To establish the diagnosis of alveolar hemorrhage (AH) in cells recovered by bronchoalveolar lavage (BAL), Golde and colleagues created a score based on the hemosiderin content of alveolar macrophages stained with Prussian blue. We used an easier method, calculating the percentage of siderophages among the total alveolar macrophages recovered by BAL. We have retrospectively studied this method in 240 BALs performed in 194 immunocompromised patients. Prussian blue staining was performed on each BAL sample, and the Golde score was calculated for 47 samples chosen at random. The methods were compared for diagnosing AH. The percentage of siderophages correlated well with the Golde score. AH was defined by at least 20% siderophages. This definition was validated by comparison with the method of Kahn and coworkers. AH was present in 87 (36%) of the samples and was significantly associated with four parameters: thrombocytopenia (< 50,000/mm3), other abnormal coagulation parameters, renal failure (creatinine > or = 2.5 mg/dl), and a history of heavy smoking. The diagnosis of AH did not correlate with either the cause or the outcome of pneumonia. AH was seen more frequently in cardiac transplant patients (75%). In our experience, (1) a percentage of siderophages > or = 20% is sufficient and is an easier determinant of the diagnosis of AH than the Golde score; and (2) AH is rarely the sole cause of lung injury and is usually associated with other causes of pneumonia. AH may be considered more as a sign than as a distinct disease in this population.