Predicting future hospital antimicrobial resistance prevalence using machine learning

Abstract

AbstractObjectivesPredicting antimicrobial resistance (AMR), a top global health threat, nationwide at a hospital level could help target interventions. Using machine learning, we exploit historical AMR and antimicrobial usage to predict future AMR.MethodsAntimicrobial use and AMR prevalence in bloodstream infections in hospitals in England were obtained per hospital group (Trust) and financial year (FY, April-March) for 22 pathogen-antibiotic combinations (FY2016-2017-FY2021-2022). XGBoost model predictions were compared to previous value taken forwards, difference between the previous two years taken forwards and linear trend forecasting (LTF). XGBoost feature importances were calculated to aid interpretability.ResultsRelatively limited year-to-year variability in AMR prevalence within Trust-pathogen-antibiotic combinations meant previous value taken forwards achieved a low mean absolute error (MAE). XGBoost models performed similarly, while difference between the previous two years taken forwards and LTF were consistently worse. XGBoost considerably outperformed all other methods in Trusts with a larger change in AMR prevalence from FY2020-2021 (last training year) to FY2021-2022 (held-out test set). Feature importance values indicated that complex relationships were exploited for predictions.ConclusionYear-to-year resistance has generally changed little within Trust-pathogen-antibiotic combinations. In those with larger changes, XGBoost models could improve predictions, enabling informed decisions, efficient resource allocation, and targeted interventions.