Introduction: Triage is an essential process used to adequately allocate resources and thus increase chances of survival in case of mass-casualty incidents (MCIs). Several triage scales are currently used, but data regarding their performance remain scarce. The objective was to compare the performance of two prehospital triage algorithms (Sieve versus SwissPre) using a validated physiological simulator.

Methods: This was a web-based, randomized open-label study. A real-time evolutive simulator based on a heart-lung-brain interaction model embedding functional blocks was used to simulate the evolution of vital parameters. Participants, who were randomly allocated to either algorithm, were asked to triage 30 patients in random order. The primary outcome was the triage score (each correct decision was awarded one point). The "Immediate patients" were defined as those who would die within the first hour according to the physiological model. The secondary outcome was the duration of patient triage.

Results: Out of 71 participants, 67 (94.4%) were included in the final analysis. The Sieve group achieved a mean score of 17.1 out of 30 (95%CI, 16.3 to 17.8). The SwissPre group scored 15.5 out of 30 (95%CI, 14.5 to 16.5). The mean difference between groups was 1.6 points (95%CI, 0.4 to 2.8; P = .011) in favor of the Sieve algorithm. Triage duration did not differ significantly between the Sieve (mean 43 minutes, SD = 10) and SwissPre (mean 46 minutes, SD = 23) groups, with a mean difference of three minutes (95%CI, -12 to 6; P = .507).

Conclusions: The simpler Sieve algorithm may slightly outperform the more complex SwissPre in accurately categorizing critically injured patients who would likely die within 60 minutes if left untreated. No significant difference was observed in triage speed. However, these exploratory findings should be interpreted cautiously, considering the mean difference was modest and the controlled simulated setting, limiting generalizability.