A Better START for Low-acuity Victims: Data-driven Refinement of Mass Casualty Triage. (3rd April 2015)

• Record Type: Journal Article
• Authors: Cross, Keith P.; Petry, Michael J.; Cicero, Mark X.
• Is Part Of: Prehospital emergency care. Volume 19:Number 2(2015:Apr./Jun.)
• Page Start: 272
• Page End: 278
• Publication Date: 2015-04-03
• Subjects: Disaster -- mass casualty -- triage -- START -- Sacco -- recursive partitioning -- trauma -- registry -- acuity; 362.18
• DOI: 10.3109/10903127.2014.942481
• Languages: English
• ISSNs: 1090-3127

Abstract

Abstract: Objective. Methods currently used to triage patients from mass casualty events have a sparse evidence basis. The objective of this project was to assess gaps of the widely used Simple Triage and Rapid Transport (START) algorithm using a large database when it is used to triage low-acuity patients. Subsequently, we developed and tested evidenced-based improvements to START.Methods. Using the National Trauma Database (NTDB), a large set of trauma victims were assigned START triage levels, which were then compared to recorded patient mortality outcomes using area under the receiver-operator curve (AUC). Subjects assigned to the "Minor/Green" level who nevertheless died prior to hospital discharge were considered mistriaged. Recursive partitioning identified factors associated with of these mistriaged patients. These factors were then used to develop candidate START models of improved triage, whose overall performance was then re-evaluated using data from the NTDB. This process of evaluating performance, identifying errors, and further adjusting candidate models was repeated iteratively.Results. The study included 322, 162 subjects assigned to "Minor/Green" of which 2, 046 died before hospital discharge. Age was the primary predictor of under-triage by START. Candidate models which re-assigned patients from the "Minor/Green" triage level to the "Delayed/Yellow" triage level based on age (either for patients >60 or >75), reduced mortality in the "Minor/Green" group from 0.6% to 0.1% and 0.3%, respectively. These candidate START models also showed net improvement in the AUC for predicting mortality overall and in select subgroups.Conclusion. In this research model using trauma registry data, most START under-triage errors occurred in elderly patients. Overall START accuracy was improved by placing elderly but otherwise minimally injured—mass casualty victims into a higher risk triage level. Alternatively, such patients would be candidates for closer monitoring at the scene or expedited transport ahead of other, younger "Minor/Green" victims.