The possibility to automate medical triage is appealing as it could decrease the burden on humans. In instances of mass casualty events, this could allow for much faster triage as the process would occur in parallel instead of in sequence. The greatest efficacy could be achieved with a system that relies solely on remote sensors, which would necessitate an adaptation of existing triage algorithms that rely on human observations. A policy capturing method is proposed to demonstrate the possibility to mimic medical experts’ decision-making model of triage based only on observable parameters sampled by wearables: heart rate, respiration rate, heart rate variability, and blood oxygen saturation. Two medical experts classified simulated cases from these five parameters and sex in regards of four potential outcomes: Delayed green, Urgent yellow, Immediate red, or Expectant black. Seven model types were trained to replicate the decision pattern of the experts. Overall, the decision pattern was best captured by a decision tree (test set accuracy of 92% and 76% for Raters 1 and 2, respectively). Interestingly, common physiological differences were found across the four classifications for both experts. The model was optimized during a workshop with the experts. We discuss the implications of using such a model to support medical triage, especially for military contexts.