The burden of malaria in regions of high endemicity frequently overwhelms hospitals' capacity to provide effective care. A rapid, simple method of identifying children who are at highest risk is vital to reduce mortality among hospitalized children. Multiple regression analysis identified prognostic variables predicting mortality in severely ill children admitted to a Ghanaian teaching hospital. These variables were compared in children with and without malaria. A total of 1492 (90.2%) of 1654 severely ill children referred for assessment had evaluable outcomes. Low Blantyre coma score (BCS), high blood lactate level, and high body mass index were independent predictors of mortality among children with malaria (area under the receiver operating characteristic curve [AUC/ROC], 0.84). In children without malaria, BCS and lactate level also predicted mortality, but the addition of respiratory distress and hematocrit improved the model (AUC/ROC, 0.77). Predictors of mortality in children with malaria differ from those for other severe illnesses and reflect differences in underlying pathophysiological processes.