BACKGROUND: The question ‘would you be surprised if this patient died in the next 12-months’ is widely used for identifying adult patients in the last year of life. However, this has not yet been studied in children. AIM: To assess the prognostic accuracy of the surprise question when used by a multidisciplinary team to predict survival outcomes of children with life-limiting conditions over a 3 and 12 month period. DESIGN: A prospective cohort study. SETTING/PARTICIPANTS: Six multidisciplinary team members working in a children’s hospice answered a 3 and 12 month surprise question about 327 children who were either newly referred or receiving care at the hospice between 2011 and 2013. RESULTS: The prognostic accuracy of the multidisciplinary team for the 3 (and 12)month surprise question were: sensitivity 83.3% (83.3%), specificity 93.2% (70.7%), positive predictive value 41.7% (23.6%), negative predictive value 99% (97.5%) and accuracy 92.6% (71.9%). Patients with a ‘no’ response had an increased risk of death at 3 (hazard ratio, 22.94, p 0.001) and 12 months (hazard ratio, 6.53, p 0.001). CONCLUSION: The surprise question is a highly sensitive prognostic tool for identifying children receiving palliative care who are in the last 3 and 12 months of life. The tool is accurate at recognising children during stable periods demonstrated through a high negative predictive value. In practice, this tool could help identify children who would benefit from specialist end of life care, act as a marker to facilitate communications on advance care planning and assist in resource allocation.