International Journal of Nursing Studies 51 (2014) 198–207

Contents lists available at SciVerse ScienceDirect

International Journal of Nursing Studies

journal homepage: www.elsevier.com/ijns

Clinical prediction of violence among inpatients with

schizophrenia using the Chinese modified version of
Violence Scale: A prospective cohort study
Shing-Chia Chen a,*, Hai-Gwo Hwu b,c, Fu-Chang Hu d,e
a
School of Nursing, College of Medicine and National Taiwan University Hospital, National Taiwan University, Taipei, Taiwan
b
Department of Psychiatry, College of Medicine and National Taiwan University Hospital, National Taiwan University, Taipei, Taiwan
c
Graduate Institute of Epidemiology, College of Public Health, National Taiwan University, Taipei, Taiwan
d
Graduate Institute of Clinical Medicine and School of Nursing, College of Medicine, National Taiwan University, Taipei, Taiwan
e
International-Harvard Statistical Consulting Company, Taipei, Taiwan

A R T I C L E I N F O A B S T R A C T

Article history: Background: A standard measure to assess and predict violence is important for psychiatric
Received 28 September 2012 services. No prospective study has examined the history of violence and heterogeneity of
Received in revised form 4 June 2013 violence in predicting specific types of violence among inpatient with schizophrenia.
Accepted 5 June 2013
Objectives: This study aimed to prospectively examine the accuracy of prediction of types
of violence using the Chinese modified version of Violence Scale (VS-CM) among inpatients
Keywords:
with schizophrenia based on their past history of violence and the real occurrence of
Chinese modified version of Violence Scale
violence during hospitalization.
Predictive ability
Violence Design: A prospective cohort study design.
Schizophrenia Setting and participants: A total of 107 adult patients with schizophrenia spectrum
disorders, consecutively admitted to an acute psychiatric ward of a university hospital in
Taiwan, were recruited.
Method: In addition to data about demographics and clinical illness, count records of the
history of violence within one month prior to admission by interview and the actual
occurrence of violence during the whole course of hospitalization by participant
observation were collected using the VS-CM. Multivariate logistic analysis and area
under the Receiver Operating Characteristic curve (AUC) analysis were applied to examine
the predictive ability of the VS-CM.
Results: A patient’s history of violence assessed by the VS-CM predicted the actual
occurrence of violence during hospitalization with the Odds Ratio of 17.5 (p = 0.001). The
predictive accuracy of the VS-CM had high sensitivity (97.0%), moderate positive
predictive value (71.4%), and high negative predictive value (87.5%); however, the
specificity was relatively low (35.0%). The AUC was 79.5% using the total scale of the VS-
CM and 70.7–74.5% using the subscales in predicting corresponding types of violence.
Conclusion: The VS-CM is a valid and reliable measure of potential violence. It can be
applied to assess and predict specific types of violence among inpatient with
schizophrenia.
ß 2013 Elsevier Ltd. All rights reserved.

* Corresponding author at: School of Nursing, College of Medicine, National Taiwan University, 1 Jen-Ai Road, Section 1, Zhongzheng District, Taipei 100,
Taiwan. Tel.: +886 2 2312 3456x88902; fax: +886 2 2321 9913.
E-mail addresses: shinchia@ntu.edu.tw (S.-C. Chen), haigohwu@ntu.edu.tw (H.-G. Hwu), fuchang.hu@gmail.com (F.-C. Hu).

0020-7489/$ – see front matter ß 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijnurstu.2013.06.002
 S.-C. Chen et al. / International Journal of Nursing Studies 51 (2014) 198–207
What is already known about this topic?
 Incidents of violence by psychiatric patients are preva-
lent and a measure for assessment and prediction is
important for psychiatric services.
 A history of violence is a good predictor, but there is
limited prospective study which examines the issue of
predictive accuracy of a measure of risk for violence.
 Criminological and forensic research has demonstrated
some evidence that violence may be predicted by
identifying particular types of offenses.
What this paper adds
 Using the Chinese modified version of Violence Scale
(VS-CM), potential violence toward property, other
persons, and self can be differentiated and different
histories of violence can be helpful in predicting specific
risks of violence in clinical settings.
 The VS-CM yields high sensitivity, relatively low
specificity, moderate positive predictive value, and high
negative predictive value in the prediction of risk of
violence and different types of violence among inpatients
with schizophrenia.
 The VS-CM is a useful clinical tool for risk assessment and
it could be utilized as a standard measure of potential
violence in psychiatric services.
1. Introduction
The risk of violence and its prevention are important
issues for psychiatric services. Evidence has shown an
incidence rate of 62.6% for aggressive behavior among
inpatient schizophrenics (Chen et al., 2011) and there
were 5.7 incidents of violence per week on acute
psychiatric wards (Foster et al., 2007). Risk assessment
and prediction of violence are necessary for appropriate
psychiatric care.
Violence can be expressed in many forms including
verbal or physical patterns with differential targets, such
as outward violence toward property and other persons,
and inward violence toward the self (Volavka and Citrome,
2008). The accuracy of prediction of violence varies
depending on what constitutes an episode of violence.
One common problem in studies of violence has been a
lack of a unified definition of violence for the purpose of
measurement (Barratt and Slaughter, 1998; Bo et al.,
2011; Morrison, 1988). A standard measure for assess-
ment with a clear operational definition of violence is
crucial in a clinical setting. AUC analysis has been applied
in medicine to describe the accuracy of a screening test for
detecting a disease. Buchanan (2008) stressed that AUC
analysis is applicable for the prediction of violence by
psychiatric patients, for it is less dependent on the base
rate of violence than are other statistics; however, the
method has not been applied in previous studies of
violence.
The original Violence Scale (VS) was a satisfactory
measure for assessing the frequency of aggressive acts
(Morrison, 1993). The Chinese modified version of Violence
Scale (VS-CM) was translated from the VS and rearranged
199
the items from the randomly ordered sequence into the
graded-severity order. Besides, the scaling of the VS-CM
was modified from the Likert scale of the VS into the count
scale for practical application in psychiatric settings (Chen
and Hwu, 2009). It has three subscales of specific types of
violence (toward property, other persons, and self) with a
comprehensive description of aggressive behaviors, and it
provides practical examples as criteria to define and
measure counts of overt aggressive acts by clinical
observation (Chen and Hwu, 2009). It has been applied
to explore the clinical manifestations of violence in an
acute psychiatric ward (Chen et al., 2011); however, its
predictive accuracy for violence has not been determined.
There is only limited evidence that a history of violence
predicts inpatient violence during the whole course of
hospitalization, although it has been reported to be a good
predictor of subsequent aggressive behavior (Bobes et al.,
2009; Lanza, 1996; McNiel and Binder, 1989; Tardiff et al.,
1997; Walsh et al., 2004). Criminological and forensic
research has demonstrated some evidence that violence
could be predicted by identifying particular types of
offenses from the histories; however, few data are
available for psychiatric samples (Buchanan, 2008).
Moreover, it is necessary to focus on a particular group
of psychiatric patients to study violence. Patients with
schizophrenia have been shown to have a relatively high
risk of violence (Tam et al., 1996; Walsh et al., 2004).
Nevertheless, the contribution of schizophrenia to violence
is heavily debated and the conclusions still vary across
studies (Bo et al., 2011). In previous studies of the accuracy
of prediction of violence, the participants were usually
from a global group of mental disorders or forensic patients
(Dolan et al., 2008; Douglas et al., 1999, 2003; Hartvig et
al., 2006; Suchy and Bolger, 1999). These nosological
problems did limit comparisons among the studies of
violence (Bo et al., 2011).
This study aimed to prospectively examine the accuracy
of prediction of types of violence by using the VS-CM in
inpatients with schizophrenia based on their past history
of violence and the real occurrence of violence during their
whole course of hospitalization in an acute psychiatric
ward.
2. Methods
2.1. Setting and participants
Adult patients over the age of 16, consecutively
admitted to the 35-bed acute psychiatric ward of a
university-affiliated teaching hospital in Taiwan were the
potential study sample. Those with a primary diagnosis of
schizophrenia spectrum disorder, including schizophre-
nia, schizoaffective disorder, or schizotypal personality
disorder according to the criteria of the Diagnostic and
Statistical Manual, 4th edition (American Psychiatric
Association, 1994) at admission were asked to participate,
and those whose diagnoses were confirmed at discharge
became the final sample (n = 107). Any patients with
mental retardation, antisocial personality disorder, or-
ganic brain syndromes, or alcohol or substance abuse
were excluded.
200 S.-C. Chen et al. / International Journal of Nursing Studies 51 (2014) 198–207
2.2. Instrument
The Chinese modified version of the Violence Scale (VS-
CM) has 18 items with descriptions and examples of verbal
and non-verbal violent threats and violent acts, and each
item is rated using the actual count of overt aggressive acts.
Based on the targets of violence, the VS-CM assesses three
types of violence: toward property (P, 5 items), toward other
persons (O, 8 items), and toward self (S, 5 items). Counts of
the aggressive acts of P and O are combined as ‘‘outward
violence’’ (PO), S acts as ‘‘inward violence’’, and POS as the
sum of PO and S. The VS-CM has fair internal consistency,
content validity, construct validity, and predictive validity
(r = 0.51, p < 0.001) for aggressive acts occurring during the
initial-week after admission among inpatients with
schizophrenia (Chen and Hwu, 2009).
2.3. Data collection
This study was approved by the internal review board
of the University hospital. Written consent was obtained
and all recruited patients were allowed to withdraw from
the study at any time. Participation was independent of
their treatment benefits or quality of care. Study data
were coded anonymously and personal data were kept
confidential. An experienced head nurse on the acute
psychiatric ward collected the data for a one year after a
pilot study with intensive supervision by the principal
investigator. Data of participants’ histories of violence
were evaluated to predict the actuarial occurrence of
violence during their hospitalization only for the purpose
of this study.
For each participant, the actual occurrence of any type
of violence in conjunction with collateral information as
accurate as possible was monitored and recorded using the
VS-CM on a daily basis during their hospitalization. The
detailed data collection methods were described previ-
ously (Chen et al., 2011).
At the time of admission, each participant’s history of
violence in the last month prior to admission was recorded
by the VS-CM after an interview with participants and their
family caregivers. The larger numbers were recorded when
there were inconsistent reports from different informants
in order to avoid under reporting.
2.4. Statistical analyses
SPSS software, version 16.0 (SPSS Inc., Chicago, IL,
U.S.A.), was used for statistical analysis. Two-tailed p
values less than or equal to 0.05 were considered
statistically significant. Odds ratios (ORs) in a multivariate
logistic regression were computed to determine the
association between the demographic data (age, gender,
religion, education, married status, and employed status),
the clinical data (type of schizophrenia spectrum, mean
age of onset), the presence of a history of violence and later
actual occurrence of violence.
For the accuracy of predictions, the Area under the ROC
(Receiver Operating Characteristic) Curve (AUC) method
based on a cut-off value of zero (absence of violence) and
non-zero (presence of violence) was applied to test the
ability of the total scale and the three subscales to predict
specific types of violence. Four indicators of accuracy were
computed. Sensitivity (true positive/[true positive + false
negative]) indicated the percentage of the cases where
violence occurred during hospitalization among those who
had a history of violence. Specificity (true negative/[true
negative + false positive]) indicated the percentage of
those who had no episodes of violence during hospitaliza-
tion among those who had no history of violence. The
positive predictive value (true positive/[true positive + -
false positive]) indicated the percentage of those who had a
history of violence among those who had episodes of
violence during hospitalization. The negative predictive
value (true negative/[true negative + false negative]) indi-
cated the percentage of those who had no history of
violence among those who had no episodes of violence
during hospitalization.
The AUC for the probability of a violent history in those
who committed specific types of violence during hospital-
ization (sensitivity) and the probability of the same violent
history in those who did not commit specific types of
violence during hospitalization (1 specificity), was ap-
plied to examine predictive ability. This evaluation index
reflected the concordance between the predictions from
past history and the actual occurrence of specific types of
violence in rank order. An AUC from 51% to 100% indicated
a good to perfect ability to discriminate. An AUC of 50% or
less indicated that the discriminatory capacity was no
better than chance (Rice and Harris, 1995).
The probability ratios for sensitivity and
(1 specificity) were further computed. If the value of
this ratio were greater than 10, the rating scale can be used
to rule in a case. If the ratio were below 0.1, it could rule out
a case. If the ratio were around 1, it would be uninforma-
tive (Hatcher et al., 2005, p. 114, 115).
3. Results
3.1. Participants’ demographic and clinical illness data
Detailed participants’ demographics and clinical ill-
ness profiles were presented in a previous publication
(Chen et al., 2011). In summary, the study participants
had a mean age of 33.4  11.9 years (mean  SD). Most
were unmarried (68.2%), unemployed (73.8%), female
(69.2%), with eastern religions of Folklore and Buddhism
(57.9%) and an education level of senior high school (41.4%).
Their clinical diagnoses were paranoid schizophrenia
(50.5%), disorganized schizophrenia (25.2%), undifferentiat-
ed schizophrenia (15.0%), residual schizophrenia (1.9%),
schizoaffective disorder (3.7%), and schizotypal personality
disorder (3.7%). Their duration of hospitalization varied from
one week to twelve weeks.
3.2. Incidence of past history and actual occurrence of
violence
The incident rates and the mean counts of participants’
specific types of violent acts one month prior to admission
and actual occurrences during hospitalization are pre-
sented in Table 1. There were 91 patients (85.0%) with a
 S.-C. Chen et al. / International Journal of Nursing Studies 51 (2014) 198–207 201

Table 1
Incident rates and mean counts of specific types of violence history prior to admission and actual occurrence during hospitalization (n = 107).

Types of violence Past history Actual occurrence

Subject Count Subject Count

n % Mean SD Range n % Mean SD Range

Pa 48 44.9 7.9 8.5 1–45 32 29.9 7.8 9.8 1–40

Oa 75 70.1 18.2 20.8 1–122 54 50.5 18.3 19.1 1–81
Sa 35 32.7 7.9 12.3 1–63 16 15.0 4.9 5.5 1–20
POb 80 74.8 21.8 26.4 1–145 58 54.2 21.4 23.8 1–112
POSb 91 85.0 22.2 26.2 1–145 67 62.6 19.7 23.4 1–112
a
P, O, and S: aggressive acts toward property, other persons, and self.
b
PO and POS: outward aggressive acts and total aggressive acts.

history of violence and 67 patients (62.6%) had violent acts
post admission. The incidence of outward violence (PO)
and inward violence (S) was 74.8% and 32.7% pre
admission, and 54.2% and 15.0% post admission, respec-
tively. The incident rates per person for specific types of
violence declined markedly after admission; however, the
mean incident counts were varied but did not decline
markedly after admission, except for violence toward self.
The mean incident counts for the three types of violence
ranged from 7.9 to 18.2 and 4.9 to 18.3 in both pre-
admission and post-admission periods, respectively.
3.3. Odds ratios of predicting the actual occurrence of
violence based on a history of violence
When all demographic, clinical, and historical violence
items were included in a multivariate logistic regression for
predicting the presence of total violence post admission, only
the item of history of violence reached statistical significance
with an OR of 17.5 (p = 0.001) (95% C.I.: 3.72–82.44).
3.4. Predictive accuracy of specific types of actual violence
using a history of violence
Any singular type of violent history toward property (P),
other persons (O), or self (S) could best predict the
corresponding type of violence post admission with AUC of
72%, 72.6%, and 70.7%, respectively (Fig. 1). The accuracy
indicators for sensitivity (81.3%, 88.9%, and 93.8%),
specificity (70.7%, 49.1%, and 78.7%), positive predictive
value (54.2%, 64.0%, and 42.9%), and negative predictive
value (89.8%, 81.3%, and 98.6%) determined the pairs of
corresponding predictions (Table 2). The probability ratios
were 2.8, 1.7, and 4.4, respectively. Additionally, the
singular types of P and O pre admission could also identify
the composite types of PO post admission (ratio = 1.5 and
2.0) better than they could identify the singular type of S
post admission (ratio = 0.4 and 1.0).
A combination of outward violence (PO) and total
violence (POS) in the history of violence could well predict
the corresponding types of violence post admission with
the AUC of 76.3% and 79.5% (Fig. 2). The accuracy indicators
for sensitivity (93.1% and 97.0%), specificity (46.9% and
35.0%), positive predictive value (67.5% and 71.4%), and
negative predictive value (85.2% and 87.5%) are presented
as pairs of corresponding predictions (Table 3). The
probability ratios are 1.8 and 1.5, respectively. Outward
violence (PO) pre admission predicted a majority of O and
POS post admission with AUC of 76.3% and 72.1% (Fig. 2a).
Total violence (POS) pre admission predicted a majority of
O and PO post admission with AUC of 76.0% and 74.5%,
respectively (Fig. 2b).

Fig. 1. The percentage and 95% confidence interval of the estimated area under the Receiver Operating Characteristic (ROC) curve using past singular types of
violence history toward (a) property, (b) other persons, and (c) self to predict the actual occurrence during hospitalization.
202 S.-C. Chen et al. / International Journal of Nursing Studies 51 (2014) 198–207

Table 2
Sensitivity (SN), specificity (SP), positive predictive value (PPV), and negative predictive value (NPV) of predicting actual occurrence of singular types of
violence during hospitalization using past specific types of violence history.

Past history Actual occurrence

Pa Oa Sa

SN SP PPV NPV SN SP PPV NPV SN SP PPV NPV

Pa 81.3 70.7 54.2 89.8 59.3 69.8 66.7 62.7 31.3 52.7 10.4 81.4
Oa 90.6 38.7 38.7 90.6 88.9 49.1 64.0 81.3 62.5 28.6 13.3 81.3
Sa 15.6 60.0 14.3 62.5 29.6 64.2 45.7 47.2 93.8 78.7 42.9 98.6
POb 96.9 34.7 38.8 96.3 94.4 45.3 63.8 88.9 75.0 25.3 15.0 85.2
POSb 96.9 20.0 34.1 93.8 98.1 28.3 58.2 93.8 100.0 17.6 17.6 100.0
a
P, O, and S: aggressive acts toward property, other persons, and self.
b
PO and POS: outward and total aggressive acts.

Fig. 2. The percentage and 95% confidence interval of the estimated area under the Receiver Operating Characteristic (ROC) curve using composite types of
(a) outward and (b) total violence history to predict the actual occurrence of specific types of violence during hospitalization.

Table 3
Sensitivity (SN), specificity (SP), positive predictive value (PPV), and negative predictive value (NPV) of predicting actual occurrence of composite types of
violence during hospitalization using past specific types of violence history.

Past history Actual occurrence

POb POSb

SN SP PPV NPV SN SP PPV NPV

a
P 60.3 73.5 72.9 61.0 55.2 72.5 77.1 49.2
Oa 86.2 49.0 66.7 75.0 80.6 47.5 72.0 59.4
Sa 27.6 61.2 45.7 41.7 35.8 72.5 68.6 40.3
POb 93.1 46.9 67.5 85.2 88.1 47.5 73.8 70.4
POSb 96.6 28.6 61.5 87.5 97.0 35.0 71.4 87.5
a
P, O, and S: aggressive acts toward property, other persons, and self.
b
PO and POS: outward and total aggressive acts.

4. Discussion hospitalization course with an Odds ration of 17.5. It can

also be applied to assess and predict specific types of
The VS-CM is a valid measure for assessing the past violence among inpatient with schizophrenia. It has a
history of violence which can predict violence during high sensitivity (97.0%), moderate high predictive value
 S.-C. Chen et al. / International Journal of Nursing Studies 51 (2014) 198–207
(71.1%) and high negative predictive value (87.5%). The
high sensitivity and negative predictive value are of
clinical value in clinical application. Most interestingly,
there is relatively high predictive power (AUC 70.7–
74.5%) in predicting subtypes of violence using corre-
sponding subscales of the VS-CM. This is also of
important meaning in clinical practice in the acute
psychiatric ward.
4.1. Predictive ability of the VS-CM
This was the first prospective validation study for the
predictive ability of the VS-CM among inpatients with
schizophrenia in an acute psychiatric ward. The risk
assessment of a history of violence using the VS-CM
generally yielded a high sensitivity, moderate positive
predictive value, and high negative predictive value in
predicting any actual occurrence of violence, but specifici-
ty was relatively low.
High sensitivity of the total VS-CM (97.0%) in predicting
violence is clinically important to alert professional staff to
identify patients at higher risk for subsequent aggressive
acts during hospitalization. The low specificity of the total
violence prediction (35.0%) is understandable as the
incidents per person decreased markedly after admission.
The process of hospitalization itself is of therapeutic value
in reducing aggressive acts.
Adoption of the identification of specific types of
violence from the past history using the subscales of the
VS-CM could enhance specificity, especially in violence
toward property and toward self. A plausible explana-
tion for the lower specificity of violence toward other
persons may be an interaction with situational factors
(Josefsson et al., 2007). The sensitivity and specificity of
the subscale toward self were both relatively high in
predicting a corresponding violent act. This could alert
clinical professionals to the persistent nature of violence
toward the self and lead clinical service teams to prevent
it. This is consistent with a history of inner aggres-
siveness as stated by Plutchik (1995) and Plutchik et al.
(1989). We suggest that identifying patients with
specific types of violent histories at admission could
allow the inpatient psychiatric service team to have an
early opportunity to recognize and avoid triggering
situations relevant to the specific types of aggressive
behavior. This clinical strategy can also be helpful for
establishing a new homeostasis on the psychiatric
service (Buchanan, 2008; Plutchik, 1995). This is similar
to Mayne and Ambrose’s (1999) emphasis on turning
down the rheostat for individual anger or impulsive
reactions. Thus, this study result provides the potential
way for preventing violence in the psychiatric ward. The
effects of violence prevention by implementing the VS-
CM in daily practice for decreasing the time of
hospitalization and facilitating better social adjustment
as suggested by Yen et al. (2002), deserved to be studied
further.
The values of moderately positive and highly negative
predictive accuracy could have been affected by the
phenomenon of the relatively low incidence of aggres-
sive acts among psychiatric inpatients. Statistically, as
203
the incidence was low, the positive predictive value
would decrease relatively and the negative predictive
value would increase (Hatcher et al., 2005, p. 114).
Clinically, the moderately positive predictive value could
also be the result of marked decrease in incidents per
person after admission due to the treatment effects of
the process of hospitalization. The highly negative
predictive value also reassures clinical professionals
that the absence of a violent history might result in a low
risk of violence after admission. Further study with
larger samples deserves consideration.
According to the ROC analysis, the ability of the total
scale of the VS-CM (POS) (AUC = 79.5%) to predict
subsequent violence of the patients with schizophrenia
was generally similar to other extensively used risk
assessment measures for other patient groups, such as
patients discharged from acute general psychiatric
facilities with AUC of 71–77% using the Preliminary
Scheme (PS) (Hartvig et al., 2006); discharged forensic
patients with a mental disorder with AUC of 63% using
the Historical-Clinical-Risk Management 20-item scale
(HCR-20) (Douglas et al., 1999, 2003); a sample of
forensic inpatients with AUC of 68.8% using the Violence
Risk Scale, 2nd edition (VRS-2) (Dolan et al., 2008); and
the psycho-geriatric inpatients with AUC of 80–84%
using the Behavioral Dyscontrol Scale (BDS) (Suchy and
Bolger, 1999).
The VS-CM used in this study can further provide
rather comprehensive data for assessment and predic-
tion in clinical practice, and can avoid the problem of
under-reporting due to patients’ own arbitrary state-
ments and the social desirability of minimizing their
risk for violence (Roaldset and Bjorkly, 2010). Moreover,
the subscales of P, O, S, and PO on the VS-CM in our
study were formulated to provide the ability to predict
the corresponding types of subsequent violence
(AUC = 72.0%, 72.6%, 70.7%, and 76.3%). Identifying
particular types of violence from past history is
important for reducing subsequent violence in clinical
practice. In contrast to the better prediction of outward
violence toward property and other persons, inward
violence toward the self shows lower predictive power.
The heterogeneous nature of violence needs further
exploration.
4.2. Predictive ability of a history of violence
In our study, a history of violence was the best
predictor of subsequent violence. The predictive ORs of
the VS-CM using multivariate logistic statistics were
high (up to 17.5), whereas a relatively low probability
ratio (not over 10) was found using the AUC. Our study
provides evidence that a violent history is an important
predictor, but it also warns that prediction needs to
incorporate other related risk factors in addition to the
historical data. For example, personal traits, psychiatric
symptoms and contextual variables, and situational and
structural factors as also emphasized by Fresán et al. (2007),
Monahan et al. (2001) and Josefsson et al. (2007),
respectively, could be considered for further improving
the predictive probability ratio.
204 S.-C. Chen et al. / International Journal of Nursing Studies 51 (2014) 198–207
4.3. The incidence of violence
We found that violence among inpatients with schizo-
phrenia had the characteristics of declining incident rates
per person but some high mean counts of violence did
not go down. Those cases with a violent pattern of
fluctuating-decrease or non-remission tended to be
refractory to regular treatment and had a longer
hospitalization (Chen et al., 2011). It is necessary to
identify, as early as possible, these patients with
refractory patterns of violence, and to modify the
treatment protocol for them. The MacArthur Violence
Risk Assessment study indicated that the incidence of
violence among treated patients was much less than that
among those not taking medication (Monahan et al.,
2001). The importance of treatment in reducing the
aggressive acts of patients with schizophrenia was also
stressed by Torrey (2006).
4.4. Strengths and limitations
Some methodological issues in this study deserve
attention. Prospective studies allow researchers to draw
conclusions of a rather causal nature from their findings.
The strength of this study lies in applying the VS-CM as
an operational measurement tool to assess the occur-
rence of aggressive acts in a prospective design, but this
prospective study included the use of recalled data from
history-taking as the predictor. Bo et al. (2011)
supported the idea that such a method is especially
relevant for incidence/prevalence-rate and frequencies
of certain characteristics in a population at a particular
point in time. Methodological limitations inherent in
this study included the violent base rate and the
treatment effect. Besides, this is a relatively small
sample in a specific population and therefore the
parameter estimates will be relatively imprecise and
their generalizability uncertain. Alternative studies
deserve further validation.
5. Conclusions and clinical implications
This prospective study demonstrated that the VS-CM
can be adopted as a valid and reliable clinical measure for
the risk assessment of violence in psychiatric services. It
also revealed that violence of inpatients with schizophre-
nia over the course of hospitalization can be fairly well
predicted by their history of violence, especially for the
corresponding types of violence. Risk factors other than a
violent history should also be considered. The findings of
this study with a sample group of schizophrenics could
serve as a reference for further comparative studies on
psychiatric inpatients with other disorders.
This study provides the empirical data to advocate for
risk assessment for violence as part of the admission routine.
Identification of a high risk for specific types of violence will
alert the clinical team to monitor behavior more closely and
implement an efficient strategy of medication and psycho-
social intervention to manage that risk.
Conflict of interest
None.
Funding
This research was supported by the National Science
Council of Taiwan (NSC 91-2314-B-002-328, NSC 93-2314-
B-002-297).
Ethical approval
This study was approved by the internal review board
of the National Taiwan University Hospital.
Acknowledgment
This study was supported by the research grants from
the National Science Council of Taiwan.
 S.-C. Chen et al. / International Journal of Nursing Studies 51 (2014) 198–207 205

Appendix A
206 S.-C. Chen et al. / International Journal of Nursing Studies 51 (2014) 198–207
 S.-C. Chen et al. / International Journal of Nursing Studies 51 (2014) 198–207
References
American Psychiatric Association, 1994. Diagnostic and Statistical Man-
ual of Mental Disorder: DSM IV, 4th ed. Washington, DC: Author.
Barratt, E.S., Slaughter, L., 1998. Defining, measuring, and predicting
impulsive aggression: a heuristic model. Behavioral Sciences and
the Law 16, 285–302.
Bo, S., Abu-Akel, A., Kongerslev, M., Haahr, U.H., Simonsen, E., 2011. Risk
factors for violence among patients with schizophrenia. Clinical
Psychology Review 31 (5) 711–726.
Bobes, J., Fillat, O., Arango, C., 2009. Violence among schizophrenia out-
patients compliant with medication: prevalence and associated fac-
tors. Acta Psychiatrica Scandinavica 119, 218–225.
Buchanan, A., 2008. Risk of violence by psychiatric patients: beyond the
‘‘actuarial versus clinical’’ assessment debate. Psychiatric Services 59
(2) 184–190.
Chen, S.C., Hwu, H.G., 2009. Psychometric evaluation of Chinese version of
Violence Scale for objective rating among inpatients with schizophre-
nia. Journal of Clinical Nursing 18 (13) 1889–1896.
Chen, S.C., Hwu, H.G., Hsiung, P.C., 2011. Clinical manifestations of
aggressive acts by schizophrenic inpatients: a prospective study.
Perspectives in Psychiatric Care 47 (3) 110–116.
Dolan, M., Fullam, R., Logan, C., Davies, G., 2008. The Violence Risk Scale
Second edition (VRS-2) as a predictor of institutional violence in a
British forensic inpatient sample. Psychiatry Research 158, 55–65.
Douglas, K.S., Ogloff, J.R.P., Hart, S.D., 2003. Evaluation of a model of
violence risk assessment among forensic psychiatric patients. Psy-
chiatric Services 54 (10) 1327–1379.
Douglas, K.S., Ogloff, J.R.P., Nicholls, T.L., Grant, I., 1999. Assessing risk for
violence among psychiatric patients: the HCR-20 violence risk as-
sessment scheme and the Psychopathy Checklist: Screening Version.
Journal of Consulting and Clinical Psychology 67 (6) 917–930.
Foster, C., Bowers, L., Nijman, H., 2007. Aggressive behaviour on acute
wards: prevalence, severity and management. Journal of Advanced
Nursing 58 (2) 140–149.
Fresán, A., Apiquian, R., Nicolini, H., Cervantes, J.J., 2007. Temperament
and character in violent schizophrenic patients. Schizophrenia Re-
search 94 (1) 74–80.
Hartvig, P., Alfarnes, S.A., Skjonberg, M., Moger, T.A., Ostberg, B., 2006.
Brief checklists for assessing violence risk among patients discharged
from acute psychiatric facilities: a preliminary study. Nordic Journal
of Psychiatry 60 (3) 243–248.
Hatcher, S., Butler, R., Oakley-Browne, M., 2005. Critical appraisal of
psychiatric rating scales, diagnostic tests, and clinical decision rules.
In: Evidence-based Mental Health Care. Elsevier, Churchill Living-
stone, pp. 114–115 (Chapter 9).
Josefsson, K., Sonde, L., Robins-Wahlin, T.B., 2007. Violence in municipal
care of older people in Sweden as perceived by registered nurses.
Journal of Clinical Nursing 16, 900–910.
Lanza, M.L., 1996. The relationship of behavioral cues to assaultive
behavior. Clinical Nursing Research 5 (1) 6–27.
207
Mayne, T.J., Ambrose, T.K., 1999. Research review on anger in psycho-
therapy. Journal of Clinical Psychology 55, 353–363.
McNiel, D.E., Binder, R.L., 1989. Relationship between preadmission
threats and later violent behavior by acute psychiatric inpatients.
Hospital and Community Psychiatry 40 (6) 605–608.
Monahan, J., Steadman, H.J., Silver, E., Appelbaum, P.S., Robbins, P.C.,
Mulvey, E.P., Roth, L.H., Grisso, T., Banks, S., 2001. Testing clinical
risk factors. In: Rethinking Risk Assessment: The MacArthur Study of
Mental Disorder and Violence. Oxford University Press, Madison
Avenue, NY, pp. 61–90, 145–161 (Chapter 4).
Morrison, E.F., 1988. Instrumentation issues in the measurement of
violence in psychiatric inpatients. Issues in Mental Health Nursing
9 (1) 9–16.
Morrison, E.F., 1993. The measurement of aggression and violence in
hospitalized psychiatric patients. International Journal of Nursing
Studies 30 (1) 51–64.
Plutchik, R., 1995. Outward and inward directed aggressiveness: the
interaction between violence and suicidality. Pharmacopsychiatry
2 (28) 47–57.
Plutchik, R., van Praag, H.M., Conte, H.R., 1989. Correlates of suicide and
violence risk: III. A two-stage model of countervailing forces. Psychi-
atry Research 28, 215–225.
Rice, M., Harris, G.T., 1995. Violent recidivism: assessing predictive
validity. Journal of Consulting and Clinical Psychology 63,
737–748.
Roaldset, J.O., Bjorkly, S., 2010. Patients’ own statements of their future
risk for violent and self-harm behaviour: a prospective inpatient and
post-discharge follow-up study in an acute psychiatric unit. Psychia-
try Research 178 (1) 153–159.
Suchy, Y., Bolger, J., 1999. The Behavioral Dyscontrol Scale as a predictor of
aggression against self or others in psychogeriatric inpatients. Clinical
Neuropsychologist 13 (4) 487–494.
Tam, E., Engelsmann, F., Fugere, R., 1996. Patterns of violent incidents by
patients in a general hospital psychiatric facility. Psychiatric Services
47 (1) 86–88.
Tardiff, K., Marzuk, P., Leon, A., Portera, L., 1997. A prospective study of
violence by psychiatric patients after hospital discharge. Psychiatric
Services 48, 678–681.
Torrey, E.F., 2006. Violence and schizophrenia. Schizophrenia Research
88, 3–4.
Volavka, J., Citrome, L., 2008. Heterogeneity of violence in schizophrenia
and implications for long-term treatment. International Journal of
Clinical Practice 62 (8) 1237–1245.
Walsh, E., Gilvarry, C., Samele, C., Harvey, K., Manley, C., Tattan, T., Tyrer,
P., Creed, F., Murray, R., Fahy, T., 2004. Predicting violence in
schizophrenia: a prospective study. Schizophrenia Research 67,
247–252.
Yen, C.F., Yeh, M.L., Chen, C.S., Chung, H.H., 2002. Predictive value of
insight for suicide, violence, hospitalization, and social adjustment for
outpatients with schizophrenia: a prospective study. Comprehensive
Psychiatry 43 (6) 443–447.