Predicting 90-Day Mortality After Geriatric Hip Fracture Using Combined Preoperative and Perioperative Risk Factors.

Abstract

Hip fractures in the elderly are associated with high 90-day mortality rates. Most existing predictive models rely solely on preoperative variables. This study aimed to develop two parallel risk score models integrating preoperative and perioperative factors to predict 90-day mortality. Medical records of patients aged ≥65 years who underwent hip fracture surgery at our institution between January 2018 and December 2024 were retrospectively reviewed. <b>Five preoperative risk factors were evaluated: Prognostic Nutritional Index <40, age ≥85 years, ≥4 comorbidities, surgical delay >5 days, and intracapsular fracture.</b> The >5-day surgical delay threshold was determined empirically through receiver operating characteristic analysis, as the conventional 48-hour cut-off lacked discriminative capacity in this cohort where 86.8% of patients exceeded it. Model performance was assessed using logistic regression and receiver operating characteristic analysis. Of 388 patients enrolled, 387 constituted the analytical sample (60 deaths). The 90-day mortality rate was 15.5%. Individually, each factor showed weak to moderate discrimination. <b>The preoperative five-factor score achieved an AUC of 0.722, and the combined six-factor score - which additionally incorporates ICU stay ≥5 days - achieved an AUC of 0.737.</b> Each unit score increase raised mortality odds approximately 2.68-fold. In patients scoring ≥4, mortality exceeded 50%. The Hosmer-Lemeshow test confirmed adequate calibration for both models. A two-stage scoring approach combining admission-available and postoperative variables demonstrated acceptable discrimination and calibration for 90-day mortality <b>prediction in elderly hip fracture patients, supporting its use as a practical risk stratification tool at both preoperative and postoperative clinical decision points.</b>