Using data from NORA, we used 18 potential risk factors in a classification and regression tree analysis to build two algorithms. These algorithms correctly identified postmenopausal women between the ages of 50 and 64 years who were at the highest risk of osteoporotic fracture within 36 months.
Introduction: The objective was to develop algorithms that best predict short-term fracture risk (3 years) in postmenopausal women 50-64 years old.
Methods: Data were from 91,652 women who responded to follow-up surveys as part of National Osteoporosis Risk Assessment (NORA) study. Peripheral bone mineral density (BMD) and risk factors obtained at baseline; incident osteoporotic fractures obtained from follow-up surveys. Eighteen risk factors were entered into a classification and regression tree analysis to build two algorithms, one with and one without BMD.
Results: Two thousand and seven (2.2%) women reported new osteoporotic fractures. Prior fracture, a peripheral BMD T-score <or= -1.1, and self-reported fair/poor health status were the most important determinants for short-term fracture and were associated, respectively, with 7.2%, 3.1%, and 2.4% fracture risk within 3 years. This algorithm with three risk factors correctly classified 65% of the women who experienced an incident fracture and 59% of the women who did not experience an incident fracture. Without BMD T-score, the most important determinants for fracture prediction were previous fracture, self-reported fair/poor health status, and no current use of postmenopausal hormone therapy.
Conclusions: NORA-based algorithms may be useful for health care providers to guide further assessment and management decisions to prevent fractures in younger postmenopausal women.