Testing the Use of Symptom-Based Terrorism Triage Algorithms with Hospital-Based Providers

Introduction: A set of symptom-based, all-hazards, decision-making algorithms was designed to aid the first-contact provider during early patient presentations after a terrorist incident.Objective: The primary objective was to assess the usability of these algorithms. A secondary objective was to assess the psychometric properties of the testing scenarios.Methods: This was a written, usability assessment of the algorithms employing a convenience sample of hospital-based, healthcare providers who had not taken any specific training in the use of the algorithms. A series of 26 paragraph-length, moderately difficult scenarios was created to reflect possible agents, means of attack, and types of patients. Each of the 26 scenarios requires that one make a triage choice on the “attack” algorithm (the trunk algorithm), then proceed to one of four other branch algorithms (dirty resuscitation, chemical agents, biological agents, bomb/blast/radiation dispersal device) to make a final triage choice. Conditional scores based on getting both the attack and final card correct were calculated for each algorithm.Results: Nineteen attending physicians, 50 emergency medicine residents, and 41 nurses took the assessment. The total score was 45% correct for all participants. The score on the attack algorithm was 66% correct. Dirty resuscitation, biological, chemical, and bomb/blast scores were 46%, 54%, 46%, and 51% respectively. The probability of guessing the correct answer on the attack algorithm was 1/7 or 14%. The conditional probability of guessing both the attack algorithm and the final card correct ranged from 4.7% for the biological, chemical, and bomb/blast algorithms to 2.4% for the dirty resuscitation algorithm. Item discrimination, item difficulty, and Cronbach's alpha were acceptable for the overall test. Certain individual items had item difficulty levels suggesting they were too difficult and should be replaced in future versions of the test.Conclusions: Performance on the test suggests that participants did substantially better than would have been expected by chance alone. Future efforts will revise the algorithms with the goal of simplification. Revision of the testing instrument and testing algorithm use after instruction also are needed.