We investigated the outcome and potential cost savings of a system designed to limit physician access to induced sputum analysis (ISA) for the diagnosis of Pneumocystis carinii pneumonia (PCP). Four respiratory medicine physicians screened all requests for ISA to determine the pretest likelihood of PCP. Twenty-two of 102 requests for ISA over a 1-year period were denied due to a low clinical suspicion for PCP. Seven individuals had a definitive alternative diagnosis confirmed and the remaining 15 were empirically treated for a presumptive diagnosis. All individuals were followed for at least 60 days or until death. None of the 22 individuals developed PCP during the follow-up period. We estimate that this approach saved $27,474, avoided exposure of health care workers to Mycobacterium tuberculosis, and was educational for the referring physicians.