Next Generation Imminent Fracture Risk Assessment Using AI

Osteoporosis-related fractures are a significant cause of morbidity and loss of independence, particularly in older adults. While traditional tools like FRAX predict long-term fracture risks, they lack precision in identifying Imminent Fracture Risk, defined as the risk of fractures within a two-year period following an initial incident. This study develops a machine learning model that integrates demographic and clinical factors to address this gap, aiming to improve short-term fracture risk predictions and enable timely interventions.The model was trained on a dataset of 32,677 patients from the Ontario Osteoporosis Strategy’s Fracture Screening and Prevention Program. It leverages an ensemble learning framework that combines Support Vector Machine, Decision Tree, Logistic Regression, and AdaBoost classifiers. Using GridSearchCV for hyperparameter tuning, the model achieved an accuracy of 76%, a recall of 76%, and an AUROC of 0.73, highlighting its potential for clinical application.Despite these promising results, limitations such as the absence of Bone Mineral Density data and incomplete patient-reported information restricted the model’s generalizability. Future research should focus on expanding the dataset, incorporating real-time data from wearable devices, and utilizing advanced natural language processing techniques to handle unstructured data effectively.This study demonstrates the potential of machine learning in predicting imminent fracture risk, offering a complementary tool to traditional methods like FRAX. By improving the early identification of high-risk individuals, this approach could significantly reduce both personal and economic burdens associated with osteoporotic fractures.