Mental Illness and Violence: Proof or Stereotype?

Next | Previous | Table of Contents

Appendix A: Annotated Bibliography

Introduction:

The following bibliography contains annotations to key quantitative and qualitative articles bearing on the central study questions relating to mental illness and violence. For ease of reading, articles have been grouped according to major topic headings. Topic areas have been grouped so as to reflect the various lines of inquiry and evidence that have accumulated in this area. The areas are:

• Community Studies
• Studies of Psychiatric Patients
• Studies of Incarcerated offenders
• Other Empirical Studies of Interest
• Review Articles and Key Position Papers.

Within each group, the annotations are presented alphabetically by author and title. A brief glossary of key technical terms is contained in Appendix B for reference.

To assist the non-scientific reader in drawing pertinent comparisons across articles, all empirical studies have been critically reviewed and summarized according to a standardized abstract structure: (a) purpose and objective; (b) research design; (c) study setting; (d) study subjects; (e) measures used; (f) main findings; (g) conclusions; (h) methodological critique; and (i) causality. A comment section is also included in the “Other Empirical Studies of Interest” category. In the “Review Articles and Key Position Papers” category, key points relevant to the critical review are highlighted, and, where appropriate, commentary is provided.

Causality:

The term ‘causality' refers to whether or not the study could be used to arrive at a causal determination regarding the relationship between mental illness and violence. Causality was judged strictly according to epidemiological criteria. At the outset, it is important to point out that no study provided strong evidence in support of a causal relationship between mental illness and violence.

Community Studies: Empirical Studies Based on General Population Samples:

Barring other methodological problems, empirical studies based on representative samples of the general population are generally preferred for drawing causal inferences because they avoid “selection bias” (Rothman, 1986). Selection bias is a systematic error that creeps into studies when subjects are chosen from treated or institutional populations that are known to be at higher risk of violence. When rates of violence in selected groups are compared to general population, the selected populations will often appear to have higher rates of violence. This is usually attributable to the fact that they were more violence-prone to begin with, not because mental illness causes violence. Because they are time-consuming and complex, population-based studies are generally few in any area of inquiry. They are also costly because large samples are required to obtain large numbers of persons with mental illness and violence to support statistical analysis.

Studies of police-citizen encounters have been included in this section to acknowledge the pivotal role of police decision-making in determining what will happen to those persons with a mental illness who, while living in the community, may engage in disordered or violent behaviour. They may either be processed through the criminal justice system, or they may be taken to a psychiatric facility for assessment and treatment. Given police discretionary power, persons who are thought to be mentally ill and who are violent may be arrested more often than non-mentally ill offenders. This could account for statistical relationships between violent criminality and mental illness in incarcerated populations. The possibility of shifting mentally ill offenders into the criminal justice system has been referred to as the “criminalization of the mentally ill”. This poses serious selection problems for studies that (a) use arrest or conviction for a violent offense as a measure of violence, or (b) select subjects for study from arrested or incarcerated populations.

Arboleda-Flórez, J. and Holley, H.L. (1988) Criminalization of the mentally ill: Part II. Initial detention. Canadian Journal of Psychiatry, 33, 87-95.

See also: Holley, H.L. and Arboleda-Flórez, J. (1988). Criminalization of the mentally ill: Part I. Police perceptions. Canadian Journal of Psychiatry, 33, 81-86.

Purpose and Objective: Compare police-identified ‘mentally ill' with police-identified ‘normals' on the basis of sociodemographic, legal, clinical, and outcome variables. The authors set out to test the hypothesis that the groups would differ in terms of important socio-clinical characteristics.

Research Design: Prospective longitudinal study.

Setting: Calgary, Alberta, Canada.

Study Subjects: Study subjects were 350 persons who came into contact with the police during the latter two weeks of October, 1984.

Measures: Police were asked to rate subjects' observable behaviour on a continuum from normal (score of 0) to severely abnormal (1-7). Then police were asked to attribute what they believed the cause of this behaviour to be, alcohol, drugs, mental illness or other, and to indicate whether they believed a psychiatric examination was warranted. Arrest information was provided by the police reports.

Main Findings: A total of 89 individuals were identified by police as acting abnormally during the arrest; 261 were considered to be acting normally, given the situation. Police-identified mentally ill did not record a greater number of crimes against persons, property, or miscellaneous charges compared to police-identified normals. Police-identified mentally ill recorded fewer victimless crimes and slightly more motor traffic violations.  Police-identified mentally ill were slightly more likely to be recommended by police for detention compared to normals but the difference was not large. They were no more likely to be detained compared to normals.

Conclusions: Because any differences noted between the study groups were small, the authors accepted the conclusion that the groups did not differ in important social, clinical, or legal respects.

Methodological Critique: It is not clear to what extent misclassification of mentally disordered offenders by police could account for the lack of differences noted across the groups. Also, because this study focussed on individuals who came into contact with the police, findings cannot be used to draw etiological inferences concerning causal relationships between mental illness and violence. It is interesting to note, however, that despite a different methodology, these findings are consistent with those reported by Teplin (1985) in her study of police-citizen contacts.

Causality: A causal inference cannot be made on the basis of these findings.

Bland, R. and Orn, H. (1986). Family violence and Psychiatric disorder. Canadian Journal of Psychiatry, 31, 129-137.

Purpose and Objective: To study the relationship between family violence and psychiatric disorder in the general population.

Research Design: Cross-sectional survey.

Setting: Edmonton, Alberta, Canada.

Study Subjects: A representative sample of 1200 residents living in the community.

Measures: Psychiatric disorder was measured using the Diagnostic Interview Schedule (DIS), a structured questionnaire for lay-interviewers that is computer-coded to generate DSM-III/R diagnostic categories. Lifetime prevalence of three diagnoses were examined (1) antisocial personality disorder, (2) major depression, and (3) alcohol abuse/ dependence. Any individual meeting the symptoms for these disorders at any time in their lives would be recorded as positive for mental disorder. Measures of family violence were derived from DIS items covering the subject's relationship with a spouse or partner, violence or neglect towards children, and violence outside of the family. This questionnaire focused on physical abuse (rather than emotional or psychological abuse) measured by questions about behaviours such as hitting or throwing. Questions were worded to reflect lifetime prevalence of these behaviours.

Main Findings: Nearly 55% of those with a diagnosis were involved in violent behaviour compared to 15.5% of those with no diagnosis. Persons with one or more of the three diagnoses studied were 6.5 times more likely to be involved in violence than those without a disorder. The risk of violence was greatly elevated among those diagnosed with a comorbid alcohol abuse disorder. For example, where alcohol was combined with antisocial personality and/or  depression, 80-93% were involved in violence.

Conclusions: The authors do not propose a simple cause and effect explanation for their findings. Rather, they suggest that their findings should alert clinicians that, when seeing persons with psychiatric disorders, family violence may be a possibility. This particularly applies to those patients with comorbid alcoholism.

Methodological Critique: As the authors point out, findings from this study cannot be used to infer a causal relationship between mental disorder and violence. Because the study used cross-sectional data and lifetime prevalence figures for both mental disorder and family violence, no assumptions can be made about the temporal ordering of these factors. Therefore, the violence may actually have preceded the mental illness. Further, DSM-III diagnoses of antisocial personality disorder and alcoholism are two of the disorders that are most likely to be defined in terms of violent behaviours. Therefore, any observed association between these and violence is likely to be a result of the way in which the disorders have been defined.

Causality: A causal inference cannot be made on the basis of these findings.

Bonovitz, J.C. and Bonovitz, J.S. (1981). Diversion of the mentally ill into the criminal justice system: The police intervention perspective. American Journal of Psychiatry, 138(7), 973-976.

Purpose and Objective: This study is an evaluation of the effects of the Pennsylvania Mental Health Procedures Act passed in 1976. The authors hypothesized that the police would be asked to handle more incidents involving mentally ill individuals after the passage of the Act and that officers would use the Criminal Code to expedite the removal of these individuals from the community.

Research Design: Although the authors do not name their design, the data collection procedures are consistent with a historical cohort study.

Setting: A suburban police department serving a population of 100,000 in Upper Darby Township, Pennsylvania, USA.

Study Subjects: The authors studied an undisclosed number of police incidents involving mentally ill individuals.

Measures: Archival data from 1975-1979 from the police files were used to identify reports in which subjects were clearly identified as “mentally disturbed”. During a six-month period in 1979, 248 incidents were studied in more detail as to their outcome.

Main Findings: Mental illness-related incidents increased 227.6% from 1975 to 1979 whereas non-mental illness-related incidents decreased 9% during this time (excluding felonies which rose to 5.6%). There was an 82% increase in the number of disorderly conduct offenses during this time period, in part because of a change in the  way these were classified. Considering the 248 incidents that occurred during six months in 1979, police officers made 13 arrests.

Conclusions: The authors did not comment on the increased number of police-mentally disordered incidents during the time period but did note that their data did not support the hypothesis that noncommittable mentally ill were being arrested and jailed as an expedient means of removing them from the community.

Methodological Critique: This study is short on methodological detail, for example the total number of incidents studied. Nevertheless, it does provide evidence to indicate that mentally disordered offenders have increasingly come into contact with the police for socially disruptive behaviours and that police discretion plays an important role in their disposition. The police in this community had taken specialized training to recognize signs of mental health problems, and to improve their techniques for resolving such incidents. Perhaps, as a result, they firmly believed that these individuals should not be arrested or held responsible for minor criminal offenses. These findings have important implications for studies focussing on incarcerated offenders because they clearly demonstrate the importance of police decision-making and highlight the selected nature of incarcerated samples.

Causality: A causal inference cannot be made on the basis of these findings.

Hodgins, S. (1992). Mental disorder, intellectual deficiency, and crime: Evidence from a birth cohort. Archives of General Psychiatry, 49, 476-483.

Purpose and Objective: To examine the relationship between mental disorder and crime and the relationship between intellectual deficiency and crime.

Research Design: The author describes the research design as a longitudinal prospective study of a birth cohort to age 30 years. However, data were collected from central registry files in 1983. Therefore, this study could more properly be described as a historical cohort study design.

Setting: Stockholm, Sweden.

Study Subjects: Study subjects were identified from among the 15,117 persons born in Stockholm in 1953 and still residing there in 1963.

Measures: Mentally ill were defined as those having had a psychiatric admission (N=603). Intellectually handicapped subjects (N=192) were defined as those who were placed in special high school classes but never admitted to a psychiatric ward. The comparison group consisted of all those who had never been admitted to a psychiatric ward or to an institution or class for the intellectually handicapped. Crime was measured by criminal conviction data obtained from a central registry. Violent crimes included all offenses involving the use or threat of physical violence (e.g. assault, rape, robbery, unlawful threat, and molestation).

Main Findings: Compared to their “normal” controls, men with major disorders were 4.16 (95% CI, 2.23 to 7.78) times more likely, and women were 27.45 (95% CI, 9.80 to 76.88) times more likely to have been convicted of a violent offense. Note that the wider confidence intervals for women reflects their smaller sample size.

Conclusions: The author considered that the result of this study supported the view that aggressivity is associated with mental illness.

Methodological Critique: Findings are biased in favour of a relationship between mental illness and violence because they were based on a subgroup of the mentally ill (those admitted to a psychiatric facility) who are known to be more likely to be dangerous. As such, these findings cannot be used to draw etiological inferences concerning the general relationship between mental illness and violence. Also, while the study design would have allowed for temporal ordering of events, it is not clear whether criminality occurred subsequent to the mental illness in every case. The author does comment that the criminality of subjects who eventually developed major mental disorders often appeared in early adolescence, well before the mental disorder was diagnosed. This statement indicates that study factors have not been temporally ordered.

In a Letter to the Editor, Weiler [1994, Archives of General Psychiatry, 51, 71] points out that 43% of women and 40% of the men with major mental disorders also had a substance abuse diagnosis. Because those with substance abuse problems were 20 (men) and 32 (women) times more likely to be convicted, the tendency to commit crime may be a function of substance abuse and not major mental illness. If the substance abusers were removed from this group, it would be plausible to expect that the rate of crime would be lower among those with major mental illness.

Causality: A causal inference cannot be made on the basis of these findings.

Monahan, J., Caldeira, C. and Friedlander, H.D. (1979). Police and the mentally ill: A comparison of committed and arrested persons. International Journal of Law and Psychiatry, 2, 509-518.

Purpose and Objective: To provide information describing how police use their discretionary power vis-à-vis persons with a mental illness.

Research Design: Although not specified by the authors, the data collection procedures used are consistent with a cross-sectional survey.

Setting: Orange County, California, USA.

Study Subjects: The study subjects were 100 police officers from various cities. The authors do not describe how the officers were selected, but make reference to a random sample in their conclusions. As no sampling strategy is described, it is unclear whether results can be considered to be representative of all police officers or  whether they simply reflect the beliefs of the 100 that were studied.

Half of the officers were interviewed subsequent to their having petitioned a person for involuntary civil committal and half were interviewed after having booked a person for arrest on a criminal charge. It is not clear whether the latter group consisted of specific experiences with mentally ill offenders booked on a criminal charge.

Measures: The officers reported their perceptions of the 100 persons they arrested or committed with respect to severity of mental illness, dangerousness, and grave disability.

Main Findings: Committed persons were more likely to be perceived as mentally ill compared to those arrested. In addition, police perceived committed persons to be slightly more likely to be violent toward others, 5 times more likely to be gravely disabled, and 20 times more likely to be harmful to themselves. In 30% of the arrests, the police could have petitioned for civil committal. They did not because they did not believe the degree of mental illness, grave disability, or violence was sufficient to sustain a committal. Similarly, in 30% of the committals, police could have proceeded with an arrest. Police chose to commit because they believed that the individuals lacked criminal intent or that they would benefit from treatment.

Conclusions: The authors believe that their study has produced little evidence to support the hypothesis that seriously mentally ill persons are criminalized or that serious law breakers are being psychiatricized.

Methodological Critique: As the authors did not have an independent measure of mental illness against which to judge the validity of police perceptions of mental illness, their conclusion that mentally ill persons are not criminalized or criminals are not psychiatricized goes beyond the scope of their data. The main point of interest for the present review is that detention in mental hospitals or jails is the final outcome of a series of decisions made by various authorities. Therefore, samples of mental patients or criminals are not representative of all individuals and may be selected into these institutions based on the gatekeeper perceptions of the severity of mental illness or propensity toward violence.

Causality: A causal inference cannot be made on the basis of these findings.

Swanson, J.W. (1993). Alcohol abuse, mental disorder, and violent behavior: An epidemiologic inquiry. Alcohol, Health & Research World, 17(2), 123-132.

Purpose and Objective: The aim of this study is to test five hypotheses that have been postulated concerning the relationship between mental illness, alcohol abuse, and violence:

1. The alcohol abuse hypothesis postulates that  the relationship between mental disorder and violence can be largely explained by the association between alcohol abuse and violence.
2. The sociodemographic hypothesis postulates that the relationship between mental illness/alcohol abuse and violence is a result of the common socio-demographic correlates in these groups, particularly gender, age, and socio-economic status.
3. The comorbidity hypothesis suggests that the combination of alcohol abuse and mental disorder results in a substantially greater risk of violence than does either condition alone and occurs more often in men of lower socio-economic status.
4. The institutional selection hypothesis postulates that the apparent relationship between mental illness and violence can largely be explained by a bias in the populations studied. Most studies focus on persons who have been involuntarily treated or incarcerated which are more likely to be violent.
5. The psychopathology hypothesis suggests that the increased rates of violence observed among institutionally defined populations is a result of the greater psychiatric and alcohol abuse symptomatology of those individuals most likely to be found in these settings.

Research Design: Cross-sectional survey.

Setting: Data reflect two of the Epidemiologic Catchment Area (ECA) sites in the United States, Durham and Los Angeles.

Study Subjects: A representative sample survey of 7,053 adults.

Measures: As in previous ECA studies, psychiatric disorder was measured using the Diagnostic Interview Schedule (DIS), a structured questionnaire for lay-interviewers that is computer-coded to generate DSM-III/R diagnostic categories. A one-year period prevalence of psychiatric disorder was used such that a person was counted as a case if they met the DSM-III criteria for a given disorder during the 12 months preceding the interview. Four items from the DIS were also used to define violence: (1) using a weapon in a fight since age 18, (2) having been in more than one fight in which blows were swapped since age 18, (3) ever hitting or spanking a child hard enough to injure, and (4) persons who were married or had lived with someone as if married were asked if they had ever hit or thrown things at their wife or partner. For the first four questions, respondents were asked when was the last time they did any of these things. Based on these answers, a four-item violence index was developed reflecting violent behaviour toward others during the one-year period prior to the interview. A respondent was counted as “violent” if one or more of the items was answered positively.

Main Findings: A total of 193 individuals were identified as violent using the measures described above. Five statistical models were tested—one relating to each of the study hypotheses. Results failed to support the alcohol abuse hypothesis or the socio-demographic hypothesis. Modest support was found for the remaining three hypotheses.

Conclusions: Major mental illnesses that were uncomplicated with alcohol abuse were associated with an increased risk of violence. Persons with comorbid mental and substance abuse disorders were at even greater risk of violence. The apparent increase in violence among younger, lower socio-economic males was found to be largely due to the increased prevalence of alcohol abuse and comorbidity in this group. A history of arrest and psychiatric hospitalization was also found to be associated with an increased probability that a person would be violent.

Methodological Critique: The hypotheses tested in this study are clearly conceptualized and described. The cross-sectional nature of this study precludes us from identifying the temporal ordering of study factors and, therefore, from drawing clear causal inferences from the findings. However, results are interesting and suggestive, particularly given the controversy in the literature regarding the role of alcohol abuse as a causal factor in violence among persons with a mental illness, and the uncertainty about how to deal with socio-economic status. The authors attempted to control for the fact that the questions used to identify violence were among those used to diagnose some mental illnesses by excluding those from the analysis. What remains unclear is the extent to which current diagnostic practices, in general, reinforce the relationship between mental illness and violence through the DSM nosology.

Causality: Although well-executed and analyzed, this study does not provide sufficient evidence to draw a causal inference.

Swanson, J.W., Holzer, C.E., Ganju, V.K. and Jono, R.T. (1990). Violence and psychiatric disorder in the community: Evidence from the Epidemiologic Catchment Area Surveys. Hospital and Community Psychiatry, 41(7), 761-770.

Purpose and Objective: To examine the relationship between violence and psychiatric disorder among adults living in the community.

Research Design: Cross-sectional survey.

Setting: The Epidemiological Catchment Areas in the United States (New Haven, Baltimore, St. Louis, Raleigh-Durham, and Los Angeles).

Study Subjects: Representative sample surveys of the adult household resident population were surveyed in 1980 and 1983 with 3,000-5,000 residents per site. Data for this analysis were pooled from the Baltimore,  Raleigh-Durham, and Los Angeles samples for a sample size of 10,059. Data were weighted based on respondents' probabilities of selection.

Measures: Psychiatric disorder was measured using the Diagnostic Interview Schedule (DIS), a structured questionnaire for lay-interviewers that is computer-coded to generate DSM-III/R diagnostic categories. A one-year period prevalence of psychiatric disorder was used such that a person was counted as a case if s/he met the DSM-III criteria for a given disorder during the 12 months preceding the interview. Five items from the DIS were also used to define violence: (1) hitting or throwing things at a spouse or partner; (2) spanking or hitting a child hard enough to cause bruises; (3) fist fighting since age 18 with someone other than spouse; (4) using a weapon (e.g. stick, knife, or gun) since age 18; and (5) getting into physical fights while drinking. Individuals were counted as positive for violence if they had responded positively to one or more items and said that the behaviour had occurred within the year prior to the interview. A measure of socio-economic status (based on occupation, educational level, and household income) was also used.

Main Findings: Individuals in the community meeting the criteria for a DIS/DSM-III psychiatric disorder were more likely to engage in violent and assaultive behaviour compared to those that did not meet the criteria for a DIS/DSM-III disorder. For example, more than half of the 368 individuals reporting violent behaviour in the preceding year met the criteria for a psychiatric disorder compared to 19.6% of non-violent offenders. The highest percentages of violence were reported among substance abusers (19.2% to 34.7% depending on the category). With the exception of phobias which showed a prevalence of violence of 5.0%, all other diagnostic categories (e.g., schizophrenic disorders, depressive disorders, and anxiety disorders) showed prevalences ranging from 10.7% to 12.7%. With one exception, each separate disorder category studied (without regard for multiple diagnoses) was associated with a higher risk. Persons with non-comorbid anxiety disorder had the same rate of violence as those with no disorder. Those with affective disorder were at only slightly increased risk. Those with schizophrenia were at somewhat elevated risk but not as great as public perceptions would suggest. Conversely, those with substance abuse disorders were at greatly elevated risk and also appeared to commit more severe acts of violence with greater frequency.

The percentage of respondents reporting violent behaviour increased from 2.1% among those with no diagnosis, to 6.8% among those with 1 diagnosis, 17.5% for 2 diagnoses, and 22.4% for 3 or more diagnoses.

Conclusions: Individuals in the community with psychiatric disorders are more likely to engage in assaultive behaviour than those who are not mentally ill. Alcohol and drug abuse and the presence of more than one diagnosis increase the risk of violence substantially.  Public fear of persons with schizophrenia living in the community is largely unwarranted, though not totally groundless. Findings indicate that persons are at much greater risk of being assaulted by a person with a substance abuse disorder than by someone suffering from a major mental illness such as schizophrenia.

Methodological Critique: While this study represents state-of-the-art psychiatric epidemiology, it can be used only to suggest a statistical association between mental illness and violence. Because data were drawn from a cross-sectional survey, the temporal ordering of factors could not be established. Both mental disorder and violence were measured during the same one-year period. Therefore, findings cannot be used to infer causality. Second, as the authors point out, the measures of violence were less than ideal because they reflected DIS questions that were used, in part, to arrive at a diagnoses of mental disorder. Thus, some level of association would be expected by definition.

Causality: A causal inference cannot be made on the basis of these findings.

Teplin, L.A. (1985). The criminality of the mentally ill: A dangerous misconception. American Journal of Psychiatry, 142(5), 593-599.

Purpose and Objective: To assess the relative criminality of mentally disordered persons by focussing on the initial point in the criminal justice system to avoid biases associated with police discretionary powers of arrest.

Research Design: Unspecified by author. Subjects were observed within the context of the police encounter, then arrest data were collected. This is consistent with a short-term follow-up design.

Setting: Large metropolitan city in the United States (population over one million persons).

Study Subjects: Police officers (N=283) were randomly selected and observed, first hand, in their day-to-day interactions with the public. Observations were conducted during all hours of the day, on all days. Evenings and weekends were oversampled. Data were collected in two precincts reflecting a wide range of socio-economic profiles.

Measures: It was not feasible for researchers to conduct full-scale structured diagnostic assessments within the context of police-citizen encounters. Therefore, a symptom checklist that listed characteristics of severe mental disorder (confusion or disorientation, withdrawal or unresponsiveness, paranoia, inappropriate or bizarre speech and/or behaviour, self-destructive behaviours) was used to measure mental illness. Subjects were defined as suffering from a mental disorder if they possessed at least one of these traits and received a global rating of severe mental disorder by the fieldworker. Results from this checklist were validated against findings from 61 randomly selected jail detainees who underwent the Diagnostic Interview Schedule with 93.4% agreement on the definition of severe mental disorder. Criminality was measured by arrest. Crimes were divided into violent personal crimes, interpersonal conflicts, major property crime, minor property crime, public health, safety or decency offenses, and public order offenses.

Main Findings: Police encounters with severely mentally disordered persons occurred infrequently (4% of 2,122 persons). Mentally disordered were more likely to be subjects of concern or objects of assistance and only slightly more likely than non-mentally disordered to be considered suspects. For those who were suspects, type of crime was not related to the presence or absence of mental disorder. The mentally ill did not commit serious crimes at a rate disproportionate to their numbers.

Conclusions: The pattern of crime among mentally disordered suspects is substantially similar to non-mentally disordered suspects.

Methodological Critique: The major strength of this study is that a representative sample of all detected violations (i.e. contacts) were examined regardless of police officer's dispositions. This permitted researchers to control any potential bias arrest rates resulting from police officer's propensity to arrest mentally ill more frequently than non-mentally ill, once initial contact had been made. A second strength of this study is the reliance on observational data, rather than archival or secondary data to measure mental illness and arrest. Finally, because a symptom checklist was used, researchers have succeeded in deriving a definition of mental illness that is largely independent of criminality and have avoided the ‘confounding by definition' that has predominated studies using DSM-III/R criteria. In short, this study provides compelling evidence to support the conclusion that there are no important differences between mentally disordered and non-mentally disordered persons who contact the police with respect to the type of crimes they have been alleged to have perpetrated.

Causality: A causal inference cannot be made on the basis of these findings.

Next | Previous | Table of Contents