An accurate on-scene assessment of a trauma patient’s injury by emergency medical service providers is critical to determining the right hospital for prompt care delivery. However, literature and anecdotal evidence from trauma professionals in the Midwestern U.S. suggest that mistriages (i.e., an error in assessing a severely injured patient as having non-severe injuries, and vice versa) often occur in practice. Due to a lack of approaches that consider such mistriages while designing the trauma network, we propose a stochastic nested multi-level, multi-transportation capacitated facility location model, which explicitly considers mistriages in injury assessment, to maximize patient safety by determining the number and locations of major trauma centers (MTCs). Assessment-related mistriages are modeled using a Bernoulli random variable for each of the two injury groups, moderate (ISS 9–15) and severe (ISS>15). We propose a Simheuristic approach that integrates Monte Carlo Simulation with a genetic algorithm. We convert infeasible solutions into feasible ones during offspring generation and maintain a superior population pool to balance quality against computational time. Our findings based on a realistic data set suggest that assessment-related mistriages in the ISS>15 group can substantially decrease patient safety and may lead to a clustering of MTCs, while mistriages in the ISS 9–15 group tend to have a dispersed distribution of MTCs in the resulting network. Our case study using actual 2019 data from a midwestern state of the U.S. resulted in an over 20% reduction in the mistriages, further providing evidence for incorporating injury assessment-related mistriages in trauma network design.