AI-Powered Tool Fracture-ML Shows High Accuracy in Predicting Hip Fracture Risk in Sweden

Researchers led by Kristian Axelsson and Mattias Lorentzon in Sweden have developed an AI-based clinical decision support tool called Fracture-ML that predicts hip fracture risk using national register data. The study analyzed data from over 3.5 million individuals aged 50 and above, tracking 142,327 hip fractures up to 2021. Fracture-ML demonstrated excellent predictive accuracy with area under the curve (AUC) values of 0.89 after one year, 0.88 after two years, and 0.85 after five years.

Unlike existing tools such as FRAX, which rely on individual patient data and limit population-level screening, Fracture-ML uses 139,980 potential predictors from registries, enabling a resource-efficient, large-scale risk assessment without patient examinations. A simplified version with 35 predictors also maintained strong performance.

Importantly, Fracture-ML identified nearly seven times more individuals at risk for hip fractures within two years compared to current strategies focused on patients with recent fractures, while maintaining specificity. This advancement supports improved primary prevention efforts, such as targeted bone density testing and osteoporosis treatment.

The authors conclude that integrating register data into AI-powered clinical decision support can enable precise, early detection of high-risk individuals, potentially reducing morbidity, mortality, and healthcare costs associated with hip fractures in Sweden.