645

R EVIEW A RTICLE
The validity and utility of violence risk assessment
tools to predict patient violence in acute care
settings: An integrative literature review
Manonita Ghosh,1 Di Twigg,1,2 Yvonne Kutzer,1 Amanda Towell-Barnard,1,2 Gideon De
Jong1 and Mary Dodds2
1
School of Nursing and Midwifery, Edith Cowan University, Joondalup, and 2Centre for Nursing Research,
Sir Charles Gairdner Hospital, Nedlands, Western Australia, Australia

ABSTRACT: To examine risk assessment tools to predict patient violence in acute care settings.
An integrative review of the literature. Five electronic databases – CINAHL Plus, MEDLINE,
OVID, PsycINFO, and Web of Science were searched between 2000 and 2018. The reference list
of articles was also inspected manually. The PICOS framework was used to refine the inclusion
and exclusion of the literature, and the PRISMA statement guided the search strategy to
systematically present findings. Forty-one studies were retained for review. Three studies
developed or tested tools to measure patient violence in general acute care settings, and two
described the primary and secondary development of tools in emergency departments. The
remaining studies reported on risk assessment tools that were developed or tested in psychiatric
inpatient settings. In total, 16 violence risk assessment tools were identified. Thirteen of them were
developed to assess the risk of violence in psychiatric patients. Two of them were found to be
accurate and reliable to predict violence in acute psychiatric facilities and have practical utility
for general acute care settings. Two assessment tools were developed and administered in general
acute care, and one was developed to predict patient violence in emergency departments. There is
no single, user-friendly, standardized evidence-based tool available for predicting violence in
general acute care hospitals. Some were found to be accurate in assessing violence in psychiatric
inpatients and have potential for use in general acute care, require further testing to assess their
validity and reliability.
KEY WORDS: acute care, integrative review, nurse, violence prevention, violence risk assessment
tool.

Correspondence: Amanda Towell-Barnard, School of Nursing and Midwifery, Edith Cowan University, 270 Joondalup Drive, Joondalup,
WA 6027, Australia. Email: [email protected]
Funding statement: This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Conflict of interest: No conflict of interest has been declared by the authors.
Author contributions: MG, DT, YK: Made substantial contributions to conception and design, or acquisition of data, or analysis and inter-
pretation of data; MG, YK, DT, ATB, GDJ, MD: Involved in drafting the manuscript or revising it critically for important intellectual con-
tent; DT: Given final approval of the version to be published. Each author should have participated sufficiently in the work to take public
responsibility for appropriate portions of the content; DT: Agreed to be accountable for all aspects of the work in ensuring that questions
related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
Manonita Ghosh, MSS, MA, PhD.
Di Twigg, RN, RM, PhD.
Yvonne Kutzer, MSc.
Amanda Towell-Barnard, RN, DCur, MCur.
Gideon De Jong, RN, MSc, PhD.
Mary Dodds, RMHN, MN (Nse Pract), MACMHN.
Accepted July 30 2019.

© 2019 Australian College of Mental Health Nurses Inc.

VIOLENCE RISK ASSESSMENT TOOLS IN ACUTE CARE 1249

found that nurses experienced post-traumatic stress dis-

INTRODUCTION
order following a violent incident. In their sample of
Violence in hospitals compromises patient, visitor, and 230 emergency nurses, 94% of them displayed at least
staff safety (Spencer et al. 2013). Identifying patients one of the post-traumatic stress symptoms – intrusion
who are at a high risk of committing violent acts is the (such as intrusive thought, nightmares, re-experienc-
first step towards the development of effective violence ing), avoidance (such as numbing, avoidance of feel-
prevention programs (NICE 2015; Policy & Strategic ings), and hyperarousal (such as irritability, anger, and
Project Division 2005). However, most violence risk difficulty concentrating) which also affected the nurses’
assessment tools have been developed for use in psy- work productivity (Gates et al. 2011).
chiatric settings (Daffern, 2007; Dolan & Doyle, 2000; In order to reduce exposure of healthcare personnel
Singh et al. 2011) and may not be useful for the gen- to violence, an effective workplace violence prevention
eral hospital environment. This review examines the and management plan is required. A number of risk
current evidence to determine the most effective vio- assessment tools are described in the research litera-
lence risk assessment tools with potential for adaptation ture; however, the majority of these have been devel-
for use in general acute care settings. oped for psychiatric settings (Daffern 2007; Dolan &
Doyle 2000; Singh et al. 2011) rather than general hos-
pital wards. This review identifies those evidence-based
BACKGROUND
tools that may have practical utility in general acute
Inpatient violence is a widely recognized hazard for care hospitals.
nursing staff (Edward et al. 2016; Johnson 2009). A
meta-analysis of 136 international studies of nurse expo-
AIM
sure to violence reported that 36.4% of nurses experi-
enced physical violence, 66.9% non-physical violence, The purpose of this literature review was to examine
39.7% bullying, and 25% were subjected to sexual risk assessment tools to predict patient violence in gen-
harassment (Spector et al. 2014). Of those who experi- eral acute care hospitals.
enced violence, 32.7% reported having been physically
injured in an assault (Spector et al. 2014). Higher levels
METHODS
of patient violence were experienced by the majority of
the acute care nurses within the Australian health care An integrative review was selected to appraise, analyse,
(Gilchrist et al. 2011; Policy & Strategic Project Division and integrate literature. This methodology allowed the
2005). A survey of 94 medical–surgical wards from 21 inclusion of studies with diverse data collection meth-
hospitals in two Australian states found that 14–38% of ods including experimental, non-experimental, quantita-
the nurse experienced physical violence, threats of vio- tive, qualitative, and mixed-methods design
lence and emotional abuse during their last five shifts (Whittemore & Knafl 2005) to obtain a comprehensive
worked (Roche et al. 2010). Patient-related violence understanding of the violence risk assessment tools and
against nurses is often not reported or under reported their potential utility for the general acute care facili-
in Australia (Lyneham 2000). ties. This methodological combination in integrative
Exposure to violence is a significant stressor within review plays a significant role in evidence-based prac-
the work environment and can result in numerous tice in nursing contributing to policy development on
physical health consequences for nurses including assessing patients (de Souza et al. 2010; Whittemore &
physical injury from assault, disability, and other physi- Knafl 2005). To enhance the rigour of the review, the
cal symptoms. A systematic review of sequelae follow- reviewers followed the five stages of Whittemore and
ing workplace violence conducted by Lanct^ ot and Guay Knafl’s (2005) systematic framework which included: (i)
(2014) found 29 studies which examined the physical problem identification; (ii) literature search; (iii) data
consequences of workplace violence. Consequences evaluation; (iv) data analysis; and (v) presentation of
included physical injuries such as bruises, bites, and findings. In combination with the integrative methodol-
lacerations with life-threatening injuries and permanent ogy, the PICOS framework (Population; Intervention;
disability present in a small percent of victims. Violent Comparator; Outcome; Study design) was used to
incidents impact on the mental health of those who refine the inclusion and exclusion of the literature
experience them and the potential for emotional stress (Schardt et al. 2007). Further, the PRISMA (Moher
following a violent incident is high. Gates et al. (2011) et al. 2009) enabled the reviewers to structure the

© 2019 Australian College of Mental Health Nurses Inc.

1250 M. GHOSH ET AL.

review and systematically present findings by identify- did not evaluate tools in hospitals (n = 128), did not
ing and screening potentially eligible studies and measure patient violence (n = 64), measured violence
including the final number of studies. in children and/or adolescent (n = 22). Some of the
articles (n = 31) were discussion papers or expert opin-
ions, and one article was not written in English. A total
Problem identification
of 74 articles were selected for full-text assessments.
As outlined in the introduction, this review examined After a follow-up discussion between the reviewers, 33
the evidence-based patient violence risk assessment studies were excluded because they did not examine
tools which may have utility in general acute care hos- the validity and reliability of the tools (n = 27), four of
pitals. For the purpose of this review, patient violence them were literature reviews and two included particu-
was defined as any violence incidents conducted by larly forensic patients. In total, 41 studies were
adult patients and are assigned to one of four cate- included in this review. Of them, 32 evaluated one or
gories: verbal aggression; physical aggression against more existing tools, three studies described the primary
objects; physical aggression against self; and physical development and assessment of a tool, two described
aggression against people (Yudofsky et al. 1986). the development of a tool without further testing, and
another four described the modification of an existing
tool. Thirty-six of the included studies reported on
Literature search
tools used to predict violence in psychiatric hospitals.
The second step in the integrative review was data col- Only three studies developed or tested tools in general
lection through literature search. A computerized data- acute care, and two described the development of a
base search of the Cumulative Index of Nursing and tool in emergency departments.
Allied Health Literature (CINAHL Plus with full text),
MEDLINE, OVID, PsycINFO, and Web of Science
Data evaluation
was performed in August 2018. The search strategy flo-
wed from a combination of MeSH terms and keywords, Data evaluation or quality appraisal in an integrative
such as, ‘violence risk assessment’, ‘inpatient violence’, review is the third stage. The assessment tools were
‘violence screening’ ‘violence checklist’, ‘psychopathy examined in terms of their predictive validity and relia-
checklist’, ‘predict hospital aggression’, ‘predict hospital bility, and practical utility. Assessing values for sensitiv-
violence’, and ‘violence checklist’. The literature search ity, specificity, positive, and negative predictive
was restricted to English language research articles provided gold standard for the predictive accuracy of
which were published in academic journals between the tools (Parikh et al. 2008). The Area under the
2000 and 2018 in order to review contemporary, evi- Receive Operating Characteristics (ROC) Curve (AUC)
dence-based violence risk assessment tools. Studies values plots the true positive rate (sensitivity) against
investigating lateral violence in which nurses experi- the false positive rate (specificity) at different threshold
enced violence from co-workers and/or violence from levels also indicated the predictive validity of the tools
patient’s family were excluded from the review. Grey and determined how well the risk assessment tool dis-
literature and studies that examined violence risk in criminated between violent and non-violent patients
community, prison or paediatric hospital settings were (Daffern 2007; Singh et al. 2011). AUC ranging from
also excluded. Hand-searching reference lists of 0.50 (chance prediction) to 1.0 (perfect positive predic-
retrieved articles, previous systematic reviews, and tion), of more than 0.90 are considered to be excellent,
commentary articles was conducted to ensure maxi- 0.80–0.89 are good, and 0.75 is considered the lower
mum coverage. boundary of a useful tool (Dolan & Doyle 2000). Inter-
rater reliability information identified the degree of
consistency between data collectors and, therefore,
Search outcome
determined that data collected in the studies were cor-
A total of 383 records were identified through the rect representations of the violence measured
search strategy and exported into EndNote X7 library (McHugh 2012). The practical utility of the tools was
(Fig. 1). The number of records was reduced to 320 determined by assessing the risk of violence within 12–
after duplicates were removed. Of these, 246 records 24 hours; completing the tool within 5–15 min, com-
were excluded after reviewing the title and abstract of pleted by nurses without knowing patients’ history of
the studies. These articles were excluded because they violence or medical conditions, and completed by

© 2019 Australian College of Mental Health Nurses Inc.

VIOLENCE RISK ASSESSMENT TOOLS IN ACUTE CARE 1251

nurses without having disease-specific knowledge and the reviewed study such as name of the tool, source of
expertise. development, development type, population type, sam-
There is no gold standard for completing a literature ple size, sensitivity and specificity, positive and negative
review using an integrative review quality appraisal predictive values, inter-rater reliability, and AUC-ROC
tools due to the inclusion of diverse methodologies, value were recorded (Table 1). Second, the descriptive
which results in a lack of homogeneity in research information was then compared and pattern estab-
design (Whittemore & Knafl 2005). Therefore, the lished. Third, the process of the comparative analysis
quality appraisal of the studies was conducted based on was examined by two reviewers, and information was
study design, sample size and characteristics, objective put together to recognize the key concept. Finally, the
measurement of outcome, statistical analysis, and rep- major groups were further scrutinized to identify sub-
resentativeness of the study. groups of information. In addition, the risk items used
in each tool were combined and summarized in Table 2.
Data extraction and analysis
The fourth stage in this integrative review was data
RESULTS
extraction and analysis which involved an iterative pro-
cess between the stated research question and the data In total, 16 violence risk assessment tools were identi-
(Beyea & Nicoll 1998). A data extraction matrix was fied from the 41 reviewed studies. The tools are cate-
developed to systematically organize and synthesis gorized into four major groups based on their follow-
information. To enhance rigour during this stage, a up assessment periods and clinical settings in which
four-step systematic analytic method was adopted the tools were administered. The four major groups are
(Pentland et al. 2011). First, descriptive summary of as follows: (i) Tools developed or administered in long-

Records identified through Additional records identified through other

Identification

database searching: CINAHL sources:

MEDLINE, OVID, PsycINFO Reference list (n = 4)
and Web of Science (n = 379)

Duplicates removed (n = 63)

Not hospital settings (n = 128)
Screening

Not patient violence (n = 64)

Records screened Not research articles (n = 31)
(n = 383) Paediatric/juvenile patients (n = 22)
Article not in English (n = 1)

Records further excluded:

Not assessed validity of the tool (n =
Full text articles assessed for
Eligibility

27)
eligibility
Review articles (n = 4)
(n = 74)
Developed/assessed in forensic
setting (n = 2)
Inclusion

Studies included in narrative

synthesis
(n = 41)

FIG. 1: Study selection process. Adapted from the PRISMA flow diagram (Moher et al. 2009).

© 2019 Australian College of Mental Health Nurses Inc.

 TABLE 1: Attributes of the reviewed studies
1252
Positive Negative
Assessment Tool developed/ Outcome Population & predictive predictive Inter-rater Area under the
Authors’ details tool administered measured sample size Sensitivity Specificity values values reliability ROC curve

Tools developed or administered in long-term psychiatric wards

Arbach-Lucioni HCR-20 & Tools evaluated Physical Psychiatric HCR = 0.69–0.77
et al. (2011), PCL-SV through incident aggression offenders, PCL = 0.61–0.70
Spain review and field against others n = 78
observation or objects
Bjørkly and PS33 Modified the Physical/verbal Psychiatric 0.86
Moger (2007), existing tool using aggression against patient
Norway patients vignettes others or objects vignettes,
n = 15 & 8
medical staff
Chagigiorgis BRC Tool evaluated Physical/verbal Psychiatric
et al. (2013), through incident aggression against patients,
Canada review in a others or objects, n = 121
cohort study self-harm, breaking
rules, or exploiting
others
Dolan and HCR-20 Tool evaluated Direct return to Psychiatric HCR = 0.86
Blattner through incident psychiatric hospital patients, H = 0.59C = 0.91
(2010), UK review from community n = 72 R = 0.86
after serious
offence
Doyle et al. VRAG, Tools evaluated Physical/verbal Psychiatric PCL = 78% PCL = 50% PCL = 0.76,
(2002), UK PCL-SV & through incident aggression offenders, VRAG = 51% VRAG = 76% VRAG = 0.71
Historical part review in a against others n = 87 H10 = 75% H10 = 51% H10 = 0.70
10 of HCR-20 cohort study or objects
Eriksen et al. PS33 Tool evaluated Physical/verbal Psychiatric 89% 59% 29% 97% 0.81
(2016), through incident aggression patients,
Norway review in a against n = 541
cohort study others
Gunenc et al. HCR-20 Tool evaluated Physical/verbal Psychiatric 0.65–0.68
(2015), UK through incident aggression patients,
review against others n = 613
Hartvig et al. PS33 Tool evaluated Physical/verbal Psychiatric 81% 73% 24% 97% 0.83
(2011), through incident aggression patients,
Norway review in a against others n = 1017
cohort study
Langton et al. HCR-20, VRS, HCR-20 Physical/verbal Psychiatric Property
(2009), UK Static-99, Risk evaluated aggression offenders, damage = 0.73,
Matrix 2000, & through incident against n = 44 Physical
PCL-R review others or aggression = 0.80
objects

(Continued)

© 2019 Australian College of Mental Health Nurses Inc.

M. GHOSH ET AL.
 TABLE 1: (Continued)

Positive Negative
Assessment Tool developed/ Outcome Population & predictive predictive Inter-rater Area under the
Authors’ details tool administered measured sample size Sensitivity Specificity values values reliability ROC curve

Lynch and ROVA Tool developed Physical/verbal Psychiatric 0.32–9.29

Noel (2010), and evaluated aggression patients,
USA through incident against others n = 161
review or objects, &
self-harm
McDermott COVR, Tools evaluated Physical Psychiatric COVR = 0.73
et al. PCL-R, through incident aggression patients, PCL = 0.73
(2011), USA HCR-20, & review against others n = 146 HCR = 0.73
VRAG VRAG = 0.72
McDermott PCL-R & Tools evaluated Physical Psychiatric Impulsivity:
et al. (2008), HCR-20 in a cohort study aggression patients, PCL = 0.62–0.72,
USA against others: n = 238 HCR = 0.60–0.71;
impulsive, Predatory:
predatory, and PCL = 0.69–0.77,
psychotic HCR = 0.69–0.89;
Psychotic:

© 2019 Australian College of Mental Health Nurses Inc.

PCL = 0.44–0.83,
HCR = 0.58–0.76
Neufeld et al. RHO-CAP Tool evaluated Physical Psychiatric
VIOLENCE RISK ASSESSMENT TOOLS IN ACUTE CARE

(2012), through incident aggression patients,

Canada review in a cohort against others n = 6425
study
Nonstad et al. START Tool evaluated in Physical Psychiatric 0.77
(2010), a cohort study aggression patients,
Norway against others n = 47
O’Shea et al. START Tool evaluated Physical/verbal Psychiatric Self-neglect = Vulnerability =
(2016), UK through incident aggression offenders, 0.83, Self-harm 0.58–0.64,
review against n = 200 & physical Strength = 0.57–0.69
others, aggression = 1.0
self-harm,
substance
misuse &
vulnerability
O’Shea et al. HCR-20 Tool evaluated Physical/verbal Psychiatric HCR = 0.66–0.72,
(2014), UK through aggression offenders, H = 0.47–0.54
incident against others n = 505 C = 0.70–0.76
review R = 0.68–0.70
Rosenfeld et al. FRST & FRST developed Physical/verbal Psychiatric FRST = FRST = FRST = 0.81
(2017), USA HCR-20 and evaluated aggression patients, 92.8% 53.8%
through patient against others n = 159
interview or objects
and other
methods

(Continued)
1253
 TABLE 1: (Continued)

Positive Negative 1254

Assessment Tool developed/ Outcome Population & predictive predictive Inter-rater Area under the
Authors’ details tool administered measured sample size Sensitivity Specificity values values reliability ROC curve

Rotter and FRST & Modified Physical/verbal Psychiatric FRST = 81% FRST = 19%
Rosenfeld, HCR-20 existing FRST aggression against patients,
(2018), USA others or objects n = 103
Snowden et al. VRAG & Tools evaluated Physical/verbal Psychiatric COVR = 0.57–0.73,
(2009), UK COVR through incident aggression against offenders, VRAG = 0.76–0.77
review others or objects n = 52
Starzomski and Imminent Risk Tool evaluated Physical/verbal Psychiatric Physical = Physical = Physical = Physical = 0.91 Physical = 0.57–0.74
Wilson (2015), Rating Scale through incident aggression against offenders, 44.4–70%, 68.7–81.6% 13.3–33.3% 91.9–95.6%, Verbal = 0.60–0.80
Canada review in a cohort others or objects n = 121 Verbal = Verbal = Verbal = Verbal =
study 39.3–67.7% 73.4–91% 43.4–75% 75.4–88.6%
Teo et al. HCR-20 Tool evaluated Physical/verbal Psychiatric 0.78–0.81 0.52–0.70
(2012), USA through incident aggression against patients,
review in a others, or objects, n = 301
case-control study & self-harm
Vitacco et al. PCL-SV, PCL-SV evaluated Physical/verbal Psychiatric PCL = instrumental
(2009), USA WARS through incident aggression against offenders, (0.52–0.87);
& BPRS review and patient others, or objects n = 152 aggression (0.48–0.64)
interviews
Wilson et al. START & Tool evaluated Physical/verbal Psychiatric HCR = 0.79–0.92
(2013), HCR-20 through incident aggression against offenders, START = 0.73–0.89
Canada review in a others n = 30
case-control study
Tools developed or administered in 24-hour psychiatric wards
Abderhalden BVC-VAS Modified the existing Physical/verbal Psychiatric BVC = 61% BVC = 93% BVC = 0.88
et al. (2006), tool and evaluated aggression against patients, VAS = 68% VAS = 95% VAS = 0.90
Switzerland in a cohort study others, or objects, n = 300
& self-harm
Almvik et al. BVC Tool evaluated Physical/verbal Psychogeriatric 0.94
(2007), through incident aggression against & dementia
Norway review others, or objects, patients,
& self-harm n = 82
Almvik et al. BVC Tool evaluated Physical/verbal Psychiatric 63% 92% 0.44 0.82
(2000), through incident aggression against patients,
Norway review others, or objects, n = 109
& self-harm
Chu et al. BVC, Tool developed Physical/verbal Psychiatric BVC = BVC = BVC = BVC = BVC = 0.77
(2013), DASA & in a case-control aggression against offenders, 3–67%, 80–97%, 3–21%, 97–99%, DASA = 0.76,
Australia Clinical scale study others, or objects, n = 70 DASA = DASA = DASA = DASA = HCR-C = 0.68
(C-5) of & self-harm 23–87%, 42–96%, 4–15%, 98–99%,
HCR-20 HCR-C = HCR-C = HCR-C = HCR-C =
4–93% 18–98% 3–7% 97–99%

(Continued)

© 2019 Australian College of Mental Health Nurses Inc.

M. GHOSH ET AL.
 TABLE 1: (Continued)

Positive Negative
Assessment Tool developed/ Outcome Population & predictive predictive Inter-rater Area under the
Authors’ details tool administered measured sample size Sensitivity Specificity values values reliability ROC curve

Clarke et al. BVC Tool evaluated in Any incidents Psychiatric

(2010), a cohort study recorded patients,
Canada n = 48,
& 6 nurses
Griffith et al. DASA-IV Tool evaluated Physical/verbal Mental health 68.10% 70% 0.71
(2013), in a case- aggression against nurses,
Australia control study others, or objects n = 105
Hvidhjelm BVC Tool evaluated in Physical/verbal Psychiatric 65.6% 99.70% 37.20% 99.89% 0.915
et al. (2014), a cohort study aggression against patients,
Denmark others, or objects, n = 156
& self-harm
Lantta et al. DASA Tool evaluated Physical/verbal Mental health 0.86–0.92
(2016), through field aggression against patients,
Finland observation in a others or objects n = 72,
feasibility study & 64 nurses
Ogloff and DASA, DASA was Physical/verbal Mental health DASA = 0.82,

© 2019 Australian College of Mental Health Nurses Inc.

Daffern (2006), BVC, & developed aggression against offenders, BVC = 0.83,
Australia Clinical scale and evaluated others, or objects, n = 100 HCR-C = 0.73
(C-5) of through incident & self-harm
VIOLENCE RISK ASSESSMENT TOOLS IN ACUTE CARE

HCR-20 review and field

observation
Rechenmacher BVC-VAS Tool evaluated Physical/verbal Psychiatric 64.70% 95.10% 14.30% 99.10% 0.93
et al. (2014), in a prospective aggression against patients,
cohort study others, or objects, n = 232
Austria & self-harm
Vojt et al. DASA-IV Modified the Physical/verbal Psychiatric 0.74 (0.70–0.79)
(2010), existing tool and aggression patients,
Scotland evaluated against others, n = 20
through incident or objects, &
review in a self-harm
prospective
cohort study
Woods et al. BVC Tool evaluated Physical/verbal Mental health
(2008), in a pilot study aggression against patients,
Canada others, or objects, n = 93
& self-harm
Yao et al. BVC Tool evaluated Physical/verbal Psychiatric 78.50% 88.20% 14.60% 99.40% 0.41–0.76 0.85
(2014), through incident aggression against patients,
China review and case others, or objects, n = 281
report in a & self-harm
prospective
cohort study

(Continued)
1255
 TABLE 1: (Continued)

Positive Negative 1256

Assessment Tool developed/ Outcome Population & predictive predictive Inter-rater Area under the
Authors’ details tool administered measured sample size Sensitivity Specificity values values reliability ROC curve

Tools developed or administered in general acute care

Ideker et al. M55 Tool evaluated in Physical or any Medical– 41% 99%
(2011), USA a prospective aggression surgical
cohort study summoning patients,
security guards n = 2063
Kim et al. ABRAT Tool evaluated Physical or any Medical– 71.40% 89.30% 15.7% 99.1% 0.65 0.82
(2012), USA in a prospective aggression surgical
cohort study summoning patients,
security guards n = 2063
Kling et al. M55 Tool developed Physical/verbal Acute care 71% 94%
(2006), and evaluated aggression patients,
Canada through incident against others n = 268,
chart review & 18 staff
in a mixed-
methods study
Tools developed or administered in emergency departments
Chapman STAMP/ Modified the Nurse perceptions Nursing
et al. (2009), ED AR existing tool of patient violence staff,
Australia through survey in hospitals n = 113
and interviews
without testing
the tool
Luck STAMP Tool developed Physical/verbal Emergency
et al. (2007), through literature aggression against nursing staff,
Australia review, interviews, others, or objects n = 29
and field
observations
without testing
the tool

Violence Risk Assessment Tool: ABRAT, Aggressive Behaviour Risk Assessment Tool; BRC, Brockville Risk Checklist; BVC, Brøset Violence Checklist; COVR, Classification of Vio-
lence Risk; DASA, Dynamic Appraisal of Situational Aggression; FRST, Fordham Risk Screening Tool; HCR-20, Historical-15, Clinical-5, Risk Management-5; IRRS, Imminent Risk Rat-
ing Scale; M55, Violence Risk Assessment Tool (M55 form); PCL-R, Psychopathy Checklist-Revised; PCL:SV, Psychopathy Checklist: Screening Version; PS33, Preliminary Scheme 33;
RHO-CAP, Risk of Harm to Others Clinical Assessment Protocol; ROVA, Risk of Violence Assessment; STAMP, Staring, Tone, Anxiety, Mumbling, and Pacing; STAMP-EDAR, Staring,
Tone, Anxiety, Mumbling, Pacing, Emotions, Disease process, Assertive/non-assertive, and Resources; START, Short-Term Assessment of Risk and Treatability; VRAG, Violence Risk
Appraisal Guide.

© 2019 Australian College of Mental Health Nurses Inc.

M. GHOSH ET AL.
 TABLE 2: Summary of risk items in each tool

RHO- STAMP-
Group items ABRAT BRC BVC COVR DASA FRST HCR-20 IRRS M55 PCL-R PCL-SV PS33 CAP ROVA STAMP EDAR START VRAG

Access to victim/weapon X2 X
Anxiety/stress X2 X X2 X X X X X X X X
Behaviour/attitude X X2 X X2 X2 X2 X
Childhood/adolescent X X X X X
violence
Confusion X X X X X2 X X
Crime/offense X X X3
Demanding X X
Denial/Withdrawn X X X X2 X X X
Early separation X
Exposure to X X X X X X X X
destabilisers
Fantasizing X X X X X
History of physical X X X X X X X X
violence
Impulsivity X X X X X X X X

© 2019 Australian College of Mental Health Nurses Inc.

Irresponsibility X X2
Irritability X X2 X X2 X
VIOLENCE RISK ASSESSMENT TOOLS IN ACUTE CARE

Lack of compliance X2 X X X3 X X X3 X X2 X
Lack of empathy/guilt X X X2 X X
Lack of support X X X X2 X
Leave X X X X
Manipulative X2 X
Mental illness X X2 X2 X3 X X X X3 X X X X2
Occupational X X X X X
Pacing X X X
Physical aggression/ X X2 X X X X X X2 X
threats
Property damage/ X X
attack
Psychopathy X X X X
Relationship instability X X X X X X
Self-harm/neglect X X2 X2
Staring X X X
Substance abuse X X2 X X X2 X X X
Suicide attempts/ X2 X X
threats
Tone of voice X X X2 X
Unable to plan X X X

(Continued)
1257
1258 M. GHOSH ET AL.

term psychiatric wards; (ii) Tools developed or adminis-

VRAG

Violence Risk Assessment Tool: ABRAT, Aggressive Behaviour Risk Assessment Tool; BRC, Brockville Risk Checklist; BVC, Brøset Violence Checklist; COVR, Classification of Vio-

CAP, Risk of Harm to Others Clinical Assessment Protocol; ROVA, Risk of Violence Assessment; STAMP, Staring, Tone, Anxiety, Mumbling, and Pacing; STAMP-EDAR, Staring, Tone,

assessment tool. For example, two risk items such as ‘access to victim’ and ‘access to weapon’ are grouped into one category. Brockville Risk Checklist comprises both of the risk items
lence Risk; DASA, Dynamic Appraisal of Situational Aggression; FRST, Fordham Risk Screening Tool; HCR-20, Historical, Clinical, Risk Management-20; IRRS, Imminent Risk Rating

Anxiety, Mumbling, Pacing, Emotions, Disease process, Assertive/non-assertive, and Resources; START, Short-Term Assessment of Risk and Treatability; VRAG, Violence Risk Appraisal
Guide. The risk items from each tool were combined into Group items (column 1). Each column represents the number of violence risk items is included in the Group items for each
Scale; M55, Violence Risk Assessment Tool (M55 form); PCL-R, Psychopathy Checklist-Revised; PCL:SV, Psychopathy Checklist: Screening Version; PS33, Preliminary Scheme 33; RHO-
tered in 24-hour psychiatric wards; (iii) Tools devel-
oped or administered in general acute care; and (iv)
START

Tools developed or administered in emergency depart-

ments. The strengths and limitations of the tools are
X

analysed in terms of their predictive validity, reliability,

STAMP-
EDAR

simplicity, and feasibility for use in general acute care

settings. To measure imminent violence, these tools
X

included items broadly categorized into static or

STAMP

dynamic factors or a combination of both. Static factors

X

for violence, such as psychopathy or history of violence,

ROVA

are not subject to change through implementation of

intervention over time. On the other hand, dynamic
factors are behavioural characteristics, such as impul-
RHO-
CAP

sivity, which are subject to change spontaneously either

through changes in the patients’ mental state or other
PS33

circumstances (Douglas & Skeem 2005). Many of these

X2

tools employ assessment approaches in which violence

PCL-SV

is predicted based on evidence-based risk factors asso-

(represents as X2), whereas Preliminary Scheme 33 includes one risk ‘access to dangerous weapon’ (represents as X).

ciated with violence (actuarial approach) or clinicians’

knowledge and experience (clinical judgement risk
PCL-R

assessment approach), allowing clinicians to conduct a

systematic, consistent and yet case-specific assessment
(Douglas & Skeem 2005).
M55

X
IRRS

Tools developed or administered in long-term

X

psychiatric wards
HCR-20

Four violence risk assessment tools were identified to

predict violence in inpatient psychiatric settings with a
3- to 12-month follow-up period. Another seven tools
FRST

were also identified which assessed violence risk within

X

2–6 weeks in psychiatric wards.

DASA

Psychopathy checklist-revised (PCL-R)/screening version

COVR

(PCL-SV)
The PCL-R includes 20-risk items, which are divided
into four domains: interpersonal traits; psychopathy;
BVC

chronically unstable lifestyle; and past antisocial beha-

X

viour (Hare 2003). Each item is scored on a three-

BRC

X2

point rating scale as 0 (absent), 1 (possibly present),

X

and 2 (definitely present) with a maximum score of 40,

ABRAT

with a score exceeding 30 indicating the presence of

psychopathy (Dolan & Doyle, 2000). To evaluate the
PCL-R in McDermott et al. (2008) study, patient files
TABLE 2: (Continued)

were reviewed by a multidisciplinary team, and aggres-

Verbal aggression/

sive behaviour was categorized as impulsive (unplanned

aggression), predatory (planned aggression), and psy-
Victimization
Group items

chotic (delusions/hallucinations-related aggression). The

threats

authors found that the PCL-R score was weakly associ-

ated with imminent impulsive aggression but was

© 2019 Australian College of Mental Health Nurses Inc.

VIOLENCE RISK ASSESSMENT TOOLS IN ACUTE CARE 1259

highly associated with predatory and psychotic aggres- encompasses both static and dynamic factors and has
sion. Vitacco et al. (2009) examined the PCL-SV, a some advantages over the PCL-R/SV, which exclusively
modified version of the PCL-R, and also found that the assesses static factors; however, the R5 items involve
scores were weakly associated with patients’ unplanned an evaluator’s subjective prediction of whether a
aggression, rendering the PCL-R/SV unsuitable to patient is likely to become violent.
assess patients’ unstable mental state in acute care.
Violence risk appraisal guide (VRAG)
Historical, clinical, risk management (HCR-20) The VRAG consists of 12 items including the PCL-R
The HCR-20 was designed to provide evidence-based Score, history of violent and non-violent offences, and
structured clinical guidance for assessing violence in mental disorder (Harris et al. 1993). The item with the
patients with mental disorders (Webster et al. 1997). highest weighting is the PCL-R score. The overall
The tool comprises 20 static and dynamic risk factors VRAG scores range from 27 to +35 and are used to
distributed across three subscales. The Historical sub- assign a patient to one of nine risk categories associated
scale (H10) measures psychopathy and history of vio- with a probability of becoming violent. Patients in cate-
lence as measured by the PCL-R/SV. The Clinical gory 1 have the lowest score and are considered to be
subscale (C5) measures dynamic risk factors through at lower risk of violence than those patients in category
observation of a patient’s current mental state and atti- 9 (Cooke et al. 1999). Two studies by Doyle et al.
tudes. The Risk Management subscale (R5) measures (2002) and Snowden et al. (2009) found the VRAG
risk-related factors such as exposure to destabilisers with moderate predictive validity and significantly cor-
(i.e. access to substances). Each item is rated as 0, 1, 2 related with PCL-SV and HCR-20. As the PCL-R/SV
in a similar manner to the PCL-R/SV and summed for needs to be administered to obtain an overall VRAG
each subscale. Total scores range from 0 to 40 with score, the VRAG will, therefore, have similar practical
higher scores indicating higher risk of violence. After implementation issues to the PCL-R/SV in acute care
rating the 20 items, evaluators then identify any clini- setting.
cal/historical factors and consider the relevance of each
factor to a particular patient and make a final risk Short-term assessment of risk and treatability (START)
judgement to estimate the risk as low, moderate, or The START is a structured tool to assess seven interre-
high for future violence. lated domains: physical violence against others, suicide,
In three studies (Arbach-Lucioni et al. 2011, Dolan self-harm, victimization, substance use, unauthorized
& Blattner 2010, and Langton et al. 2009), the predic- leave, and self-neglect in patients with mental illness
tive accuracy of the HCR-20 was reported as moder- (Webster et al. 2006). The START is one of the few
ate-to-good (AUC range 0.69-0.86) for predicting instruments that assess both risks to self and others.
violence in psychiatric patients. However, the C5 and The seven domains consist of a total of 20 dynamic risk
R5 of the HCR-20 showed stronger predictive accuracy items which are scored as positive and negative, and
than the H10 in these studies. The pattern was con- are rated for strength and vulnerability on a 3-point
firmed by O’Shea et al. (2014), Gunenc et al. (2015) scale from 0 to 2. Based on the item ratings, evaluators
and Teo et al. (2012) who found that the C5 and R5 estimate risk in a similar manner to the HCR-20 as
had significant predictive accuracy. The item ‘psy- low, moderate, or high for each of the seven domains.
chopathy’ was excluded from the assessment in these The evaluators then identify any presence of additional
three studies due to the additional time and expertise risk factors, including mental and historical, before
required to assess psychopathy. Gunenc et al. (2015) making an overall judgment concerning the patient’s
reported that ‘impulsivity’, ‘negative attitudes’, and risk of committing a violent act. Higher strength total
‘non-compliance with medication’ in the C5 subscale scores predict lower risk, whereas higher vulnerability
were the best predictors for verbal aggression in psy- total scores indicate higher risk.
chiatric inpatients, whilst Teo et al. (2012) observed The vulnerability and strength scores of the START
that the predictive accuracy of the HCR-20 largely showed moderate predictive validity in the study by
depended on clinicians’ level of knowledge and experi- Nonstad et al. (2010), but the AUC scores were not
ence. In all examined studies, the HCR-20 like the sufficient to yield a moderate effect size in the study
PCL-R/SV required comprehensive file reviews by a by O’Shea et al. (2016). Wilson et al. (2013) compared
clinician team to assess violence risk, which may not be the START with the HCR-20 and found that the
time-feasible for acute care nurses alone. The HCR-20 START and the C5 of the HCR-20 both had good

© 2019 Australian College of Mental Health Nurses Inc.

1260 M. GHOSH ET AL.

predictive validity, indicating that dynamic risk factors complete within 5 min. The PS33 has some historical
are critical for predicting imminent violence. As is the items requiring patients’ background information to be
case with the HCR-20, the assessment process in these collected before the assessment and, therefore, may
studies involved patient file review by a multidisci- not be applicable for general acute care settings.
plinary team who decided which of the 20 items were The Risk of Violence Assessment (ROVA), devel-
to be defined as critical risk factors. oped by Lynch and Noel (2010), is a 13-item checklist
across four domains: clinical disorders, personality dis-
Brockville Risk Checklist (BRC), InterRAI Risk of Harm orders, psychosocial stressors, and risk assessment and
to Others Clinical Assessment Protocol (RHO-CAP), intervention. Whilst the researchers reported that the
Imminent Risk Rating Scale (IRRS), Preliminary scale took less than 10 min to complete, they found
Scheme 33(PS33), Risk of Violence Assessment (ROVA), that not all risk items were associated with violent inci-
and Classification of Violence Risk (COVR) and the dents nor did they possess satisfactory inter-rater relia-
Fordham Risk Screening Tool (FRST) bility.
Another seven tools assessed the risk of violence within McDermott et al. (2011) administered the Classifica-
2–6 weeks following admission in acute psychiatric set- tion of Violence Risk (COVR), a 44-item computerized
tings. The Brockville Risk Checklist (BRC), developed program with various algorithms customized for each
by Chagigiorgis et al.(2013) is a 41-item checklist over- patient. They reported that it took approximately
lapping across four dynamic risk subscales to assess 20 min to complete the program in which all questions
harm to others, harm to self, risk of neglect, and risk of were linked to each other, and depending on how one
exploitation by others. The researchers reported that question was answered the subsequent question varied.
the ‘harm to others’ subscale predicted non-aggressive The researchers compared the tool with the PCL-R,
incidents rather than aggressive incidents, whilst the HCR-20, and VRAG and reported no statistical differ-
remaining three risk subscales were not associated with ences between them.
any aggressive or non-aggressive outcomes. Further The Fordham Risk Screening Tool (FRST) was
investigation to establish the predictive accuracy of the developed by Rosenfeld et al. (2017) to provide a
tool is required. structured screening approach for the assessment of
Neufeld et al. (2012) investigated the InterRAI Risk psychiatric patients. Two studies, Rosenfeld et al.
of Harm to Others Clinical Assessment Protocol (2017) and Rotter and Rosenfeld (2018) examined this
(RHO-CAP) – a risk assessment algorithm incorporat- tool, which is used to determine whether a more com-
ing a comprehensive mental health assessment includ- prehensive violence risk assessment using a validated,
ing patients’ past and present violence. In their study, existing risk assessment instrument (e.g. the HCR-20),
the risk of violence was two times higher among is required for a particular service user. The FRST
patients with high RHO-CAP compared with patients examines recent (in the previous 6 months) and sev-
with moderate or low scores. Starzomski and Wilson ere violent behaviour, threats, or suicidal ideation.
(2015) administered the seven-item Imminent Risk The FRST demonstrates high sensitivity and moderate
Rating Scale (IRRS) which combined historical, clini- specificity in identifying individuals who subsequently
cal, and contextual factors. Although the inter-rater scored high for risk for violence (based on the case
reliability demonstrated a high level of agreement, the prioritization risk rating of the HCR-20v3). However,
AUC score did not show sufficient predictive accuracy. whilst the FRST is relatively easy to administer, it
Three studies investigated the Preliminary requires trained personnel, and furthermore, its focus
Scheme 33 (PS33) – a 33-item checklist across three on static factors and history of violence would pre-
subscales: historical, clinical, and risk management. clude it from utilization in general medical–surgical
Bjørkly and Moger (2007) examined inter-rater reliabil- acute care.
ity of the tool and found high level of agreement for
the historical and clinical subscales only. The PS33 was
Tools developed or administered in 24-hour
reduced to 10 items and evaluated by Hartvig et al.
psychiatric wards
(2011) and Eriksen et al. (2016) who found the tool as
effective, with AUC = 0.83 for predicting violence in Two risk assessment tools were developed to predict
psychiatric patients. They also changed the name of violence within 24-hour time frame in acute psychiatric
the tool as V-Risk-10 after reducing the 33 checklists to wards. These tools were further examined for their pre-
10. Both researchers reported the tool as easy to dictive accuracy and inter-rater reliability.

© 2019 Australian College of Mental Health Nurses Inc.

VIOLENCE RISK ASSESSMENT TOOLS IN ACUTE CARE 1261

Brøset violence checklist (BVC) moderate risk, and 10–12 high risk. The authors con-
In addition to long-term violence assessment researchers ducted their study in a development and a validation
have focused on developing risk assessment tools for phase. Whilst considerable differences were found
predicting day-to-day aggression in acute psychiatric within the BVC-VAS, the AUC values for the original
wards. The BVC is the most frequently cited and evalu- BVC were consistent in both phases (AUC 0.87 devel-
ated tool identified for predicting violence in psychiatric opment phase, and 0.86 validation phase) indicating
units within 8–24 hours. Developed and examined by that the inclusion of the VAS did not advance the accu-
Almvik et al. (2000), the BVC comprises six items (con- racy of the original BVC. The BVC-VAS was further
fusion, irritability, boisterousness, physical threats, verbal tested by Rechenmacher et al. (2014), who reported
threats, and attacks on objects) each of which is scored that by choosing a cut-off point of ≥6 for the BVC-VAS
as 0 (absent) or 1 (present) with a total score of six. the sensitivity and the specificity was 64.7% and 95.1%,
According to the researchers, a total score of 0 sug- whereas, a cut-off point of ≥7 decreased the sensitivity
gests that the risk of violence is small; scores of 1–2 to 58.8% and increased the specificity to 96.8%, recom-
indicate a moderate risk of violence, and preventative mending further research on the BVC-VAS.
measures are recommended. A score of 3 or more
indicates a high risk of violence requiring immediate Dynamic appraisal of situational aggression (DASA)
preventative measures and the activation of appropri- The DASA was developed and examined by Ogloff and
ate strategies to handle an attack. In their initial Daffern (2006) to assess the risk for imminent aggres-
study, Almvik et al. (2000) reported that the tool had sion on a daily basis in a psychiatric hospital in Aus-
good psychometric properties with an AUC value of tralia. Of the seven risk factors of the tool, two are
0.82. Since then, the BVC has been evaluated in a derived from the HCR-20: negative attitudes and
number of studies in which the predictive validity of impulsivity; two from the BVC: irritability and verbal
the tool has consistently been strong with an AUC of threats; and another three items from the researchers’
0.85 (Yao et al., 2014), 0.92 (Hvidhjelm et al. 2014), previous study: sensitive to perceived provocation,
and 0.94 (Almvik et al. 2007). The BVC has been easily angered when requests are denied, and unwill-
reported to be quick and easy to administer by nurses ingness to follow directions. The score ranging from 0
(Almvik et al., 2007; Clarke et al. 2010; Woods et al. to 7 is calculated to obtain an overall score to assess a
2008; Yao et al. 2014). Clarke et al. (2010) and Hvid- patient’s likelihood for imminent aggression with a
hjelm et al. (2014) reported that ‘irritability’ was the score of 0 indicating low risk, 1–3 as moderate risk and
strongest predictor of the total BVC score, and most a score of 4 or above suggesting high risk for aggres-
violent incidents were triggered by the denial of sion. The researchers recommended implementing pre-
something requested by a patient. ventive measures when a patient scores at 6–7. In their
Inter-rater reliability for the BVC has been study, the DASA, which was administered by nurses
reported in four studies. In the Yao et al. (2014) who scored each item for its presence or absence
study, the inter-rater reliability for single item ranged based on their observations during the past 12–
from ICC = 0.41–0.76 with a total ICC = 0.84. Simi- 24 hours, had a good predictive accuracy with
larly, Almvik et al. (2000) reported Kappa values rang- AUC = 0.82.
ing from 0.48 to 1.0 for single item with 100% The tool has been further validated in another four
agreement for the ‘attacking objects’ item, and studies by Lantta et al. (2016), Chu et al. (2013), Grif-
r = 0.64–1.0 reported by Abderhalden et al. (2006). fith et al. (2013), and Vojt et al. (2010) with moderate
Whilst statistical analysis was not performed, Clarke to strong predictive validity. These studies further
et al. (2010) stated that ‘scores were remarkably simi- reported that the predictive validity of the DASA was
lar for all staff’ between nursing staff and students not significantly different from that of the BVC (Chu
(2010, p. 617). et al. 2013; Ogloff & Daffern 2006) and that the tool
In a prospective cohort study, Abderhalden et al. took less than 5 min to complete (Chu et al. 2013;
(2006) translated the BVC into German and extended Griffith et al. 2013; Lantta et al. 2016; Ogloff & Daf-
it with a 10-cm long Visual Analog Scale marked by ‘no fern 2006). Vojt et al. (2010), however, reported that
risk’ and ‘very high risk’ at each endpoint of the scale. the predictive power of the DASA was consistent for
The BVC-VAS provides a total score of 12, with scores incidents of verbal aggression only, and not for physical
of 0–3 indicating very low risk, 4–6 low risk, 7–9 aggression or all other aggressive incidents. However,

© 2019 Australian College of Mental Health Nurses Inc.

1262 M. GHOSH ET AL.

their study used a small sample size of 20 patients lim- STAMP/EDAR (staring, tone/volume of voice, anxiety,
iting the generalizability of the findings. mumbling, and pacing)/(emotions, disease process,
assertive/non-assertive, resources)
The STAMP is the only risk assessment tool to predict
Tools developed or administered in general
patient violence in emergency departments. Developed
acute care
by Luck et al. (2007), it stems from a qualitative study
Only two screening tools have been developed and involving 290 hours of participant observation, 16 semi-
examined in three studies (Kling et al. 2006, Ideker structured interviews, and 13 informal interviews with
et al. 2011, and Kim et al. 2012) to identify patients at nurses and clinicians, the acronym STAMP includes
risk for violence within 24 hours of admission in gen- observed behavioural cues across five domains: staring,
eral acute care settings. tone/volume of voice, anxiety, mumbling, and pacing.
The researchers reported that as the number of the
Violence risk assessment tool (M55) behavioural cues exhibited by the patients increased,
The M55 was developed and evaluated by Kling et al. the risk of violence increased accordingly in their
(2006) to flag potentially violent patients admitted to study. Chapman et al. (2009) extended the STAMP by
an acute care hospital. The tool includes 11 items and adding another four domains: emotions, disease pro-
has two screening levels to assess violence risk. The cess, assertive/non-assertive, and resources as potential
risk is rated as high if a patient has a history of vio- predictors for violent behaviour. The STAMP/EDAR is
lence or physical aggression, is physically aggressive or expected to be easy to administer by nurses working in
threatening, or is verbally hostile or threatening. The general acute care with no prior knowledge of the
risk is also rated as high with the presence of three or patient’s history (Chapman et al., 2009); however, the
more of the tool’s items. The M55 showed initial relia- predictive validity of the tool is not known, and a scor-
bility and validity with acceptable sensitivity and speci- ing procedure is yet to be developed.
ficity at 71–95%, respectively (Kling et al. 2006). These
results are very different to those reported by Ideker
DISCUSSION AND RECOMMENDATIONS
et al. (2011) who found that the tool predicted a small
percentage of patients identified as at risk of becoming The purpose of this paper was to examine the current
violent compared with those who actually became vio- evidence concerning risk assessment tools predicting
lent in medical–surgical units with lower sensitivity at violent inpatient behaviour in general acute care hospi-
41% and higher specificity at 99%. tals. Only three studies (Ideker et al., 2011; Kim et al.,
2012; Kling et al., 2006) developed or tested tools in
Aggressive behaviour risk assessment tool (ABRAT) general acute care settings, and two (Chapman et al.,
Another tool developed by Kim et al. (2012) is the 2009; Luck et al., 2007) described the primary and sec-
Aggressive Behaviour Risk Assessment Tool (ABRAT), ondary development of tools in emergency depart-
a 10-item checklist combining items from the research- ments. The remaining studies reported on risk
er’s own investigation, the M55, and from the STAMP assessment tools that were developed or tested in psy-
(Luck et al. 2007) with a total score rated on a 3-point chiatric inpatient settings. There were 16 violence risk
scale from 0 (low risk) to 2 (high risk). In this study, assessment tools were found. Of them, two were devel-
the ABRAT had good predictive validity with AUC of oped for general acute care settings, and one was for
0.82 with acceptable sensitivity (71.4%) and specificity emergency departments. The rest were developed and
(89.3%) for identifying violence within 24 hours of examined in psychiatric inpatients.
admission. The ABRAT was also found to be simple This review reveals that despite decades of research
and easy to administer with an inter-rater reliability of on psychiatric inpatient violence there is no single,
j = 0.658 between two nurses. user-friendly, standardized, and evidence-based tool
available for predicting violence in acute care hospitals.
None of the tools developed for use in long-term psy-
Tools developed or administered in emergency
chiatric wards offer support to nursing staff in the
departments
assessment of day-to-day inpatient violence. These tools
Two studies (Luck et al. 2007, and Chapman et al. consist of static risk items such as psychopathy and his-
2009) mentioned about developing a risk assessment tory of violence, which are relatively stable and not
tool in general emergency departments. amenable to deliberate intervention in acute care

© 2019 Australian College of Mental Health Nurses Inc.

VIOLENCE RISK ASSESSMENT TOOLS IN ACUTE CARE 1263

facilities where violence is more likely to be unplanned. may require patient file reviews to some extent by the
As these tools involve intensive clinical interviewing ward nurses. The STAMP, which is the only tool devel-
and patient file review, the administration of these tools oped to measure violence in emergency departments
is time-consuming and the scoring procedure is and has been recommended by researchers (Calow
lengthy, resulting in limited utility in the identification et al. 2016; Pich et al. 2010), is still a foundational work
of violence in general acute care settings where nurses in the early identification of violent behaviour. These
have limited time for risk assessments. These tools tools require to be evaluated for predictive validity and
require multidisciplinary team input to assess the risk reliability in general acute care facilities.
of violence, and so cannot be administered by nurses
alone. Nurses working in general care may have limited
Limitations
mental health expertise and appropriate training to
assess violence risk using these tools, mainly due to a Rigorous methods were undertaken for this review
lack of clinical education hours in mental health including an exhaustive and robust literature search.
(McCann et al. 2009). One exception is the clinical However, there are some limitations. The limitation of
subscale of the HCR-20, which utilizes dynamic risk this integrative review is associated with inclusion of
items to reflect patients’ current mental states, which experimental and non-experimental research studies
could potentially be suitable for assessing risk by ward which might lead to the lack of objective data. The
nurses alone in general acute care settings. The predic- search might also have excluded relevant non-English
tive validity of the C5 was found comparable with that research studies. The practical utility of the risk assess-
of the BVC and DASA (Chu et al. 2013; Ogloff & Daf- ment tools was referred to as simplicity of the tool and
fern 2006). However, the HCR-20 including the C5 was not determined by a valid instrument or a statisti-
was later revised (Douglas 2014; Douglas et al. 2014) cal test. Tools in acute health may need to consider
and, thus, requires further investigation to evaluate the other contributing factors such as different types of risk
new items for its predictive reliability. items (e.g. pain or acquired brain injury) which are not
The BVC and DASA are structured with strong pre- necessarily covered by the tools examined. The absence
dictive validity within 24 hours and have been recom- of a valid practical utility instrument might result in
mended for short-term psychiatric units (Bjorkdahl subjective observations and recommendation by the
et al. 2006; Daffern 2007). Whilst these tools were reviewers. Whilst the included studies were evaluated
developed for and evaluated within psychiatric inpa- in consultations between the reviewers to minimize
tients settings, they have potential to aid prediction of bias, utilizing approved assessment checklists for vari-
imminent violence in general acute care facilities. As ous research designs could further improve the quality
reported in the reviewed studies, the BVC and DASA of research evidence and strengthen the paper. Simi-
are easy to administer by ward nurses and take approx- larly, the use of a valid data extraction matrix could
imately 5 min to complete (Daffern et al. 2009). The have enhanced the rigour of data extraction and data
BVC and DASA comprise risk items that are dynamic analysis. All of the reviewers examined the presentation
and, therefore, can capture fluctuations in the patient’s of findings and conclusion thoroughly, yet the conclu-
mental state. These risk items are indicative of a sions drawn from the research evidence can still be
patient’s present state, not past behaviours, and so subjective.
ward nurses are not required to know about the history
of a patient. The risk information that the tools provide
CONCLUSION
can be used for communication between healthcare
staff for treatment planning and risk management. As This review examined current evidence for predicting
such, both the BVC and DASA have potential to use in violence in acute care hospitals. The main focus of the
medical–surgical care settings; however, they need to review was on the predictive accuracy and practical
be evaluated in acute care hospitals. utility of these tools in general acute care facilities.
There are only two risk assessment tools, the M55 These results of this review show that there is no sin-
and ABRAT, which were specifically developed for gle, user-friendly, standardized, and evidence-based
medical–surgical acute care units to predict inpatient tool available for predicting violence in general acute
violence within a 24-hour period. Despite the potential care hospitals. The BVC and DASA which were found
benefits of their simplicity, both tools are compromised to be accurate in assessing violence in psychiatric inpa-
by their inclusion of patients’ history of violence, which tients and have potential for use in general acute care,

© 2019 Australian College of Mental Health Nurses Inc.

1264 M. GHOSH ET AL.

require further testing to assess their validity and relia- Calow, N., Lewis, A., Showen, S. & Hall, N. (2016).
bility in acute care hospitals. The M55, ABRAT, and Literature synthesis: Patient aggression risk assessment
STAMP/EDAR which were developed particularly for tools in the Emergency Department. Journal of
Emergency Nursing, 42, 19–24.
general acute and emergency departments also need to
Chagigiorgis, H., Michel, S. F., Seto, M. C., Laprade, K. &
be thoroughly evaluated to establish their accuracy and Ahmed, A. G. (2013). Assessing short-term, dynamic
reliability before administering for regular use. changes in risk: The predictive validity of the Brockville
Risk Checklist. International Journal of Forensic Mental
Health, 12, 274–286.
RELEVANCE FOR CLINICAL PRACTICE Chapman, R., Perry, L., Styles, I. & Combs, S. (2009).
Assessing patient violence is essential for the safety of Predicting patient aggression against nurses in all hospital
areas. British Journal of Nursing, 18, 476–483.
staff, patients, and their families in acute medical–sur-
Chu, C. M., Daffern, M. & Ogloff, J. R. P. (2013). Predicting
gical hospitals. Nurses are the prime victim for patient aggression in acute inpatient psychiatric setting using
violence in acute care. The use of an accurate and reli- BVC, DASA, and HCR-20 Clinical scale. Journal of
able risk assessment tool which can be administered by Forensic Psychiatry & Psychology, 24, 269–285.
nurses with no specialized knowledge and expertise in Clarke, D. E., Brown, A. M. & Griffith, P. (2010). The
their busy schedule can be an effective way of reducing Brøset Violence Checklist: Clinical utility in a secure
patient violence and, therefore, improve nurse safety. psychiatric intensive care setting. Journal of Psychiatric
and Mental Health Nursing, 17, 614–620.
Cooke, D. J., Michie, C. & Hart, S. D. (1999). Evaluating the
ACKNOWLEDGEMENTS screening version of the Hare Psychopathy Checklist—
Revised (PCL:SV): An item response theory analysis.
The authors wish to acknowledge Helen Myers for her Psychological Assessment, 11, 3–13.
contributions to the manuscript. Daffern, M. (2007). The predictive validity and practical
utility of structured schemes used to assess risk for
aggression in psychiatric inpatient settings. Aggression and
REFERENCES Violent Behavior, 12, 116–130.
Abderhalden, C., Needham, I., Dassen, T., Halfens, R., Daffern, M., Howells, K., Hamilton, L., Mannion, A.,
Haug, H.-J. & Fischer, J. (2006). Predicting inpatient Howard, R. & Lilly, M. (2009). The impact of structured
violence using an extended version of the brøset-violence- risk assessments followed by management
checklist: Instrument development and clinical application. recommendations on aggression in patients with
BMC Psychiatry, 6, 17. personality disorder. The Journal of Forensic Psychiatry &
Almvik, R., Woods, P. & Rasmussen, K. (2000). The Brøset Psychology, 20, 661–679.
Violence Checklist: Sensitivity, specificity, and interrater De Souza, M. T., Da Silva, M. D. & De Carvalho, R. (2010).
reliability. Journal of Interpersonal Violence, 15, 1284– Integrative review: What is it? How to do it? Einstein, 8,
1296. 102–106.
Almvik, R., Woods, P. & Rasmussen, K. (2007). Assessing Dolan, M. & Blattner, R. (2010). The utility of the Historical
risk for imminent violence in the elderly: The Brøset Clinical Risk -20 Scale as a predictor of outcomes in
Violence Checklist. International Journal of Geriatric decisions to transfer patients from high to lower levels of
Psychiatry, 22, 862–867. security-A UK perspective. BMC Psychiatry, 10, 1–8.
Arbach-Lucioni, K., Andres-Pueyo, A., Pomarol-Clotet, E. & Dolan, M. & Doyle, M. (2000). Violence risk prediction:
Gomar-So~ nes, J. (2011). Predicting violence in psychiatric Clinical and actuarial measures and the role of the
inpatients: A prospective study with the HCR-20 violence psychopathy checklist. British Journal of Psychiatry, 177,
risk assessment scheme. Journal of Forensic Psychiatry & 303–311.
Psychology, 22, 203–222. Douglas, K. S. (2014). Version 3 of the Historical-Clinical-Risk
Beyea, S. C. & Nicoll, L. H. (1998). Writing an integrative Management-20 (HCR-20 V3): Relevance to violence risk
review. Association of periOperative Registered Nurses assessment and management in forensic conditional release
Journal, 67, 877–880. contexts. Behavioral Sciences & The Law, 32, 557–576.
Bjorkdahl, A., Olsson, D. & Palmstierna, T. (2006). Nurses’ Douglas, K. S. & Skeem, J. L. (2005). Violence risk
short-term prediction of violence in acute psychiatric assessment: Getting specific about being dynamic.
intensive care. Acta Psychiatrica Scandinavica, 113, 224–229. Psychology, Public Policy, and Law, 11, 347–383.
Bjørkly, S. & Moger, T. A. (2007). A secondstep in Douglas, K. S., Hart, S. D., Webster, C. D., Belfrage, H.,
development of a checklist for screening risk for violence Guy, L. S. & Wilson, C. M. (2014). Historical- Clinical-
in acute psychiatric patients: Evaluation of interrater Risk Management-20, Version 3 (HCR-20V3):
reliability of the Preliminary Scheme 33. Psychological Development and overview. International Journal of
Reports, 101 (3 Pt 2), 1145–1161. Forensic Mental Health, 13, 93–108.

© 2019 Australian College of Mental Health Nurses Inc.

VIOLENCE RISK ASSESSMENT TOOLS IN ACUTE CARE 1265

Doyle, M., Dolan, M. & McGovern, J. (2002). The validity of Kling, R., Corbiere, M., Milord, R., Morrison, J. G., Craib,
North American risk assessment tools in predicting in- K. & Saunders, S. (2006). Use of a violence risk
patient violent behaviour in England. Legal and assessment tool in an acute care hospital: Effectiveness in
Criminological Psychology, 7, 141–154. identifying violent patients. Workplace Health and Safety,
Edward, K. L., Stephenson, J., Ousey, K., Lui, S., Warelow, 54, 481–487.
P. & Giandinoto, J. A. (2016). A systematic review and Lanct^ot, N. & Guay, S. (2014). The aftermath of workplace
meta-analysis of factors that relate to aggression violence among healthcare workers: A systematic literature
perpetrated against nurses by patients/relatives or staff. review of the consequences. Aggression and Violent
Journal of Clinical Nursing, 25, 289–299. Behavior, 19, 492–501.
Eriksen, B. M. S., Bjørkly, S., Færden, A., Friestad, C., Langton, C. M., Hogue, T. E., Daffern, M., Mannion, A. &
Hartvig, P. & Roaldset, J. O. (2016). Gender differences in Howells, K. (2009). Prediction of institutional aggression
the predictive validity of a violence risk screening tool: A among personality disordered forensic patients using
prospective study in an acute psychiatric ward. International actuarial and structured clinical risk assessment tools:
Journal of Forensic Mental Health, 15, 186–197. Prospective evaluation of the HCR-20, VRS, Static- 99,
Gates, D. M., Gillespie, G. L. & Succop, P. (2011). Violence and Risk Matrix 2000. Psychology, Crime & Law, 15, 635–
against nurses and its impact on stress and productivity. 659.
Nursing Economic$, 29, 59–67. Lantta, T., Kontio, R., Daffern, M., Adams, C. E. &
Gilchrist, H., Jones, S. C. & Barrie, L. (2011). Experiences of V€alim€aki, M. (2016). Using the Dynamic Appraisal of
emergency department staff: Alcohol-related and other Situational Aggression with mental health inpatients: A
violence and aggression. Australasian Emergency Nursing feasibility study. Patient Preference and Adherence, 10,
Journal, 14, 9–16. 691–701.
Griffith, J. J., Daffern, M. & Godber, T. (2013). Examination Luck, L., Jackson, D. & Usher, K. (2007). Stamp:
of the predictive validity of the Dynamic Appraisal of Components of observable behaviour that indicate
Situational Aggression in two mental health units. potential for patient violence in emergency departments.
International Journal of Mental Health Nursing, 22, 485– Journal of Advanced Nursing, 59, 11–19.
492, 488. Lynch, D. M. & Noel, H. C. (2010). Integrating DSM-IV
Gunenc, C., O’Shea, L. E. & Dickens, G. L. (2015). Factors to predict violence in high-risk psychiatric
Prevalence and predictors of verbal aggression in a secure patients. Journal of Forensic Sciences, 55, 121–128.
mental health service: Use of the HCR-20. International Lyneham, J. (2000). Violence in New South Wales emergency
Journal of Mental Health Nursing, 24, 314–323. departments. Australian Journal of Advanced Nursing, 18,
Hare, R. D. (2003). The Hare Psychopathy Checklist - 8–17.
Revised Manual, 2nd edn. Toronto, ON: Multi Health McCann, T. V., Moxham, L., Usher, K., Crookes, P. A. &
Systems. Farrell, G. (2009). Mental health content of
Harris, G. T., Rice, M. E. & Quinsey, V. L. (1993). Violent comprehensive pre-registration nursing curricula in
recidivism of mentally disordered offenders: The Australia. Journal of Research in Nursing, 14, 519–530.
development of a statistical prediction instrument. McDermott, B. E., Quanbeck, C. D., Busse, D., Yastro, K. &
Criminal Justice and Behavior, 20, 315–335. Scott, C. L. (2008). The accuracy of risk assessment
Hartvig, P., Roaldset, J. O., Moger, T. A., Østberg, B. & instruments in the prediction of impulsive versus
Bjørkly, S. (2011). The first step in the validation of a new predatory aggression. Behavioural Sciences & The Law,
screen for violence risk in acute psychiatry: The inpatient 26, 759–777.
context. European Psychiatry, 26, 92–99. McDermott, B. E., Dualan, I. V. & Scott, C. L. (2011). The
Hvidhjelm, J., Sestoft, D., Skovgaard, L. T. & Bjorner, J. B. predictive ability of the Classification of Violence Risk
(2014). Sensitivity and specificity of the Brøset Violence (COVR) in a forensic psychiatric hospital. Psychiatric
Checklist as predictor of violence in forensic psychiatry. Services, 62, 430–433.
Nordic Journal of Psychiatry, 68, 536–542. Mchugh, M. L. (2012). Interrater reliability: The kappa
Ideker, K., Todicheeney-Mannes, D. & Kim, S. C. (2011). A statistic. Biochemical Medicine, 22, 276–282.
confirmatory study of Violence Risk Assessment Tool Moher, D., Liberati, A., Tetzlaff, J., Altman, D. G. & The
(M55) and demographic predictors of patient violence. Prisma Group (2009). Preferred reporting items for
Journal of Advance Nursing, 67, 2455–2462. systematic reviews and meta-analyses: The PRISMA
Johnson, S. L. (2009). International perspectives on statement. PLoS Med, 6, e1000097.
workplace bullying among nurses: A review. International Neufeld, E., Perlman, C. M. & Hirdes, J. P. (2012).
Nursing Review, 56, 34–40. Predicting inpatient aggression using the InterRAI risk of
Kim, S. C., Ideker, K. & Todicheeney-Mannes, D. (2012). harm to others clinical assessment protocol: A tool for risk
Usefulness of Aggressive Behaviour Risk Assessment Tool assessment and care planning. The Journal of Behavioral
for prospectively identifying violent patients in medical Health Services & Research, 39, 472–480.
and surgical units. Journal of Advanced Nursing, 68, 349– NICE (2015). Violence and aggression: short-term
357. management in mental health, health and community

© 2019 Australian College of Mental Health Nurses Inc.

1266 M. GHOSH ET AL.

settings. National Institute for Health and Care Rotter, M. & Rosenfeld, B. (2018). Implementing a violence
Excellence guideline [NG10). [Cited 7 August 2016]. risk screening protocol in a civil psychiatric setting:
Available from: https://www.nice.org.uk/guidance/ng10/ Preliminary results and clinical policy implications.
resources/violence-and-aggression-shortterm-management- Community Mental Health Journal, 54, 245–251.
in-mental-health-health-and-community-settings- Schardt, C., Adams, M. B., Owens, T., Keitz, S. & Fontelo,
1837264712389 P. (2007). Utilization of the PICO framework to improve
Nonstad, K., Nesset, M. B., Kroppan, E. et al. (2010). searching PubMed for clinical questions. BMC Medical
Predictive validity and other psychometric properties of the Informatics and Decision Making, 7, 16.
Short-Term Assessment of Risk and Treatability (START) Singh, J. P., Grann, M. & Fazel, S. (2011). A comparative
in a Norwegian high secure hospital. International Journal study of violence risk assessment tools: A systematic
of Forensic Mental Health, 9, 294–299. review and metaregression analysis of 68 studies involving
O’Shea, L. E., Picchioni, M. M., Mason, F. L., Sugarman, P. 25,980 participants. Clinical Psychology Review, 31, 499–
A. & Dickens, G. L. (2014). Differential predictive validity 513.
of the Historical, Clinical and Risk Management Scales Snowden, R. J., Gray, N. S., Taylor, J. & Fitzgerald, S.
(HCR–20) for inpatient aggression. Psychiatry Research, (2009). Assessing risk of future violence among forensic
220, 669–678. psychiatric inpatients with the Classification of Violence
Ogloff, J. R. P. & Daffern, M. (2006). The dynamic appraisal Risk (COVR). Psychiatric Services, 60, 1522–1526.
of situational aggression: An instrument to assess risk for Spector, P. E., Zhou, Z. E. & Che, X. X. (2014). Nurse
imminent aggression in psychiatric inpatients. Behavioral exposure to physical and nonphysical violence, bullying,
Sciences and The Law, 24, 799–813. and sexual harassment: A quantitative review.
O’Shea, L. E., Picchioni, M. M. & Dickens, G. L. (2016). International Journal of Nursing Studies, 51, 72–84.
The predictive validity of the Short-Term Assessment of Spencer, S., Stone, T. & McMillan, M. (2013). Violence
Risk and Treatability (START) for multiple adverse and aggression in mental health inpatient units: an
outcomes in a secure psychiatric inpatient setting. evaluation of Aggression Minimisation Programs. HNE
Assessment, 23, 150–162. Handover: For Nurses and Midwives, 3, 42–48. [Cited 7
Parikh, R., Mathai, A., Parikh, S., Sekhar, G. C. & Thomas, August 2018]. Available from: https://pdfs.semanticsc
R. (2008). Understanding and using sensitivity, specificity holar.org/095b/af8285c1b4dd6703b2f936b3223a8615f
and predictive values. Indian Journal of Ophthalmology, 405.pdf
56, 45–50. Starzomski, A. & Wilson, K. (2015). Development of a
Pentland, D., Forsyth, K., Maciver, D. et al. (2011). Key measure to predict short-term violence in psychiatric
characteristics of knowledge transfer and exchange in populations: The Imminent Risk Rating Scale.
healthcare: Integrative literature review. Journal of Psychological Services, 12, 1–8.
Advanced Nursing, 67, 1408–1425. Teo, A. R., Holley, S. R., Leary, M. & McNiel, D. E. (2012).
Pich, J., Hazelton, M., Sundin, D. & Kable, A. (2010). The relationship between level of training and accuracy of
Patient-related violence against emergency department violence risk assessment. Psychiatric Services, 63, 1089–
nurses. Nursing and Health Sciences, 12, 268–274. 1094.
Policy and Strategic Project Division (2005). Occupational Vitacco, M., Rybroek, G. J. V., Rogstad, J. E., Yahr, L. E.,
violence in nursing: An analysis of the phenomenon of Tomony, J. D. & Saewert, E. (2009). Predicting short-
code grey/black events in four Victorian hospitals. Policy term institutional aggression in forensic patients: A multi-
and Strategic Project Division, Department of Human trait method for understanding subtypes of aggression.
Service, Victoria, Australia. [Cited 7 August 2018]. Law and Human Behavior, 33, 308–319.
Available from: https://www2.health.vic.gov.au/about/ Vojt, G., Marshall, L. A. & Thomson, L. D. G. (2010). The
publications/researchandreports/occupational-violence-in- assessment of imminent inpatient aggression: A validation
nursingcode-grey-black-events study of the DASA-IV in Scotland. Journal of Forensic
Rechenmacher, J., M€ uller, G., Abderhalden, C. & Schulc, E. Psychiatry & Psychology, 21, 789–800.
(2014). The diagnostic efficiency of the extended German Webster, C., Douglas, K. S., Eaves, D. & Hart, S. D. (1997).
Brøset Violence Checklist to assess the risk of violence. HCR-20: Assessing risk for violence. V2.Canada: Mental
Journal of Nursing Measurement, 22, 201–212. Health, Law, and Policy Institute, Simon Fraser
Roche, M., Diers, D., Duffield, C. & Catling-Paull, C. University.
(2010). Violence toward nurses, the work environment, Webster, C., Nicholls, T. L., Martin, M.-L., Desmarais, S. L.
and patient outcomes. Journal of Nursing Scholarship, 42, & Brink, J. (2006). Short-Term Assessment of Risk and
13–22. Treatability (START): The case for a new structured
Rosenfeld, B., Foellmi, M., Khadivi, A. et al. (2017). professional judgment scheme. Behavioral Sciences & The
Determining when to conduct a violence risk assessment: Law, 24, 747–766.
Development and initial validation of the Fordham Risk Whittemore, R. & Knafl, K. (2005). The integrative review:
Screening Tool (FRST). Law and Human Behavior, 41, Updated methodology. Journal of Advance Nursing, 52,
325–332. 546–533.

© 2019 Australian College of Mental Health Nurses Inc.

VIOLENCE RISK ASSESSMENT TOOLS IN ACUTE CARE 1267

Wilson, C. M., Desmarais, S. L., Nicholls, T. L. & Hart, S. Yao, X., Li, Z., Arthur, D., Hu, L., An, F. R. & Cheng, G.
D. (2013). Predictive validity of dynamic factors: Assessing (2014). Acceptability and psychometric properties of
violence risk in forensic psychiatric inpatients. Law and Brøset Violence Checklist in psychiatric care settings in
Human Behavior, 37, 377–388. China. Journal of Psychiatric & Mental Health Nursing,
Woods, P., Ashley, C., Kayto, D. & Heusdens, C. (2008). 21, 848–855.
Piloting violence and incident reporting measures on one Yudofsky, S., Silver, J., Jackson, W. et al. (1986). The Overt
acute mental health inpatient unit. Issues in Mental Aggression Scale for the objective rating of verbal and
Health Nursing, 29, 455–469. physical aggression. American Journal of Psychiatry, 143,
35–39.

© 2019 Australian College of Mental Health Nurses Inc.