An index for prediction of outcome for use as a measure of the severity of illness was developed by a nonparametric multivariate analysis of cardiorespiratory data from 113 critically ill postoperative general surgical patients. This severity (predictive) index was based on a computerized algorithm that compares a given observed value with the frequency distributions of survivors and nonsurvivors. The difference in the mean values of this index for survivors and nonsurvivors was statistically significant (p less than 0.001) during each stage of shock. Sensitivity of the index in prediction of survival ranged from 70-93% depending upon stage, the specificity of the index ranged from 76-92%, and the predictive accuracy ranged from 87-96%. The severity index is used as a process measure to track the course of critically ill patients and to evaluate the efficacy of alternative therapies.