Development and Validation of a Multidimensional Predictive Model for 28-Day Mortality in Patients with...

Area of Science:

• Critical Care Medicine
• Pulmonology
• Trauma Surgery

Background:

• Post-traumatic acute respiratory distress syndrome (ARDS) is a significant cause of mortality.
• Accurate prediction of mortality is crucial for timely clinical intervention.

Purpose of the Study:

• To develop and validate a multidimensional nomogram for predicting 28-day all-cause mortality in post-traumatic ARDS patients.
• To identify key predictors of mortality in this patient population.

Main Methods:

• Retrospective analysis of 667 post-traumatic ARDS patients from the MIMIC-IV database.
• Development of a nomogram using LASSO regression and the Boruta algorithm.
• Validation using AUROC, calibration curves, decision curve analysis (DCA), and SHAP analysis.

Main Results:

• A 10-variable nomogram demonstrated superior predictive ability (AUROC = 0.848 training, 0.846 validation) compared to SOFA, APACHE II, and machine learning models.
• Lactate and platelet transfusion units were identified as core risk factors; albumin and base excess trauma as protective factors.
• The nomogram showed good calibration and clinical net benefit via DCA.

Conclusions:

• The developed nomogram is an effective and interpretable tool for predicting 28-day mortality in post-traumatic ARDS.
• This nomogram can aid clinicians in making timely interventions for improved patient outcomes.