Title : AI-augmented sepsis detection and response: A journey from alerts to action
Abstract:
Sepsis remains a leading cause of morbidity and mortality globally, with diagnostic ambiguity and delayed interventions contributing to poor outcomes. This presentation outlines a decade-long journey of digital transformation in sepsis care across Western Sydney Local Health District (WSLHD), culminating in the development and deployment of an AI-powered clinical decision support tool (CDST) for early sepsis detection.
The presentation explores the limitations of traditional screening methods such as SIRS and lactate thresholds, and the challenges posed by alert fatigue and static EMR workflows. It introduces the SAFE-WAIT dashboard, an AI-driven tool built using logistic regression and XGBoost models trained on historical ED data from four hospitals. The model stratifies risk based on initial vitals and demographic data, enabling proactive identification of patients at risk of sepsis in the ED waiting room.
Key findings from over 100,000 patient encounters demonstrate that the AI model significantly improved early antibiotic administration and reduced time to physician review for high-risk patients. Cost analysis revealed that early interventions correlated with reduced inpatient costs, highlighting the financial and clinical value of timely sepsis management.
The presentation emphasizes the importance of clinician-partnered design, iterative feedback, and augmentation over automation in building trust and ensuring successful AI adoption. It concludes with a vision for scalable, sustainable, and patient-centered digital health systems that integrate real-time analytics, genomics, and dynamic alerting to support a true learning health system.
