Machine Learning Analysis of Biomarkers and Infectious Sites in Elderly Sepsis: Distinguishing Escherichia coli from Non-Escherichia coli Infections with a Random Forest Model

Abstract

This study examines the challenge of accurately diagnosing sepsis subtypes in elderly patients, focusing on distinguishing between Escherichia coli and non-E. coli infections. Utilizing machine learning, we conducted a retrospective analysis of 119 elderly sepsis patients, employing a Random Forest model to evaluate clinical biomarkers and infection sites. The model demonstrated high diagnostic accuracy, with an overall accuracy of 87.5%, and impressive precision and recall rates of 93.3% and 87.5%, respectively. It identified infection site, Platelet Distribution Width (PDW), platelet count, and Procalcitonin (PCT) levels as key predictors, while logistic regression underscored the significance of smoking. Achieving an F1 Score of 90.3% and an ROC AUC of 88.0%, our model effectively differentiates between sepsis subtypes. This methodology offers potential for enhancing elderly sepsis diagnosis, improving patient outcomes, and contributing to the advancement of precision medicine in the field of infectious diseases.