This systematic review, published on 22 September, found 40 studies from 2010 to 2026 that built artificial intelligence or machine-learning models to predict outcomes after osteoporotic hip fracture. Most predicted mortality; fewer addressed delirium, complications, rehabilitation or surgical failure.
The models generally discriminated moderately to strongly, especially for mortality. But external validation was inconsistent, calibration was often not reported, and uncertainty was frequently missing.
For orthogeriatric teams, the gap is between a model that ranks patients well in its own data and one that gives accurate risks in a new hospital. Established scores remain the practical tools for now.
- Keep using established risk scores for hip fracture counselling.
- Before adopting any AI model, ask whether it was validated outside the hospital that built it.
- Ask whether calibration was reported; good ranking can still give wrong absolute risks.
- Mortality prediction is furthest along; delirium and function models are less mature.
Why it matters
A model that ranks risk well can still give the wrong number to a family deciding on surgery.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for orthopaedics, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free