Guidelines, algorithms and recommendations have been issued in the attempt to ensure appropriateness of transfusion practice, but the results are less than satisfactory, mainly due to the difficulty to turn paper procedures into actual practice. In our hospital we have tried to overcome this difficulty through the implementation of a quality assurance programme which includes giving the privilege of nonurgent blood prescription to a limited number of physicians and a computerized prospective audit of blood requests. The latter is performed through verification of the compliance of blood requests, which are designed to include a patient's laboratory and clinical data, with hospital guidelines for the proper use of blood.  In the 12 months since implementation of the computerized prospective audit the transfusion service has evaluated 7884 requests. Of these, 63.4% (n = 4998) were for red blood cells, 21.1% (n = 1664) for platelets and 15.5% (n = 1222) for fresh frozen plasma. The prospective audit showed that 96.8% and 98.1% of requests for red units and platelets were appropriate, respectively. Conversely, approximately 27% of plasma requests did not comply with guidelines, mainly because the evidence of coagulopathy was missing. However, inappropriateness of plasma requests for elective general surgery decreased from 39% at the onset of the programme to 14% in the last trimester considered. Moreover, the evaluation by retrospective audit of the proportion of patients transfused with both red blood cells and plasma in the perioperative period out of those transfused with red blood cells only, as an indicator of unwanted reconstitution of whole blood, showed that this proportion decreased from 47.6% (320/672) in the 12 months before implementation of computerized audit to 37.8% (244/646) in the following 12 months (difference = −9.8%, 95% confidence interval of the difference from −4.5% to −15.1%;P < 0.005 by χ2test). Our initial experience, together with the present system, shows that (1) the restriction of nonurgent blood prescription to a group of clinicians more educated in transfusion medicine than average clinicians practicing in a large multispecialty hospital is feasible; (2) prospective audit is a useful tool for assuring the