Open lung biopsy (OLB) is often performed as the definitive diagnostic procedure in patients with undiagnosed pulmonary infiltrates, but controversy exists as to the clinical utility of this practice. A retrospective review of 50 consecutive patients who underwent OLB for undiagnosed pulmonary infiltrates was done to assess the diagnostic value as well as the frequency with which these results affected therapy and mortality. Histologic tissue diagnoses were obtained in all patients. Specific pathologic diagnoses were obtained in 56% of patients, nonspecific in 44%. Lobar or lateralized infiltrates were more likely to yield a specific diagnosis (87%) than diffuse, bilateral infiltrates (42%). Thirty-four patients (68%) had previously had a nondiagnostic transbronchial biopsy; 58% of these patients had a specific diagnosis established by OLB. Twelve patients (24%) were in acute respiratory failure at the time of OLB; this group had a 50% mortality rate as compared with only 2.6% for patients not in acute respiratory failure (p less than 0.01). Therapy was altered (new specific or nonspecific treatment initiated or therapy withdrawn) in 78% of patients undergoing OLB. Thirty-day in-hospital survival was significantly higher in patients for whom either specific or nonspecific therapy was indicated and initiated versus those in whom no therapy was initiated or all therapy was withdrawn (mortality: 5.5% versus 35.7%; p = 0.01). Mortality was not related to the presence of immunosuppression or to the finding of a specific diagnosis. The overall mortality rate of 14% in this series compares favorably with mortality rates found in similar series, reflecting differences in patient populations and possibly the timing of intervention. OLB remains a clinically valuable diagnostic tool in selected patients.