Abstract
<title>Abstract</title> <p> <bold>Background</bold> The stress hyperglycemia ratio (SHR) has been linked to adverse cardiovascular outcomes; however, its prognostic significance in elderly patients with hip fracture and perioperative atrial fibrillation (AF) remains poorly defined. This study therefore aimed to investigate the nonlinear association between SHR and all‑cause mortality in this specific population and to evaluate the incremental predictive utility of SHR. <bold>Methods</bold> This retrospective cohort study consecutively enrolled elderly patients with hip fracture and perioperative AF who were admitted to the Third Hospital of Hebei Medical University between January 2019 and December 2022. Patients were followed from discharge until all‑cause death or administrative censoring at 36 months. Restricted cubic spline (RCS) curves were initially constructed to visualize the dose–response relationship between SHR and mortality. To adjust for potential confounders, generalized propensity score weighting (GPSW) was applied, and weighted RCS curves were replotted to confirm robustness. The spline basis functions were then entered as a group into a Group Lasso‑Cox regression for variable selection, followed by refitting a simplified Cox model based on the selected variables. Internal validation was performed via bootstrap resampling. To assess whether the nonlinear SHR effect provided incremental predictive value, a comparison model excluding the spline variables was established. Model discrimination was evaluated using the C‑index. Time‑dependent receiver operating characteristic (td‑ROC) analysis and time‑dependent decision curve analysis (td‑DCA) were performed at 12 and 24 months to evaluate predictive accuracy and net clinical benefit. <bold>Results</bold> A total of 340 elderly patients with hip fracture and perioperative AF were enrolled. RCS analysis revealed a J‑shaped nonlinear association between SHR and all‑cause mortality, with cutoff values of 0.74 and 1.21 (HR = 1) in the unweighted model, and 0.71 and 1.19 after GPSW. Group Lasso‑Cox variable selection identified nine predictor groups, including the SHR spline basis group, age, coronary artery disease (CAD), chronic heart failure (CHF), acute myocardial infarction (AMI), ischemic stroke, brain natriuretic peptide (BNP), left ventricular end‑diastolic diameter (LVEDD), and left ventricular ejection fraction (LVEF). The full model incorporating the SHR spline basis group achieved a C‑index of 0.7688, with time‑dependent AUCs of 0.720 at 12 months and 0.719 at 24 months. After excluding the SHR spline basis group, the reduced model yielded a C‑index of 0.7229, with time‑dependent AUCs of 0.785 at 12 months and 0.688 at 24 months. Decision curve analysis further confirmed the superior net clinical benefit of the SHR‑inclusive model at both time points. <bold>Conclusion</bold> SHR exhibited a J‑shaped relationship with long‑term all‑cause mortality in elderly hip fracture patients with perioperative AF. As an independent predictor, SHR provided good predictive efficacy and incremental prognostic value in this population. </p>