Prognostic Value of HALP and PNI Scores in Predicting 6-Month Mortality Among Geriatric Hip Fracture Patients

Introduction

Hip fractures in older adults remain a significant clinical and public health concern, as they are closely linked to high morbidity and mortality rates. Reported 1-year mortality rates range from 20% to over 30%, and recent studies continue to demonstrate high mortality among elderly patients with hip fractures, underscoring the ongoing clinical burden of these injuries [1–4]. Advanced age, multimorbidity, frailty, and diminished physiological reserve further increase susceptibility to postoperative complications and adverse outcomes [1,5]. As the global elderly population continues to grow, the demand for simple and dependable tools that support early risk assessment and guide perioperative decisions in this vulnerable group has become increasingly important.

Several prognostic models have been proposed for predicting mortality and complications in geriatric hip fracture surgery; however, many are limited by complexity, inconsistent performance, or poor applicability in emergency settings. Immunonutritional indices, including the Glasgow Prognostic Score (GPS), Geriatric Nutritional Risk Index (GNRI), Prognostic Nutritional Index (PNI), HALP (hemoglobin, albumin, lymphocyte, platelet) score, and Controlling Nutritional Status (CONUT) score, have recently emerged as simpler alternatives because they rely on routinely available laboratory variables and provide insight into both nutritional status and systemic inflammation [6–10].

Among these, HALP and PNI have gained particular attention for their ability to integrate immune, inflammatory, and nutritional components into a single composite measure [8–13]. Hemoglobin and albumin reflect anemia and protein reserves, lymphocytes indicate immune competence, and platelets are closely linked to inflammation and thrombotic activity – key determinants of outcomes in frail geriatric populations [14–16]. Importantly, although these indices were initially developed in oncologic settings, their components represent fundamental physiological markers relevant across acute and chronic diseases.

Elderly hip fracture patients frequently present with chronic anemia, hypoalbuminemia, lymphopenia, sarcopenia, and platelet activation – conditions closely aligned with the biological pathways captured by HALP and PNI [1,16,17]. Despite this theoretical relevance, evidence evaluating these scores in geriatric orthopedic trauma remains scarce, and comparative data on their prognostic performance are lacking.

The comparative evidence evaluating HALP and PNI scores specifically in geriatric patients undergoing hip fracture surgery is limited. From a clinical perspective, geriatric hip fracture patients are typically admitted through emergency pathways, where rapid and pragmatic risk stratification is essential. Complex prognostic scores can be difficult to apply in this setting, whereas immunonutritional indices derived from routine admission laboratory tests could provide immediate and actionable information. Early identification of high-risk patients may enable clinicians to intensify perioperative monitoring, initiate nutritional or geriatric co-management strategies, and anticipate postoperative complications. Accordingly, this study aimed to investigate the association between preoperative HALP and PNI scores and 6-month all-cause mortality and postoperative complications in elderly patients undergoing hip fracture surgery, and to determine which index provides the best prognostic accuracy.

Material and Methods

PATIENT SELECTION:

This retrospective study was conducted in accordance with the Declaration of Helsinki and was approved by the Local Ethics Committee of Istanbul Training and Research Hospital (approval date: 13.12.2024, approval no: 138). IRB approval was obtained before data extraction, and all data were retrieved from the hospital electronic medical record (EMR) system only after approval was granted. Because the study involved retrospective review of existing anonymized medical records with no direct patient contact or intervention, the requirement for informed consent was formally waived by the Ethics Committee. All geriatric patients (aged 65 years or older) who were diagnosed with proximal femur fractures (femoral neck, peritrochanteric, or subtrochanteric) and underwent surgical treatment at our institution between January 2021 and July 2024 were evaluated for inclusion. Surgical interventions were performed under either spinal or general anesthesia, with the choice dictated by the patient’s comorbidities and prior anticoagulant use, aiming to minimize physiological stress. All eligible patients meeting the inclusion criteria during the study period were consecutively enrolled. No additional exclusions were applied beyond the predefined exclusion criteria described below.

Patients were excluded if: (1) any admission laboratory parameter required for calculating the HALP or PNI scores was missing; (2) the fracture was an isolated greater trochanteric fracture not requiring surgery; (3) the case involved a periprosthetic fracture or a revision procedure for a previously operated hip fracture; (4) the patient had severe peripheral vascular disease, connective tissue disorders, or vasculitis, given their potential to independently influence systemic inflammation and nutritional markers; or (5) the patient had severe obesity (BMI ≥ 40 kg/m²) or advanced venous insufficiency, both of which are known to affect inflammatory status and postoperative outcomes.

A total of 587 records were reviewed, and 549 patients who met the inclusion criteria were enrolled. The 6-month follow-up period was selected as the minimum interval required to assess initial recovery and its potential impact on mortality, in alignment with our local registry protocol. Patients’ survival status at 6 months after surgery was obtained through hospital records and follow-up phone calls to patients or their caregivers. Follow-up completeness was also assessed. In the final study cohort, all patients were successfully contacted, and no cases were lost to follow-up during the 6-month follow-up period. Survival status was verified either through electronic hospital records or direct telephone communication. Patients were subsequently divided into 2 groups: a survivor group, consisting of those alive at 6 months postoperatively, and a non-survivor group, comprising those who died within 6 months after surgery. The patient inclusion and exclusion process is summarized in Figure 1.

DATA COLLECTION:

This retrospective study was conducted after obtaining approval from the Local Ethics Committee (13.12.2024, No: 138). All data were obtained from the EMR system of Istanbul Training and Research Hospital (Hospital Information System, HIS). The dataset included all consecutive patients aged ≥65 years who underwent surgical repair of proximal femoral fractures between January 1, 2021, and July 31, 2024. A total of 587 patient records were screened, and 549 patients met the eligibility criteria and were included in the final analysis.

Data extraction was performed manually by 2 independent investigators using a standardized chart review protocol to ensure accuracy. Missing values were handled by complete-case analysis; patients with missing data required for HALP or PNI calculation were excluded prior to analysis. Follow-up data were obtained from the hospital EMR and supplemented by telephone interviews with patients or caregivers. No patients were lost to follow-up at 6 months. Survival status was verified through electronic hospital records or direct telephone communication.

CLINICAL DEFINITIONS:

Fracture type was determined from radiology reports and operative notes and classified as intracapsular (femoral neck) or extracapsular (intertrochanteric or subtrochanteric). Displacement was recorded as displaced or non-displaced. Mechanism of injury was categorized as low-energy (eg, fall from standing height) or high-energy trauma. Comorbidities were defined according to documented clinical diagnoses. CKD was defined as eGFR <60 mL/min/1.73 m² for >3 months, and polypharmacy as ≥5 regular medications.

Postoperative outcomes and complications were assessed independently by 2 investigators (GC and EE) through review of electronic medical records, including pneumonia, urinary tract infection, surgical site infection, thromboembolism, reoperation, and ICU admission. Survival status at 6 months was confirmed through hospital records or telephone follow-up.

LABORATORY TESTS:

Fasting whole-blood samples were collected within 24 hours of hospital admission and transported to the central laboratory under controlled cold-chain conditions. Biochemical parameters were analyzed in the routine biochemistry laboratory of our hospital in the preoperative period using standard protocols. Preoperative evaluations included complete blood counts (hemoglobin, white blood cell count, platelet count, and neutrophil and lymphocyte counts) and biochemical parameters (albumin, creatinine, and blood glucose levels). All analyses were performed using automated hematology and biochemistry analyzers.

DEFINITIONS:

PNI, used to assess immunonutritional status (particularly in patients with malignancies), was calculated as: PNI=(10× serum albumin [g/dL]) + (0.005× total lymphocyte count [per mm3]) [18].

The HALP score, an immunonutritional biomarker introduced by Chen et al, was originally used to predict prognosis in patients with malignancies. It is determined using the formula:

STATISTICAL ANALYSIS:

All statistical analyses were performed using IBM SPSS Statistics for Windows, version 26.0 (IBM Corp., Armonk, NY, USA). Figure generation, including Kaplan-Meier survival plots, was performed using SPSS and GraphPad Prism version 9.0 (GraphPad Software, San Diego, CA, USA). Normality of variables was assessed using the Kolmogorov-Smirnov test, histograms, and probability plots. Continuous variables are reported as mean±standard deviation or median (interquartile range) depending on their distribution, while categorical variables are expressed as frequencies and percentages. Between-group comparisons for continuous variables were performed using the t test or Mann-Whitney U test, and categorical variables were compared using the chi-square or Fisher’s exact test.

Survival outcomes were analyzed using Kaplan-Meier estimates, and differences between survival curves were assessed with the log-rank test. The survival endpoint was defined as time-to-death from any cause. Each patient contributed an individual follow-up time starting from the index date (eg, diagnosis, admission, or intervention) and was followed until death or the end of the follow-up period. Death from any cause was recorded as the event and calculated as the exact number of days from baseline. Cox proportional hazards regression models were used to identify independent predictors of 6-month all-cause mortality. Baseline predictors considered in the initial univariate models included demographic variables (age, sex, BMI), comorbidities (hypertension, diabetes mellitus, coronary artery disease, congestive heart failure, cerebrovascular disease, chronic kidney disease, chronic obstructive pulmonary disease, history of cancer, hypothyroidism, dementia), perioperative factors (polypharmacy, postoperative complications, ICU stay, total hospital stay), and laboratory-based indices (HALP and PNI scores). The proportional hazards assumption for all Cox regression models was evaluated using Schoenfeld residuals, and the assumption was confirmed to be satisfied. Variables with P<0.05 in univariate analysis were entered into the multivariate Cox or logistic regression models. All variables with P<0.05 in univariate analyses were initially included in the multivariate model, which was refined using a backward stepwise elimination method (retention threshold P<0.10). Given the time-dependent nature of mortality, Harrell’s C-index was calculated to evaluate the discriminatory performance of the final survival model, along with its 95% confidence interval. In this study, potential confounders were addressed by including all clinically relevant and statistically significant variables in a multivariate Cox proportional hazards model. This approach allows simultaneous adjustment for multiple covariates, thereby minimizing confounding effects and providing adjusted hazard ratios. Before model fitting, we assessed multicollinearity among the independent variables using correlation matrices, indicating there was no meaningful multicollinearity in the final model.

To internally validate the stability and robustness of the HALP and PNI cut-off values obtained from ROC analysis, a bootstrapping procedure with 1000 iterations was performed using MedCalc Statistical Software. Bootstrap-derived 95% confidence intervals were calculated to assess the consistency of the optimal cut-off values.

In addition, multivariate binary logistic regression analysis was conducted to identify independent risk factors for 6-month mortality. Variables with P<0.05 in univariate logistic regression were included in the multivariate model, which was refined using a backward stepwise elimination method (retention threshold P<0.10). The model’s goodness-of-fit was assessed using the Omnibus test and Nagelkerke R² statistics, and results are expressed as odds ratios (OR) with 95% confidence intervals.

Results

SUBGROUP ANALYSES:

To assess the consistency of the prognostic performance of HALP and PNI across different patient profiles, age-stratified and sex-stratified ROC analyses were performed. In patients younger than 80 years, HALP showed stronger discrimination for 6-month mortality (AUC=0.720, 95% CI: 0.624–0.803, P=0.0014) compared with those aged 80 years or older (AUC=0.654, 95% CI: 0.538–0.758, P=0.0409) (Figure 3).

Similarly, HALP demonstrated superior discriminatory ability in women (AUC=0.724, 95% CI: 0.618–0.814, P=0.0010) compared with men (AUC=0.655, 95% CI: 0.551–0.750, P=0.0289) (Figure 4). PNI did not demonstrate significant discriminatory value in any subgroup.

Consistent with the ROC findings, age-stratified logistic regression analysis showed that HALP remained a significant predictor of mortality in patients <80 years (OR=0.004, 95% CI: 0.000–0.426, P=0.021), whereas the association did not reach significance in those ≥80 years (p=0.076) (Table 4).

Univariate Cox regression analyses identified several parameters significantly associated with mortality, including age, coronary artery disease (CAD), chronic kidney disease (CKD), congestive heart failure (CHF), postoperative complications, polypharmacy, HALP, and PNI (Table 5). In the multivariate Cox regression analysis, CAD (P<0.001), postoperative complications (P<0.001), and HALP (P=0.037) emerged as independent predictors of mortality. Kaplan-Meier survival analysis revealed distinct survival curves for tertiles of HALP and PNI scores, with a significant difference between groups (log-rank P<0.001). Low HALP scores were associated with reduced survival (Figures 5, 6).

In our cohort, the number of patients in the low, middle, and high HALP tertile groups were 135, 132, and 282, respectively, with corresponding 6-month survival rates of 76%, 73%, and 93%. Similarly, for PNI tertiles, group sizes were 138 (low), 273 (middle), and 138 (high), with 6-month survival rates of 74%, 85%, and 91%, respectively. The log-rank tests confirmed a significant survival difference across both HALP and PNI groups (P<0.001 for each comparison). Notably, the median survival time was not reached in any tertile within the 6-month follow-up period, as more than half of the patients in every group survived beyond this endpoint.

Given the time-to-event nature of mortality and the presence of censored data, we complemented ROC curve analysis with survival-based performance measures. In particular, we used Harrell’s concordance index (C-index) to evaluate the discriminative ability of the multivariate Cox regression model. The C-index quantifies how well the model distinguishes between patients with different survival times, but should not be interpreted as a test of proportional hazards assumptions or overall model fit. The final model, which incorporated coronary artery disease, postoperative complications, and the HALP score, achieved a C-index of 0.77 (95% CI: 0.70–0.83), indicating moderate-to-good discrimination. To quantify the added prognostic value of the HALP score, we compared a baseline model containing only conventional clinical predictors (eg, age and comorbidities) with an extended model that included HALP (Figure 7).

Discussion

LIMITATIONS:

This study has several limitations. First, its retrospective and single-center design introduces the risk of selection bias and limits the generalizability of the findings. The modest sample size may also reduce the precision of subgroup analyses. Additionally, important geriatric variables – such as frailty, functional status, and cognitive impairment – were not available, and the study did not benchmark HALP or PNI against established prognostic tools (eg, Charlson Comorbidity Index, Nottingham Hip Fracture Score), leaving room for residual confounding.

Moreover, the unexpected inverse association observed for cerebrovascular disease may reflect selection bias, as patients with severe neurological deficits are often not considered suitable surgical candidates and therefore may have been underrepresented in our cohort.

Laboratory parameters may have been influenced by transient physiological factors despite standardized preoperative timing, which could affect score variability. Furthermore, although HALP was independently associated with 6-month mortality, its overall discriminative performance was modest, underscoring that it should be used as a supportive rather than standalone prognostic marker.

Finally, prospective, multicenter studies with larger cohorts are needed to validate cut-off values, assess clinical applicability, and determine whether interventions targeting immunonutritional status can improve outcomes.

Conclusions

In this study, the HALP score emerged as a practical and readily accessible immunonutritional marker with moderate prognostic value for 6-month mortality in geriatric hip fracture patients, whereas the PNI demonstrated limited predictive utility. A preoperative HALP value ≤0.176 may help identify patients at elevated risk and could be integrated as a supplementary component of routine preoperative assessment. Patients with low HALP values may benefit from intensified perioperative monitoring and targeted optimization strategies. However, HALP should complement – not replace – established clinical risk indicators. Future multicenter prospective studies are required to determine whether modifying immunonutritional status or incorporating HALP-guided pathways leads to measurable improvements in postoperative outcomes.