At mass casualty incidents (MCIs) medical needs exceed available resources, requiring prioritization of response efforts and materials. Principles of triage have evolved since the 18th century into several modern-day algorithms that sort casualties into priority groups based on clinical parameters. It is unclear, however, if such algorithms are effective and practical during real-world MCIs. This analysis reviews the literature on use and efficacy of prehospital MCI triage algorithms.

The MEDLINE, Scopus, and Google Scholar databases were searched for peer-reviewed and grey literature on prehospital MCI medical response. Articles discussing MCI triage concepts, triage at MCIs, or algorithm efficacy were included. Articles were excluded if they described law enforcement, ethical, psychological or epidemiological perspectives without detailing the medical response.

Frequently-cited MCI triage algorithms include START (Simple Triage & Rapid Treatment); Triage Sieve; CareFlight; SALT (Sort, Assess, Lifesaving Interventions, Treatment/Transport); and RAMP (Rapid Assessment of Mentation & Pulse). They differ in the physiologic parameters assessed, inclusion of numerical measurements, and number of triage categories. Surveyed providers were less likely to have performed full triage at MCIs (16%) than in training (69%), and more likely to have performed no triage (29% vs. 1%). In retrospective trauma registry analyses, algorithms were generally poorly predictive of the need for life-saving interventions (13-58% sensitive, 72-97% specific) in one study, and variably predictive of critical injury (45-85% sensitive, 86-96% specific) in another. The Glasgow Coma Scale motor component was associated with critical injury (73% sensitive, 96% specific if <6); other physiologic variables had sensitivities under 40%. In prospective studies, algorithms were accurate for 36-52% of adults and 56-59% of children. Some suggest clinician judgment may be similarly effective.

Multiple algorithms exist for MCI triage, but they are infrequently utilized and may be inaccurate. Simpler, more realistic, scalable, and widely accepted response systems need to be instituted.