Extended Data Table 5 Performance Metrics for CONCERN Predictive Modela

From: Real-time surveillance system for patient deterioration: a pragmatic cluster-randomized controlled trial

  1. aMultinomial Gradient Boosted Machine (GBM) model built on random 12-hour time slices to predict (over the next 24 h) whether a patient is discharged alive, will still be in the hospital or has a hospital event (in-hospital mortality, cardiopulmonary arrest, sepsis, unanticipated ICU transfer). Modeling was trained on 70% of the dataset, with 30% used for 10-fold cross-validation. Average performance reported for ensemble models22.
Back to article page