Head-to-head comparison of disaster triage methods in pediatric, adult, and geriatric patients.

STUDY OBJECTIVE A variety of methods have been proposed and used in disaster triage situations, but there is little more than expert opinion to support most of them. Anecdotal disaster experiences often report mediocre real-world triage accuracy. The study objective was to determine the accuracy of several disaster triage methods when predicting clinically important outcomes in a large cohort of trauma victims. METHODS Pediatric, adult, and geriatric trauma victims from the National Trauma Data Bank were assigned triage levels, using each of 6 disaster triage methods: simple triage and rapid treatment (START), Fire Department of New York (FDNY), CareFlight, Glasgow Coma Scale (GCS), Sacco Score, and Unadjusted Sacco Score. Methods for approximating triage systems were vetted by subject matter experts. Triage assignments were compared against patient mortality at hospital discharge with area under the receiver operator curve. Secondary outcomes included death in the emergency department, use of a ventilator, and lengths of stay. Subgroup analysis assessed triage accuracy in patients by age, trauma type, and sex. RESULTS In this study, 530,695 records were included. The Sacco Score predicted mortality most accurately, with area under the receiver operator curve of 0.883 (95% confidence interval 0.880 to 0.885), and performed well in most subgroups. FDNY was more accurate than START for adults but less accurate for children. CareFlight was best for burn victims, with area under the receiver operator curve of 0.87 (95% confidence interval 0.85 to 0.89) but mistriaged more salvageable trauma patients to "dead/black" (41% survived) than did other disaster triage methods (≈10% survived). CONCLUSION Among 6 disaster triage methods compared against actual outcomes in trauma registry patients, the Sacco Score predicted mortality most accurately. This analysis highlighted comparative strengths and weakness of START, FDNY, CareFlight, and Sacco, suggesting areas in which each might be improved. The GCS predicted outcomes similarly to dedicated disaster triage strategies.