Prognostic factors and prediction model for 1-year mortality after proximal humeral fracture.

Abstract

Proximal humeral fractures (PFH) are the third most common major non-vertebral osteoporotic fractures in older persons and result in an increased mortality risk. The aim of this study was to investigate if patient characteristics can be used to predict 1-year post-fracture mortality. Retrospective study with 261 patients aged 65 and older who were treated for a PHF in University Hospitals Leuven between 2016 and 2018. Baseline variables including demographics, residential status, and comorbidities were collected. The primary outcome was 1-year mortality. A clinical prediction model was developed using LASSO regression and validated using split sample and bootstrapping methods. The discrimination and calibration were evaluated. Twenty-seven (10.3%) participants died within 1-year post-PHF. Pre-fracture independent ambulation (p < 0.001), living at home at time of fracture (p < 0.001), younger age (p = 0.006), higher BMI (p = 0.012), female gender (p = 0.014), and low number of comorbidities (p < 0.001) were predictors for 1-year survival. LASSO regression identified 6 stable predictors for a prediction model: age, gender, Charlson comorbidity score, BMI, cognitive impairment, and pre-fracture nursing home residency. The discrimination was 0.891 (95% CI, 0.833 to 0.949) in the training sample, 0.878 (0.792 to 0.963) in the validation sample and 0.756 (0.636 to 0.876) in the bootstrapping samples. A similar performance was observed for patients with and without surgery. The developed model demonstrated good calibration. The combination of 6 pre-fracture characteristics demonstrated good predictive properties for mortality within 1year of PHF. These findings can guide PHF treatment decisions.