OBJECTIVE To analyze intraoperative autologous salvage of shed mediastinal blood and subsequent transfusion in cardiac surgery. METHODS Retrospective statistical analysis. METHODS University hospital. METHODS Three thousand twenty two patients undergoing cardiac surgery from 1984 to 1988. METHODS A review of anesthesia and transfusion records of all patients who underwent intraoperative salvage of shed blood and autologous transfusion using the Sorenson Receptal Auto Transfusion System (ATS) with saline wash prior to reinfusion in cardiac surgery. RESULTS The salvaged blood volume ranged from 36 to 2,795 ml, with a mean of 321 +/- 222 ml (SD). Eighteen percent of patients did not receive any homologous blood products during their hospitalization. Patients who received only salvaged autologous transfusion were younger, had higher preoperative hemoglobin and hematocrit values, had a larger body surface area, and had shorter surgeries compared with patients who received only homologous blood or both autologous and homologous blood. More blood products were given to patients who received salvaged autologous blood compared with those who did not. Patients who underwent normovolemic hemodilution prior to extracorporeal circulation with subsequent reinfusion received significantly fewer blood products. Ten preoperative and four intraoperative variables significantly influenced the salvaged volume. Previous cardiac surgery was the most significant preoperative variable, and repair of ventricular septal defect produced by myocardial ischemia was the most significant intraoperative variable. CONCLUSIONS Considering the average salvaged volume and its current autologous transfusion-related expense, autologous blood salvage is potentially an economic benefit. Perioperative blood conservation requires a considerable commitment from surgeons, anesthesiologists, perfusionists, and intensive care physicians to be effective.