Effectiveness of Atypical Antipsychotic
Medications in Reducing Violent Behavior
Among Persons With Schizophrenia in
Community-Based Treatment
by Jeffrey W. Swanson, Marvin S. Swartz, and Eric B. Elbogen
Abstract
This study prospectively compared the effectiveness of
atypical antipsychotic medications to that of conventional
neuroleptics in reducing violent behavior among patients
with schizophrenia under "usual care" conditions in the
community. Participants (n = 229) were adults with schizophrenia spectrum disorders receiving inpatient or outpatient services in publk sector mental health systems in
North Carolina. Subjects were followed for 2 years in an
observational study using multiple methods of data collection at 6-month intervals to assess treatment, sododemographk characteristics, clinical features, and violence
outcomes. Treatment with atypical antipsychotic medications (clozapine,risperidone,or olanzapine) was found to
significantly reduce the risk of violent behavior, whereas
treatment with conventional neuroleptics did not have
this same beneficial effect A cumulative effect on reduced
violence was attributable to consistent compliance with
atypical antipsychotic medications over a 2-year period.
Concurrent reductions in psychotic symptoms, substance
abuse, and adverse medication side effects were found to
mediate the association between adherence with atypicals
and lower violence risk. Treatment with atypical antipsychotic medications should be considered as an important
component of violence risk management for schizophrenia patients at risk for violent behavior.
Keywords: Schizophrenia, atypical antipsychotic
medication, severe mental illness, violence.
Schizophrenia Bulletin, 30(l):3-20,2004.
While the great majority of persons with schizophrenia do
not commit violent acts (Swanson 1994), clinicians recognize that some individuals with this disorder may pose a
significant risk in the community—particularly those
patients with co-occurring substance abuse problems who
are noncompliant with prescribed psychiatric treatment,
and those with a history of frequent relapses resulting in
hospitalization or arrest (Geller 1992; Monahan 1992;
Kent et al. 1995; Schalock et al. 1995; Swanson et al.
1996, 1997, 2000; Borum et al. 1997; Swartz et al. 1998a,
1998&). With clinicians increasingly held liable for the
behavior of patients inadequately treated in the community, concerns about violence risk have increased (Borum
et al. 1996; Cuffel 1997; Dixon et al. 1998; Simon 1998;
Thienhaus and Piasecki 1998). The potential for violence
in even a small proportion of persons with untreated psychotic disorders increases public fear, prevents acceptance
and inclusion of persons with psychiatric disabilities in
society, disrupts continuity of care, and limits the effectiveness of community-based mental health services (Link
et al. 1987; Torrey 1994; Angermeyer and Matschinger
1996; Estroff et al. 1998; Perm and Martin 1998).
Because of the high human cost and legal liability
associated with violent acts, and because the presumption
of dangerousness is a key justification for involuntary
treatment, the assessment of the risk of violence in psychiatric patients is among the most critical judgments
clinicians make (Schopp 1996; Elbogen and Tomkins
2000). Consequently, much of the empirical research on
violence and mental disorder in recent years has been driven by the need to develop more accurate methods of violence prediction and risk assessment (Monahan and
Steadman 1994; Borum 1996; Douglas et al. 1999;
Monahan et al. 2000, 2001). However, an equal emphasis
is now needed on studies to evaluate and improve the
effectiveness of community-based treatment in reducing
violent behavior among persons with severe mental illness (Heilbrun 1997; Swanson et al. 2000).
Violence Reduction With New
Pharmacotherapies for Schizophrenia
For some patients with schizophrenia, the new class of
"atypical" antipsychotic medications—including clozaSend reprint requests to Dr. J. Swanson, Associate Professor of
Psychiatry and Behavioral Sciences, Duke University Medical Center,
Box 3071, 905 W. Main Street, Ste 23A, Durham, NC 27710; e-mail:
[email protected].
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
pine, risperidone, olanzapine, quetiapine, and ziprasidone—has the potential to significantly reduce the risk of
violence, insofar as improved control of psychotic symptomatology may reduce assaultive behavior in these individuals (Buckley 1999). Evidence from controlled trials of
atypical antipsychotic agents indicates that these medications represent a significant improvement over conventional neuroleptics, with equal or improved efficacy in
controlling positive and negative symptoms of schizophrenia, coupled with reduced adverse extrapyramidal symptoms (Marder et al. 1997; Viale et al. 1997; Barnes and
McPhillips 1998; Keck et al. 2000). In recent years, a
growing body of clinical research has also suggested that
the atypical antipsychotic medications might be more efficacious in reducing aggressive and violent behavior in
some patients with schizophrenia (Taylor et al. 1996;
Glazer and Dickson 1998; Senninger and Laxenaire 1998;
Citrome et al. 2001). To the extent that reduction in
adverse extrapyramidal symptoms improves compliance,
these agents may also reduce violent behavior indirectly
by improving patient compliance.
Much of the earlier research has focused on demonstrating the efficacy of clozapine in reducing aggression
(Ratey et al. 1993; Buckley et al. 1995; Menditto et al.
1996; Rabinowitz et al. 1996; Spivak et al. 1997, 1999;
Volavka 1999; Citrome et al. 2001). These studies generally corroborate anecdotal clinical reports that patients
with schizophrenia who take clozapine are less likely to
engage in physical and verbal aggression (Ratey et al.
1993; Rabinowitz et al. 1996) and are less likely to require
restraints and seclusion in hospital settings (Chiles et al.
1994). Patients taking clozapine have been found to obtain
lower hostility scores on the Brief Psychiatric Rating Scale
(BPRS) (Volavka et al. 1993) and lower scores on the
Overt Aggression Scale (Spivak et al. 1997). Research on
patients with dual diagnoses of psychotic and borderline
personality disorders has also shown that after starting
clozapine treatment, these patients were significantly less
likely to assault staff or peers on an inpatient ward
(Chengappa et al. 1999). No studies to date have demonstrated clozapine's effectiveness in reducing violent behavior in longitudinal community-based studies.
A few studies have examined the efficacy of risperidone in reducing violence risk in samples of schizophrenia
patients (see Buckley et al. 1997). In a retrospective study
of 139 subjects with schizophrenia, Czobor et al. (1995)
found that risperidone treatment was associated with
reduced scores on items measuring hostility in the Positive
and Negative Syndrome Scale (PANSS). Similarly, Currier
and Simpson (2001) demonstrated that risperidone liquid
concentrate lessened aggressiveness, evidenced by lower
PANSS hostility item scores, as well as lower scores on
the Clinical Global Impressions scale. Marder et al. (1997)
J.W. Swanson et al.
found similar decreases on the BPRS hostility scale for over
500 patients treated with risperidone. Most recently, electroconvulsive therapy combined with risperidone has been
shown to reduce violent behavior among male inpatients
with schizophrenia (Hirose et al. 2001). One study found
that risperidone did not reduce violence; however, this was a
very small, selected sample—10 subjects on risperidone
compared with 10 matched controls on a traditional neuroleptic, all patients in a maximum-security forensic hospital who had previously been violent (Beck et al. 1997). As
with the body of literature on clozapine, these studies are
limited in their generalizability to real-world conditions
under which community-based treatment is delivered, and
they cannot adequately address the question of whether
risperidone may, over time, actually reduce the interpersonal
violent behavior of persons with schizophrenia.
There have been no empirical studies examining the
effects of olanzapine on violent behavior in patients with
schizophrenia, although it is speculated that because
olanzapine has pharmacological properties similar to the
other atypical antipsychotic agents, it may also be effective
in reducing aggression (Keck et al. 2000). Olanzapine has
been found to reduce aggressive behavior in patients with
Alzheimer's disease (Clark et al. 2001) and in children,
adolescents, and adults with developmental disabilities
(Horrigan et al. 1997; Potenza et al. 1999). Case studies
have also found anti-aggressive effects of olanzapine for
patients with Huntington's disease (Grove et al. 2000).
One case report, however, suggests that olanzapine might
have agitating effects for some patients with schizophrenia
(Johnetal. 1998).
In general, the promise of improved compliance and
reduced violence risk among patients with schizophrenia
would seem to recommend increased use of the atypical
antipsychotic medications in treating these patients.
However, widespread adoption of these new pharmacotherapies is tempered by their significantly higher cost.
For this reason, it becomes important to establish the clinical effectiveness of these drugs under "usual care" conditions in the community—especially for patients with histories of violence who use disproportionate amounts of
high-cost services—and thus to demonstrate that the benefits may translate into cost savings sufficient to justify their
use as first-line therapies (Viale et al. 1997).
A recent randomized trial compared effects of treatment with various atypical antipsychotic medications with
effects of treatment with haloperidol on self-reported hostility in a sample of inpatients with schizophrenia (Citrome
et al. 2001). This study found that only clozapine treatment
was associated with significantly lower hostility scores.
Still, very little is known about the relationship between
actual violent behavior in the community and treatment
with atypical antipsychotic medications. It is possible, for
Atypical Antipsychotics and Reduction of Violent Behavior
example, that clozapine reduces violence by means of a
direct pharmacological effect on hostility and aggression,
while the other atypical agents reduce violence more indirectly, by means of improved compliance and, consequently, diminished substance abuse.
Further research has been needed using improved
mediodologies of assessment, larger and more generalizable samples, and longer periods of observation under community-based treatment Most of the extant research on the
effects of atypical antipsychotic medications consists of
case reports or controlled trials with relatively small sample
sizes carried out in hospital inpatient settings (Citrome et
al. 2001). Regarding the measurement of violence, most
previous studies have employed scales of aggression and
hostility as proxy measures of violence or have considered
assaultive incidents in inpatient settings. Community violence is difficult to assess accurately because it requires
gathering data from multiple sources to corroborate subjects' self-report of violent incidents (Mulvey and Lidz
1993; Swanson et al. 1999). No investigations have
prospectively examined the long-term impact of treatment
with atypical antipsychotic medications on violent behavior
measured directly in the community.
Another limitation of previous research is that, with
few exceptions (Citrome et al. 2001), studies to date have
examined treatment witfi only one or another of the atypical medications; however, it may be theoretically and
methodologically preferable to consider the effects of
these new medications as a group with respect to violence
risk reduction (Keck et al. 2000). First, studying treatment
with only one medication limits sample size, whereas
examining the effects of the atypicals as a class provides
increased statistical power. Second, as psychiatrists
attempt to optimize medication benefits, it is likely that
patients will have their medication regimens changed
often from one atypical agent to another. Therefore, in
observational effectiveness studies, it may be increasingly
difficult to find patients who are "pure types" with respect
to a single medication regimen. Third, although each
atypical agent is unique in its receptor binding and clinical profile, the atypical antipsychotics share some psychopharmacological features, including putative action on
similar neurotransmitter systems (e.g., serotonin) implicated in violent behavior (Lieberman et al. 1998; Duncan
et al. 1999; Elbogen and Huss 2000), and are generally
characterized by few extrapyramidal symptoms (i.e.,
greater tolerability) compared to traditional neuroleptics.
While it appears likely from earlier studies that the
atypicals may exert some direct pharmacological effects
that inhibit aggressive and violent behavior, it is also plausible that the atypicals may lower violence risk by indirect
means, for instance, by improving treatment adherence
and reducing substance abuse, which are risk factors for
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
violence as well as for exacerbated psychotic symptoms.
Better control of psychotic symptoms per se may reduce
violence, insofar as violent behavior, in some schizophrenia patients, is linked to paranoid delusions, auditory hallucinations, and negative affect (Cheung et al. 1997).
Treatment with atypicals may lead to improved compliance due to reduced adverse side effects as well as
reduced substance abuse. Adverse side effects are only one
of many reasons why patients with schizophrenia stop taking prescribed medication (and thus increase their risk for
violence). Substance abuse may reinforce a pattern of nonadherence, insofar as persons dependent upon alcohol and
illicit drugs become either unwilling or unable to comply
with a prescribed regimen of antipysychotic medications.
Olfson and colleagues (2000) found that substance
abuse was the strongest predictor of medication noncompliance in 213 patients with schizophrenia or schizoaffective disorder. Other studies have found that persons with
dual diagnoses have significantly greater psychotic symptomatology than their counterparts without substance
abuse, and yet are far more likely to skip doses or stop
taking prescribed psychotropic medication (Dixon et al.
1999). Continued poor compliance with recommended
medications can lead, in turn, to exacerbated psychiatric
symptoms and overall psychiatric instability, suicidality,
and aggression (Chengappa et al. 2002). The study by
Olfson and colleagues (2000) also reported that patients
who stopped taking antipsychotic medication were more
likely to avoid outpatient treatment, and to experience
symptom relapse, homelessness, emergency room visits,
and rehospitalization. In short, heightened risk of violent
behavior is often intertwined with a complex cluster of
poor outcomes—and there are good reasons to expect that
the new generation of antipsychotic medications may
reduce violence risk by multiple causal pathways. If this
new category of drugs is shown to be more effective than
conventional neuroleptics in reducing violent behavior in
usual care settings, then closer comparisons of these individual medications can be made with respect to their variant pharmacologic properties and hypothesized direct and
indirect effects in violence risk reduction.
The primary goal of the current study was to prospectively examine the effectiveness of atypical antipsychotic
medications, as compared to treatment with typical neuroleptics, in reducing community violence among persons
with schizophrenia over a 2-year period, using multiple
sources of information to measure actual violent incidents.
The secondary goal was to test key interaction effects and
potential mediating variables, in order to better specify the
indirect pathways by which the atypicals may reduce violent behavior in this population. These interacting variables
include medication adherence, adverse side effects, psychotic symptomatology, and co-occurring substance abuse.
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
Methods
Study Design and Sample. The data used for this analysis are from the North Carolina site of the Schizophrenia
Care and Assessment Program (SCAP), a multicenter,
prospective, observational study focused on assessing the
clinical, functional, and service utilization outcomes associated with routine care, both pharmacological and nonpharmacological, for persons diagnosed with schizophrenia in five U.S. geographic regions. In North Carolina,
403 persons with schizophrenia-related disorders were
recruited from several treatment facilities in an "open"
system of care across a nine-county, mixed urban and
rural area in the north-central region of the state. These
individuals were enrolled using two recruitment strategies
simultaneously, with screening for inclusion diagnoses:
(1) sequential inpatient admissions at a regional public
psychiatric hospital, an acute psychiatric unit of a private
university hospital, and a veterans affairs hospital; and (2)
random selection of outpatients from four area mental
health programs' case rosters and one veterans affairs
medical center outpatient clinic. Enrollment began in
1997 and was completed in 1999. Subjects were followed
for 2 years, with clinical data collection at 6-month intervals. Methods included (1) structured interviews conducted in person; (2) review and abstraction of medical
records; (3) electronic retrieval of data from health care
management information systems on service utilization
and costs; and (4) retrieval of arrest records from the
North Carolina Department of Justice archival data base.
The SCAP research design was purely observational, with
the purpose of understanding routine clinical care practices and treatment outcomes, and therefore no intervention or interference with usual patterns of treatment was
done.
Diagnoses were based on review of clinical records.
Because the SCAP was designed as an observational
study of the treatment of persons diagnosed with schizophrenia under "usual care" conditions, all adult patients in
treatment with recently documented DSM-FV diagnoses
of schizophrenia, schizoaffective disorder, or schizophreniform disorder were eligible for the study. Because
the aim of the study was to examine the treatment of
patients who were thought to have or who were diagnosed
as having schizophrenia, chart diagnoses assigned in the
past year were thus accepted prima facie and not verified
by additional research diagnostic assessments upon
enrollment Participants also had to be able and willing to
provide informed consent, had to be 18 years of age or
older, and could not have participated in a clinical drug
trial within 30 days prior to enrollment. Institutional
review board approval was obtained at participating study
sites prior to the initiation of the study, and informed con-
J.W. Swanson et al.
sent was received for all participants. At the time the
analyses were conducted for this report, data were available for 229 individual patients, with a total of 438 person-period observations (i.e., the sum of the number of
individual subjects times the number of 6-month observational periods for each subject) over 2 years.
Instruments. Violent behavior was measured with a
composite index that combined three sources of data: (1)
subject self-report using the MacArthur Community
Violence Interview (MCVI) (Monahan et al. 2001); (2)
systematic review of outpatient and inpatient medical
records, including civil commitment documents and other
legal information, using a chart-abstraction instrument for
coding evidence of violence that was developed in the
Duke Mental Health Study (Swanson et al. 1999); and (3)
review of records of arrests for violent offenses documented in the North Carolina Department of Justice data
base. Violent behavior was defined operationally as any
assault or battery committed against another person
specifically involving physical contact intended to harm
(e.g., hitting, shoving, kicking, biting) or threatening
another person with a lethal weapon in hand. Assaultive
behavior meeting this definition reported from any one of
the three sources was considered a positive indicator of
violence.
Type of pharmacological treatment was coded into
three hierarchical, mutually exclusive categories for each
6-month period of observation, using data collected from
medical record abstractions: (1) atypical antipsychotic—
any prescription for clozapine, risperidone, or olanzapine
in effect for more than half the given period; (2) conventional neuroleptic—a prescription for any other antipsychotic medication (but none of the three listed above) in
effect for more than half the period; and (3) none—no
antipsychotic medication prescription in effect for more
than half the period. At the time of the study, other atypical agents such as quetiapine and ziprasidone were not in
common use. We did not have sufficient statistical power
available to examine the effects of specific atypical
agents, for instance, to compare violence outcomes
among patients prescribed olanzapine versus risperidone
versus clozapine. We also did not have sufficient data
available to allow an informative analysis of the causes or
effects of medication switching within periods, dosing
differences, polypharmacy and die potential contribution
of adjunct medications (e.g., beta-blockers, anticonvulsants, benzodiazapines), and any potential differences by
route of administration of conventional neuroleptics (i.e.,
depot vs. oral medications).
For the purpose of causal attribution in longitudinal
time-series analysis, treatment effects were lagged, with
models estimating associations between treatment type in
Atypical Antipsychotics and Reduction of Violent Behavior
a given 6-month period and violent behavior in the subsequent 6-month period. If we had not lagged the treatment
period in our analysis, it would have been impossible to
tell the direction of causality in any association between
medication and violence—that is, whether the medication
had caused a reduction in violence or, instead, whether
violence had precipitated a change to a new type of medication.
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
violence, which may covary with, and condition the
effects of, the key independent variable (medication type).
The models also adjust for lack of independence between
observations for each subject over time. Our analysis used
a specific procedure designed for an autoregressive
covariance structure, that is, a matrix in which the correlation gradually declines between baseline and successive
followup measures of the dependent variable (violence).
Medication compliance was measured at each 6month assessment using a self-report item with a 5-point
fixed response scale, coded 1 ("I never missed taking my
medicine") to 5 ("I stopped taking the medicine altogether")Mental health services utilization intensity was coded
as a simple count of the number of outpatient service
encounters during the 6-month period. This included visits for outpatient therapy, case management, and medication checks, and any other billable encounters with mental
health service providers.
Demographic covariates included age, gender, and
race (African-American vs. white/other).
Baseline clinical predictors that were examined
included recent psychiatric hospitalization history (0 vs. 1
or more admissions in the past year), Global Assessment of
Functioning (GAP) score (Endicott et al. 1976), and cooccurring substance use, assessed using clinician ratings on
the Alcohol Use Scale and the Drug Use Scale (Drake et al.
19%). These independent variables were chosen for analysis based on prior clinical and epidemiological studies of
risk factors related to violence, and to control variability in
violence risk that may be associated with severity of psychopathology, substance abuse comorbidity, prior hospital
recidivism, and sociodemographic characteristics. Because
this was an observational study without random assignment
to treatment regimen, and because treatment selection may
covary naturalistically with clinical as well as demographic
risk factors for violence, it was necessary to control for
putative predictors of both violence and treatment in order
to properly interpret longitudinal treatment effects on
reduced violence risk.
Analysis. The incidence of violence over four 6-month
periods (2 years total) was modeled using generalized
estimating equations regression analysis for repeated measures (Diggle et al. 1994; Stokes et al. 1995), with the
probability of violence estimated prospectively as a function of medication type in the previous periods, controlling for time, baseline violence, intensity of mental health
services utilization, medication compliance, baseline clinical risk factors, and demographic characteristics. These
models are designed to incorporate all available longitudinal data for each subject, and to estimate the net association of multiple fixed as well as time-varying predictors of
Results
Sample Description. At the time these analyses were
conducted, a sample of 229 SCAP subjects in North
Carolina was available with followup data sufficient to
estimate 2-year trends in violence as a function of treatment. As described above, these subjects were enrolled
using two recruitment strategies simultaneously: sequential inpatient admissions and random selection from community mental health center case rosters. The study was
designed with a recruitment goal of obtaining 50 percent
of the sample with a record of hospitalization in the prior
year. Of the 229 individuals included in the present analysis, 37 percent were from the recently hospitalized cohort,
while 63 percent were outpatients who had not been hospitalized within the preceding year.
Regarding gender distribution, 58 percent of the subjects were male and 42 percent were female. Age ranged
from 18 to 71 years, with a mean of 42.5, a median of
41.5, and a standard deviation of 10.4 years. With respect
to racial background, 77 percent of the subjects were of
African-American descent and 23 percent were self-identified as white or other. This demographic distribution
fairly reflects the population of consumers with severe
mental illness receiving services in the public mental
health care system in the north-central region of North
Carolina (Swanson et al. 1999).
Considering clinical status, baseline GAF scores
ranged from 15 to 70, with a mean of 36.7, a median of
35, and a standard deviation of 9.8. This distribution indicates that, on average upon enrollment, these subjects
were clinically rated as having major impairment in several functional areas (e.g., work, social relations, judgment, thinking). Twenty-one percent of the subjects were
identified as having co-occurring substance abuse during
the 6 months preceding enrollment.
Refusal and Retention Rates. The baseline refusal rate
was 22 percent and sample retention was approximately
80 percent over 2 years of followup. Specifically, retention of subjects declined from 83.7 percent at 6 months to
73.3 percent at 24 months. There was no significant association between attrition and baseline violence; hence,
whereas 15.3 percent of subjects were assessed as violent
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
J.W. Swanson ct al.
at baseline, the percentage with a baseline history of violence was only slightly lower—13.7 percent—among subjects retained at 18 months.
It must be noted that the data presented here were from
an ongoing study in which only a portion of subjects had
reached the due date for each followup wave; thus, only a
small proportion had been enrolled for 2 years. The subset
of data that were available for this study included 229 individuals with a combined total of 438 person-period observations distributed as follows: 6 months, n = 216; 12
months, n = 129; 18 months, n = 68; 24 months, n = 25.
Baseline Association Between Subject Characteristics
and Type of Medication Prescribed. Because our design
was observational in nature and did not involve random
assignment to medication, it was important to examine
any preexisting associations between the type of antipsychotic medication prescribed at baseline and the subjects'
demographic and clinical characteristics. To prevent
potential selection biases from distorting results, it was
also crucial to statistically control for these baseline characteristics in our multivariable time-series analyses estimating the effects of medication type on violence.
Table 1 presents the frequency distributions resulting
from a cross-tabulation of subject characteristics with the
predominant type of medication that subjects had been prescribed when they were enrolled in the study. Of the 229
subjects included in this analysis, 35 had not had any prescription in effect for an antipsychotic medication during
more than half the 6-month period preceding enrollment.
Compared to subjects who had been prescribed either conventional or atypical antipsychotics, these subjects who had
Table 1. Sample characteristics by type of antipsychotic medication prescribed during 6 months prior
to study enrollment
Predominant Type of Antipsychotic Medication tit Baseline
0: None
1: Conventional
Statistical Significance
2: Atypical
n
(%)
n
(%)
n
(%)
Total
n
Age
Below median (18-41 yrs)
Median or above (242 yrs)
28
7
(23.93)
(6.25)
50
65
(42.74)
(58.04)
39
40
(33.33)
(35.71)
117
112
Gender
Female
Male
9
26
(9.28)
(19.70)
56
59
(57.73)
(44.70)
32
47
(32.99)
(35.61)
97
132
Race
African-American
White/other
28
7
(15.82)
(13.46)
98
17
(55.37)
(32.69)
51
28
(28.81)
(53.85)
177
52
28
7
(33.33)
(4.83)
28
87
(33.33)
(60.00)
28
51
(33.33)
(35.17)
84
145
22
(18.18)
52
(42.98)
47
(38.84)
121
13
(12.04)
63
(58.33)
32
(29.63)
108
21
14
(11.60)
(29.17)
90
25
(49.72)
(52.08)
70
9
(38.67)
(18.75)
181
48
26
9
(13.40)
(25.71)
98
17
(50.52)
(48.57)
70
9
(36.08)
(25.71)
194
35
0 vs.
1
0
vs.
2
Demographic characteristics
Baseline clinical characteristics
Initial treatment setting
Inpatient
Outpatient
Global Assessment of
Functioning
Below median (<32)
Median or above
(£32 or higher)
History of substance abuse
No
Yes
History of violent behavior
No
Yes
* p < 0.05; ** p < 0.01;'" p< 0.001
*
• * *
*
•*+
• * *
1 vs.
2
Atypical Antipsychotics and Reduction of Violent Behavior
gone untreated with medication were significantly younger,
more likely to be male, and more likely to have a co-occuring substance use disorder. These respondents were also
more likely to have been recruited into the study during a
hospital admission, that is, following an illness relapse that
may have resulted from not taking medication.
As summarized in the far right column of table 1,
with the exception of race, none of the potential "selection" variables examined was significantly associated with
receiving a prescription for an atypical versus a conventional antipsychotic medication. The finding that AfricanAmerican patients were significantly less likely than
white patients to have been prescribed an atypical (28.9%
vs. 53.9%; p < 0.001) deserves further study but is beyond
the scope of the present article.
It is notable that baseline history of violent behavior
was not significantly associated with medication type.
Nevertheless, baseline violence—and each of the variables specified in table 1—was included as a statistical
control to guard against selection bias in the multivariable
analyses to be presented below.
Prevalence of Violence. At baseline, it was determined
that 15.3 percent of the subjects had committed violent
acts in the 6 months prior to enrollment (as identified by
MCVI self-report interview, medical record review, civil
commitment documents, or arrest records). The 1-year
prevalence of violence from all sources of data was 21.8
percent; however, only 9 percent of the sample had official records of being arrested for violent offenses during
the year. Of those who were violent, about 41 percent had
committed serious violent acts involving weapon use or
causing injury to another person, while the remaining 59
percent were involved in only less serious incidents of
battery (e.g., physical fights involving shoving or slapping
but without causing injury or using weapons).
Cross-Sectional Analysis of Medication Type and
Violence. Prior to conducting longitudinal analysis of violence risk trends with lagged treatment effects, we examined
the cross-sectional association between violence and predominant medication type for the first year (n - 229). The
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
prevalence of violence was found to be highest in the group
of patients without any prescribed antipsychotic during
more than half the year (n = 44; 36.4% violent); lower
among those prescribed conventional neurolepu'cs (n = 116;
21.6% violent); and lowest among those prescribed an atypical antipsychotic (n = 69; 13.0% violent). Table 2 presents a
test of the statistical significance of these rates using logistic
regression. The analysis shows that the odds of any violence
were significantly lower only in the group prescribed atypical antipsychotic medications compared to those who were
untreated with antipsychotics during more than hah0 the year
(odds ratio [OR] = 0.26; p < 0.01). In this model, about 4
percent of the variance in the rate of violence was accounted
for by differences in medication type.
When patients prescribed atypicals were compared
directly with those prescribed conventional antipsychotics, this cross-sectional analysis showed that the rate
of violence was lower, but not significantly so, among
those in the atypical group. However, a more appropriate
and powerful test of this comparison—using all available
longitudinal data and controlling for medication compliance and potential selection effects—is described below.
Longitudinal Analysis of Medication Type and
Violence. As mentioned above, a total of 438 personperiod observations were available for analysis covering 2
years of followup assessments. Table 3 presents the
results of a time-series analysis using generalized estimating equations. These models estimate the longitudinal
effects of medication on reduced violent behavior in the
subsequent period, controlling for time and key covariates, as well as adjusting for the nonindependence of
repeated measures for individuals (autoregressive covariance structure). To capture the long-term effects of consistent treatment over time, the medication variables were
recoded as the cumulative number of 6-month periods
during which the medication type (atypical vs. conventional antipsychotic) was prescribed for over half a given
period. Medication compliance and mental health service
intensity were included as separate time-varying covariates, while baseline violence, demographic characteristics,
and clinical risk factors were included as controls.
Table 2. Cross-sectional association between predominant medication type and violent behavior in
1 year: Logistic regression analysis (n = 229)
Predominant medication type during year
Violence odds ratio
95% confidence Interval
No antipsychotic £90 days (comparison)
1.00
Conventional antipsychotic
0.48
(0.23-1.03)"
Atypical antipsychotic
0.26
(0.10-0.67)"
Note.—Chi-square = 8.57 with 2 df (p = 0.01); - 2 log likelihood - 232.03. Variance explained by model: Pseudo R2 = 0.04. Rank correlation predicted to observed: C = 0.62.
" p < 0.10; " p< 0.01
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
J.W. Swanson et al.
Table 3. Longitudinal effects of medication type and covariates on reduced violent behavior In persons with schizophrenia: Time-series analysis (n = 438)
Model 1
Independent variables
Violence
odds ratio
95% confidence
Interval
p
Model 2
Violence
odds ratio
95% confidence
Interval
p
Time
6-mo assessment period
1.20
(0.81-1.78)
0.96
(0.62-1.50)
Baseline violence
Any violence in 6 mos prior to enrollment
5.65
(2.85-11.21)
5.54
(2.64-11.62)
1.03
0.30
(1.01-1.05)
(0.16-0.57)
1.03
0.35
(1.01-1.05)
(0.15-0.80)
0.96
(0.73-1.27)
1.01
(0.67-1.52)
0.64
(0.46-0.91)
1.05
(0.69-1.60)
Interaction effects of prescribed medication
by compliance
Conventional antipsychotic periods x compliance
Atypical antipsychotic periods x compliance
1.35
0.40
(0.79-2.30)
(0.17-0.96)
Demographic covariates
Younger age
Male gender
African-American race
0.97
1.50
1.76
(0.93-1.01)
(0.74-3.04)
(0.64-4.91)
Baseline clinical predictors
Hospitalization history
Global Assessment of Functioning score
Substance abuse history
1.24
0.95
3.04
(0.58-2.65)
(0.91-0.99)
(1.49-6.22)
Treatment variables
(lagged, time-varying effects)
Mental hearth service intensity
Medication compliance
Conventional antipsychotic
(current + n previous periods)
Atypical antipsychotic
(current + n previous periods)
Note.—Rank correlation of predicted probabilities to observed rates: C = 0.813 (model 1 ) ; C » 0.862 (model 2). Variance explained by
model: Pseudo R2 - 0.18 (model 1); Pseudo R2 = 0.28 (model 2). Chi-square for improvement of model fit: model 1 compared to base
model with intercept and wave only: 52.86 with 5df,p< 0.001; model 1 compared to model 2: 31.06 with Bdf,p< 0.001.
* p < 0.05; ** p < 0.01; *** p < 0.001
Model 1 shows a significant main effect for the cumulative number of 6-month periods of prescribed atypical
antipsychotic medication on reduced incidence of violence
in the subsequent period (OR = 0.64; p < 0.05), controlling
for time, baseline violence (OR = 5.54; p < 0.001), mental
health service intensity, medication compliance, and treatment with conventional antipsychotics. We also found a
strong and significant negative association between medication compliance and violence (OR = 0.30; p < 0.001)
but a positive association between violence and mental
health service intensity (number of billed service encounters per period) (OR = 1.03; p < 0.01).
In model 2, the effects of medication on violence risk
were expressed as interactions between the type of medication prescribed and the degree of compliance with that
medication. Also, controls were added to hold constant the
effects of baseline demographic and clinical risk factors—
some of which covary with medication type (table 1) and
thus might bias the results if not statistically controlled.
This model shows a statistically significant net effect of
treatment with atypicals on reduced violence. Specifically,
the product of the number of 6-month periods that a subject was prescribed an atypical antipsychotic medication,
and the degree of compliance with that medication during
each period, was negatively associated with violence risk
in the subsequent period (OR = 0.40; p < 0.05). However,
the comparable variable for treatment with a conventional
neuroleptic had no longitudinal effect on reduced violence
risk (OR = 1.35; nonsignificant), despite the fact that a
greater number of subjects received treatment with conventional medications, thus providing adequate statistical
power to test the effect. The model also continued to show
a significant main effect for medication compliance, that
is, reducing violence (OR = 0.35; p < 0.05). Increased violence risk was predicted by greater functional impairment
(lower GAP score) (OR = 0.95; p < 0.05) and the presence
10
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
Atypical Antipsychotics and Reduction of Violent Behavior
antipsychotic during any given period was having a previous prescription for an atypical antipsychotic. Also, to
ascertain whether there might be a confounding effect of
time by medication prescribed, we ran an additional version of the final model with interaction effects specified
for wave X medication type; these interaction effects
were not statistically significant, while the effects that
were shown to be significant in model 2 remained so—
even with these additional variables in the model.
Finally, figures 1 and 2 illustrate the predicted probabilities of violent behavior over 6-month intervals as a
function of the cumulative number of periods during
which each of the two types of medication was prescribed, controlling for the variables included in model 2
of table 3. Figure 1 shows that patients prescribed conventional antipsychotic medications did not show a significant decline in violence probability from the first to the
of substance abuse comorbidity (OR = 3.04; p < 0.001) as
assessed at baseline.
Given the observational design of this study and the
absence of random assignment to treatment conditions, it
is possible that selection bias could affect these results,
for example, if "less problematic" patients with lower violence risk were differentially selected for treatment with
atypical antipsychotic medications. To ascertain whether
such a bias existed, we conducted a multivariable timeseries analysis estimating effects on receipt of prescription
for atypical antipsychotic medication (model not shown).
The results showed no significant longitudinal effect of
violent behavior, substance abuse, or compliance on the
likelihood of receiving atypical antipsychotic medications
in the subsequent period, when controlling for medication
prescribed in the previous period. Rather, the only significant predictor of receiving a prescription for an atypical
Figure 1. Probability of violent behavior by antipsychotic medication type and duration: North
Carolina Schizophrenia Care and Assessment Program subjects (n = 438 person-period observations
with complete data)
0.14
Conventional antipsychotic
Atypical antipsychotic
0.12
I
X)
2 0.08
>
£ 0.06
II
X)
I
0.04
t
8 0.02
0
1
n 6-mos. periods on medication (current + previous)
Note.—Estimates of lagged treatment effects on violence in subsequent wave, using generalized estimating equation time-series
regression models controlling for wave, baseline violence, mental health services utilization, demographics (age, sex, race), and baseline
clinical predictors (Global Assessment of Functioning, substance abuse, hospitalization). Subjects were coded for predominant medication type if prescription covered more than half a given 6-month period (290 days).
11
J.W. Swanson ct al.
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
Figure 2. Probability of violent behavior by antipsychotic medication type and duration: North
Carolina Schizophrenia Care and Assessment Program subjects (n = 259 person-period observations
with complete data and reported medication compliance)
0.12
• Conventional antipsychotic
Atypical antipsychotic
I
0.08
o
0.06
4>
s
a, 0.04
"S
1
0.02
0
n 6-mos. periods on medication (current + previous)
Note.—Estimates of lagged treatment effects on violence in subsequent wave, using generalized estimating equation time-series
regression models controlling for wave, baseline violence, mental health services utilization, demographics (age, sex, race), and baseline
clinical predictors (Global Assessment of Functioning, substance abuse, hospitalizatton). Subjects were coded for predominant medication type if prescription covered more than half a given 6-month period (£90 days).
took conventional antipsychotic medications did not show a
decline in violence risk. However, patients who were compliant with atypical antipsychotic medications showed reduced
rates of violent behavior starting in the first period of observation. Based on these models, the predicted probability of violence after four periods of compliance with prescribed atypical antipsychotic medications was negligible (0.0006).
fourth periods of observation (0.1203 to 0.1027).
However, patients consistently taking atypical antipsychotic medications did show reduced probability of violence over time (0.0924 to 0.0109).
Figure 2 presents the same analysis, but for only subjects
who reported being compliant "most of the time" with their
prescribed medication. Patients who were administered and
12
Atypical Antipsychotics and Reduction of Violent Behavior
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
Conversely, self-reported compliant patients taking typical
neuroleptic medications actually showed an increase in the
predicted probability of violence after four treatment periods,
although the increase was not statistically significant
Individual subject analysis. It may also be informative to consider these results in terms of individual subjects,
in addition to person-period observations. Overall, there
were 60 individual patients who were prescribed mostly
atypical antipsychotic medications and reported being compliant with their medications "most of the time" during each
period. There were 88 patients who were prescribed mostly
conventional antipsychotics and reported being compliant
"most of the time." The rate of any violence over all followup waves was significantly lower in the atypical/compliant group than in the conventional/compliant group
(5.0% vs. 15.9%; chi-square = 4.16 with 1 dfi p < 0.05).
However, the remaining 73 subjects who were noncompliant with their medications or spent most of the study
period with no prescription for any antipsychotic medication in effect were the most likely to be violent: 23
(31.5%) of these (untreated or noncompliant) individuals
were assessed by one or more sources to have committed
at least one violent act during the study's followup period.
Mediating variables. The analysis presented in table
4 tests key mediating variables, which may partly account
for the effect of the atypicals on reduced violence. Change
scores from baseline to current 6-month period were
examined for three potential mediators—psychotic symptomatology, co-occurring substance abouse, and adverse
side effects of medication. We found a significant bivariate association between each of these change scores and
reduced risk of violent behavior. Specifically, the odds of
any violence were reduced with fewer psychotic symptoms (OR = 0.86; p < 0.05), reduced substance abuse (OR
0.56; p < 0.01), and reduced adverse side effects of medication (OR = 0.83; p < 0.05). However, there was intercorrelation between these three change scores, and when
they were entered simultaneously into a multivariable
model, only psychotic symptom change remained statistically significant, as shown in Model 2 of table 4. A final
model was specified (Model 3 in table 4) in which the
interaction effects between each of these change scores
were found to significantly decrease violence risk. In this
final model, the direct effect on violence reduction attributable to atypical antipsychotic treatment (atypical periods by compliance) was rendered statistically nonsignifi-
Table 4. Longitudinal effects of antipsychotic medication and mediating variables on violent behavior
In persons with schizophrenia (n = 438 person-person ovservatlons)
Model 1
Model 2
Model 3
Odds ratio
95%
confidence
interval
Odds ratio
95%
confidence
Interval
0.79-2.31
1.25
0.74-2.10
1.25
0.74-2.09
0.16-0.96*
0.42
0.18-1.00
0.48
0.21-1.09
0.77
0.73
0.87
0.65-0.92"
0.50-1.05
0.72-1.04
0.50
0.55
0.40
0.28-0.88*
0.37-0.83**
0.22-0.74**
Odds ratio
95%
confidence
Interval
1.35
0.39
Medication treatment variables
(lagged, time-varying effects)
Typical antipsychotic periods
x compliance
Atypical antipsychotic periods
x compliance
Mediating variables
A. Improvement in psychotic symptoms
B. Improvement in substance abuse
C. Improvement in adverse side effects
Interaction of mediating variables
AxB
AXC
BxC
Note.—Rank correlation of predicted probabilities to observed rates: C - 0.86 (model 1); C = 0.89 (model 2); C - 0.90 (model 3).
Variance explained by model: Pseudo R2 = 0.28 (model 1); Pseudo R2 = 0.33 (model 2); Pseudo R2 « 0.35 (model 3). Chl-square for
Improvement of model fit: model 1 compared to base model with intercept and wave only: 372.71 with 15 df, p < 0.001; model 1 compared to model 2:61.88 with 3 df, p<0.001; Model 2 compaired to Model 3:1.10 with 4 df, p < 0.001. All models are adjusted for effects
of time, inter-subject correlation, outpatient treatment intensity, baseline violence, demographic variables (age, sex, race), baseline clinical variables (hospitalization history, substance abuse, psychotic symptoms, functional impairment), baseline adverse medication side
effects, the main effects of specified interactions, and three-way interaction of mediators In final model.
* p< 0.05; " p< 0.01
13
J.W. Swanson et al.
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
Figure 3 displays graphically the predicted probabilities of violence for subjects treated with typical vs.
atypical antipsychotic medications, stratified by adherence level and the number of mediating variables on
which subjects showed concurrent clinical improvement above average for the sample. The graph shows
generally that diminished violence risk was associated
with treatment with atypicals vs. typicals, but the
effect was conditioned by adherence among subjects
who showed concurrent improvement at least two of
three clinical domains (psychotic symptoms, substance
abuse, and adverse side effects). Specifically, among
those subjects who showed broad clinical improvement without regular adherence to medication, the type
of medication prescribed did not apper to affect violence risk.
cant, and the fit of the model improved significantly over
the model without the interaction effects of mediators.
The analysis shown in table 4 provides some evidence
that die superior effect of treatment with atypical antipsychotic medications in reducing violence is due in part to a
set of indirect effects, i.e., reductions in key risk factors for
violence. Moreover, these risk factors operate jointly and
exert reciprocal effects. In particular, co-occurring substance
abuse is the most powerful risk factor predicting violent
behavior in this sample (OR = 3.04; p < 0.001) and treatment with atypicals significantly reduced substance abuse
(OR = 0.40; p < 0.05), controlling for medication nonadherence (OR = 3.37; p < 0.0001) and adverse side effects (OR
= 1.47; p < 0.05) (model not shown). In turn, substance
abuse was significantly associated with elevated psychotic
symptomatology it = 1.99 with 2df;p< 0.05).
Figure 3. Probability of any violent behavior in 6 months by medication type, compliance, and clinical
Improvement mediators1
• Improvement in less than 2 areas
• Improvement in 2 or more areas
(«) (18) (66)
Antipsychotic rConventional
(30) (44)
(34) (111)
(24) (73)
Atypical
Conventional
Atypical
Not adherent most of the time
Adherent most of the time
Areas of clinical improvement: psychotic symptoms, substance abuse, adverse side effects of
medication.
14
Atypical Antipsychotics and Reduction of Violent Behavior
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
ment methods, researchers have had more difficulty consistently demonstrating the effects of treatment and risk
management strategies in reducing violence. Harris and
Rice (1997) conclude that very few effective techniques
exist for reducing aggressive behavior in persons with
psychiatric disorders, although some behavioral therapies
have shown promise (Shloss et al. 1989; Beck et al. 1991;
see generally Paul and Lentz 1979). Given the strength of
the findings in the present study, the administration of
atypical antipsychotic medication appears to be an important ingredient in effective clinical treatment—and violence risk management in particular—for schizophrenia
patients residing in the community.
The findings may have especially significant implications for reducing the long-term cost of treatment for
schizophrenia. Among all those who suffer from this disease and other serious psychiatric disorders, a small proportion who become dangerous or engage in violent
behavior account for a disproportionately large share of
the cost of providing the most expensive services in the
most restrictive settings (Borum et al. 1996). The study
found a significant positive association between violence
and mental health service intensity, which suggests that
patients who were likely to be violent over time tended to
have greater severity and complexity of need for mental
health services.
In North Carolina, where this study was conducted,
the large majority of state mental hospital admissions are
involuntary commitments, which often are precipitated by
threatening or assaultive behavior in patients with psychoses who experience exacerbation or relapse of their illness following a period of nonadherence with medication
(Swanson et al. 1999). Partly as a result, the North
Carolina Division of Mental Health, Developmental
Disabilities, and Substance Abuse Services spent $285
million in fiscal year 2000—71 percent of its adult mental
health care budget—on institutional (hospital) care (Stein
2001). Moreover, during the same year, the North
Carolina Department of Corrections spent an estimated
$117 million for the incarceration of about 5,000 individuals with primary psychiatric disorders (excluding substance abuse)—on average, more than $23,000 per inmate
with mental illness. It is estimated that only about 9 percent of these persons with mental illness received psychotropic medication while incarcerated. Given timely
treatment with appropriate medications and other mental
health interventions, it seems likely that many of these
offenders would never have committed the crimes that
landed them behind bars (Swanson 2001).
This study may be somewhat limited in its generalizability, insofar as the demographic distribution of our
sample (patients in the public mental health care system
in North Carolina) differs from the population of all persons with this disorder (Keith et al. 1991; Kendler et al.
Discussion
This is the first prospective observational study to examine the long-term effectiveness of treatment for schizophrenia with atypical antipsychotic medications under
"usual care" conditions, which specifically assessed the
reduced risk of violence in the community as an outcome of treatment. The key finding of this study is that
schizophrenia patients who were prescribed atypical
antipsychotic medications (including clozapine, risperidone, or olanzapine) and who complied with these medications over time had a significantly lower risk of violent behavior compared to patients treated with
conventional neuroleptics or no antipsychotic medications. By comparison, even long-term treatment with
conventional antipsychotic medications did not significantly reduce violence risk. These results were found
using a multivariable repeated-measures time-series
analysis with lagged treatment effects, controlling for
salient demographic and clinical risk factors. Consistent
with previous research, increased violence risk was also
significantly associated with substance abuse, greater
functional impairment (severity of psychiatric disturbance as measured by the GAF), poor medication compliance, and previous history of violence. Furthermore,
the findings showed both a cumulative effect of treatment over the four periods of observation, and an interaction effect with compliance: the longer patients were
prescribed atypical antipsychotic medications, and the
more consistently they adhered to the prescribed regimen, the lower the patients' probability of violent behavior. Finally, an analysis of mediating variables suggested
that the superior effectiveness of atypicals in reducing
violence involves a complex set of indirect and intercorrelated effects: decrease in adverse side effects of medication, decrease in substance abuse, and better control
of psychotic symptoms.
The current study corroborates previous reports suggesting that atypical antipsychotic medications may
reduce aggression and violence in patients with schizophrenia (Taylor et al. 1996; Glazer and Dickson 1998;
Volavka 1999). The results of this study have important
clinical implications for both inpatient and outpatient
mental health providers regarding violence risk management of patients with a diagnosis of schizophrenia and
related disorders.
Over the past few decades, an extensive body of
research has developed on violence risk assessment in
psychiatric settings (Grisso and Tomkins 1996). The purpose of this research has largely been to establish empirically validated risk factors (Monahan and Steadman 1994;
McNiel 1998) and to develop actuarial risk assessment
tools (Borum 1996; Monahan et al. 2000; Monahan et al.
2001). However, despite improved violence risk assess-
15
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
J.W. Swanson et al.
1996). Also, the study's measure of medication compliance—a single interview item assessed by self-report at
each wave—may be limited.
Although this study provides the strongest evidence
to date that consistent treatment with atypical antipsychotic medications can reduce the risk of community violence among persons with schizophrenia, there is still little evidence regarding how the atypicals reduce violence.
At least one recent study suggests that atypical antipsychotics (particularly clozapine) may exert antihostility
effects independent of the drugs' effects on psychosis and
sedation (Citrome et al. 2001).
The atypicals my reduce violence risk by means of
several alternative mechanisms in different subgroups of
patients with schizophrenia. At a basic biological level,
neurotransmitters (e.g., dopamine, serotonin, norepinephrine) have been implicated generally in aggressive behavior
(Senninger and Laxenaire 1998). Pharmacological effects
on aggression may be specific or global, and they may
reduce risk through inhibition of violent impulses or by
relieving symptoms such as paranoid delusions and hostility. Our paper provides some evidence that these medications reduce violence by direct as well as indirect means.
Reduced extrapyramidal symptoms may allow improved
compliance, leading to more consistent control of positive
and negative psychotic symptoms, diminished substance
abuse, increased participation in psychosocial rehabilitation
and other supportive outpatient services, and ultimately
improved social functioning—all of which, in turn, may
help prevent assaultive behavior (Menditto et al. 1996).
Our findings are also consistent with previous studies
showing a link between medication noncompliance, substance abuse, and violence in persons with severe mental illness (SMI) (Owen et al. 19%; Swartz et al. 1998b; Swanson
et al. 2000). It is well known that persons suffering from
psychotic symptoms often fail to adhere to conventional neuroleptic medication regimens—in part because the common
extrapyramidal symptoms of these medicines may be intolerable (Barnes and McPhillips 1998; Marder 1998). Traditional
antipsychotic drugs (e.g., haloperidol) have a well-documented profile of adverse effects, particularly extrapyramidal
symptoms and tardive dyskinesia associated with long-term
use. Noncompliance can indicate a decision by patients that
the discomfort associated with these drugs outweighs therapeutic benefits (although, clearly, noncompliance may be a
function of other factors as well—such as lack of insight and
awareness of illness).
Substance abuse disorders have been found to be
highly associated with violent and aggressive behaviors,
including assaultiveness, suicide, hostility, and criminality
(Dixon 1999; RachBeisel et al. 1999; Monahan et al.
2001; Miles et al. 2003). Use of alcohol and illicit drugs
by persons with severe mental illness may sometimes represent an attempt to self-medicate psychotic symptoms and
distress. Substance abuse may, in turn, lead to violent
behavior in several ways: by stimulating or disinhibiting
aggressive impulses; by exacerbating delusions of threat as
well as real conflict in domestic and other social relationships; by exposing the user to criminogenic environments
in which violence is commonplace; and by increasing economic stress and survival demands (which are already considerable for persons with severe psychiatric disorders). In
theory, then, the improved side effect profile of the atypical
antipsychotic medications could indirectly reduce the risk
of violent behavior among SMI individuals—principally by
means of better compliance and reduced substance abuse
comorbidity (Owen et al. 1996).
However, we also found evidence that medication
noncompliance is an important independent risk factor for
violence, i.e., even controlling for substance abuse. This
finding has important implications for clinical management of violence risk. If patients are compliant with atypical antipsychotics, violence risk is dramatically reduced.
Insofar as compliance is determined by drug tolerability
and the atypicals are associated with fewer side effects, it
is possible that the atypicals may reduce violence indirectly via better compliance. Clearly, however, medication
compliance in schizophrenia treatment is a complex phenomenon, and is affected by multiple independent variables, including insight, severity of psychotic symptoms,
substance abuse, adverse social environments, availability
of supportive caregivers, and the quality of the collaborative bond between the patient and clinician—in addition to
the well-document effects of unpleasant side effects
(Valenstein et al. 1998; Olfson et al. 2000; Menzin et al.
2003). Moreover, while patients taking atypicals may
experience fewer adverse side effects, some side effects
may persist. Thus, for a number of reasons, noncompliance
may continue to be a problem for many patients even if
they are prescribed atypical antipsychotic medications; in
such cases, other strategies should be considered.
Several scientific reviews have concluded that depot
antipsychotics modestly improve medication compliance
and other clinical outcomes (Kane 1983; Fenton et al.
1997; Adams and Eisenbruch 2000). At present, risperidone
is the only atypical antipsychotic medication with an
approved long-acting injectable form of administration.
Hence, clinicians may need to consider use of depot conventional neuroleptics or other community-based interventions designed to improve treatment adherence, such as
Assertive Community Treatment programs (Burns and
Santos 1995). Similarly, where substance abuse persists
during a course of atypical antipsychotic therapy, additional
treatment targeting substance abuse directly is still needed.
Given that violence among individuals with serious psychiatric disorders may result from multiple etiologies—with
complex interactions between developmental, cognitive,
clinical, and social-environmental risk factors (Swanson et
16
Atypical Antipsychotics and Reduction of Violent Behavior
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
al. 19%; Swanson et al. 1998; Silver et al. 1999; link et al.
1999)—the role of improved pharmacotherapies in managing violence may vary considerably across subgroups of
patients. Better specification of treatment effects, targeting
different clusters of modifiable risk factors for assaultive
behavior, may lead ultimately to significantly reduced violence in populations of persons with schizophrenia and
related disorders residing in the community.
Buckley, P.F.; Ibrahim, Z.Y.; Singer, B.; Orr, B.;
Donenwirth, K.; and Brar, P.S. Aggression and schizophrenia: Efficacy of risperidone. Journal of the American
Academy of Psychiatry and the Law, 25(2): 173-181,
1997.
Burns, B.J., and Santos, A.B. Assertive community treatment: An update of randomized trials. Psychiatric
Services, 46(7):669-675, 1995.
Chengappa, K.N.; Ebeling, T; Kang, J.S.; Levine, J.; and
Parepally, H. Clozapine reduces severe self-mutilation
and aggression in psychotic patients with borderline per-
References
Adams, C.E., and Eisenbruch, M. Depot fluphenazine for
sonality disorder. Journal of Clinical Psychiatry,
schizophrenia. Cochrane Database of Systematic
60:477^184, 1999.
Reviews, 2:CD000307 20257438s, 2000.
Angermeyer, M., and Matschinger, H. The effect of violent attacks by schizophrenic persons on the attitude of
the public toward the mentally ill. Social Science &
Medicine, 43:1721-1728, 1996.
Barnes, T.R., and McPhillips, M.A. Novel antipsychotics,
extrapyramidal side effects and tardive dyskinesia.
International Clinical Psychopharmacology, 13(Suppl
3):S49-57, 1998.
Chengappa, K.N.R.; Vasile, J.; Levine, J.; Ulrich, R.;
Baker, R.; Gopalini, A.; and Schooler, N. Clozapine: Its
impact on aggressive behavior among patients in a state
psychiatric hospital. Schizophrenia Research, 53:1—6,
2002.
Cheung, P.; Schweitzer, I.; Crowley, K.; and Tuckwell, V.
Violence in schizophrenia: Role of hallucinations and
delusions. Schizophrenia Research, 26:181-190, 1997
Chiles, J.A.; Davidson, P.; and McBride, D. Effects of
clozapine on use of seclusion and restraint at a state hospital. Hospital and Community Psychiatry, 45:269-271,
1994.
Beck, N.C.; Greenfield, S.R.; Gotham, H.; Menditto,
A.A.; Stuve, P.; and Hemme, C.A. Risperidone in the
management of violent, treatment-resistant schizophrenics
hospitalized in a maximum security forensic facility.
Journal of the American Academy of Psychiatry and the
Law, 25(4):461^68, 1997.
Beck, N.C.; Menditto, A.A.; and Baldwin, L. Reduced
frequency of aggressive behavior in forensic patients in a
social learning program. Hospital and Community
Psychiatry, 42:750-752, 1991.
Chrome, L.; Volavka, J.; Czobor, P.; Sheitman, B.;
Lindenmayer, J.P.; McEvoy, J.; Cooper, T.B.; Chakos, M ;
and Lieberman, J.A. Effects of clozapine, olanzapine,
risperidone, and haloperidol on hostility among
patients with schizophrenia. Psychiatric
Services,
52(ll):1510-1514, 2001.
Clark, W.S.; Street, J.S.; Feldman, P.D.; and Breier, A.
The effect of olanzapine in reducing the emergence of
psychosis among nursing home patients with Alzheimer's
disease. Journal of Clinical Psychiatry, 62(1 ):34—40,
2001.
Borum, R. Improving the clinical practice of violence risk
assessment. American Psychologist, 51:945-956, 1996.
Borum, R.; Swanson, J.; Swartz, M.; and Hiday, V.
Substance abuse, violent behavior, and police encounters
among persons with severe mental disorders. Journal of
Contemporary Criminal Justice, 13:236-249, 1997.
Cuffel, B. Disruptive behavior and the determinants of
costs in the public mental health system. Psychiatric
Services, 48:1562-1566, 1997.
Borum, R.; Swartz, M.; and Swanson, J. Assessing and
managing violence risk in clinical practice. Journal of
Practical Psychiatry and Behavioral Health, 2:205-215,
1996.
Currier, G.W., and Simpson, G.M. Risperidone liquid
concentrate and oral lorazepam versus intramuscular
haloperidol and intramuscular lorazepam for treatment of
psychotic agitation. Journal of Clinical Psychiatry,
62(3): 153-157, 2001.
Buckley, P.; Bartell, J.; Donenwirth, K.; Lee, S.; Torigoe,
F.; and Schulz, S.C. Violence and schizophrenia:
Clozapine as a specific antiaggressive agent. Bulletin of
the American Academy of Psychiatry and the Law,
23(4):607-611, 1995.
Czobor, P.; Volavka, J.; and Meibach, R.C. Effect of
risperidone on hostility in schizophrenia. Journal of
Clinical Psychopharmacology, 15:243-249, 1995.
Buckley, P.F. The role of typical and atypical antipsychotic medications in the management of agitation and
aggression. Journal of Clinical Psychiatry, 60(Suppl
10):52-60, 1999.
Diggle, P.J.; Liang, K-Y.; and Zeger, S.L.: Analysis of
Longitudinal Data. Oxford, U.K.: Clarendon Press,
1994.
17
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
J.W. Swanson et al.
Dixon, L. Dual diagnosis of substance abuse in schizophrenia: Prevalence and impact on outcomes.
Schizophrenia Research, 35:S93-S100, 1999.
Grove, V.E.; Quintanilla, J.; and DeVaney, G.T.
Improvement of Huntington's disease with olanzapine and
Dixon, L.; Adler, D.; Berlant, J.; Dulit, R.; Ellison, J.;
Montgomery, J.; New, A.; Sins, S.; Sonis, W.; and Stanley,
J.S. Confronting violence. Committee on Psychopathology
of the Group for the Advancement of Psychiatry.
[Editorial]. Psychiatric Services, 49(7):865, 1998.
343(13):973-974, 2000.
valproate. New England Journal of Medicine,
Douglas, K.S.; Cox, D.N.; and Webster, C D . Violence
risk assessment: Science and practice. Legal and
Criminological Psychology, 4<2):149-184, 1999.
Harris, G.T., and Rice, M.E. Risk appraisal and management of violent behavior. Psychiatric
Services,
48:1168-1176, 1997.
Heilbrun, K. Prediction versus management models relevant to risk assessment: The importance of legal decisionmaking context. Law and Human Behavior, 21:347-359,
1997.
Drake, R.E.; Mueser, K.T.; and McHugo, G.J. Clinical
rating scales: Alcohol Use Scale (AUS), Drug Use Scale
(DUS), Substance Abuse Treatment Scale (SATA). In:
Sederer, L.I., and Dickey. B., eds. Outcomes Assessment
in Clinical Practice. Baltimore, MD: Williams & Wilkins,
1996. pp. 113-116.
Hirose, S.; Ashby, C.R.; and Mills, MJ. Effectiveness of
ECT combined with risperidone against aggression in
schizophrenia. Journal of ECT, 17(l):22-26, 2001.
Horrigan, J.P.; Barnhill, L.J.; and Coirvoisie, H.E.
Olanzapine in PDD. Journal of the American Academy of
Child and Adolescent Psychiatry, 36(9): 1166-1167, 1997.
Duncan, G.E.; Zom, S.; and Lieberman, J.A. Mechanisms
of typical and atypical antipsychotic drug action in relation to dopamine and NMDA receptor hypofunction
hypotheses of schizophrenia. Molecular
Psychiatry,
4<5):418-428, 1999.
John, V.; Rapp, M.; and Pies, R. Aggression, agitation and
mania with olanzapine. [Letter]. Canadian Journal of
Psychiatry, 43:1054, 1998.
Kane, J.M. Problems of compliance in the outpatient
treatment of schizophrenia. Journal of Clinical
Psychiatry, 44:3-6, 1983.
Keck, P.E.; Strakowski, S.M.; and McElroy, S.L. The efficacy of atypical antipsychotics in the treatment of depressive symptoms, hostility, and suicidality in schizophrenia.
Journal of Clinical Psychiatry, 61(Suppl 3):4-9, 2000.
Elbogen, E.B., and Huss, M.T. The role of serotonin in
violence and schizophrenia: Implications for risk assessment. Journal of Psychiatry & Law, 28:19-48, 2000.
Elbogen, E.B., and Tomkins, A.J. From the hospital to the
community: Integrating conditional release and contingency management. Behavioral Sciences & the Law,
18(4):427^t44, 2000.
Keith, S.J.; Regier, D.A.; and Rae, D.S. Schizophrenic
disorders. In: Robins, L.N., and Regier, D.A., eds.
Psychiatric Disorders in America: The Epidemiologic
Catchment Area Study. New York, NY: Free Press, 1991.
pp. 33-52.
Kendler, K.S.; Gallagher, TJ.; Abelson, J.M.; and Kessler,
R.C. Lifetime prevalence, demographic risk factors, and
diagnostic validity of nonaffective psychosis as assessed in a
U.S. community sample: The National Comorbidity Survey.
Archives of General Psychiatry, 53:1022-1031, 19%.
Endicott, J.; Spitzer, R.; Fleiss, J.; and Cohen, J. The
Global Assessment Scale: A procedure for measuring
overall severity of psychiatric disturbances. Archives of
General Psychiatry, 33:766-771s, 1976.
Estroff, S.; Swanson, J.; Lachicotte, W.; Swartz, M.; and
Bolduc, M. Risk reconsidered: Targets of violence in the
social networks of people with serious psychiatric disorders. Social Psychiatry and Psychiatric Epidemiology,
33:95-101, 1998.
Kent, S.; Fogarty, M.; and Yellowlees, P. A review of
studies of heavy users of psychiatric services. Psychiatric
Services, 46:247-253, 1995.
Fenton, W.S.; Blyler, C.R.; and Heinssen, R.K.
Determinants of medication compliance in schizophrenia:
Empirical and clinical findings. Schizophrenia Bulletin,
23(4):637-651, 1997.
Lieberman, J.A.; Mailman, R.B.; Duncan, G.; Sikich, L.;
Chakos, M.; Nichols, D.E.; and Kraus, J.E. Serotonergic
basis of antipsychotic drug effects in schizophrenia.
Biological Psychiatry, 44(ll):1099-1117, 1998.
Geller, J. A report on the "worst" state hospital recidivists
in the U.S. Hospital and Community
Psychiatry,
43(9):904-908, 1992.
Link, B.; Cullen, F ; Frank, J.; and Wozniak, J. The social
rejection of former mental patients: Understanding why
Glazer, W.M., and Dickson, R.A. Clozapine reduces violence and persistent aggression in schizophrenia. Journal
of Clinical Psychiatry, 59:8-14, 1998.
labels matter. American Journal of Sociology,
92:1461-1500, 1987.
Grisso, T., and Tomkins, A. Communicating violence risk
assessments. American Psychologist, 51:928-930, 1996.
Link, B.G.; Monahan, J.; Stueve, A.; and Cullen, FT. Real
in their consequences: A sociological approach to under-
18
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
Atypical Antipsychotics and Reduction of Violent Behavior
Perm, D., and Martin, J. The stigma of severe mental illness: Some potential solutions for a recalcitrant problem.
Psychiatric Quarterly, 69:235-247, 1998.
standing the association between psychotic symptoms and
violence. American Sociological Review, 64(2):316-332,
1999.
Marder, S.R. Facilitating compliance with antipsychotic
medication. Journal of Clinical Psychiatry, 59(Suppl
3):21-25, 1998.
Potenza, M.N.; Marc, N.; Holmes, J.P.; Kanes, SJ.; and
McDougle, C.J. Olanzapine treatment of children, adolescents, and adults with pervasive developmental disorders:
An open-label pilot study. Journal of Clinical
Psychopharmacology, 19(l):37-44, 1999.
Marder, S.R.; Davis, J.M.; and Chouinard, G. The effects
of risperidone on the five dimensions of schizophrenia
derived by factor analysis. Journal of Clinical Psychiatry,
58(12):538-546, 1997.
Rabinowitz, J.; Avnon, M.; and Rosenberg, V. Effect of
clozapine on physical and verbal aggression.
Schizophrenia Research, 22:249-255, 1996.
McNiel, D.E. Empirically based clinical evaluation and
management of the potentially violent patient. In:
Kleespies, P.M., ed. Emergencies in Mental Health
Practice: Evaluation and Management. New York, NY:
Guilford Press, 1998. pp. 95-116.
Ratey, J.J.; Leveroni, C ; Kilmer, D.; Gutheil, C ; and
Swartz, B. The effects of clozapine on severely aggressive
psychiatric inpatients in a state hospital. Journal of
Clinical Psychiatry, 54(6):219-223, 1993.
Schalock, R.L.; Touchstone, F.; Nelson, G.; Weber, L.;
Shehan, M.; and Stull, C. A multivariate analysis of mental hospital recidivism. Journal of Mental Health
Administration, 22:358-367, 1995.
Menditto, A.A.; Beck, N.C.; Stuve, P.; Fisher, J.A.; Stacy,
M.; Logue, M.B.; and Baldwin, L.J. Effectiveness of
clozapine and a social learning program for severely disabled psychiatric inpatients. Psychiatric
Services,
47(1):46-51, 1996.
Schopp, R.F. Communicating risk assessments: Accuracy,
efficacy, and responsibility. American
Psychologist,
51:939-944,1996.
Monahan, J. Mental disorder and violent behavior:
Perceptions and evidence. American
Psychologist,
47:511-521, 1992.
Senninger, J.L., and Laxenaire, M. Risperidone in the
treatment of aggressiveness in schizophrenics. Annals
Medico-Psychologiques, 156(3):210-213, 1998.
Monahan, J., and Steadman, H., eds. Violence and Mental
Disorder: Developments in Risk Assessment. Chicago, IL:
University of Chicago Press, 1994.
Shloss, P.J.; Smith, M.; and Santora, C. A respondent conditioning approach to reducing anger responses of a
dually diagnosed man with mild mental retardation.
Behavior Therapy, 20:459-464, 1989.
Silver, E.; Mulvey, E.; and Monahan, J. Assessing violence risk among discharged psychiatric patients: Toward
an ecological approach. Law and Human Behavior,
23(2):237-255, 1999.
Simon, R. Psychiatrists' duties in discharging sicker and
potentially violent inpatients in the managed care era.
Psychiatric Services, 49:62-67, 1998.
Spivak, B.; Lamschtein, C ; Talmon, Y.; Guy, N.; Mester,
R.; Feinberg, I.; Kotler, M.; and Weizman, A. The impact
of clozapine treatment on serum lipids in chronic schizo-
Monahan, J.; Steadman, H.; Silver, E.; and Appelbaum, P.
Rethinking Risk Assessment: The MacArthur Study of
Mental Disorder and Violence. New York, NY: Oxford
University Press, 2001.
Monahan, J.; Steadman, H.J.; Appelbaum, P.S.; Robbins,
P.C.; Mulvey, E.P.; Silver, S.; Roth, L.H.; and Grisso, T.
Developing a clinically useful actuarial tool for assessing
violence risk. British Journal of Psychiatry, 176:312-319,
2000.
Mulvey, E.P., and Lidz, C.W. Measuring patient violence
in dangerousness research. Law and Human Behavior,
17:277-288, 1993.
Olfson, M.; Mechanic, D.; Hansell, S.; Boyer, C.A.;
Walkup, J.; and Weiden, P.J. Predicting medication noncompliance after hospital discharge among patients with
schizophrenia. Psychiatric Services, 51(2):216-222, 2000.
phrenic patients. Clinical
Neuropharmacology,
22:98-101, 1999.
Owen, R.R.; Fischer, E.P.; Booth, B.M.; and Cuffel, B.J.
Medication noncompliance and substance abuse among
patients with schizophrenia. Psychiatric
Services,
47(8):853-858, 1996.
Spivak, B.; Mester, R.; Wittenberg, N.; Maman, Z.; and
Weizman, A. Reduction of aggressiveness and impulsiveness during clozapine treatment in chronic neurolepticresistant schizophrenic patients. Clinical Neuropharmacology, 20:442-446, 1997.
Paul, G.L., and Lentz, R.J. Psychosocial Treatment of
Chronic Mental Patients: Milieu Versus Social-Learning
Programs. Cambridge, MA: Harvard University Press,
1979.
Stein, J. The mental health system and its financing and
use. Journal of Common Sense, 6:3, 2001. Available at
http://www.common-sense.org/Publications/Journal/system.html.
19
Schizophrenia Bulletin, Vol. 30, No. 1, 2004
J.W. Swanson et al.
Stokes, M.E.; Davis, C ; and Koch, G.G. Categorical
Data Analysis Using the SAS System. Cary, NC: SAS
Institute, 1995.
risperidone: A review. British Journal of Psychiatry,
169:21-30, 1996.
Thienhaus, O., and Piasecki, M. Assessment of psychiatric patients' risk of violence toward others. Psychiatric
Services, 49:1129-1147, 1998.
Swanson, J. Mental disorder, substance abuse, and community violence: An epidemiological approach. In:
Monahan, J., and Steadman, H., eds. Violence and Mental
Disorder: Developments in Risk Assessment. Chicago, IL:
University of Chicago Press, 1994. pp. 101-136.
Torrey, E.F. Violent behavior by individuals with serious
mental illness. Hospital and Community Psychiatry,
45:653-662, 1994.
Swanson, J. Correctional mental health in North Carolina:
An expensive non-solution. Journal of Common Sense,
6:3, 2001. Available at http://www.common-sense.org/
Publications/Journal/system.html.
Valenstein, M.; Barry, K.L.; Blow, F.C.; Copeland, L.;
and Ullman, E. Agreement between seriously mentally ill
veterans and their clinicians about medication compliance. Psychiatric Services, 49:1043-1048, 1998.
Swanson, J.; Estroff, S.; Swartz, M.; Borum, R.;
Lachicotte, W.; Zimmer, C ; and Wagner, R. Violence and
severe mental disorder in clinical and community populations: The effects of psychotic symptoms, comorbidity,
and lack of treatment. Psychiatry, 60:1-22, 1997.
Viale, G.; Mechling, L.; Maislin, G.; Durkin, M.;
Engelhart, L.; and Lawrence, B.J. Impact of risperidone
on the use of mental health resources. Psychiatric
Services, 48(9): 1153-1159, 1997.
Volavka, J. The effects of clozapine on aggression and
substance abuse in schizophrenic patients. Journal of
Clinical Psychiatry, 60(Suppl 12):43-46, 1999.
Swanson, J.W.; Borum, R.; Swartz, M.S.; and Hiday, V.A.
Violent behavior preceding hospitalization among persons
with severe mental illness. Law and Human Behavior,
23(2): 185-204, 1999.
Volavka, J.; Zito, J.; Vitrai, J.; and Czobor, O. Clozapine
effects on hostility and aggression in schizophrenia.
Journal of Clinical Psychopharmacology, 17:84—87,
1993.
Swanson, J.W.; Borum, R.; Swartz, M.S.; and Monahan,
J. Psychotic symptoms and disorders and the risk of violent behavior in the community. Criminal Behaviour and
Mental Health, 6:317-338, 1996.
Swanson, J.W.; Swartz, M.S.; Borum, R.B.; Hiday, V.A.;
Wagner, H.; and Burns, B.J. Involuntary out-patient commitment and reduction of violent behaviour in persons
with severe mental illness. British Journal of Psychiatry,
176:324-331,2000.
Acknowledgment
Funding for the Schizophrenia Care and Assessment
Program (SCAP) was provided by Eli Lilly and Company
as an initiative of the company's Health Outcomes
Evaluation Group, United States Medical Division. Core
SCAP data collection was coordinated by the MEDSTAT
Group, Inc., of Washington, DC, under contract to Eli
Lilly and Company. Funding to support additional data
collection and analysis related to violence outcomes in
schizophrenia treatment in North Carolina was provided
by Eli Lilly and Company through a contract with Duke
University Medical Center.
The authors thank H. Ryan Wagner, Ph.D., for biostatistical consultation and Barbara J. McCauley, M.A., and
Mimi Kim, M.A., for editorial assistance in preparation of
the manuscript
Swanson, J.W.; Swartz, M.S.; Estroff, S.E.; Borum, W.R.;
Wagner, H.R.; and Hiday, V.A. Psychiatric impairment,
social contact, and violent behavior. Social Psychiatry
and Psychiatric Epidemiology, 33:86-94, 1998.
Swartz, M.; Swanson, J.; Hiday, V.; Borum, W.R.;
Wagner, H.; and Burns, B. Taking the wrong drugs:
Substance abuse, medication nonadherence, and violence
in severely mentally ill individuals. Social Psychiatry and
Psychiatric Epidemiology, 33:75-80, 1998a.
Swartz, M.S.; Swanson, J.W.; Hiday, V.A.; Borum, R.;
Wagner, H.R.; and Burns, B.J. Violence and severe mental
illness: The effects of substance abuse and nonadherence
to medication. American Journal of Psychiatry,
The Authors
155:226-231, 19986.
Jeffrey W. Swanson, Ph.D., Marvin S. Swartz, M.D., and
Eric B. Elbogen, Ph.D., are all with the Services
Effectiveness Research Program, Department of
Psychiatry and Behavioral Sciences, Duke University
Medical Center, Durham, NC.
Taylor, P.J.; Butwell, M.; Gray, C ; Daly, R.; Delal, B.;
Ferraro, D.; Gibb, R.; Heads, T; Huckstep, B.; Larkin, E.;
Leese, M.; Shetty, G.; Winton, M.; Tidmarsh, D.; and
Williams, D. Schizophrenia, violence, clozapine and
20