Study Objective: We sought to retrospectively measure the accuracy of multiple-casualty incident (MCI) triage algorithms and their component physiologic variables in predicting adult patients with critical injury. Methods: We performed a retrospective review of 1,144 consecutive adult patients transported by ambulance and admitted to 2 trauma centers. Association between first-recorded out-of-hospital physiologic variables and a resource-based definition of severe injury appropriate to the MCI context was determined. The association between severe injury and Triage Sieve, Simple Triage and Rapid Treatment, modified Simple Triage and Rapid Treatment, and CareFlight Triage was determined in the patient population. Results: Of the physiologic variables, the Motor Component of the Glasgow Coma Scale had the strongest association with severe injury, followed by systolic blood pressure. The differences between CareFlight Triage, Simple Triage and Rapid Treatment, and modified Simple Triage and Rapid Treatment were not dramatic, with sensitivities of 82% (95% confidence interval [CI] 75% to 88%), 85% (95% CI 78% to 90%), and 84% (95% CI 76% to 89%), respectively, and specificities of 96% (95% CI 94% to 97%), 86% (95% CI 84% to 88%), and 91% (95% CI 89% to 93%), respectively. Both forms of Triage Sieve were significantly poorer predictors of severe injury. Conclusion: Of the physiologic variables used in the triage algorithms, the Motor Component of the Glasgow Coma Scale and systolic blood pressure had the strongest association with severe injury. CareFlight Triage, Simple Triage and Rapid Treatment, and modified Simple Triage and Rapid Treatment had similar sensitivities in predicting critical injury in designated trauma patients, but CareFlight Triage had better specificity. Because patients in a true mass casualty situation may not be completely comparable with designated trauma patients transported to emergency departments in routine circumstances, the best triage instrument in this study may not be the best in an actual MCI. These findings must be validated prospectively before their accuracy can be confirmed. [Garner A, Lee A, Harrison K, Schultz CH. Comparative analysis of multiple-casualty incident triage algorithms. Ann Emerg Med. November 2001;38:541-548.]