Limited accuracy of current mortality prediction scales in extracapsular hip fractures: A case-control study

Mortality prediction tools are commonly used in hip fracture care, yet most were developed in heterogeneous hip fracture populations and may not adequately reflect the risk profile of older adults with extracapsular hip fractures.

We performed a retrospective case-control study including patients aged >65 years who underwent surgery for extracapsular hip fractures. For the 12-month analysis, 57 patients who died within 12 months were matched 1:1 with 57 survivors by age, sex, and surgery date. The study design was chosen to compare discriminatory performance efficiently in a fracture-specific cohort and was not intended to estimate incidence or absolute risk. We assessed O-POSSUM, Nottingham Hip Fracture Score (NHFS), Almelo Hip Fracture Score (AHFS), and SERNBO. The primary performance metric was discrimination, quantified by area under the receiver operating characteristic curve (AUC) with 95% confidence intervals.

The cohort had a mean age of 86 years and 73.7% were women. Discrimination for 12-month mortality was modest for all scores (AUCs 0.63-0.66). NHFS showed the highest AUC (0.66, 95% CI 0.56-0.75), followed by O-POSSUM (0.65, 95% CI 0.55-0.75) and SERNBO (0.64, 95% CI 0.54-0.74). Combined-score models produced only a small, non-significant increase in apparent 12-month discrimination, which attenuated after bootstrap internal validation.

In extracapsular hip fractures, commonly used mortality prediction scores showed only modest discriminatory ability and appear insufficient for individual-level clinical decision-making. These findings support the development of fracture-specific prognostic models that better capture geriatric domains such as cognition and baseline functional status.