A predictive index based on cardiorespiratory-monitored values of an earlier series of postoperative critically ill patients was tested in prospective clinical trials and found to be reasonably accurate, sensitive, and specific. The hypothesis was tested that the median values of patients who survived life-threatening postoperative conditions, rather than the norms of unstressed, healthy volunteer subjects, constitute a first approximation to the optimal therapeutic goals for critically ill postoperative patients. In a prospective series of 223 consecutive, critically ill postoperative patients, normal values were used as the therapeutic goals of the control patients, whereas the median values of surviving patients were used as the goals of therapy for the protocol group. The clinical conditions of the protocol group were at least as severe as those of the control group, but the mortality was significantly less in the protocol group (12.5 percent) than in the control group (35 percent); the number of life-threatening complications were also greater in the control group. These data suggest that at least half and possibly as much as two thirds of postoperative deaths may be due to physiologic problems that can be identified, described, predicted, and prevented. Therapy for the critically ill patient should be defined by physiologic criteria, and administration of therapy should be monitored to attain prophylactically optimal physiologic goals rather than giving therapy after a deficiency has occurred to attain normal values.