Systematic review of risk prediction models for sepsis-associated brain dysfunction

Area of Science:

• Critical Care Medicine
• Neurology
• Biostatistics

Background:

• Sepsis-associated brain dysfunction (SABD) is a common complication of sepsis.
• Accurate risk prediction models are crucial for timely intervention and improved patient outcomes.
• Current SABD risk prediction models face challenges in reproducibility and clinical translation.

Purpose of the Study:

• To systematically review outcome constructs, modeling characteristics, and methodological quality of existing SABD risk prediction models.
• To identify reasons for the difficulty in reproducing and translating current models into practice.
• To propose standardized directions for future SABD risk prediction research.

Main Methods:

• Systematic review of studies developing or validating SABD risk prediction models.
• Searched multiple databases (PubMed, Embase, Web of Science, etc.) up to April 2025.
• Assessed model characteristics using CHARMS checklist and methodological quality using PROBAST tool.

Main Results:

• Twelve studies with 24 risk prediction models were included, focusing on sepsis-associated encephalopathy (SAE) or delirium (SAD).
• Significant heterogeneity observed in outcome definitions, modeling strategies, and variable selection.
• All studies exhibited a high risk of bias, with limited internal and external validation, and applicability concerns.

Conclusions:

• Current SABD risk prediction models are exploratory, with high bias and insufficient validation limiting clinical use.
• Future research must follow PROBAST and TRIPOD guidelines for standardized modeling and validation.
• Multicenter prospective studies are recommended to improve the reliability of SABD risk prediction.