Most studies of modifying test ordering have focused on costs. Questions not addressed are whether programs to reduce testing lead to a higher proportion of clinically indicated tests and is underutilization an adverse outcome of such programs? To investigate this, we studied the house staff's ordering of three common laboratory tests at baseline and after educational and administrative interventions. Over a 2-year period, 3,603 urine cultures, sputum cultures, and admission urinalyses were reviewed. A lecture emphasizing the indications for these tests followed by chart audit and weekly feedback increased the proportion of clinically indicated tests. Subsequently, an administrative intervention requiring the intern to list the reason for ordering the test on the laboratory request form further improved test ordering. Underutilization, defined as a failure to order a potentially indicated test, was assessed during two representative periods. The "underutilization rate" (omitted tests per 100 patients) was no worse during maximal intervention than it was 9 months after the last intervention (7.7 vs. 11.1, NS). No immediate adverse consequences resulted from tests not ordered. Our findings indicate that it may be possible to selectively reduce the ordering of unnecessary tests without sacrificing quality of care.