Abstract
Objective
To investigate independent risk factors for 1-year mortality after acute hip-fracture surgery in patients aged ≥80 years and construct a risk-stratification nomogram to identify high-risk patients for intensified postoperative monitoring and preventive interventions.
Methods
This retrospective study enrolled patients aged ≥80 years undergoing hip-fracture surgery (2017–2024). Candidate predictors included baseline, surgical, and laboratory variables. Independent predictors were selected via univariate-multivariate logistic and LASSO regression. Nomogram performance was assessed using ROC analysis, calibration curves, and decision-curve analysis (DCA).
Results
In total, 463 patients were randomly split into 7:3 training-validation cohorts. Sex, age, comorbidity count, preoperative PNI, hemoglobin, creatinine, and postoperative NLR were independent predictors of 1-year all-cause mortality. The derived nomogram showed favorable discrimination, calibration, and clinical net benefit.
Conclusion
This internally validated nomogram integrating PNI and other risk factors predicts 1-year postoperative mortality in patients aged ≥80 years with hip fracture and may support clinical risk-stratification decisions.
Keywords: patients aged ≥80 years, hip fracture, prognostic nutritional index, risk factors, nomogram
Introduction
Against the backdrop of accelerating global population aging, hip fracture has emerged as a prominent clinical concern threatening the health of older adults and constitutes a major global public-health burden.1,2 A large-scale multinational investigation covering multiple countries and regions analysed a total of 4,115,046 hip-fracture cases derived from 20 databases, demonstrating that more than 60% of these injuries occurred among individuals aged 80 years and older.3 Hip fracture is well-recognised for its substantial disability- and mortality-related consequences;4 published epidemiological data report 1-year disability and mortality rates as high as 50% and 12–37%, respectively.5 The 1-year mortality risk among older adults sustaining hip fracture exceeds 3.5-fold that of age-matched peers without hip-fracture injury,6 lending to hip fracture the well-known moniker“the last fracture in life”. As a nutrition- and inflammation-related prognostic marker, the Prognostic Nutritional Index (PNI) has demonstrated promising clinical utility and has been widely validated as an independent predictor of poor survival among postoperative cancer patients.7,8 It simultaneously reflects host nutritional status and systemic inflammatory burden and exhibits prognostic value across diverse disease contexts. Prior studies have further confirmed that PNI acts as an predictor of adverse postoperative outcomes in patients undergoing hip-fracture surgery.9
Multiple prognostic prediction models for hip-fracture-related mortality have been reported in existing literature. Conventional risk-scoring instruments such as the Nottingham Hip Fracture Score (NHFS) are primarily developed for estimating short-term mortality risk within one month after surgery.10 Although machine-learning algorithms have been increasingly implemented for constructing clinical predictive models,11,12 their application for prognostic evaluation in older hip-fracture patients remains relatively limited. The nomogram, by contrast, delivers an intuitive and user-friendly approach for individualised postoperative mortality-risk estimation. For the high-risk subgroup of patients aged ≥80 years, there remains a practical need for prognostic tools built upon routinely-accessible demographic and laboratory indices, without relying on specialized multidomain geriatric assessment resources.
This retrospective study analysed clinical data from patients aged ≥80 years who underwent hip-fracture surgery. We screened independent risk factors for 1-year all-cause postoperative mortality and explored the prognostic performance of PNI as a surrogate marker reflecting combined nutritional-inflammatory status. Importantly, serum albumin was not incorporated into our multivariable regression model; accordingly, this study was not designed to substitute albumin with PNI. We further constructed and internally validated a nomogram-based predictive model. The present work aims to provide clinical evidence assisting clinicians in early identification of high-risk individuals, enabling timely targeted interventions and optimised perioperative risk-stratified management, so as to improve long-term patient prognosis.
Methods
Patients
This was a retrospective cohort study. Demographic data, pre-fracture general conditions, surgical data, and preoperative and postoperative laboratory indicators were extracted from the electronic medical record system of the Third Affiliated Hospital of Guangzhou Medical University for super-elderly patients (≥80 years) who underwent surgery for fresh hip fractures between September 2017 and January 2024. All procedures were performed in accordance with the tenets of the Declaration of Helsinki, and the prediction model construction and validation process fully follow the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD).
Inclusion criteria: (1) Diagnosed with hip fracture based on medical history, physical examination, and imaging findings, including femoral neck fracture, intertrochanteric fracture, and subtrochanteric fracture; (2) Super-elderly patients aged ≥80 years.
Exclusion criteria were as follows: (1) Old fracture with an interval of more than 21 days between injury and surgery (delayed surgery may alter baseline physiological status and prognostic profiles); (2) Open hip fracture, pathological hip fracture, or periprosthetic fracture following hip replacement (these fracture subtypes have distinct aetiologies and prognostic trajectories); (3) Patients sustaining multiple fractures (polytrauma confounds mortality risk assessment for isolated hip fracture); (4) Patients receiving non-operative conservative management only (the present model was developed exclusively for surgically treated patients); (5) Patients who died during hospitalisation or required repeated surgical interventions secondary to severe perioperative complications; (6) Patients with incomplete core clinical variables or confirmed loss to 1-year follow-up. Notably, these exclusion rules remove subjects at particularly high mortality risk, which may introduce selection bias; accordingly, the developed model cannot be generalised to patients meeting any of the above-listed exclusion conditions.
The selection criteria for study subjects are shown in Figure 1. Applied censoring for patients lost to follow-up;
Figure 1.
Flowchart of study subject enrollment.
Data Collection
Patient-level statistical data were extracted from electronic medical records. Collected variables comprised demographic and comorbidity information: sex, age, body mass index (BMI), interval from admission to surgery, length of hospital stay, and comorbidity count (hypertension, diabetes mellitus, coronary heart disease, and chronic obstructive pulmonary disease; because the retrospective database did not allow complete retrieval of all Charlson or CIRS-CI items, so a simple comorbidity count was used; simple count cannot reflect disease severity); surgery-related parameters: hip-fracture type, American Society of Anesthesiologists (ASA) physical status classification, anesthesia type, surgical procedure, operation duration, and intraoperative blood loss; and laboratory test results: white blood cell count (WBC), neutrophil count (NEU), neutrophil percentage, total lymphocyte count (LY), hemoglobin (HGB), platelet count (PLT), alanine aminotransferase (ALT), albumin (ALB), and creatinine (Cr). The PNI was calculated according to the following formula:
 |
All blood-laboratory assays were processed by the hospital central laboratory and all laboratory testing followed hospital-standard clinical SOP and manufacturer specifications. For patients who underwent multiple hematological examinations, only the results of the first admission and the first postoperative blood test (within 24 hours after surgery/postoperative day 1) were analyzed. Patient outcomes, including 1-year survival information after surgery, were obtained through outpatient follow-up, telephone follow-up, and the Guangdong Provincial Population Overall Management System (the follow-up workflow: survival/death status and date were first retrieved from the Guangdong Provincial Population Overall Management System; when information was incomplete, outpatient records were reviewed; telephone follow-up of family members was used only when the first two sources were insufficient. We define loss to follow-up as the inability to confirm survival status at the 1-year time point through any of the three sources.). The follow-up deadline was 1 year after surgery. Data collection was approved by the institutional review board.
Statistical Analysis
Eligible patients were randomly assigned to a training cohort (326 patients) and a validation cohort (137 patients) at a ratio of 7:3. Statistical analyses were performed using SPSS 27.0 and R software (version 4.5.2). Categorical variables were presented as frequencies (percentages), and between-group comparisons were conducted using the chi-square test or Fisher’s exact test. Ordinal categorical variables were compared using the rank-sum test. The optimal cutoff values of preoperative NLR and postoperative NLR were determined using receiver operating characteristic curves. Univariate logistic regression was first performed for all pre-specified clinically relevant variables in the training cohort. Variables with univariate P < 0.20 or those judged clinically important for 1-year postoperative mortality were retained as candidates. Lasso regression was then applied for dimensionality reduction and predictor selection, and multivariate logistic regression was fitted using the selected variables to identify independent risk factors for 1-year postoperative mortality among patients aged ≥80 years with hip fracture. A nomogram was constructed to predict the 1-year postoperative mortality risk in super-elderly patients with hip fractures. Added bootstrap internal validation with 1000 resamples (consistently labelled 30% hold-out dataset as internal split-sample validation). The discrimination and diagnostic performance of the predictive model were assessed using ROC curves and AUC values. Calibration curves and decision curve analysis (DCA) were applied to evaluate the calibration and clinical applicability of the model. A P value < 0.05 was considered statistically significant.
Results
Baseline Comparison
During the study period, a total of 536 elderly patients with hip fracture who underwent surgical treatment were enrolled. According to the exclusion criteria, 463 patients were finally included for analysis, including 326 patients in the training cohort and 137 patients in the validation cohort. As shown in Table 1, there were no statistically significant differences in demographic data, pre-fracture general status, surgical data, and preoperative and postoperative laboratory indicators between the two groups (P>0.05), suggesting that the training and validation cohorts were randomly assigned and suitable for subsequent analysis. The grouping method for continuous variables is detailed in Table 1.
Table 1.
Comparison of Baseline Data on Factors Related to Modeling and Validation Groups
| Variables | Training Group (n=326) | Validation Group (n=137) | Z/χ2/F | P value |
|---|---|---|---|---|
| Sex, n (%) | χ2=3.57 | 0.06 | ||
| Female | 251 (76.99) | 94 (68.61) | ||
| Male | 75 (23.01) | 43 (31.39) | ||
| Age, M (Q1, Q3) | 86.00 (83.00, 90.00) | 86.00 (84.00, 89.00) | Z=−0.50 | 0.62 |
| BMI, M (Q1, Q3) | 20.50 (18.42, 22.90) | 20.28 (18.37, 22.81) | Z=−0.11 | 0.92 |
| Preoperative duration, n(%) | χ2=0.03 | 0.86 | ||
| ≤48h | 57 (17.48) | 23 (16.79) | ||
| ≥48h | 269 (82.52) | 114 (83.21) | ||
| Length of hospital stay, M (Q1, Q3) | 15.00 (11.00, 18.00) | 14.00 (11.00, 18.00) | Z=−0.11 | 0.92 |
| Survival outcome, n(%) | χ2=1.53 | 0.22 | ||
| Survival | 242 (74.23) | 94 (68.61) | ||
| Death | 84 (25.77) | 43 (31.39) | ||
| Number of comorbidities, n(%) | χ2=0.25 | 0.62 | ||
| ≤2 | 228 (69.94) | 99 (72.26) | ||
| ≥3 | 98 (30.06) | 38 (27.74) | ||
| Fracture type, n(%) | F=1.28 | 0.48 | ||
| Femoral neck fracture | 132 (40.49) | 60 (43.80) | ||
| Intertrochanteric fracture | 193 (59.20) | 76 (55.47) | ||
| Subtrochanteric fracture | 1 (0.31) | 1 (0.73) | ||
| ASA grade, n(%) | χ2=5.31 | 0.07 | ||
| Grade II | 105 (32.21) | 30 (21.90) | ||
| Grade III | 212 (65.03) | 104 (75.91) | ||
| Grade IV | 9 (2.76) | 3 (2.19) | ||
| Type of anesthesia, n(%) | χ2=0.18 | 0.67 | ||
| General anesthesia | 68 (20.86) | 31 (22.63) | ||
| Spinal anesthesia | 258 (79.14) | 106 (77.37) | ||
| Surgical procedure, n(%) | χ2=0.42 | 0.52 | ||
| Internal fixation group | 182 (55.83) | 72 (52.55) | ||
| Hip replacement group | 144 (44.17) | 65 (47.45) | ||
| Operation time (min), M (Q1, Q3) | 100.00 (80.00, 120.00) | 100.00 (85.00, 120.00) | Z=−0.59 | 0.56 |
| Intraoperative blood loss (ml), M (Q1, Q3) | 80.00 (50.00, 100.00) | 80.00 (50.00, 100.00) | Z=−0.17 | 0.87 |
| Pre-PNI, Mean ± SD | 38.40 ± 6.12 | 38.25 ± 5.80 | t=0.25 | 0.81 |
| Pre-WBC (109/L), M (Q1, Q3) | 9.09 (7.17, 11.62) | 8.90 (7.52, 11.15) | Z=−0.21 | 0.83 |
| Pre-NLR, n(%) | χ2=2.77 | 0.10 | ||
| ≤6.88 | 158 (48.47) | 78 (56.93) | ||
| >6.88 | 168 (51.53) | 59 (43.07) | ||
| Pre-NUE%, M (Q1, Q3) | 80.00 (74.00, 85.07) | 78.30 (72.90, 84.00) | Z=−1.29 | 0.20 |
| Pre-HGB (g/L), Mean ± SD | 102.14 ± 17.97 | 101.30 ± 19.78 | t=0.45 | 0.66 |
| Pre-PLT (109/L), M (Q1, Q3) | 195.00 (152.00, 236.50) | 201.00 (166.00, 250.00) | Z=−0.96 | 0.34 |
| Pre-ALT (U/L), M (Q1, Q3) | 11.80 (8.50, 16.20) | 12.30 (9.60, 17.00) | Z=−1.26 | 0.21 |
| Pre-Cr (μmol/L), M (Q1, Q3) | 87.00 (66.25, 112.00) | 86.00 (64.00, 112.00) | Z=−0.67 | 0.50 |
| Post-PNI, M (Q1, Q3) | 32.40 (29.41, 34.56) | 33.00 (30.05, 35.55) | Z=−1.78 | 0.08 |
| Post-WBC (109/L), M (Q1, Q3) | 9.11 (7.53, 11.29) | 9.44 (7.64, 10.99) | Z=−0.72 | 0.47 |
| Post-NLR, n(%) | χ2=0.34 | 0.56 | ||
| ≤9.08 | 162 (49.69) | 64 (46.72) | ||
| >9.08 | 164 (50.31) | 73 (53.28) | ||
| Post-NUE%, M (Q1, Q3) | 82.20 (77.30, 86.70) | 82.60 (76.70, 86.70) | Z=−0.19 | 0.85 |
| Post-HGB (g/L), M (Q1, Q3) | 88.00 (79.25, 99.00) | 92.00 (84.00, 102.00) | Z=−2.44 | 0.01* |
| Post-PLT (109/L), M (Q1, Q3) | 197.50 (154.50, 253.75) | 210.00 (157.00, 257.00) | Z=−0.70 | 0.48 |
| Post-ALT (U/L), M (Q1, Q3) | 11.75 (8.20, 16.70) | 13.00 (9.00, 18.00) | Z=−1.52 | 0.13 |
| Post-Cr (μmol/L), M (Q1, Q3) | 83.00 (66.00, 115.75) | 84.00 (66.00, 103.00) | Z=−0.77 | 0.44 |
Abbreviations: BMI, Body Mass Index; ASA, American Society of Anesthesiologists; PNI, Prognostic Nutritional Index; WBC, White Blood Cell; NLR, neutrophils-Lymphocyte Ratio; NUE%, neutrophils percentage; HGB, Hemoglobin; PLT, Platelet; ALT, Alanine Aminotransferase; Cr, Creatinine.
Univariate and Multivariate Logistic Regression Analysis
Subsequently, a prediction model was constructed using the training cohort. As shown in Table 2, univariate analysis revealed that sex, age, BMI, number of comorbidities, ASA classification, surgical approach, preoperative PNI, preoperative NLR, preoperative NEU%, preoperative HGB, preoperative Cr, postoperative NLR, postoperative NEU%, postoperative PNI, and postoperative Cr were significantly associated with 1-year mortality after surgery in super-elderly patients with hip fracture (all P<0.05). To further exclude the influence of confounding factors, preoperative duration was included in the multivariate Logistic analysis. Collinearity diagnosis was performed first (Table 3), we now accurately report the VIF values (eg, 2.457 for preoperative NLR and 2.290 for preoperative neutrophil percentage) and other VIF values were less than 2 which states that these indicate mild correlation without severe multicollinearity. Then, preoperative duration together with other statistically significant variables (P<0.05) was included in the multivariate Logistic regression analysis to identify independent predictors of mortality (Table 2). For 1-year mortality risk in super-elderly patients with hip fracture, the independent predictors were: sex (OR=3.125, 95%CI: 1.524–6.408), age (OR=1.148, 95%CI: 1.064–1.239), number of comorbidities (OR=2.470, 95%CI: 1.233–4.948), preoperative PNI (OR=0.900, 95%CI: 0.845–0.958), preoperative HGB (OR=0.948, 95%CI: 0.928–0.970), preoperative Cr (OR=1.007, 95%CI: 1.001–1.013), and postoperative NLR (OR=2.146, 95%CI: 1.126–4.092). The formula for multivariable logistic regression is as follows:−
 |
Table 2.
Univariate and Multivariate Analyses of One-Year Mortality Risk in the Modeling Group
| Variables | Univariate Analysis | Multivariate Analysis | ||
|---|---|---|---|---|
| OR (95% CI) | P value | OR (95% CI) | P value | |
| Sex | 3.334 (1.923~5.779) | <0.001** | 3.125 (1.524~6.408) | 0.002** |
| Age | 1.127 (1.065~1.193) | <0.001** | 1.148 (1.064~1.239) | <0.001** |
| BMI | 0.909 (0.84~0.984) | 0.019* | ||
| Length of hospital stay | 1.013 (0.963~1.067) | 0.609 | ||
| Fracture type | 1.618 (0.958~2.732) | 0.072 | ||
| Number of comorbidities | 1.891 (1.123~3.184) | 0.017* | 2.470 (1.233~4.948) | 0.011* |
| ASA grade | ||||
| Grade II | 0.002** | |||
| Grade III | 2.938 (1.56~5.536) | 0.001** | ||
| Grade IV | 5.200 (1.244~21.732) | 0.024* | ||
| Preoperative duration | 1.558 (0.765~3.17) | 0.222 | ||
| Surgical procedure | 0.582 (0.348~0.975) | 0.040* | ||
| Operation time | 1.001 (0.994~1.008) | 0.830 | ||
| Type of anesthesia | 1.052 (0.569~1.947) | 0.871 | ||
| Intraoperative blood loss | 1.002 (0.999~1.005) | 0.252 | ||
| Pre-PNI | 0.837 (0.792~0.884) | <0.001** | 0.900 (0.845~0.958) | 0.001** |
| Pre-WBC | 0.987 (0.916~1.064) | 0.739 | ||
| Pre-NLR | 2.165 (1.293~3.627) | 0.003** | ||
| Pre-NUE% | 1.033 (1.001~1.066) | 0.040* | ||
| Pre-HGB | 0.938 (0.921~0.956) | <0.001** | 0.948 (0.928~0.970) | <0.001** |
| Pre-PLT | 1.001 (0.999~1.004) | 0.346 | ||
| Pre-ALT | 1.003 (0.979~1.027) | 0.834 | ||
| Pre-Cr | 1.012 (1.007~1.017) | <0.001** | 1.007 (1.001~1.013) | 0.029* |
| Post-PNI | 0.909 (0.852~0.97) | 0.004** | ||
| Post-WBC | 1.014 (0.943~1.091) | 0.710 | ||
| Post-NLR | 2.664 (1.578~4.496) | <0.001** | 2.146 (1.126~4.092) | 0.020* |
| Post-NUE% | 1.063 (1.02 ~1.107) | 0.003** | ||
| Post-HGB | 0.986 (0.969~1.003) | 0.103 | ||
| Post-PLT | 1.001 (0.998~1.004) | 0.460 | ||
| Post-ALT | 1.001 (0.976~1.026) | 0.947 | ||
| Post-Cr | 1.011 (1.006~1.016) | <0.001** | ||
Notes: *P<0.05, **P<0.001.
Abbreviations: OR, Odds Ratio; CI, Confidence Interval; BMI, Body Mass Index; ASA, American Society of Anesthesiologists; PNI, Prognostic Nutritional Index; WBC, White Blood Cell; NLR, neutrophils-Lymphocyte Ratio; NUE%, neutrophils percentage; HGB, Hemoglobin; PLT, Platelet; ALT, Alanine Aminotransferase; Cr, Creatinine.
Table 3.
Collinearity Diagnosis
| Variables | B | SE | t value | P value | Tolerance | VIF |
|---|---|---|---|---|---|---|
| Sex | 0.162 | 0.051 | 3.192 | 0.002 | 0.898 | 1.113 |
| Age | 0.014 | 0.005 | 2.775 | 0.006 | 0.852 | 1.174 |
| BMI | −0.003 | 0.006 | −0.51 | 0.611 | 0.893 | 1.119 |
| Number of comorbidities | 0.104 | 0.047 | 2.219 | 0.027 | 0.893 | 1.120 |
| ASA grade | 0.024 | 0.043 | 0.555 | 0.580 | 0.864 | 1.157 |
| Preoperative duration | 0.026 | 0.055 | 0.476 | 0.635 | 0.936 | 1.069 |
| Surgical procedure | −0.008 | 0.046 | −0.181 | 0.856 | 0.798 | 1.253 |
| Pre-PNI | −0.011 | 0.004 | −2.673 | 0.008 | 0.692 | 1.445 |
| Pre-NLR | 0.017 | 0.064 | 0.271 | 0.787 | 0.407 | 2.457 |
| Pre-NUE% | 0.002 | 0.003 | 0.663 | 0.508 | 0.437 | 2.290 |
| Pre-HGB | −0.007 | 0.001 | −4.719 | 0.001 | 0.630 | 1.587 |
| Pre-Cr | 0.001 | 0.000 | 2.664 | 0.008 | 0.825 | 1.213 |
Abbreviations: B, Regression coefficient; SE, Standard error; VIF, Variance inflation factor.
Lasso Regression Analysis
Lasso-Logistic regression analysis was applied for variable selection and model construction with respect to survival outcome in the training cohort (Figure 2), variable coding continuous standardized in LASSO. The dependent variable was binary, and a total of 28 clinical and laboratory indicators were included as independent variables. With 10-fold cross-validation, optimal regularization parameters were determined based on minimum cross-validation error (lambda.min) and the one-standard-error rule (lambda.1se), respectively. The results showed that 13 variables were enrolled into the final model under the λ.min criterion. In contrast, only 7 variables retained non-zero coefficients under the more conservative λ0.1se criterion: sex (0.47), age (0.04), number of comorbidities (0.01), preoperative PNI (−0.06), preoperative HGB (−0.03), preoperative Cr (0.01), and postoperative NLR (0.15). Notably, preoperative PNI and postoperative NLR showed strong negative and positive associations under the two criteria, respectively, suggesting that they may serve as important indicators for predicting survival outcome.
Figure 2.
Lasso-logistic regression model for mortality risk within 1 year after surgery in very elderly patients with fresh hip fracture.
Notes: (A) Lasso coefficient path plot; (B) Cross-validated binomial deviance plot.
For clinical application of the prediction model, the model corresponding to λ0.1se is more robust with fewer variables and avoids overfitting. Therefore, it was recommended as the final model. Based on λ.min corresponding to the minimum binomial deviance, λ0.1se under the 1-standard-error rule was selected, and 7 variables with non-zero coefficients identify independent risk factors for 1-year postoperative mortality in super-elderly patients with hip fractures. finally Independent risk factors were identified by two independent modelling strategies — Lasso-Logistic regression analysis and Univariate and Multivariate Logistic Regression Analysis — which yielded an identical set of predictors.
Development and Validation of a Nomogram
Based on independent factors identified by LASSO regression and multivariate logistic regression analysis, we developed a nomogram incorporating the Prognostic Nutritional Index for predicting 1-year postoperative mortality risk in super-elderly patients with hip fractures (Figure 3). The results of receiver operating characteristic (ROC) curve analysis in the training and validation cohorts are presented in (Figure 4). The area under the curve (AUC) values were 0.870 (95% CI: 0.823–0.918) and 0.840 (95% CI: 0.774–0.907) for the training and validation cohorts, respectively, indicating favorable predictive performance of the nomogram. Calibration curves showed acceptable agreement between predicted probabilities and actual observed probabilities (Figure 5). Decision curve analysis (DCA) demonstrated positive net benefit of the model, suggesting that the model established in this study has significant clinical applicability (Figure 6).
Figure 3.
Nomogram for mortality risk within 1 year after surgery in very elderly patients with fresh hip fracture; to apply the nomogram, locate and mark the patient-specific value or status on each predictor axis; determine the corresponding points on the top points scale for all predictive variables. Sum these points to obtain the total score. Mark this total score on the total points axis, then project a perpendicular line downward. The value at the intersection with the risk indicates the predicted probability of mortality risk within 1 year after surgery in very elderly patients with fresh hip fracture.
Figure 4.
ROC curve the training AUC of 0.870 and the validation of 0.840 indicate excellent performance.
Figure 5.
Calibration curves: (A) Training group, (B) Validation group. The dashed line (ideal) denotes the theoretical reference for perfect calibration, where predicted risk aligns with the observed event frequency. The solid line (Bias-corrected) represents the model’s performance adjusted via 1000 bootstrap resamples, while the dotted line (apparent) shows the original performance. The closer the solid line is to the ideal dashed line, the higher the predictive accuracy.
Figure 6.
DCA analysis: (A) Training group, (B) Validation group; the red solid line (nomogram model) represents the net benefit of using this model to guide clinical decisions. The light grey line (all) represents the strategy of treating all patients, and the black horizontal line (none) represents the strategy of treating no patients. A clinical net benefit is achieved when the model’s curve stays above both the “all” and “none” lines across a wide range of threshold probabilities.
Discussion
In accordance with geriatrics guidelines and the Chinese consensus on hip fracture research, the present study defines super-elderly patients as those aged ≥80 years. Clinical epidemiological data demonstrate that the annual incidence of hip fractures is substantially higher among patients aged ≥80 years relative to other elderly subgroups,13,14 accompanied by elevated postoperative mortality. These observations underline the clinical need for prognostic prediction studies specifically targeting this high-risk age stratum. Through a retrospective cohort analysis of 463 surgically-treated hip-fracture patients within this age group, we identified that the preoperative Prognostic Nutritional Index (PNI) may act as an independent predictor of 1-year postoperative mortality among patients aged ≥80 years with hip fracture. Unlike prior investigations focusing merely on the prognostic performance of isolated biomarkers, the present model integrated PNI with multiple routinely-collected clinical and laboratory variables. Although the Prognostic Nutritional Index (PNI), calculated from serum albumin and lymphocyte counts, is regarded as a pragmatic and easily-accessible surrogate marker of combined nutritional-inflammatory status, it has inherent limitations and cannot comprehensively evaluate nutritional status or geriatric vulnerability, as it still carries biases introduced by inflammation and capillary-leakage effects that confound the interpretation of albumin. Accumulating evidence indicates that multidimensional nutritional assessments, body-composition metrics, the HIP-G Index, and anthropometric measurements such as arm circumference can provide more comprehensive prognostic information for hip-fracture patients.10 These multidimensional variables were not available within our retrospective dataset; consequently, formal head-to-head comparisons against established risk scores such as the Nottingham Hip Fracture Score15 could not be performed. Further studies incorporating more comprehensive, multi-domain variables are therefore warranted. Nevertheless, the mortality-risk nomogram developed in this retrospective work achieved favourable discrimination and acceptable calibration based on internal split-sample and bootstrap validation. The clinical value of PNI as a surrogate marker has been preliminarily validated in the present study.
Although increasing evidence has confirmed that malnutrition is closely associated with adverse outcomes in patients with hip fracture and constitutes a key risk factor for postoperative mortality,16,17 the limitations of albumin, a widely-used conventional nutritional marker in traditional nutritional assessment, have become increasingly apparent. As a classic inflammatory-metabolic marker,18 albumin undergoes a series of abnormal changes under systemic inflammatory stress, including suppressed synthesis, increased capillary leakage, and accelerated degradation, which prevent it from accurately reflecting the patient’s true nutritional status. This inherent deficiency contributes to the unsatisfactory predictive performance reported in previous hip-fracture mortality-prediction models incorporating albumin,19 leaving substantial room for improvement. Unlike earlier studies that focused solely on single nutritional biomarkers, the present study incorporated the Prognostic Nutritional Index (PNI) into risk-stratification analyses for postoperative mortality among patients aged 80 years and older with hip fracture. Notably, serum albumin itself was not entered into our multivariable model; therefore, we cannot draw conclusions that PNI is prognostically superior to albumin. Initially developed to evaluate perioperative immune function and surgical risk in patients with gastrointestinal diseases,20,21 PNI is calculated using serum albumin concentration and peripheral-blood lymphocyte counts.22 This marker has subsequently been widely applied to assess nutritional and immune status in patients with cancer, cardiovascular disease, and other perioperative populations, offering a useful reference for prognostic prediction.23–26 Nevertheless, because PNI is mathematically derived from albumin and lymphocytes, it still carries the inflammatory-related biases that constrain the interpretation of albumin and should be viewed merely as a pragmatic surrogate marker of combined nutritional-inflammatory status, rather than a comprehensive tool for nutritional or geriatric-vulnerability assessment.
Although the molecular mechanisms underlying the association between PNI and clinical outcomes remain incompletely understood, PNI serves as a quantitative biomarker reflecting malnutrition, impaired immune function, or both. Malnutrition and compromised immune status may be major contributors to poor prognosis in patients with hip fractures.16,27 Many previous studies have consistently demonstrated a close link between nutritional status, the development of complications, and higher mortality.16,17,28 A low PNI indicates poor general condition, inadequate nutritional status, and heavy disease burden, resulting in reduced tolerance to fracture trauma and subsequent surgical intervention. Furthermore, the coexistence of malnutrition and impaired immune function may interact or exacerbate each other, further deteriorating the patient’s overall health status.29,30 This leads to increased susceptibility to adverse clinical outcomes. Nevertheless, other potential mechanisms by which PNI influences prognosis may exist beyond nutritional and immune causes, which warrant further investigation.31
In recent years, several investigators have preliminarily explored the association between PNI levels and prognosis in hip fracture patients. However, most studies have focused only on the general elderly population aged 65–79 years and failed to draw subgroup-specific conclusions for patients aged 80 years and older. In a prospective cohort study including 705 patients undergoing hip fracture surgery, Chen et al28 confirmed that higher PNI levels were positively correlated with postoperative fracture healing and functional recovery. Similarly, in a large observational study involving 3351 hip fracture patients, Wang et al9 demonstrated that patients in the middle and high PNI groups had significantly lower 2-year postoperative mortality risk compared with the low PNI group (hazard ratio [HR] = 0.66, 95% confidence interval [CI]: 0.49–0.88; HR = 0.61, 95% CI: 0.42–0.88, respectively). Notably, neither of these two studies performed subgroup analyses for patients aged 80 years and older, preventing them from capturing the unique characteristics of this vulnerable population.
Compared with the aforementioned studies, the key breakthrough of the present study lies in its unique focus on the super-elderly population aged 80 years and older, filling a critical gap in age-stratified research. Our findings are consistent with and extend previous results: elevated preoperative PNI was identified as a protective factor against 1-year postoperative mortality in super-elderly hip fracture patients aged 80 years and older (OR = 0.900, 95%CI: 0.845–0.958), indicating that each 1-unit increase in PNI was associated with approximately 10% lower odds of 1-year postoperative mortality (OR = 0.90), rather than an absolute 10% reduction in mortality risk. This finding clarifies the core value of PNI in the early warning of mortality risk among super-elderly patients. Clinically meaningful, patients aged 80 years and older exhibit significant physiological decline, a high comorbidity burden, and severely impaired nutrient absorption and metabolic function, where nutritional status exerts a decisive impact on postoperative recovery.32,33 However, no mortality prediction model incorporating PNI has been specifically developed for hip fracture patients in this age group. Accordingly, This study used logistic regression analysis together with Lasso regression to jointly identify independent risk factors and further validated the optimal model. The prediction model constructed based on the finally screened independent risk factors can effectively handle high-dimensional clinical data, balance prediction errors, accurately identify core predictive variables, and possesses the advantage of low multicollinearity. This model is not only easy to translate into clinical practice; the high weighting of PNI in the nomogram further confirms the role of this indicator in predicting mortality risk in elderly hip fracture patients aged 80 years and older.
The core motivation of this study was to address the unmet clinical need for accurate prediction of 1-year postoperative mortality in elderly hip fracture patients aged 80 years and older, a subgroup with significantly higher rates of postoperative complications and mortality than the general elderly population with hip fracture. By systematically analyzing the association between preoperative PNI and 1-year postoperative mortality, this study established a tailored prediction model for this specific population, providing a scientific basis for individualized clinical decision-making. However, the limitations of this study must be objectively acknowledged. As a single-center retrospective study, selection bias may exist. Some potential confounding variables, such as preoperative nutritional interventions, differences in postoperative rehabilitation programs and individual osteoporosis status, were not included in the analysis. In addition, the model has not undergone external validation, which limits its generalizability. Future studies should focus on addressing these limitations by conducting multicenter, large-sample prospective cohort studies to verify and optimize the predictive variables in the model, thereby further enhancing its clinical value and feasibility for widespread application.
Conclusion
To meet the demand for prognostic assessment in elderly hip fracture patients aged ≥80 years, this study demonstrated via multivariate stratified analysis that the preoperative Prognostic Nutritional Index (PNI) was associated with 1-year all-cause postoperative mortality, and this marker showed favorable predictive performance within the mortality risk prediction system. Further variable screening using multivariate logistic regression and Lasso-logistic regression suggested that male sex, advancing age, ≥3 comorbidities, low preoperative PNI, low preoperative hemoglobin (HGB), postoperative neutrophil-to-lymphocyte ratio (NLR) > 9.08, and elevated preoperative serum creatinine (Cr) were potential independent risk factors for 1-year all-cause mortality in super-elderly patients with hip fracture. These predictors reflect multiple prognostic domains encompassing physiological status, nutritional and immune status, and postoperative inflammatory response.
The nomogram prediction model constructed based on the seven identified independent risk factors was evaluated through internal split-sample validation, calibration curve analysis and decision curve analysis (DCA). This internally established model showed good discriminative ability, acceptable calibration performance, and considerable clinical net benefit. The nomogram developed in this study can achieve individualized quantitative assessment of 1-year postoperative mortality risk for hip fracture patients aged 80 years and above. On this basis, clinicians can consider strengthening perioperative nutritional support, carrying out geriatric specialist consultation, implementing close postoperative inflammatory monitoring, arranging early multidisciplinary rehabilitation, and conducting transparent prognostic communication with patients’ family members. For patients identified as low-risk, management can be implemented in accordance with standard clinical pathways. However, since this study only completed internal validation of the model, external validation based on independent cohorts is still needed before the model can be applied clinically. This model can provide objective and practical reference support for clinicians to formulate perioperative management strategies stratified by risk.
More importantly, through systematic screening of prognostic risk factors and construction and validation of the prediction model, this study initially improved the prognostic risk assessment system for hip fracture patients aged 80 years and above. It is expected to provide a replicable methodological reference for future prospective multi-center cohort studies that adopt standardized research procedures.
Funding Statement
There is no funding to report.
Data Sharing Statement
The original data in the study will be made available by the corresponding authors, further inquiries can be directed to the corresponding author.
Ethics Approval and Consent to Participate
This study received retrospective ethics approval from the Ethics Committee of The Third Affiliated Hospital of Guangzhou Medical University (approval ID: YL-20260409000379), and was conducted in accordance with the principles of the Declaration of Helsinki. The ethics committee also authorized data linkage with the Guangdong Provincial Population Overall Management System. This retrospective analysis utilized archived medical records, with patient data collected between January 1, 2017 and January 31, 2024. A total of 463 patients were enrolled. Of these, 204 patients provided broad written informed consent during index hospitalization, and 259 patients gave consent via electronic checkbox confirmation. In accordance with ethical approval, consent was waived for deceased patients and for the retrospective reuse of existing medical records in this investigation.
Author Contributions
Shikai Li, Sheng Lu, Long Zhang, Delong Yin, Zhenbo Fan obtained the data, conceived the manuscript, Delong Yin and Zhenbo Fan participated in the study design and reviewed the manuscript. Shikai Li analyzed the data and wrote the first draft of the manuscript, Weihuang Wen and Shangbin Zhang help to analyze the data. Zhenbo Fan, Sheng Lu and Long Zhang made contributions during the manuscript revision period. All the authors have read and approved the manuscript. All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.
Disclosure
The authors declare no competing interests.
References
- 1.Wei S, Xiang X, Zhou S, et al. Development and multicenter validation of a dynamic nomogram for early postoperative neurocognitive disorder in older adults with hip fracture. BMC Geriatr. 2026;26(1):428. doi: 10.1186/s12877-026-07125-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Zhao ZJ, Yang Y, Wei SR, Zhang ZQ, Yao M, Ni HD. Repetitive transcranial magnetic stimulation to prevent postoperative delirium in elderly arthroplasty patients: study protocol for a single-centre, prospective, randomized controlled trial. BMC Geriatr. 2026. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Sing CW, Lin TC, Bartholomew S, et al. Global epidemiology of hip fractures: secular trends in incidence rate, post-fracture treatment, and all-cause mortality. J Bone Miner Res. 2023;38(8):1064–14. [DOI] [PubMed] [Google Scholar]
- 4.Farahmand BY, Michaëlsson K, Ahlbom A, Ljunghall S, Baron JA. Survival after hip fracture. Osteoporos Int. 2005;16(12):1583–1590. doi: 10.1007/s00198-005-2024-z [DOI] [PubMed] [Google Scholar]
- 5.Chen Q, Hao P, Wong C, Zhong X, He Q, Chen Y. Development and validation of a novel nomogram of 1-year mortality in the elderly with hip fracture: a study of the MIMIC-III database. BMJ Open. 2023;13(5):e068465. doi: 10.1136/bmjopen-2022-068465 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Lystad RP, Cameron CM, Mitchell RJ. Mortality risk among older Australians hospitalised with hip fracture: a population-based matched cohort study. Arch Osteoporos. 2017;12(1):67. doi: 10.1007/s11657-017-0359-7 [DOI] [PubMed] [Google Scholar]
- 7.Wang X, Wang Y. The prognostic nutritional index is prognostic factor of gynecological cancer: a systematic review and meta-analysis. Int J Surg. 2019;67:79–86. doi: 10.1016/j.ijsu.2019.05.018 [DOI] [PubMed] [Google Scholar]
- 8.Tu X, Ren J, Zhao Y. Prognostic value of prognostic nutritional index in nasopharyngeal carcinoma: a meta-analysis containing 4511 patients. Oral Oncol. 2020;110:104991. doi: 10.1016/j.oraloncology.2020.104991 [DOI] [PubMed] [Google Scholar]
- 9.Wang Y, Jiang Y, Luo Y, et al. Prognostic nutritional index with postoperative complications and 2-year mortality in hip fracture patients: an observational cohort study. Int J Surg. 2023;109(11):3395–3406. doi: 10.1097/JS9.0000000000000614 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Ceolin C, Cannavò D, Sella S, et al. Hip fracture geriatric (HIP-G) index: predicting 12-month mortality in older adults with hip fracture. J Am Med Dir Assoc. 2025;26(12):105925. [DOI] [PubMed] [Google Scholar]
- 11.Liu F, Liu C, Tang X, Gong D, Zhu J, Zhang X. Predictive value of machine learning models in postoperative mortality of older adults patients with hip fracture: a systematic review and meta-analysis. Arch Gerontol Geriat. 2023;115:105120. doi: 10.1016/j.archger.2023.105120 [DOI] [PubMed] [Google Scholar]
- 12.Liu J, Ma Y, Xie W, et al. Lasso-based machine learning algorithm for predicting postoperative lung complications in elderly: a single-center retrospective study from China. Clin Interventions Aging. 2023;18:597–606. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Leung MTY, Marquina C, Turner JP, Ilomaki J, Tran T, Bell JS. Hip fracture incidence and post-fracture mortality in Victoria, Australia: a state-wide cohort study. Arch Osteoporos. 2023;18(1):56. doi: 10.1007/s11657-023-01254-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Koseki H, Sunagawa S, Noguchi C, et al. Incidence of and risk factors for hip fracture in Nagasaki, Japan from 2005 to 2014. Arch Osteoporos. 2021;16(1):111. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Sun L, Liu Z, Wu H, Liu B, Zhao B. Validation of the nottingham hip fracture score in predicting postoperative outcomes following hip fracture surgery. Orthop Surg. 2023;15(4):1096–1103. doi: 10.1111/os.13624 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Malafarina V, Reginster JY, Cabrerizo S, et al. Nutritional status and nutritional treatment are related to outcomes and mortality in older adults with hip fracture. Nutrients. 2018;10(5):555. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Foo MXE, Wong GJY, Lew CCH. A systematic review of the malnutrition prevalence in hospitalized hip fracture patients and its associated outcomes. JPEN J Parenter Enter nutr. 2021;45(6):1141–1152. doi: 10.1002/jpen.2211 [DOI] [PubMed] [Google Scholar]
- 18.Manolis AA, Manolis TA, Melita H, Mikhailidis DP, Manolis AS. Low serum albumin: a neglected predictor in patients with cardiovascular disease. Eur J Intern Med. 2022;102:24–39. doi: 10.1016/j.ejim.2022.05.004 [DOI] [PubMed] [Google Scholar]
- 19.Pan L, Ning T, Wu H, et al. Prognostic nomogram for risk of mortality after hip fracture surgery in geriatrics. Injury. 2022;53(4):1484–1489. doi: 10.1016/j.injury.2022.01.029 [DOI] [PubMed] [Google Scholar]
- 20.Zhang L, Ma W, Qiu Z, et al. Prognostic nutritional index as a prognostic biomarker for gastrointestinal cancer patients treated with immune checkpoint inhibitors. Front Immunol. 2023;14:1219929. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Li Z, Zhang D, Mo C, Zhu P, Fan X, Tang T. The prognostic significance of prognostic nutritional index in gastrointestinal stromal tumors: a systematic review and meta-analysis. Medicine. 2022;101(47):e32067. doi: 10.1097/MD.0000000000032067 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Buzby GP, Mullen JL, Matthews DC, Hobbs CL, Rosato EF. Prognostic nutritional index in gastrointestinal surgery. Am J Surg. 1980;139(1):160–167. doi: 10.1016/0002-9610(80)90246-9 [DOI] [PubMed] [Google Scholar]
- 23.Zhang X, Zhang Y, Zhang J, Yuan J, Zhu S, Li X. Immunotherapy of small cell lung cancer based on prognostic nutritional index. Front Immunol. 2025;16:1560241. doi: 10.3389/fimmu.2025.1560241 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Tobing E, Tansol C, Tania C, Sihombing AT. Prognostic nutritional index (PNI) as independent predictor of poor survival in prostate cancer: a systematic review and meta-analysis. Clin Genitourin Cancer. 2024;22(5):102142. doi: 10.1016/j.clgc.2024.102142 [DOI] [PubMed] [Google Scholar]
- 25.Xu W, Zhang L, Yang Q, et al. Associations of prognostic nutritional index with cardiovascular all-cause mortality among CVD patients with diabetes or prediabetes: evidence from the NHANES 2005-2018. Front Immunol. 2025;16:1518295. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Zhang X, Su Y. Low prognostic nutritional index predicts adverse outcomes in patients with heart failure: a systematic review and meta-analysis. Angiology. 2024;75(4):305–313. doi: 10.1177/00033197231159680 [DOI] [PubMed] [Google Scholar]
- 27.Keusch GT. The history of nutrition: malnutrition, infection and immunity. J Nutr. 2003;133(1):336s–340s. doi: 10.1093/jn/133.1.336S [DOI] [PubMed] [Google Scholar]
- 28.Chen Y, Bei M, Liu G, et al. Prognostic nutritional index (PNI) is an independent predictor for functional outcome after hip fracture in the elderly: a prospective cohort study. Arch Osteoporos. 2024;19(1):107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Li S, Zhang J, Zheng H, Wang X, Liu Z, Sun T. Prognostic role of serum albumin, total lymphocyte Count, and mini nutritional assessment on outcomes after geriatric hip fracture surgery: a meta-analysis and systematic review. J Arthroplasty. 2019;34(6):1287–1296. doi: 10.1016/j.arth.2019.02.003 [DOI] [PubMed] [Google Scholar]
- 30.Sahin S, Tasar PT, Simsek H, et al. Prevalence of anemia and malnutrition and their association in elderly nursing home residents. Aging Clin Exp Res. 2016;28(5):857–862. [DOI] [PubMed] [Google Scholar]
- 31.Inoue T, Maeda K, Nagano A, et al. Undernutrition, Sarcopenia, and frailty in fragility hip fracture: advanced strategies for improving clinical outcomes. Nutrients. 2020;12(12):3743. doi: 10.3390/nu12123743 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Rizzoli R, Biver E, Brennan-Speranza TC. Nutritional intake and bone health. Lancet Diabetes Endocrinol. 2021;9(9):606–621. doi: 10.1016/S2213-8587(21)00119-4 [DOI] [PubMed] [Google Scholar]
- 33.Lai WY, Chiu YC, Lu KC, Huang IT, Tsai PS, Huang CJ. Beneficial effects of preoperative oral nutrition supplements on postoperative outcomes in geriatric hip fracture patients: a PRISMA-compliant systematic review and meta-analysis of randomized controlled studies. Medicine. 2021;100(47):e27755. doi: 10.1097/MD.0000000000027755 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The original data in the study will be made available by the corresponding authors, further inquiries can be directed to the corresponding author.