A logical sequence of testing in evaluating prognosis early in acute myocardial infarction (AMI) would be to use clinical data first, then add noninvasive data and finally add invasive data. The incremental prognostic information concerning 1-year survival obtained from such a sequence in 107 patients with AMI was studied using logistic regression and receiver-operating characteristic curves. Cardiac mortality was 24% at 1 year. Clinical data obtained soon after admission (prior myocardial infarction, heart rate, blood pressure, age) were 78 +/- 5% accurate in the prediction of 1-year survival. The addition of radionuclide-estimated left ventricular ejection fraction or invasive hemodynamic data to the clinical model at this time improved prognostic accuracy to 84 +/- 5% (p = 0.05) and 87 +/- 4% (p = 0.007), respectively. The further addition of invasive data to the model containing clinical and left ventricular ejection fraction data provided a further increment in prognostic accuracy to 89 +/- 4%, whereas no significant increase in accuracy was seen on addition of left ventricular ejection fraction to the model containing clinical and invasive data. It is concluded that clinical data provide important prognostic information concerning late survival early in the course of AMI. This may be improved by the logical application of noninvasive and invasive studies at this time.