Introduction: Trauma is the leading cause of death in children globally. Approximately, 10% of admissions to pediatric intensive care units (PICUs) are trauma related. Establishing a common method of evaluating outcomes in PICUs is a critical factor in improving the quality of patient care. In the pediatric population, there is no certainty whether critical illness scores or trauma scores are the best predictors of mortality. Materials and methods: Retrospective review of the trauma registry of the PICU of eliminar Ricardo Gutiérrez Children's Hospital, which includes deidentified data from all admissions to the PICU between 2005 and 2017. The scores evaluated were pediatric trauma score (PTS), injury severity score (ISS), new injury severity score (NISS), revised trauma score (RTS), trauma injury severity score (TRISS), and pediatric index of mortality (PIM2). Results: In total, 360 patients (56% male) with an average age of 60 months, 30 to 116 (median, interquartile range) were included. In total, 73% suffered traumatic brain injury, 26% had injuries to the extremities, 19% to the chest, 14% to the abdomen, 6% to the pelvis, and 5% to the spine or spinal cord. In total, 43% had injuries in one body region, 29% in two regions, and 28% in more than two. The most frequent combination was a brain injury with an injury to the extremities (16%). The main causes of trauma were falls (42%), followed by vehicular collision against pedestrians (20%) and injuries to motor passengers (7%). The overall mortality was 6.0%. We evaluated the discriminative ability of each score using receiver operating characteristic curve (ROC) with mortality as the outcome of interest (area under the curve and 95% CI): NISS 0.749 (0.63–0.86), ISS 0.788 (0.69–0.89), PTS 0.899 (0.84–0.96), RTS 0.912 (0.84–0.98), TRISS 0.933 (0.86–0.99), and PIM2 0.973 (0.93–1.0). Conclusion: In this sample from a single institution, the PIM2 score had a superior discrimination ability than trauma scores using mortality as the outcome variable. These results need to be tested in a study with a larger population sample.