Background: Hip fractures in the elderly are associated with advanced comorbidities and high mortality rates. Mortality prediction models can support clinicians in tailoring treatment for medical decision making in frail elderly patients. The aim of this study was to develop and internally validate the Brabant Hip Fracture Score, for 30-day (BHFS-30) and 1-year mortality (BHFS-365) after hip fracture.
Material and methods: A cohort study was conducted in 2 hospitals on operatively treated patients of 65 years and older with a hip fracture. Manual backward multivariable logistic regression was used to select independent predictors of 30-day and 1-year mortality. Internal validation was performed using bootstrapping techniques. Model performance was assessed with: (1) discrimination via the area under the receiver operating characteristic curve (AUC); (2) explained variance via Nagelkerke's R2; (3) calibration via Hosmer-Lemeshow (H&L) test and calibration plots.
Results: Independent predictors of 30-day mortality were: age, gender, living in an institution, Hb, respiratory disease, diabetes and malignancy. In addition, cognitive frailty and renal insufficiency, were selected in the BHFS-365. Both models showed acceptable discrimination after internal validation (AUC = 0.71 and 0.75). The Hosmer-Lemeshow test indicated no lack of fit (p > 0.05).
Discussion: We demonstrated that the internally validated and easy to use BHFS in surgically treated elderly patients after a hip fracture showed acceptable discrimination and adequate calibration. In clinical practice a cutoff of BHFS-30 ⩾ 24 could identify frail elderly patients at high risk for early mortality and could support clinicians, patients and families in tailoring treatment for medical decision making.
Keywords: Clinical prediction model; elderly; hip fracture; mortality.