Antimicrobial resistance analysis of Klebsiella pneumoniae bloodstream infections based on a random forest algorithm: a longitudinal study based on data from tertiary hospitals in China from 2012 to 2023

This large retrospective study of more than 109,000 Klebsiella pneumoniae bloodstream infections in China shows a high and increasing burden of antimicrobial resistance, with nearly one-third of isolates resistant to carbapenems and over 40% resistant to third-generation cephalosporins and fluoroquinolones. The dominant carbapenem-resistant strain was ST11 carrying the blaKPC gene, which—together with blaNDM—strongly predicted carbapenem resistance, while ICU admission was the main predictor of 30-day mortality. Patients with carbapenem-resistant infections had almost double the mortality compared with susceptible cases. A machine-learning model accurately predicted resistance trends and patient outcomes, underscoring the value of genomic surveillance and data-driven tools to guide targeted treatment and reduce excess mortality from resistant K. pneumoniae infections.