Criminal Justice

Can Mental Illness Predict Violence? Researchers Identify 106 Risk Factors

  • Print.

Does mental illness predict violent behavior? Over the years, psychiatrists have looked at the question and found that mental illness alone doesn’t predict an above average risk of being violent.

Those with severe mental illness, however, are slightly more likely to commit violent acts—such as rape, assault, arson and homicide—than the general population, the Washington Post reports. The risk increases further when severe mental illness is combined with substance abuse.

The studies include:

• A 2001 study of 35,000 adults that found that 2.9 percent of those who had suffered from severe mental illness in the past year reported three years later that they had committed a violent act. The violence rate rose to 9.4 percent among those who reported both severe mental illness and substance abuse problems in the prior year. For those without mental illness or substance abuse problems, the rate of violence is only .8 percent.

•Violent risk and schizophrenia was the topic of an analysis of 20 studies published three years ago. It found the risk of schizophrenics committing homicide was 0.3 percent, more than 10 times greater than the risk for the average person.

University of Virginia law professor and psychologist John Monahan wanted a better risk assessment, according to the Post story. He helped develop a questionnaire based on 106 risk factors that was given to patients being discharged from psychiatric units in Massachusetts and Pennsylvania. An online executive summary says risk factors included prior violence and serious childhood abuse.

Ninety percent of those judged to be low-risk based on their answers did not commit violent acts, compared to 49 percent of those judged to be high-risk. “From our research, we could quickly distinguish between a patient whose chance of being violent was 1-in-10 from one whose was 1-in-2,” Monahan told the Post.

Give us feedback, share a story tip or update, or report an error.