Model predicted frailty risk in older emergency patients
The model achieved a discrimination score of 0.940 in 143 patients set aside for testing, in a retrospective study of older emergency patients.
Research Square
In a retrospective cohort study, researchers analysed records from 474 patients aged 60 or over who attended a tertiary hospital’s emergency department in Shandong Province. Frailty means reduced ability to withstand illness or other stresses. They built a machine-learning model using seven clinical features, including age, breathing rate and blood markers of inflammation and nutrition. The model’s area under the curve, a measure of how well it distinguished frailty risk, was 0.940 in the 143-patient test group.
An external analysis used a different model with 19 clinical features to predict death in hospital within 28 days. It scored 0.788 on the same measure in a 6,801-patient group from a separate database. Markers of inflammation and nutrition were leading predictors in both cohorts.
Why it matters
Frailty is associated with functional decline and death in older people. The work addresses whether clinical information can support frailty risk assessment in emergency care.
Caveats
The retrospective study did not test benefits in clinical use, and the external analysis used a different model to predict mortality rather than frailty. The paper is a preprint and has not been peer-reviewed.
The paper
Wang X, Lin Y, Han W et al.
Research Square · 6 Oct 2026 · Preprint, not peer-reviewed
