Violent Behavior

To the Editor: In their article in the September 1997 issue, Grant Harris and Marnie Rice (1) presented a succinct review of current methods to evaluate, manage, and treat violent individuals in psychiatric facilities. However, their review left me with some troubling questions.

First, they assert that "no evidence has been found that clinicians' unaided judgments are better than those of lay persons in assessing violence risk." Some of the studies reviewed in their paper (McNiel and Binder, reference 30; Lidz and associates, reference 23; and Gardner and associates, reference 25) provide evidence that clinicians can, on the basis of their clinical observations and judgments, predict violent acts by their patients with 59 percent to 69 percent accuracy, and can predict the seriousness of the violent acts with the same accuracy as researchers using actuarial tools. But where is the evidence to support their claim that laypersons can predict violence with equal accuracy?

Second, when Harris and Rice state that "major mental disorder and psychiatric disturbance are poor predictors of violence," they appear to be arguing that accurate predictions about the behavior of potentially violent patients cannot be based on psychopathological data. Yet studies they cite by Swanson and associates (reference 63) and by Link and Stueve (reference 48) indicate that symptoms of psychosis and major mental disorder are strong predictors of violence, even after taking into account comorbid substance use disorder and various demographic factors.

The findings of Swanson and Link are consistent with reviews (2,3) indicating that actuarial methods like those advocated by Harris and Rice may be relevant to large populations but that predictive statements about individual risk must take into account individual psychopathology. As Mulvey (2) explained, "An overall coefficient of association describes only the linear trends in a total data set. It does not describe the strength of the association for every individual in the sample, nor does it allow for an adequate representation of the prevalence of individuals in a sample for whom the relationship is strong enough to warrant intervention or policy concern." Furthermore, individual psychopathology may not always lend itself to quantification, and for that reason actuarial methods may discount important information relevant to individual risk (4). Monahan has observed that "denying that mental disorder and violence may be in any way associated is disingenuous and ultimately counterproductive. … the flat denial that any relationship exists between disorder and violence can no longer credibly be prefaced by 'research shows'" (5).

Third, Harris and Rice make several contradictory statements. They assert that mental symptoms are poor predictors of violence (page 1171) after having earlier stated that command hallucinations increase the risk of violence (page 1169). To support their statement about this increased risk, they cite a study by Zisook and associates (reference 50) that found command hallucinations did notincrease risk. They declare that violent patients can be managed effectively with medications (page 1171), but later, citing a study by Allan and associates (reference 90), they state that little evidence exists that violence can be managed with medications (page 1172), although the Allan study actually suggested that medications can be useful in managing violence.

And finally, Harris and Rice contend that "treatments aimed specifically at increasing self-esteem are contraindicated because good evidence exists that self-esteem is positively related to violence." This statement contradicts their earlier observation elsewhere that feelings of worthlessness, among other depressive symptoms, can lead to violent action (6). In their article, they deplore the expenditure of resources for the treatment of anything except personality and lifestyle deviations that correlate highly with violence, stating that it is impossible to justify spending scarce resources on other targets, including mental disorders associated with violence. Surely more balanced judgment would accept that mental disorders and personality-lifestyle factors both contribute to violent actions and deserve measured and appropriate responses.

Despite these problems and contradictions, Harris and Rice demonstrate convincingly that individuals at risk for violent behavior can be evaluated accurately and managed effectively by experienced clinicians who take into account both individual-psychopathological and population-actuarial risk factors. We should feel grateful to them for that lesson.