In a recent and valuable review, Ribeiro et al. conclude that self-injurious thoughts and behaviours (SITBs) only provide a marginal improvement above chance in diagnostic accuracy for later suicidal thoughts and behaviours (Ribeiro et al. Reference Ribeiro, Franklin, Fox, Bentley, Kleiman, Chang and Nock2016). This result is seemingly in contrast to a widespread belief in the field that prior SITBs are some of the most important risk factors of future SITBs, and especially prior suicidal attempts has been regarded a robust risk factor for future suicide (Ribeiro et al. Reference Ribeiro, Franklin, Fox, Bentley, Kleiman, Chang and Nock2016).

An important factor that may help to explain some of the low diagnostic accuracy of SITBs is that one will always seek to prevent suicidal behaviour, even in an observational study. It appears that this important issue has been omitted in the discussion of the results. In observational studies with an endpoint representing an adverse or unwanted event (like suicidal behaviour), it is difficult just to observe and not to try to prevent the adverse event occurring. Both the healthcare and the community will try to prevent suicidal behaviour for obvious ethical reasons, independent of an ongoing research project or not. SITBs are well known risk factors for suicidal behaviour, both among professionals and among ordinary people. If a prior or present SITB has been expressed, there will be no neutral ‘research’ observation of a possible future SITB; family/peers or professionals will try to prevent new suicidal behaviour. Thus, a successful risk management could prevent suicidal behaviour, and then turn a possible true positive prediction into a false positive prediction, and the predictive accuracy will be decreased. This implies that a good risk assessment might lead to a ‘wrong’ risk prediction, which might explain some of the ‘ceiling effect’ of risk assessment instruments (Coid et al. Reference Coid, Yang, Ullrich, Zhang, Sizmur, Farrington and Rogers2011).

In the light of these viewpoints it is important to emphasize that the interpretation of the low predictive accuracy of prior SITBs for future SITBs in the review by Ribeiro et al. should not be used to reduce the importance of SITBs as robust risk factors for suicidal behaviour. On the other hand, I fully agree with the authors’ suggestions for further research in this complex and important field.