Few large studies have evaluated risk factors for mortality among children hospitalized for pneumonia and this may contribute to suboptimal case management efficiency. To identify useful screening criteria for mortality among children hospitalized for pneumonia in a developing country setting, we conducted a population-based hospital cohort study among children less than 2 years of age admitted for pneumonia during 1999-2001 at one of three major hospitals on Lombok Island, Indonesia. Of 4351 children admitted for pneumonia, 12 per cent died before discharge. Case fatality proportions were seasonal, with peaks occurring immediately after peaks in the proportion of cases positive for respiratory syncytial virus. Children with an oxygen saturation < or = 85 per cent or age younger than 4 months were 5.6 times more likely to die than children with none of these predictive factors (95 per cent CI, 4.5-7.1); 83 per cent of children who died had one of these two risk factors. For children < 4 months old, mortality increased at an oxygen saturation < 88 per cent compared with < 80 per cent for older children. Laboratory, physical examination, and radiological findings were not associated with or did not contribute substantially to mortality prediction. Among children hospitalized for pneumonia, age less than 4 months and hypoxia were identified with those at high risk of death. Age influences cut-off levels for hypoxia.