Development and validation of a model for predicting mortality in patients with hip fracture

Abstract

Objective: to develop a user-friendly prediction tool of 1-year mortality for patients with hip fracture, in order to guide clinicians and patients on appropriate targeted preventive measures. Design: population-based cohort study from 2011 to 2017 using nationwide data from the Danish Hip Fracture Registry. Subjects: a total of 28,791 patients age 65 and above undergoing surgery for a first-time hip fracture. Methods: patient-related prognostic factors at the time of admission were assessed as potential predictors: Nursing home residency, comorbidity (Charlson Comorbidity Index [CCI] Score), frailty (Hospital Frailty Risk Score), basic mobility (Cumulated Ambulation Score), atrial fibrillation, fracture type, body mass index (BMI), age and sex. Association with 1-year mortality examined by determining the cumulative incidence, applying univariable logistic regression and assessing discrimination (area under the receiver operating characteristics curve [AUROC]). The final model (logistic regression) was utilised on a development cohort (70% of patients). Discrimination and calibration were assessed on the validation cohort (remaining 30% of patients). Results: all predictors showed an association with 1-year mortality, but discrimination was moderate. The final model included nursing home residency, CCI Score, Cumulated Ambulation Score, BMI and age. It had an acceptable discrimination (AUROC 0.74) and calibration, and predicted 1-year mortality risk spanning from 5 to 91% depending on the combination of predictors in the individual patient. Conclusions: using information obtainable at the time of admission, 1-year mortality among patients with hip fracture can be predicted. We present a user-friendly chart for daily clinical practice and provide new insight regarding the interplay between prognostic factors.