We examined early mortality (within 30 days) and survival (beyond 30 days) after acute myocardial infarction in 221 patients by screening 158 variables measured soon after the patient's admission to the hospital. Nineteen of these measurements had predictive value, but each variable alone was relatively insensitive. Therefore, we subjected groups of variables to stepwise discriminant function analysis and classification rates were estimated by calculating 95% confidence intervals using a jackknife procedure. When factors from the history, physical examination, and noninvasive assessment were combined, we identified 70% of deaths (confidence interval 48-80%) and 94% (90-98%) of survivors; when 11 selected variables including hemodynamic data were combined, we identified 86% (66-98%) of deaths and 96% (92-100%) of survivors (93% overall accuracy). We further tested the validity of this method in a subsequent series of 150 patients. Using the original discriminant functions, classification rates based on noninvasive and hemodynamic data fell within predicted limits, although the number of patients studied hemodynamically was unrepresentative and too small to allow overall predictive accuracy. Therefore, we randomly divided the entire population (371 patients) into a base sample from which we constructed new discriminant functions, with which we classified the remaining patients. The classification rates for the validation sample fell within the predicted confidence intervals. Thus, our method provides a reliable approach for predicting the risk of early death or the likelihood of survival in patients soon after acute myocardial infarction.