The abstract is read at the publisher; the summary is JClub's.
In a Danish multicentre cohort of 2·08 million hospital encounters, XGB-EWS yields only small average net benefit over NEWS for 24 h mortality, with gains concentrated in initial NEWS ≥7.
- Why it matters: Retrospective NEWS validations can be distorted by intervention bias, and it is unclear whether simpler rules are clinically sufficient or whether complex machine learning adds meaningful utility across patient risk groups.
- What they did: Adults (≥18 years) with at least one NEWS assessment in 15 Danish hospitals (Oct 16, 2018–Oct 16, 2023) were followed for all-cause death within 24 h; models were compared using propensity score weighting, decision-curve net benefit, AUROC, and calibration.
- The result: Overall, XGB-EWS vs NEWS showed a maximum net benefit difference of 1·9 additional correct identifications per 10·000 patients, while NEWS vs simplified NEWS showed 1·5; in NEWS ≥7, XGB-EWS gained 78·1, but simplified NEWS lost 60·2, whereas in NEWS 0–4 differences were minimal.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.