Real-time FT-IR typing of Klebsiella pneumoniae: a flexible and rapid approach for outbreak detection and infection control

Abstract Background Expansion of carbapenemase-producing Klebsiella pneumoniae (CP-Kp) is driven by within-hospital transmission, requiring timely typing data for effective infection control. Objectives We evaluated real-time performance and flexibility of our previously developed Fourier-transform infrared (FT-IR) spectroscopy workflow (spectra acquisition and analysis by machine-learning model). Methods All CP-Kp infection isolates (n = 136) identified at a northern Portuguese hospital (April 2022—March 2023) were tested from Columbia agar with 5% sheep blood, identified by FT-IR (KL-type/sublineage) and confirmed by reference methods (wzi sequencing, MLST and/or WGS). Results FT-IR typing from Columbia agar with 5% sheep blood showed 73% sensitivity, 79% specificity and 74% accuracy. Our method correctly identified 94% of typeable isolates, 87% of which were communicated in <24 h. Non-typeable isolates belonged to new KL-types to the model (40%) or non-recognized KL-types (60%), most of which (66%) were correctly predicted when retested from Mueller–Hinton agar. Accuracy was then higher (88%) when results from both culture media were considered, and the model retrained to incorporate new sublineages. Three K. pneumoniae sublineages (ST147-KL64, ST15-KL19, ST268-KL20) were predominant and 86% of the isolates were correctly identified. During the study, an outbreak by ST268-KL20 in the neonatal ICU was quickly recognized, and solved in 23 days. Most isolates (98%) produced KPC-3. Conclusions We demonstrate that FT-IR spectroscopy meets high performance standards in real-time and adaptability to clonal dynamics, and we provide practical guidance for integrating FT-IR into daily microbiology practices. The unique time to response (same day as bacterial identification) enables early and effective infection control interventions.