Predicting negative blood cultures from EHR using Orange Data Mining
Date:
I designed and delivered a hands-on session on reproducing a clinical machine-learning study with Orange Data Mining. Participants explored EHR-derived tabular data, missingness and class imbalance, built visual machine-learning workflows, and interpreted diagnostic metrics including sensitivity, specificity, positive predictive value, and negative predictive value.
The complete teaching materials are freely available as an open-access lesson on Zenodo: Predicting negative blood cultures from EHR using Orange Data Mining (CC BY 4.0).
The session formed part of the BioMET ‘26 pre-conference workshop Clinical AI Transforming Diagnosis, Treatment & Beyond.