Does the emphasis on risk in psychiatry serve the interests of patients or the public? No

No evidence of effectiveness

Risk assessment was not introduced into mental healthcare because of scientific progress but because of complex social and political factors, including media pressure after a small number of catastrophic events. Subsequently, no scientific evidence has emerged to show that risk assessment reduces the harms associated with mental illness.

Risk assessment does not work because it lacks the statistical power to discriminate between those who will and those who will not commit particular harms, such as suicide and severe violence, and it fails to consider the range of possible harms that people might experience.

Let me start with suicide. Suicide is more common than homicide, we know more about it than about homicide, and, unlike homicide, it is strongly associated with mental illness. When we categorise people into low and high risk groups for suicide, their subsequent behaviour places them into four groups: true positives and false negatives (who die by suicide) and false positives and true negatives (who do not).

Let us assume that we had an intervention that could decrease the likelihood of suicide. Using risk assessment to categorise people’s risk of suicide might be justified if the proportion correctly predicted to commit suicide (true positives) was sufficiently high to justify the intervention for all those regarded as at high risk (true and false positives). This would require the downsides of treating false positives to be clearly outweighed by the reduction in suicides. The intervention would have to be benign because the proportion of people who commit suicide among those considered to be at high risk is very low, often well below 1%.1 2

However, even under these circumstances, risk categorisation would not be necessary if it were feasible to provide the benign effective intervention to people at low risk as well. After all, some low risk people will, in fact, commit suicide. In psychiatric settings, people regarded as low risk are still at a much higher risk of suicide than the general community. For example, a recent study showed that people classified as low risk after presenting with self harm to emergency departments had an annual suicide rate 14 times the UK national average.3 In mental health settings, the term low risk is misleading.

Now let us turn to the risk of violence. Although there are few empirical studies of the risk factors for homicide by mentally ill people, the proportion of high risk people who will go on to kill someone is also extremely low. Using assumptions that optimise the predictive value of risk assessment, around 35 000 patients need to be detained to treat one person with schizophrenia who would otherwise kill a stranger during the next year.4 Never in the history of medicine have so many been hospitalised to protect so few.

Prejudicial judgments

But perhaps risk assessment is about more common and less serious harms. It is true that if we conflate minor violence with severe and serious violence, the proportion of true positives in high risk groups can be about as much as a third.5 Proponents of risk assessment argue that this sort of assessment provides useful information, even though any single high risk categorisation is still likely to be wrong. But this misses the point. Even if we could make reasonable judgments about the probability of a particular harm, we do not know what people will do next. When we assess the risk of violence, we make a prejudicial and premature judgment about the possibility of a violent outcome. In fact, our patients are more likely to face a wide range of other harms, including the intrinsic harms of severe mental illness, experiencing violent victimisation, misadventure, suicide, life shortening side effects of antipsychotic medications, and the infringement of their freedoms of choice and movement when they are involuntarily detained.

No current risk assessment tool can capture or estimate this diversity of unwanted outcomes.6 Although structured risk assessments for violence or suicide are known to be more accurate than clinical judgments, clinicians dislike them. This is because even an accurate estimate of the probability of a suicide or violence is of little use to them as they help patients and families with decisions about the benefits and side effects of mental healthcare.

Finally, the most important problem with risk assessment is that it encourages us to make paternalistic decisions that we would not make in general medicine. When we identify medical risk factors—for example, for cancer or heart disease—we discuss these factors with our patients and, with the patient’s wishes in mind, we treat or don’t treat. When we overemphasise risk in mental health, we make decisions about the involuntary detention of high risk people who are capable of making decisions for themselves and we fail to treat low risk patients who have lost the capacity to consent.

Mental health professionals, the courts, and governments have placed too much faith in risk assessment. It does not work. It has distracted clinicians from a broader consideration of our patients, their illnesses, and their best interests. We should downgrade risk assessment and return to our real and enduring duties as doctors and consider what people can do, what they want to do, and how we can help them recover.