Proof of concept study on early forecasting of antimicrobial resistance in hospitalized patients using machine learning and simple bacterial ecology data

Antibiotic resistance is a global health threat, exacerbated by misuse of antibiotics. A machine learning-based system predicts antimicrobial susceptibility and explains cofactors in hospitalized patients at four stages before antibiograms. Comparing state-of-the-art methods, multilayer dense neural networks and Bayesian models are suitable for early prediction, with AUROCs reaching 0.88 at positive culture and 0.92 at species identification. The system’s potential clinical applications are discussed.