BACKGROUND: The Manchester triage system (MTS) is now widely used in UK accident and emergency (A&E) departments. No clinical outcome studies have yet been published to validate the system. Safety of triage systems is related to the ability to detect the critically ill, which has to be balanced with resource implications of overtriage. OBJECTIVES: To determine whether the MTS can reliably detect those subsequently needing admission to critical care areas. METHODS: Analysis of emergency admissions to critical care areas and comparison with original A&E triage code by a nurse using the MTS at time of presentation. Retrospective coding of all cases according to the MTS by experts and case analysis to determine whether any non-urgent coding was due to the system or to incorrect coding. RESULTS: Sixty one (67%) of the patients admitted to a critical care area were given triage category 1 or 2 (that is, to be seen within 10 minutes of arrival). Eighteen cases given lower priority were due to incorrect coding by the triage nurse. Six cases were correctly coded by the MTS, of which five deteriorated after arrival in the A&E department. Only one case was critically ill on arrival and yet was coded as able to wait for up to one hour. CONCLUSIONS: The MTS is a sensitive tool for detecting those who subsequently need critical care and are ill on arrival in the A&E department. It did fail to detect some whom deteriorated after arrival in A&E. Most errors were due to training problems rather than the system of triage. Analysis of critically ill patients allows easy audit of sensitivity of the MTS but cannot be used to calculate specificity.