Development and validation of a sepsis-specific systemic inflammation classification system for mortality prediction

Authors:

Abstract
Background: The role of nutrition and systemic inflammation in the prognosis of sepsis remains unclear. This study aimed to evaluate the predictive value of existing nutrition/inflammation-based indices for mortality in sepsis patients and to develop a sepsis-specific systemic inflammation classification (SSIC) system. Methods: Using the MIMIC-IV 3.0 database, we investigated and compared the predictive value of 18 nutrition/inflammation-based indices for mortality in sepsis patients admitted to the intensive care unit (ICU). These indices were calculated from biochemical tests and anthropometric measurements at ICU admission. The SSIC system was established by combining existing indices and validated in an independent institutional cohort. Results: Analysis of 11,577 sepsis patients from the MIMIC-IV 3.0 database revealed that seven of the 18 nutrition/inflammation-based indices had predictive value for in-hospital and in-ICU mortality, albeit with low discriminative ability. The neutrophil-to-lymphocyte ratio (NLR) and glucose-to-lymphocyte ratio (GLR), both of which focus on systemic inflammation, demonstrated the best predictive performance in both indirect and direct comparisons. The SSIC system, which combines NLR and GLR, classified systemic inflammation risk into low, moderate, and high categories. This system improved discriminative ability for mortality and showed consistent predictive value across different subpopulations. Moreover, adding the SSIC system to existing severity scores enhanced their discriminative performance for sepsis mortality. The utility of the SSIC system was validated in an independent institutional cohort of 496 sepsis patients. Conclusion: This study confirmed the predictive value of 18 nutrition/inflammation-based indices for mortality in sepsis patients. The SSIC system was established and validated as a tool for mortality risk stratification and may be considered for further evaluation alongside established severity scores.
Funding Information
  • National Natural Science Foundation of China (82503220)

This publication has 44 references indexed in Scilit: