SOFA+: Applying Machine Learning to Enhance the Prognostic Power of the SOFA Score for Predicting ICU Sepsis Mortality

Sepsis remains one of the leading causes of mortality in the Intensive Care Unit (ICU), with early risk stratification playing a critical role in improving patient outcomes. While various scoring systems have been developed for this purpose, the Sequential Organ Failure Assessment (SOFA) score is among the most widely used. This study aims to enhance the predictive power of the SOFA score for ICU sepsis mortality by integrating machine learning techniques. Using data from 15,100 ICU patients in the MIMIC-IV v3.0 dataset for five-fold cross-validation, we developed the SOFA+ model, which retains the same clinical variables as the original SOFA score. The resulting model, SOFA+, outperformed the conventional SOFA score. Ten machine learning models was evaluated to compare their predictive performance in terms of AUC in the dataset extracted from MIMIC-IV database. Incorporating machine learning into the SOFA scoring system significantly enhances its ability to predict ICU sepsis mortality. The SOFA+ model offers a more efficient and accurate prognostic tool, potentially facilitating earlier and more targeted clinical interventions.