1 Introduction

The global elderly population is expanding in parallel with increasing life expectancy, resulting in a concomitant rise in the incidence of hip fractures [1]. Currently, approximately one million hip fractures occur annually worldwide, and this number is projected to increase nearly fivefold by 2050 [1–3]. Hip fractures represent a major cause of morbidity and mortality among older adults [4]. Despite continuous advancements in surgical techniques, perioperative management, and medical care, the 1-year postoperative mortality rate in this population remains unacceptably high, ranging from 20% to 40% [5]. Accordingly, timely identification of risk factors for adverse outcomes is crucial for improving survival and preserving quality of life.

Malnutrition has emerged as a particularly important determinant of poor outcomes in elderly patients with hip fractures [3,6]. It is associated with prolonged hospital stays, higher complication rates, and increased mortality [7]. However, there is no consensus on the most accurate and practical method to assess nutritional status in this population, and existing evaluation tools differ in their predictive validity.

The Geriatric Nutritional Risk Index (GNRI), first proposed by Bouillanne et al. [8], is an objective and simple prognostic tool designed to assess the risk of morbidity and mortality in elderly hospitalized patients. GNRI is calculated using serum albumin levels—a well-established marker of nutritional and inflammatory status—and the ratio of present body weight to ideal body weight, thereby integrating both biochemical and anthropometric parameters [8]. Several recent studies have confirmed GNRI as an independent and robust predictor of both short- and long-term mortality in older patients undergoing hip fracture surgery, underscoring its utility for perioperative risk stratification [9–11].

Nevertheless, evidence is still limited regarding whether GNRI maintains its prognostic value when evaluated alongside other well-established risk factors such as age, sex, fracture type, comorbidity burden, and ASA classification. Therefore, the present study aimed to investigate whether GNRI serves as an independent predictor of 1-year mortality in elderly patients undergoing hip fracture surgery. We hypothesized that lower GNRI scores would remain independently associated with an increased risk of 1-year postoperative mortality, even after adjustment for established demographic and clinical risk factors.

2 Materials and methods

3 Results

3.1 Baseline characteristics of the study population

A total of 786 patients with hip fractures were screened for eligibility. After applying the exclusion criteria, 624 patients were included in the final analysis (Fig 1).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Flow diagram of patient selection and exclusion process for the study.

https://doi.org/10.1371/journal.pone.0348117.g001

The baseline demographic and clinical characteristics of the study population are presented in Table 1. The mean age was 78.8 ± 7.1 years, and 65.4% of patients were female. When expressed as within-group mortality rates, mortality was highest among patients aged >85 years (63/121, 52.1%), those with ASA III–IV (112/141, 79.4%), those with ≥3 comorbidities (104/173, 60.1%), and those with severe or moderate malnutrition risk (GNRI ≤ 91.9; 122/183, 66.7%). In contrast, mortality was 5.6% in patients with ASA I–II (27/483) and 0.8% in those with normal nutritional status (2/240). Overall, the 1-year mortality rate was 22.3%, and 43.9% of deaths occurred within the first 3 postoperative months. Also, there was no significant association between surgical procedure type and mortality (p = 0.764). The mean serum albumin level was 3.47 ± 0.43 g/dL, and the mean GNRI score was 93.5 ± 6.35. Both lower albumin levels and lower GNRI categories were significantly associated with higher 1-year mortality (p < 0.001); specifically, 66.7% of patients at severe or moderate malnutrition risk (GNRI ≤ 91.9) died within 1 year.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Baseline demographic, clinical, and nutritional characteristics of the study population (n = 624).

https://doi.org/10.1371/journal.pone.0348117.t001

The distribution of pre-existing comorbidities in the study population is summarized in Table 2. The most common conditions were hypertension (46.2%) and diabetes mellitus (28.7%), followed by dementia (15.9%) and anemia (11.4%). Several comorbidities—including hypertension, coronary artery disease, cerebrovascular disease, chronic obstructive pulmonary disease, chronic renal failure, anemia, dementia, and malignancy—were significantly associated with increased 1-year mortality (all p < 0.05). In contrast, diabetes mellitus (p = 0.811) and thyroid disease (p = 0.623) showed no significant relationship with mortality.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Comparison of baseline characteristics and pre-existing comorbidities between survivors and non-survivors within 1 year after hip fracture surgery.

https://doi.org/10.1371/journal.pone.0348117.t002

A comparison of baseline characteristics between survivors and non-survivors is presented in Table 2. Patients in the mortality group were significantly older (83.9 ± 6.69 vs. 77.3 ± 6.57 years; p < 0.001), had longer fracture-to-surgery intervals (2.55 ± 1.40 vs. 2.13 ± 1.02 days; p < 0.001), longer surgery-to-discharge durations (7.11 ± 3.13 vs. 5.76 ± 2.49 days; p < 0.001), and longer total hospital stays (9.67 ± 3.55 vs. 7.91 ± 2.89 days; p < 0.001). The mean serum albumin level was markedly lower in the mortality group (2.91 ± 0.33 vs. 3.64 ± 0.30 g/dL; p < 0.001), and the mean GNRI score was also significantly lower (85.05 ± 4.88 vs. 95.89 ± 4.50; p < 0.001). Non-survivors were more likely to be > 85 years of age (45.3% vs. 12.0%; p < 0.001), to have intertrochanteric fractures (53.9% vs. 41.6%; p = 0.023), and to present with ASA class III–IV (81.8% vs. 6.0%; p < 0.001). The proportion of patients with ≥3 comorbidities (74.8% vs. 14.2%; p < 0.001) and those at severe or moderate malnutrition risk (87.7% vs. 12.6%; p < 0.001) was also markedly higher among non-survivors. No significant differences were observed with respect to sex, anesthesia type, or surgical procedure.

3.2 Univariate logistic regression analysis

In univariable logistic regression analyses, older age, longer fracture-to-surgery interval, longer postoperative hospital stay, lower albumin, lower GNRI, intertrochanteric fracture type, higher ASA status, and greater comorbidity burden were associated with higher 1-year mortality (Table 3). Sex, anesthesia type, and surgical procedure were not significantly associated with mortality. Very high odds ratios observed in the lowest GNRI categories should be interpreted with caution, as the reference group (GNRI ≥98) included only two deaths (Table 3).

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Univariable and multivariable logistic regression analyses for predictors of 1-year mortality after hip fracture.

https://doi.org/10.1371/journal.pone.0348117.t003

3.4 ROC curve analysis and GNRI cut-off value

ROC analysis showed that the GNRI had excellent discriminatory power for predicting 1-year mortality after hip fracture, with an AUC of 0.928 (95% CI: 0.904–0.951; p < 0.001). The optimal cut-off value, determined using the Youden Index, was 91.45, providing a sensitivity of 87.8% and a specificity of 90.3% (Youden Index = 0.781) (Fig 2).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. ROC curve showing the predictive performance of GNRI for 1-year mortality after hip fracture surgery.

https://doi.org/10.1371/journal.pone.0348117.g002

Patients with a GNRI ≤91.45 experienced nearly a 20-fold higher risk of 1-year mortality compared with those with a GNRI >91.45 (72.2% vs. 3.7%; RR: 19.32, 95% CI: 12.01–31.09; p < 0.001).

4 Discussion

In this retrospective, two-center cohort study of elderly patients undergoing surgery for acute hip fractures, three independent predictors of 1-year mortality were identified: age, ASA classification, and GNRI. Multivariate logistic regression demonstrated that each additional year of age increased mortality odds by 7%, while ASA class III–IV patients had more than a ten-fold higher odds compared with ASA class I–II. Moreover, each 1-point increase in GNRI was associated with a 15% reduction in mortality odds. Consistently, using a clinically relevant threshold, patients with GNRI ≤ 91.45 had nearly a 19-fold higher 1-year mortality risk compared with those above this cut-off. Although serum albumin also emerged as a strong predictor in the univariate analysis, it was excluded from the multivariate model due to collinearity with GNRI, which inherently incorporates albumin as a core component and thus provides a more comprehensive prognostic measure. These findings emphasize that, beyond chronological age and physiological reserve, preoperative nutritional status—as captured by GNRI—provides important prognostic information, supporting its potential as a practical and objective tool for perioperative risk stratification.

The GNRI, originally proposed by Bouillanne et al. [8], is an objective index incorporating serum albumin and body weight ratio to predict morbidity and mortality in elderly populations. In their cohort, lower GNRI values were strongly associated with adverse outcomes, with GNRI ≤ 82 linked to up to 29-fold increased mortality odds [8]. In line with this, our analysis showed that patients with GNRI < 82 had a markedly elevated 1-year mortality odds (up to 243-fold) compared with those with GNRI ≥98. This striking difference may be explained by the older mean age of our cohort, the higher prevalence of ASA class III–IV, greater multimorbidity, frequent cardiovascular and cerebrovascular comorbidities, and a considerable early postoperative mortality rate (43.9% of deaths occurred within 3 months). It should be noted, however, that the extremely high odds ratios observed for the lowest GNRI categories (e.g., < 82) likely reflect the very small number of deaths in the reference group (GNRI ≥98, n = 2), which may have inflated relative odds estimates and widened confidence intervals. Therefore, while the direction of the association is robust and consistent with prior studies, the absolute magnitude of these odds ratios should be interpreted with caution. Collectively, these findings reinforce GNRI as a valuable prognostic marker in elderly hip fracture patients.

Previous studies have also reported the prognostic utility of GNRI in predicting short-term mortality after hip fracture surgery [12,13]. For example, in a retrospective cohort of 1040 patients, GNRI was identified as a significant predictor of 30-day mortality, with ROC analysis yielding an optimal cut-off of 75.4; patients below this threshold had up to 27-fold increased mortality odds. This underscores the usefulness of GNRI in the immediate postoperative period. Extending beyond these findings, our study demonstrated that GNRI independently predicts not only early but also longer-term mortality, with an optimal cut-off of 91.45 below which patients had a nearly 19-fold increased risk. Moreover, a comparative study assessing several nutritional assessment tools—including Graz Malnutrition Screening, Prognostic Nutritional Index, GNRI, and Controlling Nutritional Status—found only limited predictive value across all scores (AUC = 0.64–0.68), with no clear superiority [13]. By contrast, our ROC analysis showed that GNRI alone provided excellent discriminatory ability for predicting 1-year mortality (AUC = 0.928), underscoring its robustness as a prognostic marker.

Yerli et al. further evaluated GNRI in relation to 90-day outcomes in a large cohort of 1345 elderly hip fracture patients [14]. They reported a 90-day mortality rate of 10.6%, with GNRI significantly associated with short-term mortality (p < 0.001) but not with early postoperative complications. However, ROC analysis to define an optimal cut-off was not performed. While their findings highlight the role of GNRI in predicting early mortality, our study expands on this by demonstrating its strong association for 1-year mortality. Specifically, we observed a higher mortality rate of 22.3% and identified an optimal GNRI threshold of 91.45, with excellent discriminatory performance (AUC = 0.928).

Similarly, Kotera et al. analyzed 607 patients (mean age 87 ± 6 years) and reported a 180-day mortality rate of 5.4% [11]. Non-survivors had significantly lower mean GNRI values (83 ± 9) than survivors (92 ± 9), with ROC analysis showing moderate discriminatory power (AUC = 0.74). In our study, the mean GNRI was 85.05 ± 4.88 in the death group and 95.89 ± 4.50 in the survivor group, with a substantially higher 1-year mortality rate of 22.3%. Importantly, our higher optimal cut-off (91.45) demonstrated excellent predictive accuracy (AUC = 0.928). These findings indicate that GNRI is not only a predictor of short- to mid-term outcomes, as suggested by Kotera et al., but also a robust marker of long-term mortality. Moreover, Yokoyama et al. [15] reported a 6-month mortality rate of 13.3% in patients with GNRI < 92 undergoing bipolar hemiarthroplasty, also noting higher complication, transfusion, and impaired gait rates. Compared with their findings, our study demonstrated a higher mortality rate, likely reflecting the longer follow-up, larger sample, and more comorbid cohort.

Furthermore, Fujimoto et al. [10], Wu et al. [9], and Liu et al. [16] consistently emphasized the prognostic significance of GNRI in hip fracture patients. Fujimoto et al. reported that each 1-point increase in GNRI reduced 1-year mortality odds by 20% [10]; Wu et al. found that low GNRI nearly doubled 1-year mortality odds and was associated with higher early postoperative complications [9]; and Liu et al., in a meta-analysis of 3959 patients, confirmed significant associations between lower GNRI, increased mortality, and higher complication rates [16]. In our two-center cohort of 624 patients, a consistent inverse association was observed, with each 1-point increase in GNRI reducing 1-year mortality odds by 15% (OR: 0.85; 95% CI: 0.81–0.91; p < 0.001). The optimal GNRI cut-off of 91.45 also effectively stratified mortality risk. Collectively, evidence from single-center, multi-institutional, and pooled analyses converges to support GNRI as a strong and independent predictor of 1-year mortality after hip fracture, reinforcing its potential role in preoperative risk stratification and in guiding targeted nutritional and multidisciplinary interventions.

In contrast, Choi et al. [17] conducted a study involving 548 patients aged ≥80 years who underwent surgery for proximal femur fractures, focusing exclusively on an “extremely elderly” population and evaluating major morbidities (including in-hospital mortality) as the primary endpoint. They reported that GNRI was not a significant predictor of morbidity (AUC = 0.561, p = 0.083), whereas serum albumin level (cut-off: 3.55 g/dL) and the Charlson Comorbidity Index (CCI) emerged as independent predictors. The overall major morbidity rate was 14.4%, and the in-hospital mortality rate was 4% [17]. By contrast, our study encompassed a broader age range (65–102 years), included patients from two centers, and focused on long-term outcomes with 1-year mortality as the primary endpoint. We observed a 1-year mortality rate of 22.3%, with 43.9% of deaths occurring within the first 3 months after surgery. GNRI was confirmed as an independent prognostic factor, and patients with GNRI ≤91.45 exhibited a 19-fold higher risk of death. Moreover, ROC analysis demonstrated excellent discriminative ability of GNRI for predicting 1-year mortality (AUC = 0.928). Taken together, while the findings of Choi et al. suggest limitations of GNRI in predicting short-term morbidity in extremely elderly patients, our results highlight its strong prognostic value for long-term mortality across a broader elderly population. This discrepancy is likely attributable to differences in follow-up duration (in-hospital vs. 1-year), the predominance of early mortality, variations in patient age distribution, and the choice of outcome measures (morbidity vs. mortality).

Additionally, Zhou et al. [18] proposed a novel combined index integrating GNRI with the Systemic Immune-inflammatory Index (SII) in a cohort of 597 elderly hip fracture patients, reporting an overall 1-year mortality of 15.1%. Patients with high SII–GNRI scores exhibited a substantially higher mortality rate (28.8%) compared with those with low scores (7.6%). This study highlights the prognostic significance of simultaneously considering malnutrition and systemic inflammation, suggesting that the integration of GNRI and SII may offer a more comprehensive and accurate tool for risk stratification. Taken together, the available evidence indicates that while GNRI alone remains a robust predictor of long-term mortality, its predictive value may be further enhanced when combined with markers of systemic inflammation.

Importantly, while our findings highlight the prognostic value of GNRI in predicting post-fracture mortality, a recent large-scale prospective study by Pan et al. [19] further demonstrated its association with fragility fracture risk in elderly patients with type 2 diabetes mellitus. In this 9-year ambispective cohort, lower GNRI levels were significantly linked to an increased incidence of fragility fractures, as well as to higher 10-year probabilities of major osteoporotic and hip fractures assessed by the Fracture Risk Assessment Tool (FRAX) [19]. Similarly, Wang et al. [20], analyzing data from the U.S. National Health and Nutrition Examination Survey (NHANES), reported that higher GNRI values were positively associated with femoral bone mineral density and negatively associated with the risk of osteoporosis in postmenopausal women. Consistently, Ji et al. [21] found that lower GNRI was independently associated with osteoporosis among elderly patients with type 2 diabetes in Northern China.

Moreover, Yoshida et al. [22] demonstrated that a low GNRI was an independent risk factor for bone fractures in patients undergoing hemodialysis, with fracture incidence being significantly higher in the low GNRI group regardless of sex. These results suggest that GNRI may serve not only as a prognostic marker after fracture but also as a predictor of bone health and fracture occurrence, underscoring its role as a comprehensive biomarker. In line with these findings, Kamioka et al. [23] showed that lower GNRI was significantly associated with osteoporosis and higher FRAX-derived fracture risk in patients with chronic liver disease, further reinforcing its value as a simple and reliable indicator of bone health.

Taken together with our findings, this growing body of evidence supports the potential of GNRI to contribute both to preventive strategies and to post-fracture mortality risk stratification in elderly populations. Nonetheless, further studies in larger and more diverse cohorts are warranted to confirm its generalizability.

This study has several strengths, including its relatively large sample size, two-center design, and comprehensive statistical analyses incorporating both multivariate regression and ROC curve evaluation, which together provide robust evidence for the prognostic utility of GNRI in elderly hip fracture patients. Nevertheless, certain limitations should be acknowledged. The retrospective design may have introduced selection bias, and the study population was regionally restricted, which may limit the generalizability of our findings. Measurement bias is also possible because some variables were obtained from routine medical records. In addition, potentially important prognostic factors, including sarcopenia, pre-fracture mobility, and cognitive status, were not uniformly available and could not be analyzed. Formal survival analysis was not performed because exact event dates and censoring times were not uniformly available for the full cohort; therefore, mortality timing was summarized descriptively. Moreover, other nutritional indices and postoperative complications were not systematically evaluated, precluding direct comparison of GNRI with alternative prognostic tools, and important confounding factors such as frailty, pre-fracture functional status, and inflammatory markers were not included in the analysis. The GNRI cut-off was derived from the same dataset and was not internally and externally validated therefore, the threshold should be interpreted as exploratory and cohort-specific.

Future large-scale, prospective, multi-institutional studies are warranted to validate our results, to assess the additive value of GNRI when combined with functional and inflammatory parameters, and to explore whether GNRI-guided nutritional and multidisciplinary interventions can improve survival outcomes in this vulnerable population.

5 Conclusions

This two-center retrospective cohort study demonstrated that GNRI is an independent and robust predictor of 1-year mortality in elderly patients undergoing surgery for acute hip fractures. A GNRI ≤91.45 was associated with nearly a 19-fold higher risk of death, while higher scores were protective. These findings highlight the importance of incorporating GNRI into routine preoperative assessment, as patients with low GNRI may benefit from closer monitoring and targeted multidisciplinary interventions. Future prospective studies are warranted to validate these results and to determine whether GNRI-guided management can improve survival outcomes in this vulnerable population.