About FREM

FREM_ML - Fracture Risk Evaluation Model with machine learning

In 2018, we developed the first version of the FREM algorithm for automatic MOF risk prediction. Overall, FREM was designed to automatically predict the imminent (one-year) risk of MOF for an individual considering sex, age as well as diagnoses and prescriptions in the past fifteen years.

Recent publications from the European osteoporosis society suggest that detection of high-risk individuals can be enhanced through models or technologies that estimate fracture risk automatically based on electronic health records or registry data. In line with these suggestions and on-going developments towards explainable artificial intelligence in the health sector, we enhanced the FREM with machine learning approaches in 2025.

FREM_ML was trained and tested in national registry data extracted for the Danish population aged ≥45 years without previous osteoporosis diagnoses and/or osteoporosis-related treatment (approximately 2.5 mio). Predictors of the imminent risk of MOFs observed in 2022, automatically extracted for a 15-year lookback period (2007-2021), included hospital diagnoses, filled medication prescriptions, and days since last redemption of 96 fall- and osteoporosis-specific risk medication as well as markers of polypharmacy and multi-morbidity.

FREM_ML is the core prediction engine in the CHOICE project, where it serves as a fully automated clinical decision support tool for early detection of individuals at high risk of fractures. The model eliminates the need for manual data input from healthcare professionals and patients and allows seamless integration into electronic health record systems.