A series of interventions at a 228-bed general hospital provided physicians with feedback at regular intervals concerning the amount of laboratory services employed in treating their patients. Case-mix-adjusted estimates of laboratory tests allowed each physician to compare use of laboratory tests with that of peers in the same department at the same hospital. Physicians with "excess" practice patterns ordered hundreds more laboratory tests than average each year. A multifaceted educational program included the following: 1) meetings were held concerning costs and unnecessary laboratory tests; 2) physicians were given descriptions of their practice patterns relative to their peers as part of both large and small departmental discussions; 3) the feedback was repeated a year later; 4) a consensus conference established guidelines for test ordering; and 5) a sample of patient records was examined for appropriateness of laboratory test ordering. A total of 37% of a sample of tests ordered during the baseline period by physicians with "excess" practice patterns was classified as inappropriate. The intervention resulted in a reduction of 1.8 tests per patient (P = 0.0005). Eight of the nine tests individually showed reductions in use. Charge data from the target hospital showed a statistically significant reduction in laboratory charges per patient in the quarter following program initiation (P = 0.02) and no evidence for change in a group of five comparison hospitals. There was no evidence for reductions in the ordering of essential tests. These results demonstrate a cost-effective approach to reducing unnecessary costs that can be implemented in hospitals with integrated data systems.