The review of risk assessment by Large & Nielssen ^{Reference Large and Nielssen1} is timely as there has been an increasing tendency to rely on structured protocols in the assessment of patients, particularly with regard to future probabilities of violence and self-harm. However, there are a number of aspects which have not been discussed, the most important of these being the concept of acceptable risk.

Politicians and service managers are happy to point to a process of risk assessment, yet they universally abrogate their duty as representatives of the community to define what level of risk is acceptable. Despite the statistical difficulties discussed by Large & Nielssen, this has been successfully embraced in aviation medicine. Acceptable failure rates in mechanical components have been used to define the risk management for pilot incapacitation. Although this approach was pioneered by cardiologists, aviation medicine is where the approach has been more widely adopted, even though the definition of base rates of risk in other areas is not as straightforward. ^{Reference Davies2}

The acceptance of a defined level of risk has important implications for services. As an example, if a patient is considered as being at risk of suicide, rather that the accepted risk being progressively increased as the bed availability declines, the service should have an obligation to provide a bed for those whose risk is considered greater than the acceptable level.

Other common areas where risk assessments are required are release of potentially violent individuals from hospital or prison, safety in driving, the ability to own a firearm and suitability for employment in areas where inappropriate behaviour would involve significant community risk. When these assessments are made, it is important that there is not only an understanding of their predictive value, but that there should be some idea of the relative and absolute risk considered acceptable by the community. Once this is defined, it automatically follows that an adverse result does not imply error. It is important that the community representatives, including coroners and politicians as well as the media, should be educated about this. Ultimately, a decision about acceptable risk levels must be explicitly made by the community in advance with regard to their cost/benefit ratio. Post hoc assessments of individual decisions are generally unhelpful.

When providing reports involving risk assessment, I always enclose a comment stating that whereas I have made my own evaluation, I would reconsider my assessment on the basis of a defined acceptable level of risk. Finally, I would not agree with Large & Nielssen that risk assessment protocols should not be used. Their importance is not that they produce a usable rating (and I would note that these are strictly ordinal rather than interval scales), but that they do document that appropriate risk factors have been considered in the clinical decisions made.