Sepsis severity score: an internationally derived scoring system from the surviving sepsis campaign database*

Area of Science:

• Critical Care Medicine
• Epidemiology
• Health Services Research

Background:

• The Surviving Sepsis Campaign requires robust patient-level data for international comparisons.
• Existing severity-of-illness scoring systems may lack the pragmatism and parsimony needed for widespread clinical application.
• Accurate mortality prediction is crucial for quality assessment in sepsis management.

Purpose of the Study:

• To evaluate a pragmatic and parsimonious severity-of-illness scoring system for patients with sepsis.
• To develop and validate a Sepsis Severity Score for accurate hospital mortality prediction.
• To enable appropriate quality comparisons using practical clinical and research applications.

Main Methods:

• Prospective, observational evaluation of data from 23,438 patients with suspected or confirmed sepsis across 218 hospitals in 18 countries (January 2005–March 2010).
• Development of the Sepsis Severity Score using maximum likelihood logistic regression, with model coefficients estimated.
• Model validation using a randomly selected 90% of patients for development and the remaining 10% for validation, supplemented by bootstrapping techniques.

Main Results:

• The Sepsis Severity Score demonstrated good calibration (Hosmer-Lemeshow p > 0.05) and discrimination (Area Under the Curve 0.736 development, 0.748 validation).
• Observed and estimated hospital mortality probabilities were both 0.334 for the total population.
• The score performed consistently across deciles of predicted mortality and international geographic regions.

Conclusions:

• The Sepsis Severity Score accurately estimates hospital mortality in severe sepsis and septic shock.
• The score exhibits robust calibration and discrimination, consistent across patient subgroups and international settings.
• This validated tool provides a practical and effective means for quality comparisons in sepsis research and clinical practice.