Sepsis remains a leading cause of morbidity and mortality worldwide, characterized by a complex interplay between host response, infection, and organ dysfunction. Early identification of individuals at high risk of poor outcomes is essential to improve clinical management and reduce mortality.
This study aims to develop and validate artificial intelligence (AI)-based prognostic models for sepsis using multi-dimensional data from the UK Biobank cohort. Specifically, we will integrate demographic, clinical, biochemical, and genetic information to identify predictors associated with sepsis incidence, severity, and 28-day mortality. Machine learning algorithms (such as gradient boosting, random forest, and deep neural networks) will be applied to construct models capable of individualized risk prediction of mortality.
Genetic data, including imputed genotypes and relevant polygenic risk scores, will be incorporated to explore host genetic susceptibility and gene-environment interactions in sepsis outcomes. Feature selection and explainable AI approaches (e.g., SHAP values) will be used to identify key determinants contributing to prediction performance. Model performance will be assessed through internal cross-validation and external validation using publicly available cohorts (e.g., MIMIC-IV).
This research will provide insights into the biological and clinical heterogeneity of sepsis, support precision medicine approaches, and facilitate early risk stratification and personalized interventions for sepsis patients.
