To develop and test the predictability of a paediatric early warning score to identify children at risk of developing critical illness. Prospective cohort study. Admissions to all paediatric wards at the University Hospital of Wales. Respiratory arrest, cardiac arrest, paediatric high-dependency unit admission, paediatric intensive care unit admission and death. Data were collected on 1000 patients. A single abnormal observation determined by the Cardiff and Vale paediatric early warning system (C&VPEWS) had a 89.0% sensitivity (95% CI 80.5 to 94.1), 63.9% specificity (95% CI 63.8 to 63.9), 2.2% positive predictive value (95% CI 2.0 to 2.3) and a 99.8% negative predictive value (95% CI 99.7 to 99.9) for identifying children who subsequently had an adverse outcome. The area under the receiver operating characteristic curve for the C&VPEWS score was 0.86 (95% CI 0.82 to 0.91). Identifying children likely to develop critical illness can be difficult. The assessment tool developed from the advanced paediatric life support guidelines on identifying sick children appears to be sensitive but not specific. If the C&VPEWS was used as a trigger to activate a rapid response team to assess the child, the majority of calls would be unnecessary.