A Computational Approach to Evaluating Empirical Antibiotic Coverage for Gram-Negative Bloodstream Infections in Pediatric Febrile Neutropenia

Background: Empirical antibacterial therapy for febrile neutropenia requires adaptation to local epidemiology, a process that is often complex, time-consuming, and prone to human error. This study aims to address this challenge by developing a practical, data-driven tool to efficiently evaluate and adapt treatment protocols. Methods: We developed a novel, open-source computational script in Python (version 3.10), aided by large language models for code revision, to analyze antibiotic susceptibility data. The script was validated using a retrospective dataset of 237 Gram-negative bloodstream infections (BSIs) from 2015 to 2024 in cancer or hematopoietic stem cell transplant recipients at a tertiary care pediatric hospital in Italy. The script calculates efficacy metrics for both single agents and two-drug combinations. Results: Among the Gram-negative BSI strains analyzed, meropenem monotherapy demonstrated the highest efficacy (median 95.4%). In contrast, piperacillin/tazobactam and cefepime showed lower efficacy (80.3% and 81.8%, respectively). On the contrary, combination therapy, particularly with amikacin, significantly increased the efficacy of beta-lactams, elevating their effectiveness to a level comparable to meropenem. Conclusions: The developed script is a valuable tool for antimicrobial stewardship programs, offering a rapid and accessible method to validate international guidelines against local epidemiological data. While meropenem shows high efficacy, its broad use should be limited to prevent resistance. The combination of piperacillin–tazobactam and amikacin is identified as a robust and effective empirical treatment choice.