Background. World Health Organization treatment guidelines recommend that adults with severe malaria be admitted to an intensive care unit (ICU). However, ICU facilities are limited in the resource-poor settings where most malaria occurs. Identification of patients at greater risk of complications may facilitate their triage and resource allocation. Methods. With use of data from a trial conducted in Southeast Asia (n = 868), a logistic regression model was built to identify independent predictors of mortality among adults with severe malaria. A scoring system based on this model was tested in the original dataset and then validated in 2 series from Bangladesh (n = 188) and Vietnam (n = 292). Results. Acidosis (base deficit) and cerebral malaria (measured as Glasgow Coma Score) were the main independent predictors of outcome. The 5-point Coma Acidosis Malaria (CAM) score was simply derived from these 2 variables. Mortality increased steadily with increasing score. A CAM score <2 predicted survival with a positive predictive value (PPV) of 95.8% (95% confidence interval [CI], 93%- 97.7%). Of the 14 of 331 patients who died with a CAM score <2, 11 (79%) had renal failure and death occurred late after hospital admission (median, 108 h; range, 40-360 h). Substitution of plasma bicarbonate as the measure of acidosis only slightly reduced the prognostic value of the model. Use of respiratory rate was inferior, but a score <2 still predicted survival with a PPV of 92.2% (95% CI, 89.1%-94.7%). Conclusions. Patients with a CAM score <2 at hospital admission may be safely treated in a general ward, provided that renal function can be monitored.