Incidence of Fractures and Development of the Fracture Risk Model (FRM) in a Population-Based Cohort Study in Abu Dhabi

Abstract Fractures in older adults cause high morbidity and mortality. This study aims to identify fracture incidence and risk factors, and to develop the Fracture Risk Model-Abu Dhabi (FRM-AD). Method This retrospective cohort study of 1757 males and females aged 40+ from Abu Dhabi, United Arab Emirates, who participated in a screening program from 2016 to 2020, followed until 2024 for 4.7 years, SD =1.8. It utilized Electronic Medical Records (EMRs) data and telephone interviews. The outcome assessed was the occurrence of fractures. Results There were 58 out of the 1149 females (5.0%) and 26 out of the 608 males (4.3%) who had at least one incident fracture. Hip and vertebral fractures accounted for 12.1% and 11.5%, respectively. There was an annual incidence of 10.5 fractures per 1,000 person-years, 10.7 cases among females and 9.9 among males, rising after age 80 to 39.1 among females and 25.3 among males. Fracture Predictors identified using Poisson regression analysis were older age, smoking, history of previous fracture(s), interaction of history of fracture in a parent with a family history of osteoporosis, positive history of hip fracture in a parent after the age of 40, lack of dyslipidemia diagnosis and the interaction of diabetes mellitus diagnosis and family history of osteoporosis. A higher Body Mass Index (BMI) increased the risk of fractures in this cohort, especially in obese males. Logistic regression of variables at the end of follow-up showed that lower latest vitamin D levels were associated with increased fracture risk. FRM-AD, derived using Poisson regression, performed well in predicting fractures, with Receiver Operating Characteristic (ROC) curves of 0.736 (0.677-0.794) and 0.742 (0.684-0.800) without and with BMD, respectively. The FRAX AUC to predict MOF and hip fracture ranged between 0.624 and 0.683. Conclusion In this Emirati cohort, the locally derived FRM-AD showed moderate discrimination for incident fracture and outperformed FRAX-AD, suggesting that a locally derived model may improve fracture risk stratification. Several risk factors and predictors were identified that can be targeted for prevention.