In Search of the Truth: Choice of Ground-Truth for Predictive Modeling of Trauma Team Activation in Pediatric Trauma.

Abstract

Assigning trauma team activation levels for trauma patients is a classification task that machine learning models can help optimize. However, performance is dependent upon the "ground-truth" labels used for training. Our purpose was to investigate two ground-truths, the Cribari matrix and the Need for Trauma Intervention (NFTI), for labeling training data. Data was retrospectively collected from the institutional trauma registry and electronic medical record, including all pediatric patients (age <18 y) who triggered a trauma team activation (1/2014 - 12/2021). Three ground-truths were used to label training data: 1) Cribari (Injury Severity Score >15 = full activation), 2) NFTI (positive for any of 6 criteria = full activation), and 3) the union of Cribari+NFTI (either positive = full activation). Of 1,366 patients triaged by trained staff, 143 (10.47%) were considered under-triaged using Cribari, 210 (15.37%) using NFTI, and 273 (19.99%) using Cribari+NFTI. NFTI and Cribari+NFTI were more sensitive to under-triage in patients with penetrating mechanisms of injury (p = 0.006), specifically stab wounds (p = 0.014), compared to Cribari, but Cribari indicated over-triage in more patients who required prehospital airway management (p < 0.001), CPR (p = 0.017), and who had mean lower GCS scores on presentation (p < 0.001). The mortality rate was higher in the Cribari over-triage group (7.14%, n = 9) compared to NFTI and Cribari+NFTI (0.00%, n = 0, p = 0.005). To prioritize patient safety, Cribari+NFTI appears best for training a machine learning algorithm to predict trauma team activation level.