Mental Illness and Violence: Proof or Stereotype?

Previous | Table of Contents

Appendix B: Glossary of Terms

The following glossary of terms is offered to assist readers in understanding the scientific and technical language used  throughout this document. Unless otherwise referenced, definitions are drawn from Last, J.M. (1988). A Dictionary of epidemiology (2nd edition). Toronto: Oxford University Press, and Rothman, K.J. (1986). Modern epidemiology. Boston: Little, Brown and Company.

Bias is used to refer to an error in study design, data collection, or interpretation that can lead to faulty conclusions. Bias may be a result of misclassification of study subjects on either exposure or outcome factors, or it may result from studying selected groups of individuals such as mental patients or incarcerated offenders. Confounding (described below) is a third source of bias that must be controlled in order to make causal inferences.

Case-control studies choose study subjects on the basis of the outcome of interest (e.g. violent versus non-violent) and then survey subjects for exposure information (e.g. history of mental illness). Temporal ordering of factors may be a problem for case-control studies because investigators must rely on subject's memory for exposure information. Also, there is the possibility that cases (those who are violent) will remember past events differently than controls, leading to recall bias. As a result, causal inferences are usually not made on the basis of case-control studies without supporting evidence from cohort investigations.

Cohort Studies follow two or more groups through time who differ with respect to some purported causal factor (termed the “exposure”). The groups are compared with respect to their outcomes. An essential element of the cohort study is that all groups are free from the outcome of interest (in this case violence) at the outset of the study. Cohort studies make it possible to establish the temporal ordering of study factors, something that is crucial in making causal interpretations. Therefore, they are considered to provide the strongest evidence supporting causal interpretations.

Confounding occurs when the effects of two or more factors become mixed in a dataset such that it is difficult to see their independent effects. Confounding factors are related to the outcome of study. If they are also differentially distributed across the study groups, they can distort study results. Statistical adjustment may reduce or control the effects of confounding. Factors that may confound the relationship between mental illness and violence include age, sex, and past history of violence.

Cross-sectional surveys collect exposure (e.g. mental illness) and outcome (e.g. violence) simultaneously on a representative sample of persons. Thus, it is impossible to temporally order factors. Cross-sectional studies are considered to be ideal for generating hypothesis but are never used as the basis for inferring causality.

DSM Nosology, (published by the American Psychiatric Association) has been the accepted standard for  psychiatric diagnoses in North America for many decades. The most recent version is DSM-IV (published in 1994). However, most studies reported in the literature use DSM-III (Revised version). The DSM involves assessment on five axes, each of which refers to a different domain of information. Axis I refers to clinical disorders and other conditions that may be a focus of clinical attention. Axis II refers to personality disorders and mental retardation. Axis III is used to code medical conditions. Axis IV describes psychosocial stressors and environmental problems, and Axis V is used to make a global assessment of the patient's social and occupational functioning.

Epidemiology studies the occurrence of diseases and health events in human populations and their relationship with co-occurring “risk” and “protective” factors in order to derive causal explanations that can be used to lessen the burden of illness on the population. In the United States of America, courts of law have determined that statements of causality in human populations come most authoritatively from studies employing the causal logic characteristic of epidemiologic studies.

Selection bias is an error due to systematic differences in the characteristics between those who are selected for study (e.g. mental patients or incarcerated offenders) and the population from which they are drawn (all mentally ill or all persons who are violent).

Statistical adjustment is used to reduce differences in composition of two groups so that they may be fairly compared with respect to an outcome of interest. Unadjusted comparisons may result in biased conclusions.

Previous | Table of Contents