To determine the effect of respiratory infections on oxyhemoglobin saturation in a high-altitude population, we recorded clinical signs, oxyhemoglobin saturation determined by pulse oximetry, and findings on radiographs of the chest of 423 children with acute respiratory infections; the children were living at an altitude of 3750 m in the Peruvian Andes. We defined hypoxemia as an oxyhemoglobin saturation value greater than 2 SD below the mean value for 153 well children in this population. Eighty-three percent of children with clinical bronchopneumonia, but only 10% of children with upper respiratory tract infection, had hypoxemia (p less than 0.001). Compared with previous studies of children living at lower altitudes, the presence of tachypnea was relatively nonspecific as a predictor of radiographically determined pneumonia or of hypoxemia, especially in infants. A history of rapid breathing was 74% sensitive and 64% specific in the prediction of hypoxemia, and performed as well as a standard World Health Organization case management algorithm in the prediction of radiographic pneumonia or hypoxemia. Radiographic pneumonia was not a sensitive predictor of hypoxemia or clinically severe illness. In contrast, the presence of hypoxemia was a useful predictor of radiographic pneumonia, with both sensitivity and specificity of 75% in infants. We conclude that acute lower respiratory tract infection in children living at high altitude is frequently associated with hypoxemia, and that oxygen should be administered to children with a diagnosis of pneumonia in these regions. Case management algorithms developed in low-altitude regions may have to be modified for high-altitude settings. In this setting, pulse oximetry is a good predictor of pneumonia. Because pulse oximetry is more objective and cheaper than radiography, its role as a clinical and investigative tool merits further exploration.