Assessing the prognostic utility of the De Ritis ratio in mortality prediction models for trauma intensive care unit patients

Abstract Background The De Ritis ratio (aspartate aminotransferase/alanine aminotransferase, AST/ALT) has been reported as a prognostic biomarker in various medical conditions and has been associated with worse outcomes in trauma patients. However, its added prognostic value for trauma intensive care unit (ICU) patients remains unclear. This study aimed to determine whether an elevated De Ritis ratio on ICU admission predicts in‐hospital mortality and improves the accuracy of established trauma prognostic models. Methods We retrospectively analyzed 2281 adult trauma ICU patients (age ≥ 20 years) at a Level I trauma center in Taiwan from 2016 through 2020. Admission AST and ALT levels were measured to calculate the De Ritis ratio. Clinical data (demographics, injury severity, and outcomes) were collected from a trauma registry. We examined the association between the admission De Ritis ratio and in‐hospital mortality. We also assessed whether adding the ratio improved the area under the receiver operating characteristic curve (AUC) of standard mortality prediction models (Acute Physiology and Chronic Health Evaluation II, Simplified Acute Physiology Score II, Multiple Organ Dysfunction Score, Logistic Organ Dysfunction System, Mortality Probability Model II). Results Among 2281 patients, 261 (11.4%) died during hospitalization. Nonsurvivors had a significantly higher mean admission De Ritis ratio than survivors (2.2 vs. 1.56 and p < 0.001). Incorporating the De Ritis ratio into the prediction models yielded minimal improvement in their predictive performance. Only one model showed a significant increase in AUC from 0.813 to 0.840 (ΔAUC = 0.027 and p = 0.003) after adding the ratio. All other models demonstrated negligible nonsignificant changes in AUC. Conclusion An elevated De Ritis ratio at ICU admission is associated with higher mortality risk in trauma ICU patients. However, integrating this biomarker into existing mortality prediction models provided only marginal improvement in predictive performance.