The abstract is read at the publisher; the summary is JClub's.
A day-7 multiclass model in a geriatric rehabilitation cohort (N=10,920) predicts minimal dependency, nursing-care dependency, or medical deterioration with AUC 0.81 overall.
- Why it matters: Early, accurate discharge-risk estimates are needed to guide timely multidisciplinary planning and monitoring, but routine-data tools that work across time are limited.
- What they did: Using multinomial regression and XGBoost trained on 80% of 2016–2024 admissions from a large national HMO and tested on the remaining 20% plus other HMOs, the models used pre–day-7 demographics, function, cognition, clinical, and lab variables linked to community records.
- The result: Multinomial regression matched FIM alone (AUC 0.81 vs 0.72) and held up in temporal validation on 2025 admissions (AUC 0.80), enabling probability-threshold alerts with PPVs of 72% (nursing-care dependency) and 69% (deterioration).
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.