Study objective: Little is known about the accuracy and reliability of current triage methods. We examined agreement among observers with regard to the need for ED care and the ability to predict at triage the need for admission to the hospital and compared these findings with admission rates after medical evaluation and management.
Methods: We used a crossover design in which each subject was subjected to nurse or computer-guided triage first, the other type of triage second, and physician triage last. Our null hypothesis: Triage methods will yield the same results. Our patients were a consecutive sample of patients at the ED of a university-affiliated county referral center. Critically ill patients were excluded. Triage categorization was examined for interobserver agreement (kappa-statistic) and prediction of admission (sensitivity, specificity, and predictive values).
Results: Of the 5,106 patients enrolled in the study, 289 (6.2%) were admitted. With regard to the agreement of triage categorizations, we found kappa-values of .452 and .185, respectively, for physician triage compared with nurse (SE +/- .012) and computer triage (SE +/- .012)(P = .001 for the difference between the kappa values). Sensitivity and specificity in predicting admission were 41.3 and 93.8, respectively, for nurses, 61.6 and 87.1, respectively, for physicians; and 68.2 and 73.6, respectively, for computer-aided triage.
Conclusion: We found great variability among physicians, nurses, and a computer program with regard to triage decisions. Comparison of the three groups' triage decisions with actual data after medical evaluation and management showed that none of the three performed well in predicting which patients required admission. Until triage methods are standardized and validated, triage decisions should not be used to determine the timeliness of access to emergency care.