Antimicrobial resistance recommendations via electronic health records with graph representation and patient population modeling

Pei Gao, Zheng Chen, Xin Liu, Peng Chen, Yasuko Matsubara, Yasushi Sakurai

Abstract: Highlights•Electronic Health Records (EHRs) are conceptualize as graph structures where nodes represent individual medical events and edges signify concurrent occurrences of these events.•Antimicrobial resistance (AMR) recommendations are decided based on the graphical similarity of EHRs for different patients learned by a deep graph neural network (GNN).•Considering population-level modeling, a G2GNN method is introduced for estimating commonalities among patients with spatial specificity and addressing the imbalance in AMR labels.•An end-to-end framework is designed for multi-antibiotic recommendations aligned with in-lab testing labels, utilizing a multi-task learning (MTL) strategy.