Related Experiment Videos
Development and validation of a multidimensional nomogram for predicting 28-day mortality in sepsis patients with
Piao Zhang1, Chengcheng Sun2, Yingjing Wu3
1Department of Anaesthesiology, West China Hospital, Sichuan University, Chengdu, Sichuan, 610041, China.
Objective:
To develop and validate a multidimensional nomogram for predicting 28-day mortality in sepsis patients with acute lung injury (ALI).
Methods:
A retrospective analysis was conducted on 4620 sepsis patients with ALI from the MIMIC-IV database, then these patients were divided into training (n = 3119) and validation (n = 1501) cohorts (7:3). LASSO regression combined with Boruta algorithm was used to screen predictive variables. Thereafter, these variables were utilized to construct a nomogram model. The model performance was evaluated by AUROC, calibration curves, and decision curve analysis (DCA), and SHAP analysis was applied to identify core predictors.
Results:
Twelve variables (e.g., SOFA score, lactate, creatinine) were selected to build the nomogram, which showed superior discriminative ability (AUROC = 0.869 in training set, 0.884 in validation set) compared with SOFA, APACHE II, and SAPS II scores. Calibration curves indicated good agreement between predicted and actual risks, and DCA confirmed stable clinical net benefit. SHAP analysis identified lactate, creatinine, and SOFA score as core risk factors, while platelet count and albumin as protective factors.
Conclusion:
The nomogram has excellent predictive efficacy and clinical interpretability for 28-day mortality in sepsis patients with ALI, providing a reliable tool for clinical precise intervention.