Presentation (CTI)

OBJECTIVES: Multidrug-resistant (MDR) Gram-negative infections in the intensive care unit (ICU) present a major evidence gap. Clinicians often make empiric treatment decisions before microbiological confirmation, while many patients are too acutely ill to enrol in conventional randomised trials. This work developed a practical framework for applying target trial emulation to ICU MDR Gram-negative infections to support credible real-world evidence generation where randomisation is impractical.
METHODS: A target trial framework was developed drawing on causal inference Hernan-Robbins principles, structured engagement with ICU and infectious disease clinicians, and review of best-practice guidance for real-world evidence in acute care. Protocol specification required explicit definition of eligibility criteria, treatment strategies, assignment procedures, outcomes, follow-up, and causal contrasts, adapted to high-acuity clinical settings. Emulation guidance addressed mapping these elements to linked electronic health record, microbiology, and medication administration data, with data quality assessed for reliability and relevance.
RESULTS: Seven design considerations were identified. First, empiric treatment decisions before culture results and definitive decisions after susceptibility confirmation represent different target trials and should not be conflated. Second, time zero should align with the clinical decision point, not microbiological confirmation, to avoid immortal time bias and ensure estimates reflect actionable choices. Third, treatment strategies should pre-specify allowable modifications, including dose adjustment, de-escalation, toxicity-driven switching, and combination regimens. Fourth, active-comparator designs are preferred to unconstrained best-available-therapy comparators, which vary across sites and resistance ecology. Fifth, linked clinical, microbiology, and medication data are required; claims data alone are insufficient. Sixth, confounding by indication must be addressed through design and analytical methods. Seventh, collaborative specification by clinicians, methodologists, and data scientists is essential.
CONCLUSIONS: Target trial emulation offers a structured approach to generating treatment evidence for ICU MDR Gram-negative infections. By making assumptions explicit and aligning studies with clinical workflows, it can complement randomised and platform trials.