Machine learning to predict antimicrobial resistance: future applications in clinical practice?

This review explores the use of machine learning (ML) in predicting antimicrobial resistance (AMR). The review included 36 studies, primarily based on hospital and outpatient data, with the majority conducted in high-resource settings. The studies focused on predicting drug resistance in infected patients, ML-assisted antibiotic prescription, and colonization with carbapenem-resistant bacteria. The most common inputs were demographic characteristics, previous antibiotic susceptibility testing, and prior antibiotic exposure. The majority of studies targeted Gram-negative bacteria (GNB) resistance prediction. The studies showed moderate to high performance, with an AUROC ranging from 0.56 to 0.93. The review concludes that ML can potentially aid in AMR prediction, but future research is needed to design, implement, and evaluate the use and impact of ML decision support systems.