We have applied neural network algorithms to estimate future vital status (dead or alive) of patients with acute myocardial infarction. A large database (Myocardial Infarction Data Acquisition System - MIDAS) including all 49,250 myocardial infarctions that occurred in the state of New Jersey in 1986 and 1987 with follow-up as long as five years was used in the development and testing of such a computer algorithm. Since the information included in the database was not sufficient to allow the exact prediction of vital status in all patients with 100% accuracy, we developed a neural network able to categorize patients according to the probability of dying within a given period of time rather than predicting categorically whether a given patient will be dead or alive at a given time in the future. A new algorithm was developed to accommodate cases where identical input vectors were associated with different outputs (vital status). In addition, a method of linear output adjustment was devised to describe the degree of confidence of each prediction. A two hidden layer perceptron, using ALOPEX, a feedback optimization algorithm, for weight updating, with 19 input patient variables of the MIDAS data set was used to predict six month mortality. The neural network was able to learn and was successful in predicting vital status at six months.