Background Suicide is a major public health issue, and a priority requirement is accurately identifying high‐risk individuals. The SAD PERSONS suicide risk assessment scale is widely implemented in clinical settings despite limited supporting evidence. This article aims to determine the ability of the SAD PERSONS scale (SPS) to predict future suicide in the emergency department. Methods Five thousand four hundred sixty‐two consecutive adults were seen by psychiatry consultation teams in two tertiary emergency departments with linkage to population‐based administrative data to determine suicide deaths within 6 months, 1, and 5 years. Results Seventy‐seven (1.4%) individuals died by suicide during the study period. When predicting suicide at 12 months, medium‐ and high‐risk scores on SPS had a sensitivity of 49% and a specificity of 60%; the positive and negative predictive values were 0.9 and 99%, respectively. Half of the suicides at both 6‐ and 12‐month intervals were classified as low risk by SPS at index visit. The area under the curve at 12 months for the Modified SPS was 0.59 (95% confidence interval [CI] range 0.51–0.67). High‐risk scores (compared to low risk) were significantly associated with death by suicide over the 5‐year study period using the SPS (hazard ratio 2.49; 95% CI 1.34–4.61) and modified version (hazard ratio 2.29; 95% CI 1.24–2.29). Conclusions Although widely used in educational and clinical settings, these findings do not support the use of the SPS and Modified SPS to predict suicide in adults seen by psychiatric services in the emergency department.