Development and Validation of a Multidimensional Predictive Model for 28-Day Mortality in Patients with Post-Traumatic Acute Respiratory Distress Syndrome

<b>Objective:</b> To develop and validate a multidimensional nomogram for predicting 28-day all-cause mortality in patients with post-traumatic acute respiratory distress syndrome (ARDS). <b>Methods:</b> A retrospective analysis was conducted on 667 post-traumatic ARDS patients from the MIMIC-IV database, divided into training (n = 466) and validation (n = 201) cohorts (7:3). LASSO regression combined with the Boruta algorithm was used to screen variables and construct a nomogram. Model performance was evaluated by AUROC, calibration curves, and decision curve analysis (DCA) with SHAP analysis to identify core predictors. <b>Results:</b> Ten variables (e.g., lactate, platelet transfusion units, D-dimer) were selected and used to construct the nomogram model. The nomogram showed superior discriminative ability (AUROC = 0.848 in training set, 0.846 in validation set) compared with SOFA, APACHE II scores, and machine learning models (XGBoost, random forest). Calibration curves confirmed good agreement between predicted and actual risks, and DCA indicated better clinical net benefit. SHAP analysis identified lactate and platelet transfusion units as core risk factors and albumin and base excess trauma as protective factors. <b>Conclusions:</b> The nomogram has excellent predictive efficacy and interpretability, providing a reliable tool for clinical intervention in post-traumatic ARDS patients.