Bacterial Infection Risk in Hospitalized COVID-19 Patients
Machine-learning tool designed to help rule out bacterial infection
in adults hospitalized for COVID-19.
Clinical Features
Laboratory Features
ng/mL
g/dL
×109/L
mg/L
This XGBoost model was developed by Muhammad Hamdan Gul, MD using the
National COVID Cohort Collaborative (N3C) dataset (n = 302,179; bacterial infection prevalence 10.8 %). Model performance: Area under the ROC curve (AUC) = 0.836 (83.6 %), internally validated.
It is intended as a decision-support aid—final clinical decisions remain with the treating team.