|Home | About | Journals | Submit | Contact Us | Français|
To review empirical studies, published since 1990, of the prevalence and incidence of violent perpetration and violent victimization among persons with serious mental illness and to compare their relative importance as a public health concern.
We searched three computerized bibliographic databases, MEDLINE, PSYCH INFO, and Web of Science, using the following keywords: (1) Violent perpetration: SMI, mental illness, mental disorder, psychiatric disorder, psychopathology, violence, violent behavior, and violent act(s); and (2) Violent victimization: SMI, mental illness, mental disorder, psychiatric disorder, psychopathology, and victimization.
The search yielded 31 studies of violent perpetration and 10 studies of violent victimization. Few studies examined perpetration and victimization in the same sample. Prevalence rates varied by the type of sample and time frame (recall period). Half of the studies of violent perpetration examined inpatients; of these, about half studied only committed inpatients; these studies reported higher rates of violent perpetration (17% – 50%) compared with other samples. Among outpatients with severe mental illness, 2% to 13% had perpetrated violence in the past 6 months to 3 years, compared with 20% to 34% who had been violently victimized in the same time frame. Studies combining outpatients and inpatients reported that 12% to 22% had perpetrated violence in the past 6 to 18 months versus 35% who had been a victim of violence in the past year.
Violent perpetration and victimization are more common among persons with severe mental illness than in the general population. Victimization is a greater public health concern than perpetration. Ironically, the discipline’s focus on the perpetration of violence among inpatients may contribute to the negative stereotypes of persons with severe mental illness.
For decades, researchers have investigated violence perpetrated by persons with severe mental illness (1–6). This research has, in part, been driven by a common perception that persons with mental illness are dangerous (7–10). Far fewer empirical studies have examined the risk of violent victimization among persons with severe mental illness (11–20), and, to our knowledge, no literature review has been published. Moreover, no literature review has weighed the relative importance of violent perpetration and violent victimization in persons with severe mental illness.
Reviewing the literatures on perpetration and victimization is timely. Severe mental illness is estimated to affect 1 in 17 persons, or 6% of adults (13.2 million people) in the United States (21). Long-term psychiatric hospitalizations are now rare; the median length of stay has been reduced from 41 days in 1971 to 5.4 days in 1997 (22). Consequently, more persons with severe mental illness now live in the community. Moreover, the recent homicides in Omaha and at the Virginia Polytechnic Institute and State University (Virginia Tech) have highlighted the importance of examining the role of mental illness in violent perpetration.
In this article, we review empirical studies conducted in the United States of violent perpetration and violent victimization in persons with severe mental illness published since 1990. We also weigh the relative importance of violent perpetration and violent victimization in persons with severe mental illness as public health concerns. Finally, we suggest directions for future research and discuss the implications of our conclusions for treatment and public health policy.
Severe mental illness refers to a subset of psychiatric disorders (psychotic disorders and major affective disorders) characterized by severe and persistent cognitive, behavioral, and emotional symptoms that reduce daily functioning (21). Symptoms, despite medication and treatment, periodically worsen such that short-term hospitalization is required (21).
All searches, restricted to studies conducted in the United States, were performed on three commonly used computerized bibliographic databases: MEDLINE, PSYCH INFO, and Web of Science. Studies were included if they met the following criteria: (1) Published empirical investigations of recent (not lifetime) prevalence or incidence of violent perpetration or violent victimization; (2) Studies of persons in treatment for severe mental illness; of special populations (e.g., homeless persons) if separate rates were reported for persons with severe mental illness; and of non-treatment (community) samples if investigators compared persons with and without severe mental disorders.
Our searches and keywords are as follows:
Incidence refers to the number of new cases of a disease that occur during a specified period of time in a population at risk for developing the disease (24). We could not find any studies that measured the incidence of violent perpetration.
Prevalence refers to the number of affected persons present in the population divided by the number of persons in the population within a given period of time (24). Table 1 lists studies of the prevalence of violent perpetration by the type of sample.
Table 1 shows that four studies examined outpatients (11,12,25,26). Prevalence of violence ranged from 2.3% (11) to 13.0% (25) and varied by the time frame (recall period) and the type of measure. The rates in the study by Brekke et al. (11) were lower than other studies’ because of their narrow definition of violence -- criminal charges for a violent crime in the past 3 years (2.3%) and contacts with police for aggression against others (6.4%). Conversely, the rates in the study by Bartels et al. (25) were likely higher than other studies’ rates because they examined self-reported violence among “the most severely disturbed patients” discharged from a state hospital. One study (26) used a sample too small (n=42) to generate reliable prevalence rates.
Table 1 shows that two studies examined psychiatric emergency room patients. Prevalence of violence ranged from 10.0% in the 2 weeks prior to patients’ emergency room visits (27) to 36.0% in the previous 3 months (28). McNiel et al. (27) may have found lower rates than other studies because they used mental health records to assess violence instead of self-report. Conversely, Gondolf et al. (28), who studied an “accidental” sample (n=389), may have found rates higher than other studies because they used self-reports and hospital records.
Table 1 shows that of the 31 published articles on violent perpetration in persons with severe mental illness, approximately half (48%, 15/31) (29–43) examined samples composed solely of inpatients. Of these, more than half (53%, 8/15) (29,34–39,41) included committed inpatients in their sample; four studies examined only committed inpatients (29,34,35,41). Prevalence rates vary widely, depending on the measure of violence and when the violence took place relative to the hospitalization.
Findings varied by time frame and by the type of illness; prevalence ranged from 14.2% among voluntary inpatients in the month before hospitalization (43) to 50.4% among committed inpatients in the 4 months prior to hospitalization (41). Committed inpatients may have higher rates of violence than other inpatients because of the national dangerousness standard used in many states’ commitment procedures, in which being “imminently” or “probably” dangerous precipitates hospitalization (44). Overall, the prevalence of violence was highest in studies of committed inpatients; those that used broader definitions of violent behavior (35,41); and those that measured self-reported violence (35,41) instead of using medical chart reviews (29,40) or official records (medical records, police records, and civil commitment forms) (31).
Prevalence rates varied from 16.0% (during the first week of hospitalization) to 23.0% (occurring any time during hospitalization). Table 1 shows that all four studies of violence during hospitalization examined patients in locked units and assessed violence using medical chart reviews (29,32,34).
Findings varied by type of sample and time frame; the lowest prevalence rates of self-reported “physical violence” (3.7%) were reported within two weeks after discharge by voluntary inpatients (42); the highest rates (27.5%) were reported in inpatients participating in the MacArthur Risk Violence Assessment Study one year after discharge, of whom over two-fifths were involuntarily committed (39). Involuntary patients were significantly more likely to be violent at follow-up than voluntary patients (45). Table 1 shows that the prevalence of violence in the MacArthur Risk Violence Assessment Study decreased with time. Of note, after controlling for substance abuse, there were no significant differences in the prevalence of violence between their sample and a control group of persons without mental disorders who lived in the community (39).
In sum, studies of inpatients with severe mental illness show that violent perpetration is most prevalent among committed patients prior to hospitalization, when violence may have precipitated their commitment. Moreover, prevalence rates are higher in studies that assess a broad range of self-reported violent acts than in those that rely solely on medical chart reviews.
Six studies combined inpatients and outpatients. All collected self-reported data; time frames varied from the past six months (46,47) to the past 18 months (48). Prevalence rates of violence ranged from 12.3% to 26.0% (46–51), lower than prevalence rates found in most studies of inpatients and higher than those found in most studies of outpatients. The highest rate (26.0%), reported by Elbogen et al. (49), combined self-reported violent behavior and any arrest (violent and non-violent), which may have inflated their rates.
Table 1 shows that only four of the 31 articles examined community samples (52–55). Data for these four articles were drawn from two multi-site community surveys of mental disorders (National Institute of Mental Health Epidemiologic Catchment Area [ECA] survey (53–55) and the National Comorbidity Survey [NCS] (52)). Because these surveys were not designed to assess violent behavior, the authors derived a dichotomous variable, any violence (yes/no), from the sections on mental disorders, physical health, and recent life events.
In studies using the ECA data (53–55), the authors used five questions from the Diagnostic Interview Schedule’s antisocial personality disorder and alcohol use disorder modules; respondents were scored as violent if they responded positively to one or more items. Items varied in severity, from “physical fighting while drinking” to “used weapon in a fight.” Among persons with severe mental illness, prevalence of any violent behavior in the past year ranged from 6.8% to 8.3% (53–55) -- up to 4 times higher than among persons who were not diagnosed with a mental disorder. Swanson et al. also examined differences by age, gender, and socioeconomic status when comparing persons with major mental disorders and persons without any disorder (54,55); however, cell frequencies were too small to estimate the effect of major mental disorder separately within sociodemographic categories (56).
In the study using the NCS data (52), respondents were scored as violent if they reported they “had serious trouble with the police or the law” or “had been in a physical fight.” Analyses focused on differences among diagnostic groups. Prevalence of violence ranged from 4.6% in the past year for a lifetime diagnosis of major depressive disorder to 16.0% for a past-year diagnosis of bipolar disorder, 2 to 8 times higher than persons without a mental disorder. Findings from this study, however, conflate violent behavior with involvement with the police, which may or may not have been precipitated by violence.
Most general population studies of crime victimization -- such as the National Crime Victimization Survey (NCVS) (23) -- examine incidence. To our knowledge, only one study of adults in treatment with severe mental illness investigated the incidence of recent violent victimization (19). Using the same instruments as the NCVS, Teplin et al. (19) examined 936 randomly selected persons with severe mental illness from a random sample of treatment facilities --outpatient, day treatment, and residential treatment -- in Chicago, Illinois. There were 168.2 incidents of violent victimization per 1000 persons per year, more than 4 times greater than general population rates. Incidence ratios remained statistically significant even after controlling for sex and race/ethnicity.
Table 2 shows that all 10 studies examined self-reported prevalence of victimization. Prevalence varies because of differences in sample sizes, time frames, and the type of sample. Some studies had samples too small to generate reliable prevalence rates of relatively uncommon events such as violent victimization (15). Studies of treatment populations with larger samples (n≥100) found prevalence rates of recent violent victimization between 8.2% (in the past 4 months) (16) and 35.0% (in the past year) (14). The largest study of homeless persons with severe mental illness (17) found that 44.0% had been victimized violently in the past 2 months. Among studies that assessed violent victimization occurring within the past year -- the same time frame as the NCVS -- prevalence rates ranged from 19.7% (19) to 35.0% (14), compared with 2.9% in the NCVS.
Prevalence rates appear to vary by type of victimization. However, these differences may be because of the way victimization was measured. For example, White et al. (20) asked only one question about victimization in the past 6 months. Other studies (13,14,17,19) collected detailed information on the type of victimization.
Prevalence rates also varied by the type of sample. For example, 19.0% of the sample of outpatients and residential treatment patients in the study by Teplin et al. (19) and 35.0% of the combined sample of inpatients and outpatients in the study by Goodman et al. (14) had been victims of physical assault in the past year. Similarly, prevalence of rape and sexual assault in the past year ranged from 2.6% (19) among outpatients to 12.7% (14) in a combined sample of outpatients and inpatients. Prevalence of victimization in homeless persons with severe mental illness is generally higher than in treatment samples (13,17). Irrespective of the type of sample and victimization, prevalence is much higher in all studies listed in Table 2 than in the general population, as found in the NCVS (23).
Are persons with severe mental illness more likely to be perpetrators of violence or victims of violence? Table 3 summarizes and compares the prevalence of violent perpetration and victimization from the studies in Table 1 and Table 2.
Only three studies assessed perpetration and victimization in the same participants. Brekke et al. (11) found that among outpatients with schizophrenic disorders, 6.4% had contact with police for “aggression against others” in the past three years compared with 34.0% who reported being violently victimized. The marked differences in rates may be because violent perpetration was counted only if the person had contact with the criminal justice system; many violent behaviors do not come to the attention of the police or culminate in formal processing (57,58). Had the authors used a broader measure of violence, the reported differences between violent perpetration and victimization might have been less dramatic. Brunette and Drake (12) had similar findings; 6.4% of their sample had been physically aggressive in the past year compared with 19.8% who had been a victim of a violent crime in the past year. In the Outpatient Commitment Study, Swanson et al. (41) found that among committed inpatients, the prevalence of violent perpetration in the four months prior to commitment ranged from 17.8% (for “serious violence”) to 50.4% (using a broader measure of violence) (41); in contrast, 8.2% reported violent victimization (16).
Why is the prevalence of violent perpetration so high in the Outpatient Commitment Study? Most likely, it was because participants were sampled soon after commitment. The authors did not indicate the proportion of their sample that was committed because of their violent behavior. Discrepancies between violent perpetration and victimization might also have occurred because of differences in the definitions of violence. Victimization was narrowly defined as self-reported “violent crimes;” violent perpetration referred to a range of violent behaviors elicited from patients and their collaterals, as well as hospital records.
The MacArthur Violence Risk Assessment Study provides some information comparing violent perpetration and violent victimization among discharged inpatients. The authors report that 13.5% had perpetrated violence (39) and 15.2% had been victims of violence (18) ten weeks after discharge from a psychiatric inpatient unit. However, because one article used a subsample (18), the rates are not directly comparable.
Other studies listed in Table 3 show that irrespective of the type of sample and regardless of the time frame, violent victimization is more prevalent than violent perpetration. For example, among outpatients and residential patients with severe mental illness, 20.0% to 34.0% (11,12,19,20) (depending on the time frame and gender) had been a victim of recent violence compared with 2.0% to 13.0% (11,12,25) who had perpetrated violence. Similarly, in samples combining outpatients and inpatients, 35.0% had reported a violent victimization in the past year (14) compared with 12.0% to 22.0% (depending on whether the time frame was 12 months or 18 months) who had reported recent violent perpetration (46–50).
Violent perpetration and victimization are more common in persons with severe mental illness than in the general population (19,53–55). Studies analyzing the Epidemiologic Catchment Area data found that approximately 2% of persons without a mental disorder perpetrated violence in the past year compared with 7% to 8% of persons with severe mental illness (53–55). For victimization, the disparity between the general population (3%) and persons with severe mental illness (25%) is even greater, as found in the NCVS (19).
Overall, our review does not support the stereotype that persons with severe mental illness are typically violent (7–10). This stereotype may persist, in part, because of researchers’ focus on inpatients. Although fewer than 17% of persons with severe mental illness in the United States are hospitalized (59), nearly half of the studies that investigate violence in persons with severe mental illness examined only inpatients (29–43). Among these, the largest and most well-cited studies focused on involuntarily committed inpatients. The Outpatient Commitment Study included only involuntarily committed inpatients. Two-fifths of the MacArthur Risk Assessment Study’s sample had been involuntarily committed, a significant predictor of subsequent violence (45). Because commitment criteria include imminent dangerousness (to self or others) (44), findings derived from samples of involuntarily committed patients are generalizable only to the most acutely disturbed patients who have required the courts to intervene.
How much violence in the United States is caused by persons with mental illness? One study found that overall, the attributable risk of mental illness to perpetrating violence in the United States is approximately 2% (52); by comparison, demographic variables -- gender and age -- are more powerful predictors of violence (52); 75% of violent crimes are perpetrated by males younger than 18 years (60).
Despite the small attributable risk of severe mental disorders on violent perpetration, negative stereotypes of persons with severe mental illness dominate the public’s view (61,62) and behavioral scientists’ focus. Among 39 studies that met our inclusion criteria, 79% (n=31) studied violent perpetration. The focus on violent perpetration extends to non-empirical articles as well. A computerized search of MEDLINE and PsycINFO yielded 283 empirical or review articles mentioning crime victimization in persons with mental illness; more than 13 times that many articles were found on violent perpetration (19).
Based on our review, we suggest the following directions for research and mental health policy.
We suggest that future studies:
We suggest the following:
Although society may regard persons with mental illness as dangerous criminals (8,10), our review of the literature shows that violent victimization of persons with severe mental illness is a greater public health concern than violent perpetration. Although some symptoms of severe mental illness are correlated with violence, serious mental disorder accounts for only a modicum of violence. Ironically, the discipline’s focus on the perpetration of violence among inpatients may contribute to the negative stereotypes of persons with severe mental illness, which are often based on the label of “mental patient,” not on observed behavior (83,84). We must balance the dual public health concerns of protecting the safety of the public and protecting persons with severe mental illness from crime victimization.
This work was supported by a MERIT award, R37MH47994, from the National Institute of Mental Health, Bethesda, MD, and by the National Institute of Mental Health grant R01MH54197 (Division of Services & Intervention Research).
The official published article is available online at: http://ps.psychiatryonline.org/article.aspx?articleid=99084
This study could not have been accomplished without the contribution of Erin G. Romero, B.S.
Disclosures: None for any author