Using Statistical and Machine Learning Methods to Evaluate the Prognostic Accuracy of SIRS and qSOFA

Area of Science:

• Critical Care Medicine
• Emergency Medicine
• Health Informatics

Background:

• Sepsis diagnosis relies on early identification for effective treatment.
• Systemic Inflammatory Response Syndrome (SIRS) and quick Sepsis-related Organ Failure Assessment (qSOFA) are commonly used criteria.
• Comparing their diagnostic performance is crucial for clinical practice.

Purpose of the Study:

• To compare the diagnostic performance of qSOFA and SIRS criteria for early sepsis detection.
• To evaluate the predictive accuracy of these criteria for 28-day in-hospital mortality.
• To utilize statistical and machine learning models for performance assessment.

Main Methods:

• Retrospective analysis of emergency department patient visits with sepsis-related diagnoses.
• Utilized odds ratios (OR) and machine learning models (decision tree, logistic regression, naïve Bayes).
• Assessed the association between qSOFA/SIRS criteria and 28-day mortality.

Main Results:

• qSOFA (≥2) showed a stronger association with mortality (OR=3.06) than SIRS (≥2) (OR=1.22).
• The area under the ROC curve for qSOFA (0.70) was significantly higher than for SIRS (0.63).
• Machine learning models demonstrated varying sensitivity and specificity for both criteria, with qSOFA generally outperforming SIRS.

Conclusions:

• qSOFA is a more effective diagnostic criterion for sepsis compared to SIRS.
• Optimizing qSOFA's sensitivity through threshold adjustments can improve sepsis prediction.
• Findings support healthcare providers in selecting appropriate risk-stratification tools for sepsis patients.