Predicting Mortality in Non-Variceal Upper Gastrointestinal Bleeding: Machine Learning Models Versus Conventional Clinical Risk Scores.

This represents a remarkable improvement over the best conventional score, the pre-endoscopic Rockall, and the difference in performance was highly significant. The ability of the ML model to attain high sensitivity and specificity simultaneously is particularly valuable in the emergency setting—it enables identification of high-risk patients without over-triaging those at low risk.

also in 132,142 other papers