BACKGROUND Surgical procedures frequently require blood transfusion. Blood and its components are frequently ordered and cross-matched excessively, without proper analysis of the real needs. The purpose of this retrospective study was to analyse the actual intraoperative requirement vs the ordering of blood, and the transfusion practices of the University Hospital. METHODS We analysed the records of all patients scheduled for surgery in July 2007, allocating them to four groups: high, medium, low and minimal risk of blood loss and transfusion. The following calculations were made: cross-match to transfusion RBC ratio (C/T); the number of patients transfused compared to those cross-matched (transfusion probability - %T); and the number of units transfused relative to the number of cross-matched patients (transfusion index - TI). The values justifying blood ordering were: C/T ratio <3.0, %T >30% and TI>0.5. RESULTS In all four study groups the limit values were not reached. The C/T ratios were 6.61 (high risk group), 13.7 (medium risk group) and 35.5 (low and minimal risk groups). The overall C/T ratio was 9. The %T values were 18.8%, 8.69% and 2.94%, respectively. The TI values were 0.27, 0.42 and 0.15. CONCLUSIONS We conclude that preoperative blood ordering is far from being related to real needs, and suggest therefore, that hospital blood ordering policy should be reassessed.