Abstract

Background: Mortality after geriatric hip-fracture surgery remains substantial. This study evaluated whether preoperative GNRI, comorbidity burden, and CBC-derived inflammatory indices were associated with one-year and long-term mortality after PFN for pertrochanteric/intertrochanteric fractures. Methods: In this single-centre retrospective cohort study of prognostic risk markers, 248 PFN records were screened. After excluding six incomplete records and 25 patients aged <65 years, 217 older fracture/surgical episodes formed the time-to-event cohort; 194 were evaluable for binary one-year mortality. Living patients/episodes with <365 days of follow-up were censored in survival analyses. Logistic and Cox models were supported by Firth penalized logistic regression, calibration assessment, and internal validation. Results: Median age was 82 years, 65.0% were female, and 68.7% were ASA III–IV. Overall, 96/217 (44.2%) died during median follow-up of 570 days. Thirty-day, 90-day, and one-year evaluable mortality were 7.4%, 15.2%, and 27.3%. GNRI < 82 identified a small high-risk subgroup: 10/13 evaluable patients/episodes (76.9%) died within one year versus 43/181 (23.8%) with GNRI ≥ 82. GNRI < 82 remained associated with one-year mortality in adjusted logistic regression, although with a wide confidence interval due to sparse subgroup size (OR 6.43, 95% CI 1.50–27.55; p = 0.012), and in Firth penalized sensitivity analysis (OR 5.51, 95% CI 1.34–22.61; p = 0.018). In Cox analysis, age, ASA III–IV, available Charlson-domain comorbidity burden, and lower continuous GNRI were associated with long-term mortality, whereas adding GNRI < 82 and NLR did not materially improve cross-validated discrimination. Conclusions: GNRI < 82 identified a small subgroup with high observed mortality after PFN. Because the subgroup was sparse and biomarker addition did not materially improve internally validated discrimination, GNRI should be treated as an alerting clinical flag rather than a stand-alone basis for patient-level risk prediction. External validation is required.

IPC Classification

Keywords