BACKGROUND: Pulmonary infiltrate in immunocompromised hosts has many infectious and non- infectios etiologies. To evaluate the diagnostic yield and therapeutic implication of two invasive diagnostic methods, such as bronchoscopy and surgical lung biopsy, we performed retrospective analysis of these patients. METHODS: All immunocompromised patients admitted to Samsung Medical Center from October 1995 to August 1998 who underwent bronchoscopy and/or surgical lung biopsy for the diagnosis of pulmonary infiltrates were included in this study. Confirmative diagnostic yield, the rate of changed therapeutic plan and patients' survival were investigated. RESULTS: Seventy-five episodes of pulmonary infiltrates developed in 70 patients(M : F = 46 : 24, median age 51). Underlying diseases of patients were hematologic malignancy(n=30), organ transplantatio n(n=11), solid tumor(n=12), connective tissue disease(n=6) and others. Confirmative diagnosis was made in total 53 cases (70.7%), of which 70.2% had infectious etiology. Diagnostic yields of bronchoscopy, bronchoalveolar lavage(BAL), transbronchial lung biopsy(TBLB) and surgical lung biopsy were 35.0%(21/60), 31.4%(16/51), 25.0%(9/36) and 80.0%(20/25). Therapeutic plan was changed in 40%(24/60) of patients after bronchoscopy and in 36%(9/25) of patients after surgical lung biopsy. More patients survived (84.4% vs 60.5%, p=0.024) when therapeutic plan was changed after invasive diagnostic study. CONCLUSION: Bronchoscopy and surgical lung biopsy are helpful for the therapeutic implication of pulmonary infiltrates in immunocompromised hosts. Large-scale prospective case-control study may further clarify their limitation and usefulness.