Sputum induction (SI) is a sensitive and specific method for diagnosing Pneumocystis carinii pneumonia (PCP) in patients with AIDS. Although less expensive than bronchoscopy with bronchoalveolar lavage (BAL), SI followed by BAL does not necessarily reduce costs compared with BAL alone. Cost analysis demonstrates that the cost of diagnosing PCP is dependent on the prevalence of PCP (PCPprevalence) in the studied population, the sensitivity of SI (SIsensitivity) for diagnosing PCP, and the relative costs of SI and BAL (SIcost and BALcost) for diagnosing PCP. In any given clinical setting, SI reduces the cost of diagnosing PCP if (PCPprevalence)(SIsensitivity) greater than SIcost/BALcost. A graphic approach relating these parameters is also presented. Evaluation of reported PCPprevalence and SIsensitivity from recent literature illustrates that SI is not always the least costly method for diagnosing PCP. Cost reduction is not the only measure of a diagnostic procedure's value, and other aspects, such as discomfort, availability, risks, and patient prognosis, must be considered. The cost analysis approach used in this study identifies those variables that can be manipulated to reduce the cost of diagnosing PCP.