Assessing suicide risk is an integral part of what a mental health professional does. It permeates out even to the training programmes of charitable organisations such as Samaritans. So much emphasis is given to suicide risk assessment because suicide is one of the worst outcomes for lost years of life, and it has a ripple effect on surrounding family and friends that cannot be understated. However, predicting whether an individual will go on to die by suicide is a tricky business, and I sympathise with the rhetoric of Mulder and colleagues ^{Reference Mulder, Newton-Howes and Coid1} in a recent edition of the British Journal of Psychiatry. They rightly identify that because the base rate of suicide is so low in the general population (10.8 in 100 000), our current formal predictive tools just do not achieve satisfactory statistical efficacy. I think it is appropriate to say not only that there is a base rate issue, but also that suicide risk is dynamic and that the validity of an initial assessment decays appreciably with increased follow-up time. All this stacks up against our attempts to formally risk assess patients. However, I would not support a retreat to relying purely on unstructured clinical judgement, for three reasons.

First, it is clear that humans, whether medically trained or not, are poor at predicting the future. This is not a new insight. 2016 marks 50 years since the Baxstrom v Herold ruling. Essentially, this ruling meant that almost 1000 inmates who were previously regarded high risk for violent reoffending were released or had their confinement level stepped down. The controversy is that although they were believed to be high risk, very few went on to be reconvicted for violent crime. While acknowledging that suicide and violence are not analogous, this exemplifies the underlying flawed nature of unstructured professional judgement. Second, by incorporating structured professional judgement (SPJ) tools into our practice, we are not dissociating ourselves from the patient but following an evidence-based structure to inform our management. SPJs are conducted as interviews, leading to a numerical score which is flexible to interpretation by the clinician. Unlike when using atheoretical actuarial assessments, the clinician is not dictated to by the score. Unlike unstructured clinical judgement, the clinician can visualise all the relevant risk factors. Finally, to have a standardised approach is ideal for audit and reduces interclinician variability: a nod to the ideals of modern practice.

I was fortunate enough to be in the audience for a lecture by Professor Robert Snowden recently at a Royal College of Psychiatrists conference on old age psychiatry. He and his team are developing an SPJ tool, to be called the Risk of Suicide Protocol (RoSP), which may answer some of the issues we currently face in this area. Overall, given that suicide is such an important issue, with lifetime prevalence for attempt at nearly 3%, can we afford to adopt a defeated rhetoric?