Machine learning-based mortality prediction in hip fracture patients using biomarkers

Area of Science:

• Geriatric Medicine
• Data Science in Healthcare
• Orthopedic Surgery

Background:

• Hip fractures are a significant cause of morbidity and mortality in older adults.
• Accurate prediction of mortality risk is crucial for patient management and resource allocation.

Purpose of the Study:

• To develop and evaluate a machine learning model for predicting mortality after hip fracture.
• To identify key demographic and laboratory variables associated with 1-, 5-, and 10-year mortality.

Main Methods:

• Retrospective analysis of 3751 hip fracture patient records from the MIMIC-IV database.
• Evaluation of 10 machine learning classification models, with a focus on LightGBM.
• Identification and analysis of top-weighted features for mortality prediction.

Main Results:

• The LightGBM model achieved 81% accuracy and an AUC of 0.79 for 1-year mortality prediction.
• Key predictors included age, glucose, red blood cell distribution width, and white blood cell count.
• Several identified biomarkers were consistent across 1-, 5-, and 10-year mortality models.

Conclusions:

• Machine learning models utilizing basic clinical data can effectively predict hip fracture mortality.
• Specific biomarkers like age and red blood cell distribution width are strong indicators of long-term survival.
• These findings can inform clinical decision-making and the development of risk stratification tools.