Chan and colleagues ^{Reference Chan, Bhatti, Meader, Stockton, Evans and O'Connor1} provide an overview of risk factors and risk scales for suicide following self-harm. However, their conclusions go beyond their review findings, and we think that discounting the potential value of risk scales on the basis of imperfect tools designed for other purposes is premature.

First, although we agree that the use of risk categories has its limitations (in particular when post hoc cut-offs are used), this can be resolved if risk prediction tools use pre-specified cut-offs, and consider reporting absolute probabilities as well as risk categories. ^{Reference Fazel, Chang, Fanshawe, Långström, Lichtenstein and Larsson2} Absolute probabilities provide greater flexibility, and could help optimise treatment allocation, waiting list prioritisation, or referral for more detailed assessments. A multicentre study in three UK centres found that only 70% of hospital episodes of self-harm receive psychosocial assessments, ^{Reference Hawton, Bergen, Casey, Simkin, Palmer and Cooper3} and hence there will likely be further challenges linking those at risk with appropriate clinical services. Clearly, psychosocial assessments are recommended for all persons who self-harm, but more personalised therapies will also involve a degree of triaging.

Second, the review identified three tools used in practice: the Beck Hopelessness Scale, Scale for Suicidal Ideation, and Suicide Intent Scale. However, none of these were developed for the purposes of risk prediction and thus critiquing the whole field on the basis of these tools goes beyond the evidence.

All risk prediction tools should be critically evaluated in terms of discrimination, calibration and reclassification – but the same high standards should also be applied to alternative approaches. What would be the performance of not using risk assessment, through purely qualitative or needs-based approaches? Without this information, this review might encourage a return to more subjective risk assessment approaches, which in the field of violence risk assessment have been shown to perform less well than structured methods. ^{Reference Ægisdóttir, White, Spengler, Maugherman, Anderson and Cook4}

Although purely qualitative and needs-based approaches have a strong intuitive appeal, risk assessment, if it can be linked to treatment, is likely to play a part in reducing suicide risk.