Limitations of XGBoost-SHAP integration for interpretable machine learning in antimicrobial resistance prediction
Yoshiyasu Takefuji · Journal of Infection · 2025
Yuan et al. investigated machine learning applications for predicting antimicrobial resistance in Enterobacterales bloodstream infections.1 Their research implemented XGBoost machine learning models to forecast resistance to seven antibiotics in bloodstream infection cases. Their study featured SHapley Additive exPlanations (SHAP) plots illustrating feature importance and their impacts on model output, particularly for amoxicillin resistance prediction at the time of blood culture sampling. Additional SHAP plots demonstrated how the time elapsed since the last resistant isolate influenced predictions of resistance to the same antibiotic.