External Validation of the TERMINAL-24 Score in Predicting Mortality in Patients with Multiple Trauma

Abstract

Objective: A prediction model: "TERMINAL-24,” was developed and internally validated for use in predicting early mortality of multiple trauma patients in the Emergency Department. In this study this model's external validity and generalizability was evaluated.Material and Methods: A retrospective cohort was used for the construction of two datasets. Temporal external validation used the dataset from the same location at a different period, and geographic external validation used the dataset from a different location.Results: In total, 1,932 patients underwent temporal external validation, with 14 (0.7%) patients dyeing within 8 hours, 35 (1.8%) patients died between 8 and 24 hours, and 1,883(97.5%) patients were alive at 24 hours. From this, 2,336 patients were eligible for geographical external validation, with 106 (4.5%) patients having died at the emergency room, 143 (6.1%) patients died in hospital and 2,087 (89.3%) patients survived. The TERMINAL-24 score was applied to both datasets, with a benchmark of 4 or higher (range 0-5). In the temporal dataset, this score showed a mortality of greater than 20% (specificity 0.97) area under the receiver operating characteristic curve (AuROC) 0.91 (95% Confidence interval (CI) 0.85-0.96); whereas, it demonstrated a mortality of greater than 60% (specificity 0.99) AuROC 0.92 (95%CI 0.89-0.94) in the geographical dataset.Conclusion: TERMINAL-24 was effective at predicting early death in the emergency room. It was successfully implemented within the same hospital; hoever, the cut-point should be adapted for application in other institutions with unspecified time of death. Prospective studies at different hospitals should be planned to generalize this scoring system for clinical practice.