Effect of competing mortality risks on predictive performance of the QFracture risk prediction tool for major osteoporotic fracture and hip fracture: external validation cohort study in a UK primary care population

Objectives: To externally evaluate QFracture for predicting major osteoporotic fracture (MOF) and hip fracture.

Design, setting and participants: Linked primary care, hospital admission and mortality data from the Clinical Research Practice Datalink GOLD. People aged 30-99 years with up-to-standard linked data for at least one year were eligible. MOF was defined as any hip, distal forearm, proximal humerus or vertebral crush fracture, ascertained from GP, hospital discharge and mortality data. QFracture 10-year predicted MOF and hip fracture risk was calculated, and performance evaluated versus observed 10-year fracture risk in the whole population, and in subgroups of age and comorbidity. QFracture calibration was examined with and without accounting for competing non-fracture mortality risk.

Results: There were 2,747,409 women with 95,598 MOF and 36,400 hip fractures, and 2,684,730 men with 34,321 MOF and 13,379 hip fractures. Incidence of all fractures was higher than in QFracture internal derivation. Competing mortality risk was more common than fracture from middle-age onwards. QFracture discrimination in the whole population was excellent or good for MOF and hip fracture (Harrell’s-C in women 0.813 and 0.918 respectively; in men 0.738 and 0.888), but was poor to moderate in age subgroups (e.g. Harrell’s-C in women and men aged 85-99 respectively 0.576 and 0.624 for MOF, and 0.624 and 0.637 for hip). Without accounting for competing risks, QFracture systematically under-predicted fracture risk in all models, more so for MOF than hip fracture, and more so in older people. Accounting for competing risks, QFracture still under-predicted in the whole population but showed considerable over-prediction in older age-groups and people with high comorbidity at high fracture risk.

Conclusion: QFracture systematically under-predicts fracture risk (because of fracture under-ascertainment) and over-predicts in older and comorbid people (because of competing mortality). The use of QFracture in its current form needs reviewing, particularly in people at high risk of death from other causes.