Background: No study has assessed the ability of pneumonia severity scores to identify the risk for early intensive care unit (ICU) transfer in patients with community-acquired pneumonia (CAP) admitted to general wards (GW). We aimed to compare the ability of CURB-65 (confusion, urea level, respiratory rate, blood pressure, and age ≥65 years) and SMRT-CO (systolic blood pressure, multilobar chest radiography involvement, respiratory rate, tachycardia, confusion, and oxygenation) scores to predict early ICU transfers in these patients. Design: Retrospective, case-control study. Cases were defined as patients admitted to GW with CAP that required ICU transfer within 48 hours. Controls were defined as patients admitted to GW with CAP that did not require ICU transfer. CURB-65 and SMRT-CO scores were calculated on presentation to emergency department (ED), and upon admission to GW. Composite scores were calculated combining data from ED and GW. Sensitivities, specificities, likelihood ratios, and areas-under-the-curve (AUC) were calculated for each score. Results: From 2003 to 2009, 115 cases and 345 controls were identified. Both groups had similar baseline characteristics. Composite scores combining data from ED and GW had better sensitivity and AUC than scores calculated only with ED or GW data (P < 0.001). A composite SMRT-CO score ≥2 had 76.5% (95% CI, 67.7 to 83.9) sensitivity, 67.5% (95% CI, 62.3 to 72.4) specificity, and 0.81 (95% CI, 0.77 to 0.85) AUC. A composite CURB-65 score ≥3 had 36.5% (95% CI, 27.7 to 46.0) sensitivity, 86.3% (95% CI, 82.3 to 89.8) specificity, and 0.66 (95% CI, 0.60 to 0.72) AUC to predict early ICU transfers. Composite SMRT-CO had higher sensitivity and AUC (P < 0.001) than composite CURB-65. Conclusions: Composite SMRT-CO had a better combination of sensitivity and specificity than CURB-65 for predicting early ICU transfers. Prospective studies to confirm our findings are needed. © 2011 Society of Hospital Medicine.