To determine predictors of mortality in immunocompromised patients with pulmonary infiltrates, we reviewed the records of all such patients admitted to two community teaching hospitals who underwent a lung biopsy over a ten-year period. We examined the consequences of advancing age, primary disease, fever, neutropenia, immunosuppressive corticosteroid therapy, previous lung radiation, roentgenographic pattern, result of lung biopsy, room air arterial oxygen pressure (Pao2), early mechanical ventilation, and the presence of a comorbid disease on eventual outcome. We identified 104 episodes in 99 patients. Sixty-seven (64%) survived and 37 died. By both discriminant analysis and logistic regression statistical methods, mechanical ventilation, the initial room air Pao2, and corticosteroid therapy were the dominant independent variables, in that order, to significantly predict mortality. No patient survived who simultaneously had a room air Pao2 less than or equal to 50 mm Hg, was on corticosteroids, and was mechanically ventilated. Eighty-three percent of survivors had either none or, at most, one of these three variables present. We conclude that hypoxia, immunosuppression by corticosteroids, and the necessity for mechanical ventilation within 72 hours of hospitalization indicate a poor prognosis in the immunocompromised patient with pulmonary infiltrates who has undergone a lung biopsy.