Machine learning algorithms for early sepsis detection in the emergency department: A retrospective study

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical Decision Support Systems

Background:

• Early sepsis recognition and treatment are critical for patient outcomes.
• Sepsis diagnosis is challenging due to non-specific clinical presentations.
• Novel screening tools are needed for timely sepsis identification.

Purpose of the Study:

• To develop and evaluate machine learning models for early sepsis risk prediction.
• To compare the performance of machine learning models against traditional sepsis screening tools.
• To assess the utility of electronic health record data for sepsis prediction.

Main Methods:

• Retrospective analysis of 133,707 emergency department visits (June 2018-May 2020).
• Development of sepsis prediction models using logistic regression, gradient boosting, random forest, and neural network algorithms.
• Comparison of model performance (AUROC, sensitivity, specificity) against qSOFA, MEWS, and SIRS using a held-out testing set.

Main Results:

• Machine learning models significantly outperformed traditional methods in sepsis prediction.
• The random forest model achieved the highest Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.931.
• Performance metrics demonstrated superior accuracy for machine learning models over qSOFA, MEWS, and SIRS.

Conclusions:

• Machine learning models show superior performance for sepsis diagnosis in emergency settings.
• These AI-driven tools offer a promising advancement over conventional sepsis screening methods.
• Further research is warranted to validate clinical utility and impact on patient outcomes.