AI for anticoagulant therapy for patients with sepsis-induced DIC: issues and future direction of identifying sepsis subclass

Abstract

Sepsis is defined as life-threatening organ dysfunction caused by a dysregulated host response to infection, and remains a global issue due to its high incidence and mortality. Many randomized controlled trials have investigated sepsis, but no specific treatments have been established. This is due to “heterogeneity,” which indicates that a diagnosis of sepsis includes various pathophysiological states. To resolve this issue, efforts are being made to identify subclasses of sepsis. Past studies have selected different variables, such as expressed genes, cytokines, and vital signs, and identified subclasses based on clinical or biomarker-expression features. We also identified subclasses of sepsis with different coagulation features, and found that a specific anticoagulant agent (recombinant human thrombomodulin) was associated with improved survival rates in only one subclass. We also developed a model that could predict the subclass. However, none of the reported subclasses had shared characteristics, and they are not in clinical use. Further research is required to identify subclasses that are strongly associated with the pathophysiological mechanism and reproducible in any cohort.