The purpose of this study was to assess: 1) the percentage of human immunodeficiency virus (HIV)-infected patients with suspected Pneumocystis carinii pneumonia (PCP) but unable to undergo the induced sputum procedure, together with the reasons involved; 2) the sensitivity and specificity of induced sputum procedure, using conventional stains and an immunofluorescence test; and 3) the cost of introducing induced sputum procedure for the diagnosis of PCP. One hundred and thirty eight HIV-infected patients with suspected PCP underwent induced sputum procedure and bronchoalveolar lavage (BAL). P. carinii was identified in induced sputum and BAL samples using conventional and immunofluorescence staining. The economic analysis took into account the direct costs of the two procedures. The induced sputum procedure was either not feasible or unsuccessful in 29% of the patients. The sensitivity of induced sputum, using conventional and immunofluorescence staining, was 0.27 and 0.56 respectively. The economic analysis showed that the two strategies (systematic BAL versus BAL only after negative induced sputum) are equivalent in cost terms when the induced sputum to BAL cost ratio is equal to the product of the prevalence of PCP by the sensitivity of induced sputum procedure. We conclude that the immunofluorescence test should be the reference technique for induced sputum samples, whilst conventional stains are more clinically relevant for BAL samples. The cost of introducing induced sputum should take into account the sensitivity of induced sputum and the prevalence of PCP in the suspected population.