AI models outperform FRAX in predicting fracture risk using EHR and DXA data

Researchers have developed and validated new models for predicting fracture risk in adults over 50, utilizing data from dual-energy X-ray absorptiometry (DXA) and electronic health records (EHR). These models, including penalized Cox regression, random survival forest, gradient-boosting survival, and XGBoost survival, demonstrated superior discrimination compared to the clinically used FRAX scores. The expanded Cox model achieved a Harrell C-index of 0.779 in internal validation and 0.714 in external validation, while gradient-boosting survival showed the highest external discrimination at 0.725. Further assessment of calibration, prospective evaluation, and implementation workflows is recommended before clinical adoption. AI

IMPACT These advanced machine learning models could improve clinical decision-making for osteoporosis management by providing more accurate fracture risk assessments than current tools.

RANK_REASON The cluster is based on a research paper published on arXiv detailing the development and validation of machine learning models for a specific medical application. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Cox proportional hazards model
• dual-energy X-ray absorptiometry
• electronic health records
• gradient-boosting survival
• Harrell C-index
• Indiana Health Information Exchange
• random survival forest
• Weill Cornell Medical Center
• XGBoost survival

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →