SepsisCalc: Integrating Clinical Calculators into Early Sepsis Prediction via Dynamic Temporal Graph Connection

Abstract

Early prediction of sepsis is crucial for improving patient outcomes in clinical settings. This paper introduces SepsisCalc, a novel approach that integrates clinical calculators into sepsis prediction through dynamic temporal graph connections. By leveraging both structured electronic health record data and clinical calculators commonly used by healthcare professionals, our model captures complex relationships between clinical variables over time. The dynamic temporal graph connection framework enables effective information flow between different data sources, leading to more accurate and interpretable sepsis predictions. Experimental results demonstrate that SepsisCalc outperforms existing methods in early sepsis detection while providing clinically meaningful insights.

Dakuo Wang

Associate Professor

My research interests include human-computer interaction, human-AI collaboration, natural language processing, health informatics, data science, and computer-supported cooperative work.