x Abu Dhabi, UAEMonday 24 July 2017

Trying to predict criminal behaviour is a bit psychotic

There is no scientific proof that severely mentally ill people are pose an elevated threat to the public in terms of violent crimes.

Imagine having the violent deaths of six innocent people on your conscience. You're indicted on 49 criminal charges and if convicted, you could face the death penalty. Would you smile?

This is the case of Jared Loughner, who earlier this year in Arizona opened fire on a US congresswoman and her supporters. Last week, Loughner walked into court with a smile on his face and pleaded not guilty. His mental competence to stand trial was discussed, with the defendant's legal team acknowledging their client has "serious mental afflictions". As Loughner's mental health gains increasing attention in this high-profile case, there is a danger in perpetuating the myth that mentally ill people are dangerous to society. One commentator has already proposed that the key questions raised in this case are: "How can we more aggressively treat mentally ill people who are becoming increasingly disruptive? How can we prevent them getting guns? Do we need to make involuntary treatment easier for authorities to invoke?"

These are valid questions. However, scientific data generally does not support the idea that severely mentally ill people pose an elevated threat to the public in terms of homicide and violence. Data from the UK, reported in the British Journal of Psychiatry in 2006, suggests that about five per cent of people incarcerated for murder are suffering from a severe mental illness. In the UK, this equates to less than 20 homicides per year. Compare this figure to the UK's 280 deaths per year from drowning, 500 deaths from fire, and 4,000 deaths from accidental falls.

Even this five per cent figure is slightly misleading, especially when you consider people with severe mental health problems are also far more likely to have problems with alcohol and narcotics. When this is taken into consideration, it becomes even harder to identify any clear causal relationship between psychosis and violence.

A recent study undertaken in Sweden looked at the risk for homicide and violent crimes among 3,743 people with severe mental illness, and a control group of 37,429 healthy people from the general population. At first glance, the findings from this study suggest that having a severe mental health problem greatly increased the odds of being involved in a homicide or violent crime. However, once the researchers controlled for drug and alcohol use, these difference vanished. In short, you are as equally unlikely to be murdered by a "normal" person as by one experiencing a severe mental health problem.

In an attempt to predict and prevent violence by people with severe mental health problems, researchers have designed complex statistical modeling techniques. These predictive models try to evaluate all the factors that might help identify which person with a severe mental illness is most likely to act violently. The factors of predictive violence for mentally ill and "normal" people are the same; they include alcoholism and a past history of violence. However, the idea behind developing these predictive tools is to help clinicians make crucial decisions about which psychiatric patients to detain in hospital (against their will if necessary) and which ones to discharge, or allow to remain in the community.

Professor George Szmukler, the director of the Institute of Psychiatry at the Maudsley in the UK, has assessed these models in terms of detecting violence in actual clinical practice. He concluded that within a population of 100 patients, where 20 are prone to violence, the best the model can do is identify 14 (70 per cent) of the 20 violent patients. Regrettably, even this is at the cost of also misidentifying and detaining 24 (30 per cent) of the non-aggressive patients.

This issue of "false positives" has serious implications for civil liberties. How many non-aggressive patients is it acceptable to detain in order to perhaps prevent one violent incident? When the same model was used to identify extreme acts of violence, the model's predictive ability dropped to the realm of uselessness, getting it wrong 97 times out of 100.

The truth is that if we cannot accurately predict violent acts, we cannot reliably prevent them, at least not without seriously infringing on the civil liberties and human rights of mental health patients. As Professor Richard Bentall writes in Doctoring the Mind: "More vigorous efforts to control...psychiatric patients might make us feel safer, but it seems unlikely they would actually make the world a safer place."


Justin Thomas is associate professor of pyschology at Zayed University