Machine learning model can predict 28-day mortality in sepsis patients

Sepsis is one of the most common and fatal syndromes encountered in intensive care units (ICUs), and acute respiratory failure (ARF) is one of the most serious complications. When respiratory failure develops, patients often experience severe hypoxemia and multiple organ failure over a short period of time, resulting in a significantly increased risk of death. Despite advances in critical care, accurately assessing short-term prognosis early after ICU admission remains a major challenge in clinical practice.

In a recent study, researchers including Dr. Jian Liu of Gansu Provincial Maternal and Child Health Hospital (Gansu Provincial Central Hospital), Technician Zi Yang of Lanzhou University First Hospital in China, Dr. Hong Guo of Gansu Provincial Maternal and Child Health Hospital (Gansu Provincial Central Hospital) in China, and other researchers developed and tested a machine learning model to predict 28-day mortality in patients with sepsis. Complicated by ARF. The results of this study were published online. Intensive Medical Journal January 10, 2026.

Commenting on the study, Dr. Liu said:This model is designed to leverage clinical information available during the early stages of ICU admission, allowing clinicians to quickly identify high-risk patients and thereby optimize treatment strategies and monitoring resource allocation.. ”

The Medical Information Mart for Intensive Care IV (MIMIC-IV, version 3.1) database was used as a development and training cohort containing adult ICU patients who met diagnostic criteria for both sepsis and ARF. To assess the applicability of the model across different hospitals and patient populations, independent external validation was performed using data from the eICU Collaborative Research Database (eICU-CRD, version 2.0). This combined “training and external validation” design increases the relevance of the results to real-world clinical practice.

During variable selection, candidate predictors were first identified based on sepsis-related international guidelines and expert clinical consensus to ensure strong clinical relevance. We then applied the Boruta feature selection algorithm and multicollinearity analysis to identify a final set of 20 key predictive features. All selected variables were routinely obtainable within the first 24 hours after ICU admission and reflected multiple clinical aspects such as oxygenation status, organ function, metabolic parameters, and disease severity.

Seven machine learning algorithms were systematically compared, including logistic regression, random forests, gradient boosting, and neural networks. Among them, the XGBoost model showed the best overall performance. In the training cohort, the model showed strong discrimination for predicting 28-day mortality, and its performance remained stable in an independent external validation cohort, indicating good generalizability. Unlike traditional “black box” predictive models, this study placed special emphasis on interpretability. Researchers applied SHapley Additive exPlanations (SHAP) to quantify the contribution of individual clinical variables to mortality risk prediction.

”Our analysis highlighted the importance of oxygenation index, serum albumin level, liver function-related indicators, and disease severity score in short-term prognosis. This transparent, interpretable framework may facilitate clinician understanding and encourage the use of the model as a decision support tool rather than a substitute for clinical judgment.” explained engineers Zi Yang and Dr. Hong Guo.

According to the research team, this model could be further integrated into bedside or web-based risk assessment tools to support early risk stratification of sepsis patients with ARF. Overall, this study demonstrates the potential of interpretable machine learning approaches in critical care medicine and provides a new technological pathway for personalized management of high-risk patients with sepsis.