FRACTURE RISK ASSESSMENT IN LONG-TERM CARE (FRAIL) PREDICTS NON-VERTEBRAL FRACTURES

We recently developed a model (FRAiL) to predict 2-year risk of hip fracture in NH residents. The purpose of this study was to determine whether FRAiL can be used to predict non-vertebral fractures. Competing risk regression was used to model associations between 14 clinical characteristics and the outcome non-vertebral fracture. Models were repeated after adding 29 medication classes associated with bone loss or falls. Among 894,584 residents (mean age 83.8 years), 4.7% experienced a non-vertebral fracture, whereas 40.6% died. In the adjusted model, 11 clinical characteristics predicted fracture (C-index=0.65). Falls (HR 1.32, 95% CI 1.29, 1.36) and wandering (HR 1.24, 95% CI 1.19, 1.29) predicted fracture, whereas dependence with Activities of Daily Living (versus independence) was protective (HR 0.53, 95% CI 0.46, 0.61). After adding medications, 9 classes remained significant (C-index 0.66). Calibration was excellent. FRAIL is a practical model to identify NH residents at greatest risk for non-vertebral fracture.