Assessment of fracture risk tools in care home residents: a multi-centre observational pilot study

Key summary pointsAim To compare the efficacy of four fragility risk assessment tools and the Timed Up and Go test (TUGT) in care home residents. Findings None of the tools was reliable for predicting falls. The QFractureScore, BMI and the Garvan nomogram were the best predictors of fractures and combined falls & fractures. In the multiple logistic regression analyses, age was the only statistically significant covariate associated with falls, fractures and combined falls & fractures. Message Three of the five tools tested predicted fragility fractures in the care home residents. Of these, the BMI is easiest to use, and is therefore most suitable for this population. Background Fragility fractures are common in care home residents but established tools have not been tested in this population. Aim:To identify the most practicable tool for use. MethodsDesign Multicentre prospective observational cohort pilot study. Setting: 18 care homes in Boston, UK. Assessments: fragility risk score at baseline with FRAX, QFractureScore, Garvan nomogram, body mass index and TUGT for each participant. Outcomes: falls, fractures, combined falls & fractures. Follow-up; 12 months. Results 217/618 (35%) residents in the 18 care homes were enrolled. 147 (68%) had mental capacity,70 (32%) did not. There were 325 falls and 10 fractures in participants during the study. At the same time there were 1671 falls and 103 fractures in residents not participating in the study. Multiple regression analyses showed that only age had a statistically significant association with falls (chi(2)(1) = 5.7775, p = 0.0162), fractures (chi(2)(1) = 4.7269, p = 0.0297) and combined falls & fractures (chi(2)(1) = 4.7269, p = 0.0297). C-statistics were: falls; FRAX 0.544, BMI 0.610, QFractureScore 0.554, Garvan nomogram 0.579, TUGT 0.656, fractures; FRAX 0.655, BMI 0.708, QFractureScore 0.736, Garvan nomogram 0.712, TUGT 0.590, combined falls and fractures, c-statistics were same as for fractures. Fifty-four participants (25%) died during follow-up. Charlson comorbidity index predicted mortality, R-2 = 0.021 (p = 0.034). Conclusions QFractureScore, BMI and Garvan nomogram were good predictors of fractures and combined falls and fractures Only age had statistically significant association with the outcomes. No tool was good predictor of falls.