# Integrated nutritional–inflammatory and frailty-based model for mortality risk stratification following hip fracture surgery: a multicentre cohort study

## TL;DR

Combining frailty and nutritional-inflammation measures improves predicting death risk after hip fracture surgery in older patients.

## Contribution

A new model integrating frailty and nutritional-inflammation indices (CALLY and GINI) improves mortality risk prediction after hip fracture surgery.

## Key findings

- Adding nutritional-inflammation indices improved 1-year mortality risk stratification (AUC increase from 0.666 to 0.673).
- Patients with low CALLY or high GINI had ~40% 1-year mortality compared to ~18% in others.
- Model 3 achieved 51.7% sensitivity and 75.0% specificity for 1-year mortality prediction at threshold 0.34.

## Abstract

Mortality after hip-fracture surgery remains high, and current prognostic tools often assess frailty or nutritional–inflammatory status separately. We hypothesised that integrating frailty with composite immune–nutritional indices, particularly the C-reactive protein–albumin–lymphocyte (CALLY) index and the Global Immuno-Nutrition Inflammation (GINI) index, would enhance mortality risk stratification in older hip-fracture patients.

This multicentre retrospective cohort study included 517 patients aged ≥ 65 years who underwent surgical treatment for hip fracture between 2018 and 2024. Baseline data included demographics, established frailty measures, fracture characteristics, and surgical delay. Nutritional–inflammatory status was assessed using routine laboratory-based indices, including albumin, C-reactive protein, Prognostic Nutritional Index, Geriatric Nutritional Risk Index, CALLY, and GINI. Multivariable logistic regression models were developed to predict in-hospital, 30-day, 90-day, and 1-year mortality. Model discrimination, calibration, and reclassification performance were evaluated using AUC, information criteria, calibration metrics (Hosmer-Lemeshow test, Brier score), and continuous net reclassification improvement. Multicollinearity was rigorously assessed using variance inflation factors (VIF). Sensitivity and specificity were evaluated across multiple probability thresholds (0.20–0.50) to optimize clinical utility.

One-year mortality was 28.8%. Frailty measures remained independent predictors of mortality, while the addition of nutritional–inflammatory indices significantly improved 1-year risk stratification (AUC increase from 0.666 to 0.673; NRI + 14.9%). Incorporation of CALLY and GINI provided further, albeit modest, improvement (AUC 0.678; NRI + 11.7%). All models demonstrated excellent calibration (Hosmer-Lemeshow p > 0.69, Brier scores 0.12–0.19). Patients in the lowest CALLY or highest GINI quartiles exhibited substantially higher 1-year mortality (~ 40%) compared with those in the highest CALLY or lowest GINI quartiles (~ 18%; p < 0.001). However, severe multicollinearity among frailty scores (VIF 12–34) and nutritional indices (VIF 139–277) resulted in paradoxical coefficient estimates. At the optimal probability threshold (0.34), Model 3 achieved sensitivity of 51.7% and specificity of 75.0% for 1-year mortality prediction.

An integrated nutritional–inflammatory and frailty-based approach, incorporating the CALLY and GINI indices derived from routine laboratory parameters, offers clinically meaningful improvement in mortality risk stratification after hip-fracture surgery. While multicollinearity limits coefficient interpretability, excellent calibration and improved risk reclassification support clinical utility. Threshold optimization (0.30–0.35) is recommended to maximize sensitivity for screening purposes. Prospective validation and nutrition-targeted interventional studies are warranted.