Zuyderland Hip Inference for Survival and Lifetime Expectancy (ZHISLE) following hip fracture surgery: validation of the model that demonstrated good predictive power.

Purpose: Proximal femoral fractures are common within the elderly population and are associated with a high risk of mortality and reduced quality of life. Hemiarthroplasty or osteosynthesis (extramedullary or intramedullary) is the primary treatment option for these fractures. However, within this fragile patient population many comorbidities, among others dementia, are seen. Therefore, predicting patients with a high mortality risk after surgery may lead to adopting alternative treatment options with less risks. This paper proposes a new model to distinguish patients with high postoperative mortality risk with adequate follow-up time in combination with a wide set of useful and available variables.

Methods: Patients treated with hemiarthroplasty or osteosynthesis for proximal femoral fractures were studied, with a follow-up period of 6 months. Patients who died within this follow-up period were compared to survivors, and predicting variables were assessed in logistic regression: The Zuyderland Hip Inference for Survival and Lifetime Expectancy (ZHISLE). The model was validated internally against a held-out dataset. Furthermore, the model performance was compared against the Almelo Hip Fracture Score (AHFS) on the same sample.

Results: Out of 2463 patients undergoing surgical treatment for proximal femoral fractures, 415 (16.8%) died within 183 days. Predictors for early mortality included old age, male sex, high heartbeat, KATZ-ADL and GFI scores, C-reactive protein and urea concentrations and low albumin concentration. Our model showed satisfactory predictive and discriminatory power (ROC curve = 0.81). Internal validation was good (ROC in validation dataset = 0.81), and better than the AHFS (ROC = 0.57).

Conclusions: The ZHISLE model demonstrates good predictive power concerning mortality risk for old patients with a proximal femoral fracture. The model could benefit patients by indicating if a conservative, non-invasive policy might be a better option for those patients.