The authors prospectively evaluated the usefulness of 12 preoperative tests (including blood type and screen) ordered selectively according to clinical status and importance of scheduled surgery. Two methods of assessing usefulness of tests were used, taking into account not simply their abnormality yield, but also their impact on patient care during the entire hospitalization: first, usefulness was assessed by anesthetists involved in care; second, usefulness was assessed by an automated analysis of specific clinical situations in which tests were a priori considered useful. In 3,866 patients 15,920 tests were performed; 30% of performed tests were abnormal. As a consequence of test results, surgery was delayed or cancelled in 19 patients and modified in 1, and a treatment was instituted or anesthetic management influenced in 347 (9%). Blood component therapy was instituted in 652 patients (17%). Anesthetists and automated analysis found at least one preoperative test useful in 931 patients (24%) and 890 patients (23%), respectively, whereas on a per-test basis 17% and 9% of tests were found useful. The discrepancy in usefulness assessment was mainly due to differences in evaluation of the usefulness of normal results. This study shows that the usefulness of selectively ordered preoperative tests may be high, but better criteria for their indications are needed, because tests other than those recommended by protocol were found useful.