Schizophrenia Research 160 (2014) 27–32

Contents lists available at ScienceDirect

Schizophrenia Research
journal homepage: www.elsevier.com/locate/schres

Factors associated with violence among Japanese patients with

schizophrenia prior to psychiatric emergency hospitalization:
A case-controlled study
Atsushi Imai a,b,⁎, Naoki Hayashi c, Akihiro Shiina d, Noriko Sakikawa a, Yoshito Igarashi b
a
Department of Psychiatry, Tokyo Metropolitan Matsuzawa Hospital, Tokyo, Japan
b
Division of Law and Psychiatry, Center for Forensic Mental Health, Chiba University, Chiba, Japan
c
Department of Psychiatry, Teikyo University School of Medicine, Tokyo Japan
d
Department of Psychiatry, Chiba University Hospital, Chiba, Japan

a r t i c l e i n f o a b s t r a c t

Article history: Background: Numerous studies have examined a wide range of risk factors associated with violence among pa-
Received 17 June 2014 tients with schizophrenia. However, risk factors linked to different socio-cultural backgrounds remain unclear.
Received in revised form 9 September 2014 Objective: The objectives were to investigate factors associated with violence among Japanese patients with
Accepted 7 October 2014 schizophrenia prior to emergency hospitalizations and to compare them with factors found in studies on other
Available online 10 November 2014
populations.
Methods: We recruited 420 Japanese patients with schizophrenia who had committed violent acts immediately
Keywords:
Violence
prior to emergency admission to a psychiatric hospital in Tokyo, during the period 1986 to 2005. Cases were com-
Schizophrenia pared with controls (non violent hospitalized patients with schizophrenia) matched for gender, age and admis-
Japanese sion year. All medical records were reviewed retrospectively. Inter-rater reliability tests of assessment were
Case–control study performed. Conditional logistic regression analysis was used to identify factors associated with violence.
Results: The symptoms of gross excitement, prior violence, auditory hallucinations, systematization of delusions,
incoherence of speech, delusions of reference, TCO symptoms, living with others and long duration of illness were
found to be associated with violence. In contrast, antisocial traits such as substance abuse and antisocial episodes
were not recognized as significant violence-associated factors.
Conclusion: Violence among Japanese patients with schizophrenia was strongly associated with elements of
schizophrenia itself, rather than antisocial traits. This study highlighted associated factors for violence among
Japanese patients with schizophrenia which differ distinctly from associated factors in other countries. This result
demonstrates that future studies assessing the risk of violence among patients with schizophrenia need to con-
sider cultural and racial differences in cohorts.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction However, when evaluating these risks it is necessary to be aware of

Great debate surrounds the definition of risk factors associated with
violence among patients with schizophrenia, from both clinical and
forensic perspectives (Monahan and Steadman, 1994; Heilbrun et al.,
2009). Since the 1980s, a number of studies have set about trying to eval-
uate risk factors in this cohort of patients. Identified factors include,
psychopathological symptoms such as TCO (Threat/Control-Override)
symptoms, command hallucinations and persecutory delusions, comor-
bidity of psychiatric problems such as, psychopathic traits, substance
abuse, and mentalizing disabilities (Bo et al., 2011).
⁎ Corresponding author at: Department of Psychiatry, Tokyo Metropolitan Matsuzawa
Hospital, 2-1-1 Kamikitazawa, Setagaya-ku, Tokyo 156-0057, Japan. Tel.: +81 3 3303
7211; fax: +81 3 3229 7586.
E-mail address: atsushi.imai0518@gmail.com (A. Imai).
how the substantial differences in basic statistics for crime and violent
incidences vary from country to country (United Nations Office on
Drugs and Crime, 2013). These confounding factors make generaliza-
tions difficult and fuel the need for research into country specific risk as-
sessments in schizophrenia (Monahan et al., 2001). This is particularly
relevant in Japan which has a very low crime rate (Tsushima, 1996).
From previous studies, it remains controversial as to whether violence
in patients with schizophrenia is linked to socio-cultural factors (Coid,
1983; Large et al., 2008). We therefore hypothesized that associated fac-
tors for violence among Japanese patients with schizophrenia may differ
from those in countries where the incidence rate of violence, antisocial
conduct, and drug use are much higher.
Currently, there is little research addressing risk factors linked to
violence among Japanese patients with schizophrenia. In the one
study identified risk factors were gender, antisocial behavior before

http://dx.doi.org/10.1016/j.schres.2014.10.016
0920-9964/© 2014 Elsevier B.V. All rights reserved.
28 A. Imai et al. / Schizophrenia Research 160 (2014) 27–32
the onset of schizophrenia, alcoholism and systematic delusions
(Yoshikawa, 1995). However, this study was limited to recidivism and
homicides, providing little insight into the wider range of risk factors.
In this study, we aimed to obtain comprehensive data on the fac-
tors associated with violence among Japanese patients with schizo-
phrenia. We performed a case–control study, comparing clinical
characteristics of patients with schizophrenia who had committed
acts of violence immediately prior to emergency hospitalization,
with gender, age and admission year-matched, non-violent patients
with schizophrenia. Then, we attempted to extract characteristics
associated with violence and to compare and contrast these traits
with international findings.
2. Methods
2.1. Study design: identification of cases and controls
All subjects in this study were recruited from patients hospitalized at
the psychiatric emergency unit of Tokyo Metropolitan Matsuzawa Hos-
pital (TMMH). The principal task of this unit is to accept patients in need
of emergency psychiatric admission, out of normal working hours (hol-
idays/night-times), in a catchment area of 3.4 to 3.8 million inhabitants.
During the study period spanning the years 1986 to 2005, the unit ad-
mitted between 254 and 521 patients per year. These patients were
characterized by severe psychiatric symptoms and problematic behav-
ior such as suicidal attempts and violence. Approximately 80% of admis-
sions were accompanied by police officers, due to fears of imminent
self-inflicted danger or danger to others.
The violence examined in this study included physical violence, sex-
ual violence and weapon use or threat with weapon. Behavior such as
verbal hostility, self-harm, and damaging property were not included.
This definition of violence is consistent with that used in the MacArthur
violent risk assessment study (Monahan et al., 2001). Additionally,
arson was included as an act of violence in this research, since it is con-
sidered a serious offense against others in Japan (Nakatani et al., 2010).
The description of physical aggression items in the Modified Overt Ag-
gression Scale (MOAS) (Kay et al, 1988) was applied to identify physical
violence. In MOAS, physical aggression was defined as violent action
intended to inflict pain, bodily harm, or death upon another, and was
scored on a 0–4 point scale, with 0 denoting no physical aggression; 1,
menacing gestures, swinging at people or grabbing at clothing; 2, strik-
ing, kicking, pushing, scratching or pulling the hair of others (without
injury); 3, attacking others, causing mild injury (bruises, sprains,
welts, etc.); 4, attacking others, causing serious injury (fracture, loss of
teeth, deep cuts, loss of consciousness). The violence scored at 2 or
over was counted as positive. Arson, sexual violence, and weapon use
or threat with weapon use, were evaluated as dichotomous variables.
The inclusion criteria for subjects were as follows. (1) principal dis-
charge diagnosis of schizophrenia with an ICD-9 code of 295 (excluding
295.4, 295.5 and 295.7) and with an ICD-10 code of F20, (2) aged be-
tween 18 and 60 years at admission, (3) a hospital stay of one week
or longer, (4) native Japanese language speaker, (5) violence committed
within 24 h prior to admission, (6) violence meets the definitions set
out in the study.
For this study, extraction of cases was conducted using the inclusion
criteria set out in points (1)–(6). We extracted cases and controls as fol-
low: First, we identified all patients with an emergency admission to
TMMH, between the years 1986 to 2005. Second, we excluded patients
who did not meet criteria (1), (2), (3) and (4). Third, we divided the se-
lected patients into cases who met criterion (5) (violent group) and
controls (non-violent group). Fourth, from the violent group, patients
who did not meet criterion (6) were excluded. Fifth, from non-violent
group, we randomly matched one control with each case using the fol-
lowing criteria: gender, age (with maximum age difference of 3 years),
and year range of admission (1986–1995 or 1996–2005) (Fig. 1).
2.2. Assessments and data collection
Items assessed on the basis of a retrospective review of medical re-
cords were as follows.
(i) Psychiatric symptoms
Psychiatric symptoms were assessed principally by descriptions
in the Present State Examination (PSE) (Wing et al., 1974). We
used PSE as a definition of psychiatric symptom evaluation from
medical records. The first group of symptoms were those identi-
fied as factors associated with violence in previous studies, name-
ly: TCO symptoms, auditory hallucinations (PSE Items 60–62),
command hallucinations (PSE Item 63), delusions of persecution
(PSE Item 74), systematization of delusions (PSE Item 93), gross
excitement (PSE Item 112) and depressed mood (PSE Item 23)
(Link and Stueve, 1994; McNiel, 1994; Yoshikawa, 1995; McNiel
et al., 2000; Swanson et al., 2006; Soyka et al., 2007; Fullam and
Dolan, 2008). In addition, typical symptoms seen in psychiatric
emergency patients were also included in a second group of
symptoms, such as delusions of reference (PSE Item 72), expan-
sive delusions (PSE Item 75), incoherence of speech (PSE Item
136), catatonic movement (PSE Item 119) and expansive mood
and ideation (PSE Items 41–42). When PSE items scored between
1 and 3, the symptom was rated as positive. In the assessment of
TCO symptoms, the definition used by the MacArthur study
(Appelbaum et al., 2000) was applied, and each symptom was
evaluated as a dichotomous variable.
(ii) Additional clinical characteristics
The presence or absence of antisocial episodes prior to illness
onset (antisocial episodes were denoted as follows: violence
as defined in this study, a criminal record, participation and
interchange with an anti-social group and use of illegal
drugs), a family history of substance abuse, schizophrenia, or
other psychiatric disorders, and disorders of unknown diag-
nosis, episodes of substance abuse, presence or absence of a
history of psychiatric treatment, presence or absence of previ-
ous hospital admissions, first episode or recurrence, duration
of illness (under 1 month, 1 month–1 year, 1 year–10 years,
over 10 years), presence or absence of prior violence (prior vi-
olence was denoted as: violent acts as defined in this study,
whether the violence occurred before and after disease onset
did not matter) and finally, the presence or absence of
autoaggression. Autoaggression was evaluated by MOAS.
In MOAS, autoaggression was defined as physical injury
toward oneself, such as self-mutilation or suicide attempts.
It was scored on a 0–4 point scale, with 0 denoting no
autoaggression; 1, picking or scratching of skin, pulling out
of hair, hitting oneself (without injury); 2, banging one's
head, hitting fists on walls, throwing oneself on the floor; 3,
inflicting minor cuts, bruises, burns, or welts on oneself; 4,
inflicting major injury on oneself or a suicide attempt.
Autoaggression scored at 2 or over was counted as positive.
(iii) Socio-demographic characteristics
Gender, age, year of admission, divorced or not divorced parents,
highest educational attainment (junior high school, high school,
vocational school, university/junior college), marital history (mar-
ried or unmarried), divorce history (having ever been divorced),
marital status at the time of admission (married or unmarried),
history of employment (having ever been employed), employ-
ment at the time of admission (employed or unemployed), and
living with others or living alone.
(iv) Descriptions of the violence
Types of violence (physical violence, arson, sexual violence, and
weapon use or threat with weapon), victims of violence (family
members, strangers, police officers/rescue workers, friends/ac-
quaintances, other individuals), location of violence (subject's
 A. Imai et al. / Schizophrenia Research 160 (2014) 27–32
Patients admitted as an emergency to Tokyo
Metropolitan Matsuzawa hospital during the period
1986 to 2005
Patients with schizophrenia
(n=1593)
Patients
who had committed violent acts 24
hours prior to admission
(n=504)
E xcluded
・violent acts that did not
fulfill study criteria (n=84)
Recruited as cases (n=420)
Fig. 1. Flowchart showing the recruitment procedure for this study.
home, street/outdoors, other residence, hospital/outpatient clinic,
other) and physical aggression score in MOAS.
The data of cases and controls were retrospectively extracted from
medical records by one of the authors, with extensive experience in
the evaluation of forensic psychiatry (A. I.). Medical records took the
form of standardized proforma to record clinical information. Medical
records were accrued from a number of different data sources including
patient's self-reporting, collateral reports and official records. We have
compiled a manual of operationally defined criteria for reference. A
rater couldn't be blinded to outcome.
Inter-rater reliability for variables except matching variables
was conducted independently on 33 cases, by two psychiatric raters
(A. I. and N. S., each with greater than nine years of clinical psychiatric
experience). They each coded data independently. Assessments show-
ing very poor reliability (κ b 0.2) were excluded from subsequent
analyses.
2.3. Statistical analyses
Of the variables that showed acceptable reliability, we excluded (iv)
four descriptions of the violence variables found only in the violent
group. Conditional logistic regression analysis was performed since we
were matching cases and controls.
Firstly, univariate conditional logistic regression analyses were car-
ried out to assess the effect of individual factors on violence. Only vari-
ables from univariate analysis where p was less than 0.05 were
entered into the multivariable model. Forward stepwise multivariate
conditional logistic regression analysis was then performed to evaluate
the associated power of related variables.
A p value of less than 0.05 was considered statistically significant. Pa-
tients for whom relevant data were missing were excluded. All statisti-
cal analyses were performed using the IBM SPSS statistics for Mac,
version 22.0 (SPSS Inc., Chicago, IL, USA).
2.4. Ethical considerations
This study was approved by the ethical review boards of Tokyo Met-
ropolitan Matsuzawa Hospital, and Chiba University.
3. Results
3.1. Clinical characteristics of the subjects
We identified 8 048 patients who were admitted to the psychiatric
emergency unit of TMMH during the study period. We selected 420
29
Excluded (n=6455)
・Patients with other diagnoses (n=3939)
・Patients aged ＜18 or > 60 (n=202)
・Patients admitted within7 days of
screening (n=2107)
・Non native Japanese speakers(n=207)
Patients
who had not committed violent acts 24
hours prior to admission
(n=1089)
Matched for gender, age
and year of admission.
Recruited as controls (n=420)
cases, and diagnosis, gender, age and admission period matched, con-
trols (Fig. 1).
Recruits consisted of 318 males and 102 females, with an average
age (SD) of 33.3 (10.6) and 33.7 (10.4), 200 admitted between 1986
and 1995 and 220 admitted between 1986 and 2005, respectively.
3.2. Characteristics of violence
We documented a total of 439 violent incidents. The breakdown was
as follows: 296 incidents with a MOAS physical aggression score of 2
points (67%) and 72 with a score of 3 points (16%), 56 acts using a weap-
on or a threat with a weapon (13%), 10 of arson (2%) and 5 of sexual as-
sault (1%). The victims of violence totaled 422 individuals, with 247
(59%) family members, 89 (21%) strangers, 46 (11%) police officers/res-
cue workers, 28 (7%) friends/acquaintances and 8 (2%) other individ-
uals. Violence occurred in a total of 421 places; 244 incidents (58%) in
the subject's home, 73 (17%) in the street or outdoors, 20 (5%) at a res-
idence other than that of the subject, 18 (4%) in a hospital or outpatients
clinic, 66 (16%) other locations.
3.3. Inter-rater reliability tests
Out of a total of 39 variables, gender, age and year of admission were
excluded, leaving 36 variables for inter-rater reliability testing. Of these,
18 variables (50%) showed excellent reliability (κ N 0.8), 10 (28%)
showed very good reliability (0.6 b κ b 0.8), 6 (17%) moderate reliability
(0.4 b κ b 0.6), and 2 (6%) (delusions of persecution, gross excitement)
showed poor reliability (0.2 b κ b 0.4). No variables showed very poor
reliability (κ b 0.2), and so none were excluded based on inter-rater re-
liability testing.
3.4. Univariate conditional logistic regression analyses
Univariate conditional logistic regression analyses were performed
for 32 variables.
Tables 1−3 show the results of univariate conditional logistic re-
gression analyses. In (i) psychiatric symptoms, TCO symptoms, auditory
hallucinations, delusions of persecution, systematization of delusions,
delusions of reference, gross excitement, and incoherence of speech
displayed positive association with violence. Additionally, depressed
mood was inversely associated with violence. In (ii) additional clinical
characteristics, such as family history of schizophrenia, history of
hospital admissions, disease recurrence, long duration of illness, and
prior violence were identified as associated factors for violence. In
converse, autoaggression was inversely associated with violence.
30 A. Imai et al. / Schizophrenia Research 160 (2014) 27–32

Table 1
Psychiatric symptoms: univariate conditional logistic regression analysis.

Violent Non violent

Variable n % n % Odd ratio (95% Cl) P value

TCOa symptoms 199/420 (47) 118/420 (28) 2.31 (1.71–3.11) b0.001

Auditory hallucinations 227/420 (54) 170/420 (41) 1.83 (1.36–2.45) b0.001
Command hallucinations 62/420 (15) 47/420 (11) 1.38 (0.92–2.07) 0.1
Delusions of persecution 312/420 (74) 226/420 (54) 2.51 (1.85–3.41) b0.001
Systematization of delusions 85/420 (20) 27/420 (5) 3.76 (2.33–6.09) b0.001
Gross excitement 387/420 (92) 197/420 (47) 12.9 (7.74–21.4) b0.001
Depressed mood 4/420 (1) 37/420 (9) 0.11 (0.04–0.30) b0.001
Delusions of reference 157/420 (37) 99/420 (24) 1.84 (1.37–2.47) b0.001
Expansive delusions 65/420 (16) 59/420 (14) 1.12 (0.77–1.62) 0.6
Incoherence of speech 237/420 (56) 120/420 (29) 3.25 (2.38–4.44) b0.001
Catatonic movement 51/420 (12) 62/420 (15) 0.80 (0.54–1.19) 0.3
Expansive mood and ideation 29/420 (7) 17/420 (4) 1.80 (0.96–3.38) 0.07

Note. CI = confidence interval.

a
TCO = Treat/Control − Override.

Finally, in (iii) the socio-demographic characteristics, living with others
was identified as a positively associated factor of violence. Being
employed at the time of admission was inversely associated with
violence.
3.5. Multivariate conditional logistic regression analysis
Table 4 shows the final multivariate conditional logistic regression
model. This model identified nine independent factors associated with
violence. They were gross excitement, prior violence, auditory halluci-
nations, systematization of delusions, incoherence of speech, delusions
of reference, TCO symptoms, living with others and long duration of ill-
ness. Autoaggressoion remained inversely associated with violence.
4. Discussion
Most of the associated risk factors identified in this study were con-
sistent with reports of previous studies on risk factors for violence in
Caucasian patients with schizophrenia.
Although findings from prior studies vary, TCO symptoms (Link and
Stueve, 1994), auditory hallucinations (McNiel, 1994), systematization
of delusions (Yoshikawa, 1995), gross excitement (Fullam and Dolan,
2008), living with others (Swanson et al, 2008) and prior violence
(Walsh et al., 2004) were found to be significant predictors of violence.
Our study confirms these relationships. There is debate as to whether
depressed mood is related to violence in patients with schizophrenia
(Soyka et al., 2007; Fazel et al., 2010; Krakowski and Czobor, 2013). Uni-
variate conditional logistic regression analysis detected an inverse rela-
tionship between depression and the risk of violence. However,
multivariate conditional logistic regression analysis was unable to de-
tect this relationship. Autoaggression showed inverse association with
violence in the final multivariate conditional logistic regression model.
As depressed mood strongly correlated with autoaggression, the associ-
ation between depressed mood and violence could disappear. In this
study, we identified incoherent speech and delusions of reference as
novel factors associated with violence.
Some of our results show important differences from those obtained
in previous risk factor association studies. While other studies reported
substance abuse and antisocial behavior prior to illness onset as risk fac-
tors, very significantly (Räsänen et al., 1998; Fullam and Dolan, 2008),
this was not the case in our Japanese cohort.
We put forward two main reasons why an association between a
history of substance abuse and violence was not found. First, we consid-
er the remarkably low rate of drug use in Japan, compared with previ-
ously studied countries to be a major factor (United Nation, 2012). For
instance, while the United States has the second highest seizure rate
for cocaine in the world (24%), Japan falls below the rankings threshold
(under 0.4%). Since the ratio of substance abuse was low (7–8%) in this
study, it could also be that the study lacked statistical power to detect
any significant differences in this area.

Table 2
Additional clinical characteristics: Univariate conditional logistic regression analysis.

Violent Nonviolent

Variable n % n % Odd ratio (95% Cl) P value

Antisocial episodes prior to illness onset 19/418 (5) 36/420 (9) 1.94 (1.08–3.49) 0.03
Family history of substance abuse 10/420 (2) 7/416 (2) 0.44 (0.53–4.21) 0.4
Family history of schizophrenia 65/419 (16) 41/416 (10) 1.73 (1.13–2.65) 0.01
Family history of other psychiatric disorders 15/420 (4) 16/416 (4) 0.93 (0.45–1.93) 0.9
Family history of unknown diagnosis 16/420 (4) 19/416 (5) 0.78 (0.39–1.56) 0.5
Episodes of substance abuse 33/420 (8) 31/420 (7) 1.07 (0.64–1.80) 0.79
History of psychiatric treatment 266/420 (63) 240/419 (57) 1.28 (0.97–1.70) 0.09
Previous hospital admissions 208/420 (50) 173/419 (41) 1.45 (1.09–1.93) 0.01
Episode 157/416 (38) 182/416 (44) 1.35 (1.00–1.80) 0.04
First episode 259/416 (61) 234/416 (56)
Recurrence
Duration of illness 29/419 (7) 54/418 (13) 1.43 (1.20–1.70) b0.001
Under 1 month 59/419 (14) 79/418 (19)
1 month–1 year 191/419 (46) 169/418 (40)
1 year–10 years 140/419 (33) 116/418 (28)
Over 10 years
Prior violence 153/419 (37) 48/420 (11) 4.4 (2.97–6.48) b0.001
Autoaggression 5/420 (1) 52/420 (12) 0.08 (0.03–0.21) b0.001

Note. CI = confidence interval.

 A. Imai et al. / Schizophrenia Research 160 (2014) 27–32 31

Table 3
Socio-demographic characteristics: univariate conditional logistic regression analysis.

Violent Non violent

Variable n % n % Odd ratio (95% Cl) P value

Divorced parents 51/420 (12) 66/418 (16) 0.75 (0.51–1.10) 0.1

Highest educational attainment
Junior high school
High school 100/418 (24) 71/416 (17)
Vocational school 189/418 (45) 189/416 (45) 0.89 (0.79–1.01) 0.07
University/Junior college 20/418 (48) 21/416 (5)
107/418 (26) 129/416 (31)
Marital history 57/420 (14) 69/419 (16) 0.77 (0.51–1.16) 0.2
Divorce history 27/420 (6) 37/419 (9) 0.71 (0.42–1.19) 0.2
Marital status at the time of admission 22/420 (5) 27/419 (6) 0.74 (0.43–1.28) 0.3
History of employment 301/419 (72) 315/413 (76) 0.75 (0.54–1.06) 0.1
Employment at the time of admission 93/420 (22) 124/420 (30) 0.65 (0.47–0.91) 0.01
Living with others 299/420 (71) 225/420 (54) 2.20 (1.62–2.96) b0.001

Note. CI = confidence interval.

The second important factor is one of cultural differences in diagno- hallucinations may increase the risk of violence. However, the lack of as-
sis. For instance, in ICD-10, there is no clear protocol as to whether pa- sociation between auditory hallucinations and violence may simply be a
tients with psychotic symptoms that persist after six months or more result of the low prevalence of this symptom in our study.
of substance use should be diagnosed with schizophrenia (F20) or Re- In addition, we found a contrasting result regarding disease duration.
sidual and late-onset psychotic disorder (F1x.7) (World Health Previous studies reported that the association between violence and psy-
Organization, 1992). In Japan, patients with psychosis caused by drugs choses appeared most prominent during first-episode psychosis, rather
or alcohol, lasting more than six months tend to be diagnosed as than in later stages of the illness (Large and Nielssen, 2011). Here, we
F1x.7. In contrast, these cases tend to be diagnosed as F20 in the coun- found that violence was linked to longer disease duration. This result
tries where previous studies were performed (Sato and Matsumoto, suggests that we should be careful regardless of stages of illness when
1999). It is therefore possible that high risk patients with schizophrenia assessing violence among Japanese patients with schizophrenia.
comorbid with substance abuse who would have been diagnosed as F20 Finally, we found that autoaggression remained inversely associated
in previously studied cohorts, would be excluded in this study. with violence. This result differed from the findings of a previous study
Next, we considered the reason why we were unable to detect a re- (Witt et al., 2014). A possible explanation would be that violence and
lationship between violence and antisocial episodes prior to illness autoaggression are distinctly different types of aggression and could
onset. One report suggests that Japanese have a lower prevalence of fac- therefore share an inverse relationship.
tors that constitute psychopathic personality traits, compared with Cau-
casians (Lynn, 2002). Second, the factorial structure of the psychopathic 5. Limitations
personality differs between Japan and the United States (Yokota, 2012).
These findings highlight the possibility that the prevalence and nature Though our findings have a number of important implications, they
of psychopathic personality traits are different between Japanese and should be treated as a preliminary hypothesis. Since this was a retro-
Caucasians. These variations may alter the elements that affect the risk spective study using medical records, causal or even temporal associa-
of violence. tion between psychopathology and violent acts are undetermined
Even though univariate conditional logistic regression analysis de- (Nolan et al., 2003). Also, it remains equivocal whether the differences
tected a relationship between violence and antisocial episodes prior to elicited in this study were derived from socio-cultural factors, or, other
illness onset, multivariate conditional logistic regression did not. If anti- as yet undetermined factors. Future studies applying different method-
social episodes after illness onset were investigated, they might be more ology will be needed to confirm our findings.
strongly related to violence, compared with antisocial episodes prior to Care should be taken when generalizing the findings of this study,
illness onset. Additional studies are clearly needed to clarify this point. because of possible sampling bias. For instance, Controls subjects
Although there is no consensus on including command hallucina- displaying autoaggression comprised 12% of participants, but this figure
tions as a risk factor for violence (Cheung et al., 1997; McNiel et al., fell to one percent in cases. In addition, we excluded patients who
2000), this study found no relationship between these variables. Con- stayed in hospital for shorter than one week because of poor informa-
sidering that whole auditory hallucinations were identified as a risk fac- tion accrual. This may well have excluded patients with schizophrenia
tor in the present study, auditory hallucinations, but not command who were violent for reasons other than their mental illness. Finally,
cases and controls might not be fully representative of the wider popu-
Table 4 lation of people with schizophrenia.
Results of final forward stepwise multivariate conditional logistic regression analysis. The violence examined in this study was predominately point 2 on
Variable B Wald df Odd ratio (95% CI) P value
the MOAS aggression scale (67%). Therefore, this study would not accu-
rately capture factors that influence the risk for extreme violence. Care
Gross excitement 2.42 52.1 1 11.2 (5.82–21.6) b0.001
should therefore be exercised when extending these findings to forensic
Prior violence 1.46 22.6 1 4.31 (2.36–7.88) b0.001
Auditory hallucinations 1.02 13.7 1 2.76 (1.61–4.73) b0.001 settings with significantly more serious acts of violence, such as murder.
Systematization of delusions 0.91 5.06 1 2.48 (1.12–5.48) 0.02 Here we compared a total of 32 variables and set the p value at 0.05.
Incoherence of speech 0.90 12.9 1 2.45 (1.50–4.00) b0.001 These parameters could lead to one or two false positive findings by
Delusions of reference 0.86 8.73 1 2.37 (1.34–4.21) 0.003
type I errors.
TCO symptoms 0.72 7.02 1 2.05 (1.21–3.47) 0.008
Living with others 0.62 5.49 1 1.86 (1.11–3.11) 0.02
Duration of illness 0.43 7.93 1 1.54 (1.14–2.08) 0.005 6. Conclusion
Autoaggression −2.75 13.9 1 0.06 (0.02–0.27) b0.001

Note. Multivariate model created using 96.3% (809/840) of the whole sample. CI = - Our data suggested that violence among Japanese patients with
confidence interval. schizophrenia was strongly associated with elements of schizophrenia
32 A. Imai et al. / Schizophrenia Research 160 (2014) 27–32
itself, such as long disease duration and psychotic symptoms, rather
than antisocial traits, such as substance abuse and antisocial episodes.
This marks a difference in the associated factor model of violence
among patients with schizophrenia between the Caucasian model and
the model as practiced in clinical psychiatry field in Japan. This study
cautions us on the need for careful interpretation of findings when we
assessing violence in patients with schizophrenia, with relation to cul-
tural and ethnic differences between populations. More research is
needed to confirm and extend these preliminary findings.
Role of the funding source
The study was unfunded.
Contributors
Dr. A. Imai designed the study, undertook the statistical analysis and interpreted the
data. Drs. Shiina, and Igarashi designed the study. Drs. Imai and Hayashi collected the
data. Dr. Sakikawa coded the data for inter-rater reliability test. Dr. Imai wrote the first
draft of the manuscript. Drs. Hayashi, Shiina, Sakikawa and Igarashi critically revised the
first draft. All authors contributed to and have approved the final manuscript.
Conflict of interest
The authors have no conflict of interest.
Acknowledgments
We thank Dr. Masanari Itokawa for his helpful advice in the manuscript and Mr.
Shigeo Ohsawa for his assistance in randomization on extracted controls.
References
Appelbaum, P.S., Robbins, P.C., Monahan, J., 2000. Violence and delusions: data from the
MacArthur Violence Risk Assessment Study. Am. J. Psychiatry 157 (4), 566–572.
Bo, S., Abu-Akel, A., Kongerslev, M., Haahr, U.H., Simonsen, E., 2011. Risk factors for vio-
lence among patients with schizophrenia. Clin. Psychol. Rev. 31 (5), 711–726.
Cheung, P., Schweitzer, I., Crowley, K., Tuckwell, V., 1997. Violence in schizophrenia: role
of hallucinations and delusions. Schizophr. Res. 26 (2–3), 181–190.
Coid, J., 1983. The epidemiology of abnormal homicide and murder followed by suicide.
Psychol. Med. 1983 (4), 855–860.
Fazel, S., Buxrud, P., Ruchkin, V., Grann, M., 2010. Homicide in discharged patients with
schizophrenia and psychoses: a national case–control study. Schizophr. Res. 123
(2–3), 263–269.
Fullam, R.S., Dolan, M.S., 2008. Executive function and in-patient violence in forensic pa-
tients with schizophrenia. Br. J. Psychiatry 193 (3), 247–253.
Heilbrun, K., Grisso, T., Goldstein, A.M., 2009. Foundations of Forensic Mental Health As-
sessment. Oxford University press, New York.
Kay, R.S., Wolkenfeld, F., Murrill, M.L., 1988. Profiles of aggression among psychiatric pa-
tients. I. Nature and prevalence. J. Nerv. Ment. Dis. 176 (9), 539–546.
Krakowski, M.I., Czobor, P., 2013. Depression and impulsivity as pathways to violence: im-
plications for antiaggressive treatment. Schizophr. Bull. 13. http://dx.doi.org/10.1093/
schbul/sbt117.
Large, M.M., Nielssen, O., 2011. Violence in first-episode psychosis: systematic review and
meta-analysis. Schizophr. Res. 125 (2–3), 209–220.
Large, M., Smith, G., Swinson, N., Shaw, J., Nielssen, O., 2008. Homicide due to mental dis-
order in England and wales over 50 years. Br. J. Psychiatry 193 (2), 130–133.
Link, B.G., Stueve, A., 1994. Psychotic symptoms and the violent/illegal behavior of mental
patients compared to community controls. In: Monahan, J., Steadman, H.J. (Eds.),
Violence and Mental Disorder: Developments in Risk Assessment. University of
Chicago Press, Chicago, pp. 137–160.
Lynn, R., 2002. Racial and ethnic differences in psychopathic personality. Pers. Individ. Dif-
fer. 32, 273–316.
McNiel, D.E., 1994. Hallucinations and Violence. In: Monahan., J., Steadman, H.J. (Eds.), Vi-
olence and Mental Disorder: Developments in Risk Assessment. University of Chicago
Press, Chicago, pp. 183–202.
McNiel, D.E., Eisner, J.P., Binder, R.L., 2000. The relationship between command hallucina-
tions and violence. Psychiatr. Serv. 51 (10), 1288–1292.
Monahan, J., Steadman, H.J., 1994. Violence and Mental Disorder: Developments in Risk
Assessment. University of Chicago Press, Chicago.
Monahan, J., Steadman, H.J., Silver, E., Applebaum, P.S., Robbins, P.C., Mulvey, E.P., Roth,
L.H., Grisso, T., Banks, S., 2001. Rethinking Risk Assessment: The MacArthur Study
of Mental Disorder and Violence. Oxford University Press, Inc., New York.
Nakatani, Y., Kojimoto, M., Matsubara, S., Takayanagi, I., 2010. New legislation for of-
fenders with mental disorders in Japan. Int. J. Law Psychiatry 33 (1), 7–12.
Nolan, K.A., Czobor, P., Roy, B.B., Plat, M.M., Shope, C.B., Citrome, L.L., Volavka, J., 2003.
Characteristics of assaultive behavior among psychiatric inpatients. Psychiatr. Serv.
54 (7), 1012–1016.
Räsänen, P., Tilhonen, J., Isohanni, M., Rantalkallio, P., Lehtonen, J., Moring, J., 1998. Schizo-
phrenia, alcohol abuse, and violent behavior: a 26-year followup study of an unse-
lected birth cohort. Schizophr. Bull. 24 (3), 437–441.
Sato, M., Matsumoto, K., 1999. Kakuseizaiizon to kanrensyougai (syoujyou keika sindan)
(in Japanese) (Dependence syndrome with methamphetamine and methamphet-
amine related disorders). In: Sato, M., Suwaki, H. (Eds.), Rinsyou seisin igaku kouza
8 [Encyclopedia of Clinical Psychiatry 8]. Nakayama syoten, Tokyo, pp. 222–235.
Soyka, M., Graz, C., Bottlender, R., Dirschedl, P., Schoech, H., 2007. Clinical correlates of
later violence and criminal offences in schizophrenia. Scizophr. Res. 94 (1–3), 89–98.
Swanson, J.W., Swartz, M.S., Van Dorn, R.A., Elbogen, E.B., Wagner, H.R., Rosenheck, R.A.,
Stroup, T.S., McEvoy, J.P., Lieberman, J.A., 2006. A national study of violent behavior
in persons with schizophrenia. Arch. Gen. Psychiatry 63 (5), 490–499.
Swanson, J.W., Swartz, M.S., Van Dorn, R.A., Volavka, J., Monahan, J., Stroup, T.S., McEvoy,
J.P., Wangner, H.R., Elbogen, E.B., Lieberman, J.A., 2008. Comparison of antipsychotic
medication effects on reducing violence in people with schizophrenia. Br. J. Psychia-
try 193 (1), 37–43.
Tsushima, M., 1996. Economic structure and crime: the case of Japan. J. Socio Econ. 25 (4),
497–515.
United Nations Office on Drugs and Crime, 2012. World Drug Report 2012, (Accessed
from http://www.unodc.org/unodc/en/data-and-analysis/WDR-2012.html June
2014).
United Nations Office on Drugs and Crime, 2013. UNODC Homicide Statistics 2013,
(Accessed from https://www.unodc.org/gsh/en/data.html June 2014).
Walsh, E., Gilvarry, C., Samele, C., Harvey, K., Manley, C., Tattan, T., Tyrer, P., Creed, F.,
Merray, R., Fahy, T., UK700 Group, 2004. Predicting violence in schizophrenia: a pro-
spective study. Schizophr. Res. 67 (2–3), 247–252.
Wing, J.K., Cooper, J.E., Sartorius, N., 1974. Measurement and Classification of Psychiatric
Symptom: An Instruction Manual for the PSE and CATEGO Program. Cambridge Uni-
versity Press, Cambridge.
Witt, K., Hawton, K., Fazel, S., 2014. The relationship between suicide and violence is
schizophrenia: analysis of the Clinical Antipsychotic Trials of Intervention Effective-
ness (CATIE) dataset. Schizophr. Res. 154 (1–3), 61–67.
World Health Organization, 1992. The ICD-10 Classification of Mental and Behavioral Dis-
orders: Clinical Descriptions and Diagnostic Guidelines. World Health Organization,
Geneva.
Yokota, K., 2012. The validity of a three-factor model in PPI-R and social dominance ori-
entation in Japanese sample. Pers. Individ. Differ. 53, 907–911.
Yoshikawa, K., 1995. Predictive factors and predictability of recidivism in schizophrenic
homicides (in Japanese). Act. Crim. Jpn. 61 (6), 216–234.