OBJECTIVE:To describe the timing of cardiac arrest detection in relation to episodes of Medical Emergency Team (MET) review and routine nursing observations. DESIGN AND SETTING:Retrospective observational study in a university-affiliated hospital. PATIENTS:279 cardiac arrests involving ward patients MEASUREMENTS AND RESULTS:Cardiac arrests were allocated to one of 24 1-h intervals (24:00-00:59, 01:00-01:59, etc.). The actual hourly rate of cardiac arrests was related to the expected average hourly rate. Peak levels of cardiac arrest detection occurred during times of routine overnight nursing clinical observations between 02:00 and 03:00 (OR 3.06) and 06:00-07:00 (OR 1.95). The lowest level of cardiac arrest detection occurred between 20:00 and 21:00 (OR 0.42). After introduction of the MET there were 162 cardiac arrests, 28% of which occurred shortly after an initial MET call. The odds ratio for risk of cardiac arrest during periods of lowest MET activation (24:00-08:00) when compared with periods of highest MET activation (16:00-24:00) was 2.26. CONCLUSIONS:Cardiac arrest detection in our hospital is episodic with peak levels corresponding to periods of overnight routine nursing observations following a period when patient review is likely to be low. After the introduction of the MET there was an inverse link between detection of cardiac arrests and levels of MET activation over the 24-h period. Increased overnight utilization and earlier MET activation may further reduce the incidence of cardiac arrests at our hospital.