Long-term mortality, refracture, and 1-year mortality risk after osteoporotic fracture: development in a public cohort and external validation in an independent hospital cohort

Background Long-term mortality and refracture are major outcomes after osteoporotic fracture, yet their prognostic patterns may differ according to fracture category and patient characteristics. We aimed to characterize long-term mortality, refracture risk, and post-fracture pathways after osteoporotic fracture and to develop and externally validate a parsimonious model for 1-year post-discharge mortality. Methods We conducted a retrospective prognostic study using a publicly available development cohort and an independent hospital-based external validation cohort. The development cohort comprised 4,773 patients with osteoporotic fractures reconstructed from a publicly released longitudinal dataset. Long-term all-cause mortality was estimated from index admission using Kaplan–Meier methods, whereas first refracture was analysed using cumulative incidence functions with death treated as a competing event. Adjusted prognostic associations were assessed using Cox, cause-specific Cox, and Fine–Gray models, and post-fracture transitions were summarized descriptively. A nine-variable Cox proportional hazards model incorporating age, sex, fracture category, body mass index, ASA class, source-provided Charlson Comorbidity Index, haemoglobin, albumin, and creatinine was developed to estimate 365-day mortality after discharge. Internal validation used 500 patient-level bootstrap samples, with multiple imputation re-estimated within each resample. The frozen model was then applied without refitting to an independent retrospective hospital cohort of 695 patients, among whom 28 deaths occurred within 365 days. Results In the development cohort, 5-year all-cause mortality was 17.64% (95% CI, 16.19–19.22), compared with a 5-year cumulative incidence of first refracture of 8.41% (95% CI, 7.30–9.51). Mortality was highest after hip/femur fracture, reaching 32.10% at 5 years (95% CI, 29.24–35.17). Female sex was associated with lower mortality (HR, 0.64; 95% CI, 0.54–0.76) but higher refracture risk (cause-specific HR, 1.47; 95% CI, 1.09–1.98; Fine–Gray sHR, 1.54; 95% CI, 1.14–2.08). Lumbar spine/pelvis and thoracic spine/rib fractures were similarly associated with lower mortality than hip/femur fracture but higher refracture risk. After optimism correction, the 365-day IPCW time-dependent AUC in the development cohort was 0.860, with a calibration slope of 0.958. In the external cohort, the AUC for 365-day mortality was 0.802 (95% CI, 0.709–0.884), the calibration slope was 0.849 (95% CI, 0.540–1.237), and the observed-to-expected ratio was 1.006 (95% CI, 0.667–1.387). Twenty of the 28 deaths (71.4%) occurred in the highest predicted-risk quintile. Findings were materially unchanged in the outcome-informed multiple-imputation sensitivity analysis. Conclusions Patients hospitalized with osteoporotic fractures experienced substantial long-term mortality and refracture burden, with distinct prognostic patterns across fracture categories and patient characteristics. A parsimonious clinical Cox model retained good discrimination and close overall calibration for 1-year post-discharge mortality when applied without refitting to an independent hospital cohort. The model may provide a practical basis for post-fracture mortality risk stratification and warrants further evaluation across broader clinical settings.