We retrospectively reviewed the medical records of all patients who were admitted to the medical and surgical intensive care units of a university center (N = 100) and its affiliated veterans' hospital (N = 46) between 1982 and 1986 to receive dialysis. The APACHE II severity-of-disease classification was used to identify the cases in which the prognosis was so poor that no long-term benefit would accrue from hemodialysis treatment. A "risk of death" was calculated for each patient. At a risk of death of 70% or greater, the system correctly predicted the demise of patients with 100% specificity regardless of what interventions were carried out. Sensitivity and predicted negative value were low in all cases, however, indicating a poor predictability of those who will survive. Withholding the average of 6 dialysis treatments that this group of patients received would probably have reduced patient suffering during a lingering terminal illness and led to a savings of about $4,500 per patient.