INTRODUCTION: Risk assessment tools are routinely used to identify patients at high risk. There is increasing evidence that these tools may not be sufficiently accurate to determine the risk of suicide of people, particularly those being treated in community mental health settings. METHODS: An outcome analysis for case serials of people who died by suicide between January 2014 and December 2016 and had contact with a public mental health service within 31 days prior to their death. RESULTS: Of the 68 people who had contact, 70.5% had a formal risk assessment. Seventy-five per cent were classified as low risk of suicide. None were identified as being at high risk. While individual risk factors were identified, these did not allow to differentiate between patients classified as low or medium. DISCUSSION: Risk categorisation contributes little to patient safety. Given the dynamic nature of suicide risk, a risk assessment should focus on modifiable risk factors and safety planning rather than risk prediction. CONCLUSION: The prediction value of suicide risk assessment tools is limited. The risk classifications of high, medium or low could become the basis of denying necessary treatment to many and delivering unnecessary treatment to some and should not be used for care allocation.