A novel approach to antimicrobial resistance: Machine learning predictions for carbapenem-resistant Klebsiella in intensive care units

Abstract

This study was conducted at Kocaeli University Hospital in Turkey and aimed to predict carbapenem-resistant Klebsiella pneumoniae infection in intensive care units using the Extreme Gradient Boosting (XGBoost) algorithm, a form of artificial intelligence. This was a retrospective case-control study involving 289 patients, including 159 carbapenem-resistant and 130 carbapenem-susceptible individuals as controls. The model's predictive analysis combined a diverse range of demographic, clinical, and laboratory data, resulting in an average accuracy of 83.0%, precision of 83%, sensitivity of 88%, F1 score of 85%, and Matthews Correlation Coefficient of 0.66. Prolonged hospitalization and intensive care unit stay were significant predictors of carbapenem-resistant Klebsiella pneumoniae infection. The role of artificial intelligence role in healthcare, particularly in ICUs for managing antibiotic-resistant infections, is a major development in medicine. This study emphasizes the potential of artificial intelligence to predict antimicrobial resistance and improve clinical decisions in resource-limited settings. The study was approved by ClinicalTrials.gov (trial registration number NCT05985057 on 02.08.2023).

Keywords: Carbapenem-Resistant Enterobacteriaceae; Intensive Care Units; Klebsiella pneumoniae; Supervised Machine Learning.