The aim of this study was to design a statistical model which will predict death or life-threatening complications in patients admitted to Coronary Care Unit using data which is available at the time of presentation. The study included 3721 consecutive admissions over a period four year period. Predictive models were developed using logistic regression analysis (with data from 1000 patients) and their performance was assessed using receiver operating characteristic (ROC) curve analysis. The most useful model included nine data items and was tested on data from 2721 patients. These could be divided into four groups according to their calculated probability of developing a serious complication. The lowest risk group had a mortality of 0.05%, compared with 3.5%, 6.4% and 18.1% respectively in the higher risk groups (p 1000 U/1) in the four groups was 14.1%, 21.2%, 46.9% and 51.5% respectively (p<0.001). The overall complication rates were 16.9%, 35.4%, 75.4% and 71.8% respectively (p<0.001).