Association Between the Sequential Organ Failure Assessment-2 (SOFA-2) Score and 90-Day and 365-Day Mortality in Patients With Sepsis: A Retrospective Cohort Study Based on the MIMIC-IV Database
Zhengchao Li, Yuzhu Wang, Yudong Gao, Yulin Tian, Fenggao Zhou, Xu Cheng, Bing Gong, Dexing YangBackground
The Sequential Organ Failure Assessment-2 (SOFA-2) score is a recently developed update of the conventional SOFA score that incorporates contemporary patterns of organ support in the intensive care unit (ICU). However, the association between SOFA-2 and long-term mortality in patients with sepsis has not been well characterized.
Methods
This retrospective cohort study extracted adult sepsis ICU data from MIMIC-IV v3.1. ICU admission SOFA-2 scores (0–24) were calculated. Boruta algorithm selected confounders; binary logistic regression and restricted cubic splines (RCS) analyzed linear association between SOFA-2 and 90-/365-day all-cause mortality. Random forest (RF) combined key predictors, SHAP interpreted feature importance, and AUC evaluated discriminative performance.
Results
8,779 patients were enrolled. Fully adjusted regression showed SOFA-2 independently elevated 90-day mortality (OR=1.17, 95%CI 1.15–1.18) and 365-day mortality (OR=1.16, 95%CI 1.14–1.18). RCS confirmed an approximately linear correlation (all nonlinearity p>0.05). Single SOFA-2 yielded modest AUCs: 0.65 (90-day) and 0.64 (365-day). The RF model improved prediction (validation AUC: 0.68 for 90-day, 0.67 for 365-day). SHAP analysis identified SOFA-2 as the most influential predictor.
Conclusions
SOFA-2 provides moderate, stable long-term mortality discrimination for sepsis patients. Further external validation and optimization are required to improve its risk stratification utility.