Predicting fracture outcomes from clinical registry data using artificial intelligence supplemented models for...

Area of Science:

• Orthopedic Surgery
• Medical Informatics
• Artificial Intelligence

Background:

• Distal radius (wrist) fractures are common hospital admissions with diverse injury patterns.
• Clinical management guidelines for wrist fractures are inconsistent, limiting evidence-based practice.
• Predictive modeling using patient and fracture characteristics can reduce care variation and improve outcomes.

Purpose of the Study:

• To utilize AI on unstructured data for enhanced description of wrist fracture characteristics.
• To assess if AI-derived fracture information improves prediction of patient outcomes compared to standard registry data.
• To reduce variation in care and improve patient outcomes for distal radius fractures.

Main Methods:

• Adult patients (16+) with wrist fractures at four Victorian hospitals will be studied.
• Data from the Victorian Orthopaedic Trauma Outcomes Registry (VOTOR) and electronic medical records (EMRs) will be used.
• A multimodal deep learning fracture reasoning system (DLFRS) will be developed to analyze EMR data, with machine learning models testing its performance.

Main Results:

• AI techniques will provide enhanced fracture characteristic information for wrist fractures.
• Prediction models incorporating AI-derived characteristics are expected to outperform standard models.
• Improved identification of key fracture characteristics will be achieved through AI analysis.

Conclusions:

• The PRAISE study will establish AI's utility in analyzing unstructured EMR data for wrist fractures.
• AI-derived fracture characteristics are anticipated to enhance the prediction of clinical and patient-reported outcomes.
• This approach promises to improve the management and outcomes of distal radius fractures.