BACKGROUND--Sudden coronary death is a major public health issue. The identification of patients at high risk should therefore be as efficient as possible. This study compares simple and more elaborate risk stratification procedures. METHODS--Risk functions for the prediction of sudden death were determined in a population of 6693 consecutive patients who had 24 hour electrocardiography for various indications. The functions were based on the clinical and electrocardiographical data on 245 patients who died suddenly during 2 year follow up and 467 patients randomly drawn from the total study population. RESULTS--The model based on history (age, sex, myocardial infarction, congestive heart failure, palpitation, syncope, use of diuretics, and use of nitrates), 12 lead electrocardiography (major intraventricular conduction defect, T wave abnormality, and ST depression > or = 0.05 mV), and standard rhythm analysis of 24 hour electrocardiography (ventricular tachycardia, frequent premature atrial complexes, sinus tachycardia (> 150 min-1), and atrial fibrillation) was almost as efficient in the prediction of sudden death as extended models that also contained information from exercise testing, echocardiography, ventriculography, and computer-aided re-analysis of 24 hour electrocardiography (QT and RR interval variability). CONCLUSIONS--These results indicate that additional information from advanced function tests does little to increase the efficiency of prediction of sudden coronary death over that of tests based on history, the standard 12 lead electrocardiogram, and 24 hour electrocardiography.