Open Access

Stefanie S. Ebel, Andreas Bayer, Stephan Prückner, Lukas Kreß, Axel R. Heller, Christina C. Bartenschlager

• Managing a mass casualty incident presents significant challenges for both emergency responders and hospitals. Consequently, hospitals must prioritize incoming patients using triage algorithms. Prior research from our working group demonstrated substantial differences in the accuracy of these triage algorithms. Therefore, improvements to existing algorithms should aim to enhance their sensitivity (SE) and specificity (SP) while ideally minimizing the effort required for their implementation. Method Building on the Berlin Algorithm (BER) for in-hospital triage and incorporating clinical experience, an adaptation was proposed for LMU University Hospital. With approval from the German Federal Office of Civil Protection and Disaster Assistance, 210 expert-validated case vignettes were utilized to analyze this Munich in-hospital triage algorithm (MUC). Each vignette was processed through the modified algorithm using a computer-assisted approach to assess its diagnostic performance. ResultsManaging a mass casualty incident presents significant challenges for both emergency responders and hospitals. Consequently, hospitals must prioritize incoming patients using triage algorithms. Prior research from our working group demonstrated substantial differences in the accuracy of these triage algorithms. Therefore, improvements to existing algorithms should aim to enhance their sensitivity (SE) and specificity (SP) while ideally minimizing the effort required for their implementation. Method Building on the Berlin Algorithm (BER) for in-hospital triage and incorporating clinical experience, an adaptation was proposed for LMU University Hospital. With approval from the German Federal Office of Civil Protection and Disaster Assistance, 210 expert-validated case vignettes were utilized to analyze this Munich in-hospital triage algorithm (MUC). Each vignette was processed through the modified algorithm using a computer-assisted approach to assess its diagnostic performance. Results Compared to the BER, SP for triage category I improved from 0.89 to 0.93 while maintaining the same SE. For triage category II, SE increased from 0.37 to 0.48, with a minor decrease in SP from 0.92 to 0.91. Both algorithms achieved identical SE and SP for triage category III. The overall test performance, measured by the Youden index, was highest for the modified algorithm, with values of 0.93 for category I and 0.39 for category II, outperforming all other examined algorithms, including the original model. Conclusion Beyond the data-driven validation of the MUC, this study demonstrates that even simple modifications to triage algorithms, such as changes in the sequence of assessment items, can significantly influence their diagnostic performance in in-hospital triage settings.…