Artificial intelligence based predictive models for early sepsis detection in intensive care units: a scoping review.

Background: Early detection of sepsis in intensive care units remains a major clinical challenge. Artificial intelligence based predictive models have emerged as promising tools to support early identification of sepsis, yet their clinical readiness and methodological robustness remain heterogeneous. Objective: To map and critically synthesize the available evidence on artificial intelligence based predictive models for early sepsis detection in intensive care units. Methods: A scoping review was conducted following the PRISMA ScR framework. Multiple databases were systematically searched to identify studies developing or validating artificial intelligence based models for early sepsis detection in adult intensive care settings. Data were extracted on study design, data sources, model type, prediction horizon, validation strategies, and reported performance. Results: Thirty seven studies were included. Most models were developed using retrospective electronic health record data and relied on machine learning techniques, with limited external validation. Reported performance varied widely, and few studies addressed clinical implementation, interpretability, or integration into real time workflows. Conclusions: Although artificial intelligence based models show potential for early sepsis detection, substantial gaps remain regarding external validation, clinical integration, and real world applicability. Future research should prioritize methodological transparency and implementation focused evaluation.