Association between Neutrophil-to-Albumin Ratio and 28-Day All-Cause Mortality in Patients with Traumatic Brain Injury : A Retrospective Analysis of the MIMIC-IV Database

Article information

J Korean Neurosurg Soc. 2025;68(6):736-748

Publication date (electronic) : 2025 August 7

doi : https://doi.org/10.3340/jkns.2025.0061

Ziming Huang^,¹

, Hengfa Ge ¹

, Ying Sun ²

¹Department of Trauma Center, The Affiliated Huaian No.1 People’s Hospital of Nanjing Medical University, Huaian, China

²Jiangsu Bioactive Substances Engineering Research Center, School of Pharmaceutical Engineering, Jiangsu Food & Pharmaceutical Science College, Huaian, China

Address for correspondence : Ziming Huang Department of Trauma Center, The Affiliated Huaian No.1 People’s Hospital of Nanjing Medical University, Beijing West Road 1, Huaian 230032, China Tel : +86-15896163441, Fax : +86-517-80872243, E-mail : hzm87@njmu.edu.cn

Received 2025 March 12; Revised 2025 April 18; Accepted 2025 May 12.

Abstract

Objective

The neutrophil-to-albumin ratio (NAR) has emerged as a novel prognostic biomarker in multiple disease contexts, including infectious and cardiovascular disorders. Its prognostic relevance in traumatic brain injury (TBI), however, remains unexamined. This study investigates the association between NAR and 28-day all-cause mortality in patients with TBI.

Methods

This retrospective study analyzed data from the Medical Information Mart for Intensive Care IV database. Neutrophil counts, serum albumin concentrations, and NAR values were recorded within the first 24 hours following TBI admission. Additional clinical and laboratory parameters were collected. The Youden index was employed to determine the optimal NAR threshold. Receiver operating characteristic (ROC) curve analysis was used to assess the discriminative capacity of NAR for all-cause mortality. Subgroup analyses were conducted to evaluate potential heterogeneity in NAR’s prognostic utility across distinct patient subsets. An external validation cohort comprising 112 TBI cases from the institutional database was included to confirm predictive performance. Kaplan-Meier survival analyses were used to compare outcomes between high- and low-NAR groups, while ROC analysis was performed across the entire TBI cohort to assess overall prognostic accuracy.

Results

A total of 213 TBI patients were included and stratified based on 28-day survival status : 180 survivors and 33 non-survivors, resulting in an overall mortality rate of 15.5%. Multivariate Cox regression identified NAR as an independent predictor of 28-day all-cause mortality (hazard ratio [HR], 3.224; 95% confidence interval [CI], 1.321–4.594; p<0.001). ROC curve analysis determined an optimal NAR cutoff of 1.2839 for discriminating between survivors and non-survivors. Kaplan-Meier survival analysis revealed significantly elevated mortality among patients with NAR ≥1.2839 compared to those with NAR <1.2839 (log-rank p<0.001). The area under the curve (AUC) for NAR reached 82.45% (95% CI, 67.02–87.50%), surpassing the predictive performance of neutrophil count (AUC, 60.27%) and serum albumin level (AUC, 60.91%) when assessed individually. Subgroup analyses showed no significant interaction effects (p for interaction, 0.302–0.908), indicating consistent predictive performance across patient subgroups. External validation reinforced the prognostic value of NAR : patients in the high-NAR group demonstrated significantly worse survival outcomes (HR, 3.611; 95% CI, 1.385–9.419; p<0.01), with comparable discriminatory accuracy (AUC, 82.91%; 95% CI, 65.13–89.59%).

Conclusion

NAR functions as an independent and robust prognostic indicator of 28-day all-cause mortality in patients with TBI. Compared to neutrophil count and serum albumin levels alone, NAR demonstrates superior predictive accuracy and may serve as a valuable biomarker for early mortality risk stratification in this population.

Keywords: Traumatic brain injury; Neutrophils; Albumins; Prognosis; ROC curve

INTRODUCTION

Traumatic brain injury (TBI) constitutes a major global public health concern, frequently resulting in long-term cognitive and behavioral impairments that significantly diminish quality of life and impose considerable socioeconomic costs [18]. Epidemiological data suggest that nearly half of the global population may experience at least one TBI during their lifetime, with over 500000 new cases reported annually worldwide [2]. Despite notable advances in therapeutic interventions aimed at mitigating secondary brain injury, mortality rates among TBI patients remain persistently high [11]. This underscores the critical need for early risk stratification to enhance prognostic accuracy and guide clinical decision-making [7].

Recent investigations have explored the prognostic utility of biomarkers derived from serum or cerebrospinal fluid, including β2-integrin, neurofilament light chain, and growth differentiation factor 15 [5,21,27]. While these central nervous system (CNS)-specific biomarkers offer valuable mechanistic insights into brain injury, their clinical utility is often constrained by limited availability, high cost, and delayed turnaround times [9]. In contrast, peripheral biomarkers reflecting systemic inflammation and immune activation—key elements of TBI pathophysiology—offer practical advantages [16]. Composite inflammatory indices, such as the neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio, have been widely studied for their prognostic relevance in critical illness [14,17]. The neutrophil-to-albumin ratio (NAR) has recently gained attention as a cost-effective and readily accessible marker for evaluating inflammation severity and predicting outcomes in infectious and cardiovascular conditions [6,20].

In the context of acute trauma, including TBI, elevated neutrophil levels may promote vascular dysfunction and disrupt blood-brain barrier integrity, thereby influencing clinical progression [15]. Neutrophils can also release pro-inflammatory cytokines such as tumor necrosis factor, which exacerbate cellular damage and secondary injury [25]. Concurrently, serum albumin— an essential plasma protein—plays a pivotal role in maintaining oncotic pressure, stabilizing microvascular function, and attenuating inflammatory responses [22]. Hypoalbuminemia following TBI impairs anti-inflammatory capacity, thereby exacerbating cerebral edema and worsening neurologic outcomes [29]. These opposing dynamics suggest that the NAR, integrating both inflammatory burden and nutritional/inflammatory status, may serve as a sensitive marker for prognosis in TBI. Despite its promise, the relationship between NAR and mortality in TBI remains poorly characterized. To address this gap, this study leveraged data from the MIMIC-IV (Medical Information Mart for Intensive Care IV; v2.2) database, comprising TBI patients admitted to the intensive care unit (ICU) between 2008 and 2019. The objective was to evaluate the association between NAR and 28-day all-cause mortality. It was hypothesized that NAR would complement existing clinical and imaging-based assessments, enhancing the accuracy of prognostic evaluation in TBI. Additionally, an external validation cohort from the institutional TBI registry was analyzed to corroborate the findings.

MATERIALS AND METHODS

Ethical approval for the use of the MIMIC-IV database (Record ID : 001107) had been obtained for both data collection and secondary analyses. In accordance with database access protocols, completion of CITI training and agreement to the Data Use Agreement (DUA) were mandatory. All access requirements were fulfilled (PhysioNet Credential ID : [12765719]).

Database introduction

The MIMIC-IV database, curated by the Massachusetts Institute of Technology, serves as an open-access, multiparametric repository of critical care data. Detailed information is presented at https://physionet.org/content/mimiciv/2.2/. It encompasses comprehensive clinical records, including ICU stay durations, laboratory results, medication administration, vital signs, and other relevant physiological parameters. To uphold patient confidentiality, all identifiable information has been anonymized and replaced with randomly generated codes, in full compliance with data privacy standards. Consequently, the use of de-identified public datasets such as MIMIC-IV exempts researchers from obtaining individual patient consent or additional ethical approval.

Population selection criteria

The database contains 431231 hospital admissions, of which 76540 involve ICU stays. TBI cases were identified using the International Classification of Diseases, 9th Revision (ICD-9) codes 800.2-804.2 and ICD-10 codes S06.1-S06.9, yielding 4369 TBI patients, including 2588 with ICU admissions. After applying exclusion criteria—namely, age below 18 years at first admission; diagnoses of end-stage renal disease, liver cirrhosis, or neoplasms related to TBI; absence of documented neutrophil or albumin values; or laboratory measurements obtained more than 24 hours post-admission—a final cohort of 213 patients was selected for analysis (Fig. 1).

Fig. 1.

Patient selection flowchart from the MIMIC-IV (Medical Information Mart for Intensive Care IV) database. Initial cohort : 4369 adult traumatic brain injury (TBI) patients (age ≥18 years). Exclusion criteria : 1) non-first intensive care unit (ICU) admission (n=1781), 2) missing neutrophil or albumin measurements (n=342), 3) patients with chronic diseases (n=422), and 4) neutrophil or albumin measurements recorded more than 24 hours after admission (n=1611). Final analytical cohort : 213 patients.

To further evaluate the predictive validity of the NAR, an external validation cohort consisting of 112 patients with TBI was analyzed using data retrieved from the institutional database. Among these cases, 22 (19.6%) resulted in mortality. Comprehensive demographic and clinical characteristics are detailed in Table 1.

Table 1.

Baseline characteristics between survivors and non-survivors in the external cohort

Data extraction and data processing

All relevant variables pertaining to blood neutrophils and serum albumin—including chart time, measured value, stay ID, subject ID, ICU admission time, and ICU discharge time—were extracted during hospitalization using PostgreSQL (version 11; PostgreSQL Core Team, Berkeley, CA, USA) in conjunction with pgAdmin (version 6.21; pgAdmin Development Team, Berkeley, CA, USA). Data corresponding to both biomarkers within identical timeframes were filtered using StataMP (version 16.0; StataCorp LLC, College Station, TX, USA), and the NAR was subsequently computed. To minimize potential bias introduced by therapeutic interventions, chart times were selected from the first day following ICU admission.

Potential confounders were categorized and extracted as follows : 1) demographics : age, gender, race, weight; 2) vital signs : heart rate (HR), systolic and diastolic blood pressure, mean arterial pressure, respiratory rate (RR), temperature; 3) clinical interventions : vasopressor administration; 4) injury types : epidural hematoma (EDH), subdural hematoma (SDH), subarachnoid hematoma (SAH); 5) laboratory parameters : serum potassium, serum sodium, red blood cells, white blood cells (WBCs), platelets, hemoglobin, blood urea nitrogen, neutrophils, lymphocytes, creatinine, blood glucose, lactate; 6) severity indices : Glasgow coma scale (GCS), Sequential Organ Failure Assessment (SOFA) score; 7) comorbidities : sepsis, chronic obstructive pulmonary disease, hypertension, diabetes mellitus, obesity; and 8) surgical interventions : craniotomy with hematoma evacuation, stereotactic aspiration, decompressive craniectomy, hybrid surgical approaches.

Grouping and endpoint event

The study cohort was stratified into two groups based on 28-day survival status: survivors (n=180) and non-survivors (n=33). The primary outcome was all-cause mortality within 28 days following hospital admission. The 28-day mortality rate was computed by dividing the total number of deaths from any cause within the defined period by the overall at-risk population during that interval.

Management of missing data and outliers

To identify and manage outliers, the "scatter" function in StataMP was utilized. Outlier values were excluded from analysis and recoded as missing. Variables with over 15% missing data were excluded to mitigate the risk of bias. For variables exhibiting 5% to 15% missingness, imputation was performed using random sampling from predicted distributions. When missingness was below 5%, the mean value of the respective variable was used to replace missing entries.

Statistical analysis

Continuous variables were presented as mean±standard deviation (SD) for normally distributed data and as median with interquartile range (IQR) for data exhibiting non-normal distributions. Categorical variables were reported as frequencies with corresponding percentages. For comparisons of baseline characteristics, continuous variables were assessed using independent t-tests or one-way analysis of variance (ANOVA) based on distributional assumptions, while categorical variables were evaluated via Pearson’s chi-squared test or Fisher’s exact test, as appropriate. Univariate Cox regression analysis was employed to screen potential risk factors for all-cause mortality. Variables with a p-value <0.1 in the univariate model were subsequently entered into multivariate Cox regression to identify independent predictors.

To assess the prognostic accuracy of neutrophils, albumin, NAR, and SOFA scores for predicting 28-day mortality, receiver operating characteristic (ROC) curve analysis was conducted. Sensitivity, specificity, and the area under the curve (AUC) were calculated for each parameter. The optimal NAR cut-off was determined using the Youden index, facilitating stratification into high and low NAR groups. Kaplan-Meier survival curves were generated for these groups, and statistical differences were examined using the log-rank test. Subgroup analyses were performed to evaluate the association between NAR and mortality across predefined categories, including age, gender, vasopressor use, hypertension, sepsis, diabetes, and injury type. All statistical procedures were conducted using Free Stat-MP software (version 1.6; StataCorp LLC), applying two-tailed tests with a significance threshold set at p<0.01.

RESULTS

Baseline demographic and clinical characteristics

The baseline characteristics of the 28-day survivor and non-survivor groups are presented in Table 2. The study cohort comprised 213 patients, including 68 females (31.9%) and 145 males (68.1%), with a mean age of 63.8 years (SD, 19.3). The overall 28-day mortality rate was 15.5%. Compared to survivors, non-survivors exhibited significantly lower systolic and diastolic blood pressure as well as reduced GCS scores. In contrast, elevated RR, HR, and SOFA scores were observed among non-survivors. Vasopressor administration occurred in 11.7% of non-survivors, and sepsis was documented in 40.4% of this subgroup. Laboratory data at admission indicated a significantly higher NAR in the non-survivor group (3.0; IQR, 2.1–4.2) compared to survivors (2.2; IQR, 1.4–3.2; p<0.001). Additionally, levels of sodium, WBC, glucose, NLR, neutrophils, lymphocytes, albumin, and lactate were all significantly elevated in non-survivors (p<0.05). No statistically significant differences were found for the remaining covariates (p>0.05).

Table 2.

Baseline characteristics between survivors and non-survivors

An external validation cohort of 112 TBI patients was also assessed, with 22 deaths (19.6%) reported during the 28-day follow-up period. Detailed demographic and clinical variables, including injury severity and NAR distribution, are provided in Table 1.

The NAR is an independent risk factor for all-cause mortality at 28-day of hospital admission

Univariate Cox regression analyses were performed using the variables listed in Table 2. The results indicated a significant unadjusted association between NAR and 28-day all-cause mortality (HR, 1.809; 95% confidence interval [CI], 1.185–2.484; p<0.001). Variables with p<0.1 from Table 3, along with known risk factors, were included in the multivariate Cox proportional hazards regression. As shown in Table 4, after adjustment for surgical intervention and GCS severity, NAR remained an independent predictor of in-hospital 28-day mortality (HR, 3.224; 95% CI, 1.321–4.594; p<0.001). These findings underscore the prognostic value of NAR as a robust, independent biomarker for mortality risk stratification in TBI patients.

Table 3.

Univariate Cox analysis of risk factors for 28-day death in patients by logistic regression analysis

Table 4.

Multivariate Cox analysis of risk factors for 28-day death in patients by logistic regression analysis

ROC curve analysis and Kaplan-Meier curve

ROC curves were generated for four indicators—NAR, neutrophils, albumin, and SOFA—to assess their predictive performance for 28-day all-cause mortality in TBI patients (Fig. 2). NAR demonstrated a superior AUC of 82.45% (95% CI, 67.02–87.50%), significantly exceeding those of neutrophils (AUC, 60.27%; 95% CI, 56.11–64.46%) and albumin (AUC, 60.91%; 95% CI, 57.31–66.32%), though marginally lower than SOFA (AUC, 83.78%; 95% CI, 77.62–88.29%) (Fig. 2A). These results highlight the strong prognostic utility of NAR relative to other individual markers. The optimal NAR cut-off value, determined to be 1.2839, yielded a sensitivity of 66.67% and a specificity of 67.78%. Based on this threshold, patients were stratified into high-NAR (≥1.2839; n=171) and low-NAR (<1.2839; n=42) groups. Kaplan-Meier survival analysis (Fig. 3A) revealed a significantly higher mortality rate in the high-NAR group (HR, 2.356; 95% CI, 1.173–4.730; log-rank p<0.001).

Fig. 2.

Receiver operating characteristic (ROC) curves for in-ICU mortality prediction. A : ROC in the MIMIC-IV (Medical Information Mart for Intensive Care IV) cohort. The neutrophil-to-albumin ratio (NAR, red solid line) shows superior discriminative performance (AUC, 0.8245; 95% CI, 0.6702–0.8750) compared to isolated neutrophil count (green solid line; AUC, 0.6027) and albumin levels (yellow solid line; AUC, 0.6091). B : ROC in the external validation cohort. The NAR (red solid line) exhibits superior discriminative performance (AUC, 0.8297; 95% CI, 0.651–0.896) compared to isolated neutrophil count (green solid line; AUC, 0.6619) and albumin levels (yellow solid line; AUC, 0.6528). SOFA : sequential organ failure assessment, ICU : intensive care unit, AUC : area under the curve, CI : confidence interval.

Fig. 3.

Kaplan-Meier survival curves for 28-day all-cause mortality stratified by neutrophil-to-albumin ratio (NAR) levels. A : NAR in the MIMIC-IV (Medical Information Mart for Intensive Care IV) cohort : the high-NAR group (≥1.2839, pink line) demonstrates significantly lower survival probability compared to the low-NAR group (<1.2839, blue line; HR, 2.356, 95% CI, 1.173–4.730; log-rank p<0.001). B : NAR in the external validation cohort : the high-NAR group (≥1.2839, pink line) exhibits significantly reduced survival probability compared to the low-NAR group (<1.2839, blue line; HR, 3.611; 95% CI, 1.385–9.419; log-rank p<0.001). HR : hazard ratio, CI : confidence interval.

In the external validation cohort, Kaplan-Meier survival curves revealed a significant separation between high- and low-NAR groups (HR, 3.611; 95% CI, 1.385–9.419; p<0.01; Fig. 3B). The NAR indicator, depicted by a red solid line, exhibited superior discriminatory capability for mortality prediction, achieving an AUC of 82.91% (95% CI, 65.1–89.6%), substantially outperforming neutrophil count (AUC, 66.19%) and albumin (AUC, 65.28%) (Fig. 2B).

Subgroup analysis

Fig. 4 presents a subgroup analysis evaluating the consistency of NAR’s association with 28-day all-cause mortality across clinically relevant strata. Forest plot analysis identified no significant interactions between NAR and stratifying variables, including age, sex, ethnicity, vasopressor use, hypertension, sepsis, diabetes, and injury type (p for interaction, 0.302–0.908), confirming the independence of NAR as a prognostic marker. The stable predictive performance across heterogeneous subgroups underscores the broad clinical utility and robustness of NAR in risk stratification for TBI-related mortality.

Fig. 4.

Forest plot of subgroup analysis for NAR-ICU mortality association. Subgroups are categorized by traumatic brain injury subtypes (EDH/SDH/SAH), age, gender, vasopressin use, high blood pressure (HBP), sepsis, and diabetes. Interaction p-values range from 0.302 to 0.908. HR : hazard ratio, CI : confidence interval, EDH : epidural hematoma, SDH : subdural hematoma, SAH : subarachnoid hematoma, NAR : neutrophil-to-albumin ratio, ICU : intensive care unit.

DISCUSSION

This retrospective cohort study systematically investigated the prognostic relevance of blood biochemical markers in critically ill TBI patients, with an emphasis on NAR as a key biomarker. The findings demonstrated that NAR independently predicted 28-day all-cause mortality following TBI admission (adjusted HR, 3.224; 95% CI, 1.321–4.594; p<0.001). In comparative performance analysis, NAR exhibited superior predictive accuracy (AUC, 82.45%; 95% CI, 67.0–87.5%) relative to neutrophil count (AUC, 60.27%) and albumin levels (AUC, 60.91%), and approached the discriminative capacity of the SOFA score (AUC, 83.78%; 95% CI, 77.62–88.29%). Kaplan-Meier survival curves derived from both the MIMIC-IV and external validation cohorts consistently revealed a significantly increased mortality risk among patients with NAR ≥1.284 (log-rank p<0.001). Subgroup analyses confirmed the stability and consistency of NAR’s prognostic value across diverse clinical strata.

TBI pathology follows a biphasic trajectory, comprising an initial primary insult and a subsequent cascade of secondary injury processes [4]. While primary injury—resulting from mechanical trauma at the time of impact—is largely irreversible, secondary injury evolves through dynamic neuroinflammatory and metabolic mechanisms within the brain [31], governed by immune activity in both the CNS and peripheral circulation [24]. Neutrophils are the first peripheral immune responders, rapidly accumulating at the injury site in response to chemokine gradients released by damaged neural tissue [23]. These cells exhibit dual functionality: promoting early debris clearance and microbial defense through phagocytosis [19], while also modulating adaptive immunity via recruitment of T and B lymphocytes [1]. However, excessive neutrophilic infiltration can exacerbate inflammation and contribute to secondary neurodegeneration. Consistent with prior reports [26], the present analysis identified significantly elevated neutrophil counts in non-survivors compared with survivors at admission (median [IQR] : 9.7 [6.7–12.1] vs. 7.75 [4.5–10.7]; p<0.001). Multivariate regression further established neutrophil elevation as an independent risk factor for 28-day in-hospital mortality (adjusted HR, 1.157; 95% CI, 0.909–1.223; p=0.003), reinforcing its prognostic significance in the acute phase of TBI.

Serum albumin exhibits a characteristic temporal decline following TBI, reaching its lowest concentration 72 hours post-injury—coinciding with its peak prognostic utility for short-term outcomes (AUC, 61%; 95% CI, 57–66%) [8,28]. The development of hypoalbuminemia in this context likely results from a confluence of systemic inflammatory activation, oxidative stress-driven protein catabolism, and increased metabolic substrate consumption. Although the precise regulatory mechanisms remain partially elucidated, a strong inverse correlation between albumin levels and systemic inflammatory burden has been reported (β=–0.43; p=0.002) [13], potentially accounting for the observed 38% increase in mortality risk per 1 g/dL decrement in serum albumin (adjusted HR, 0.52; 95% CI, 0.41–0.66; p<0.001). Admission albumin levels were significantly lower in non-survivors compared to survivors (3.0±0.7 vs. 3.3 ± 0.6 g/dL; p<0.001), affirming its role as both a marker of acute physiological deterioration and an independent predictor of adverse outcomes. However, the modest discriminatory capacity of albumin alone (AUC, 0.609) likely reflects the multifactorial pathophysiological alterations accompanying severe brain injury, such as blood-brain barrier compromise and neuroendocrine dysfunction. These findings underscore the limitations of single-variable biomarkers and highlight the necessity for integrative models that incorporate inflammatory, metabolic, and neuro-specific parameters to enhance prognostic accuracy.

The NAR, a composite biomarker combining neutrophil count and serum albumin concentration, represents an emerging multidimensional tool with validated prognostic relevance across a range of clinical settings, including infectious and cardiovascular diseases [3,12]. By capturing the bidirectional relationship between systemic inflammation and nutritional reserve, NAR offers a more comprehensive reflection of a patient’s physiological status and short-term risk profile [10,30].

In trauma-induced conditions, empirical evidence confirmed that the NAR offers predictive value comparable to established biomarkers. This study revealed that NAR values in the 28-day survival group were significantly lower than in the non-survival group (median [IQR] : 2.2 [1.4, 3.2] vs. 3.0 [2.1, 4.2]; p<0.001). Multivariate regression analysis yields an adjusted hazard ratio of 3.224 (95% CI, 1.321–4.594; p<0.001), indicating that each unit increase in NAR corresponds to a 3.22-fold increase in mortality risk. By combining neutrophil counts, which indicate systemic inflammation, with albumin levels, reflecting nutritional status, NAR enhances predictive accuracy across various clinical scenarios. Notably, this composite biomarker demonstrates superior discriminative power for mortality in TBI, achieving an AUC of 82.45% (95% CI, 67.0–87.5%), surpassing AUC values for neutrophils (0.603) and albumin (0.609) alone. As such, NAR represents a significant advancement in the prognostic assessment and management of TBI, facilitating more precise risk stratification and personalized treatment strategies. Elevated neutrophil levels signal systemic inflammation, potentially aggravating secondary brain injury by disrupting the blood-brain barrier, inducing oxidative stress, and promoting cerebral edema. Hypoalbuminemia, indicative of malnutrition, metabolic stress, and endothelial dysfunction, hinders systemic recovery and heightens susceptibility to complications such as infections and organ failure. While NAR encapsulates these systemic disturbances, its direct causal link to TBI pathophysiology remains undefined. Further mechanistic research is necessary to determine whether modulation of NAR components, such as anti-inflammatory therapies or albumin supplementation, could improve outcomes.

This study has several limitations. Firstly, inherent biases due to the retrospective design : the temporal range of the MIMIC-IV dataset (2008–2019) may introduce heterogeneity, as evolving treatment protocols (e.g., neuroprotective agents, surgical criteria) and clinical practices (e.g., neutrophil/albumin assay methods) were not accounted for. Additionally, unmeasured confounders, such as pharmacotherapy, socioeconomic status, and genetic factors, may dilute causal inferences between NAR and mortality. Secondly, the reliance on a single-timepoint NAR measurement : excluding patients without concurrent neutrophil/albumin data limits generalizability, especially for critically ill individuals with incomplete laboratory workups. While baseline NAR measurements effectively stratify mortality risk in the early phase, serial monitoring could better capture evolving pathophysiology. Future work should involve prospective cohort studies with 6-hour interval NAR measurements within the first 72 hours post-admission to 1) identify temporal fluctuations in NAR, 2) develop time-dependent mortality prediction models, and 3) pinpoint optimal windows for targeted interventions. Thirdly, this study's single-center retrospective cohort design limits its generalizability. While ongoing data collection is in progress, more research time is required. Detailed subgroup analyses (e.g., stratified Cox models for EDH, SDH, and SAH) would strengthen the findings. However, due to the low number of mortality outcomes in the SAH subgroup (n=6), such analyses risk overfitting or producing unstable estimates. Finally, the study did not account for medication use due to incomplete data, which may confound the relationship between NAR and mortality. Future directions include : 1) dynamic monitoring through prospective cohorts with serial NAR measurements (e.g., every 6 hours) to identify critical intervention windows; 2) multimodal models integrating NAR with neuroimaging and biomarker data using machine learning; 3) mechanistic studies through preclinical experiments to elucidate causal relationships between NAR components and secondary injury mechanisms; and 4) global collaboration, establishing standardized data collection platforms for multicenter validation and pharmacotherapy analysis.

CONCLUSION

This study establishes the NAR as an independent predictor of 28-day all-cause mortality in ICU-admitted TBI patients. NAR outperforms individual blood neutrophil counts and serum albumin levels in prognostic accuracy, demonstrating comparable performance to the SOFA score. This provides healthcare professionals with a more effective tool for early risk stratification and clinical decision-making, thereby improving patient outcomes. However, large-scale, multicenter prospective validation is crucial to confirm NAR’s potential as a reliable, easily accessible biomarker for TBI management.

Notes

Conflicts of interest

No potential conflict of interest relevant to this article was reported.

Informed consent

This type of study does not require informed consent.

Author contributions

Conceptualization : ZH; Data curation : ZH; Formal analysis : ZH; Methodology : HG; Project administration : ZH; Visualization : YS; Writing - original draft : ZH; Writing - review & editing : ZH

Data sharing

None

Preprint

None

References

1. Aghadoost H, Salehabadi G, Fakharian E, Mohammadabad HJ. A comparative analysis of Rotterdam score and neutrophil-to-lymphocyte ratio in predicting outcomes for patients with moderate to severe traumatic brain injury. Eur J Transl Myol 35:13443. 2025;

2. Ansari A, Zoghi S, Tavanaei R, Payman AA, Lu VM, Niakan A, et al. Characteristics and recovery trends of severe TBI patients with a favorable functional outcome at 6-month follow-up. Neurosurg Rev 48:41. 2023;

3. Cahyani KMD, Candrawati NW, Arisanti NLPE, Suryana K, Kusumawardani IAJD, Artana IGNB, et al. Relationship between neutrophil to lymphocytes ratio, red cell distribution width, procalcitonin, neutrophil to albumin ratio, and bun to albumin ratio with mortality of severe cap patients with septic shock. Int J Adv Med 11:291–298. 2024;

4. Cembran A, Eisenbaum M, Paris D, Mullan M, Crawford F, Ferguson S, et al. Contribution of brain pericytes to neuroinflammation following repetitive head trauma. J Inflamm (Lond) 22:11. 2023;

5. Chen L, Luo S, Liu T, Shuai Z, Song Y, Yang Q, et al. Growth differentiation factor 15 aggravates sepsis-induced cognitive and memory impairments by promoting microglial inflammatory responses and phagocytosis. J Neuroinflammation 22:44. 2025;

6. Cheng W, Li T, Wang X, Xu T, Zhang Y, Chen J, et al. The neutrophil-to-apolipoprotein A1 ratio is associated with adverse outcomes in patients with acute decompensated heart failure at different glucose metabolic states: a retrospective cohort study. Lipids Health Dis 23:118. 2024;

7. Corbella D, Zangari R, Biroli F, Magnone S, Cavalleri G, Passoni M, et al. Comparing survival and outcomes in isolated versus polytrauma-associated TBI: a retrospective cohort study. J Neurosurg Sci 2024;

8. Guo Y, Leng Y, Gao C. Blood urea nitrogen-to-albumin ratio may predict mortality in patients with traumatic brain injury from the MIMIC database: a retrospective study. Bioengineering (Basel) 11:49. 2024;

9. Harris M, Yue JK, Jain S, Sun X, Puccio AM, Gardner RC, et al. Effect of blood alcohol on the diagnostic accuracy of glial fibrillary acidic protein and ubiquitin carboxy-terminal hydrolase L1 for traumatic intracranial hemorrhage: a TRACK-TBI study. Acad Emerg Med 2025;

10. Hua X, Wang MD, Ni WQ, Long ZQ, Wang SF, Duan FF, et al. Development and validation of a prognostic nomogram incorporating neutrophil-to-albumin ratio for predicting overall survival in patients with nasopharyngeal carcinoma undergoing concurrent chemoradiotherapy. Heliyon 11e40881. 2024;

11. Kersey J, Alimi E, McArthur AR, Marquez H, Baum C, Skidmore E, et al. ENGAGE-TBI: adaptation of a community-based intervention to improve social participation after brain injury. Brain Inj 39:518–525. 2023;

12. Lai G, Zhao Y, Yang C, Zheng Y, Sun J, Zhao Y, et al. Association of Naples Prognostic Score with cardiovascular disease risk and its longitudinal prognostic impact on mortality in cardiovascular disease patients: evidence from NHANES. Nutr Metab Cardiovasc Dis 35:103840. 2025;

13. Lee JH, Lee D, Lee BK, Cho YS, Kim DK, Jung YH, et al. The association between lactate to albumin ratio and outcomes at early phase in patients with traumatic brain injury. Ulus Travma Acil Cerrahi Derg 29:752–757. 2023;

14. Li W, Wang Z, Gao M, Wang Y, Ke Y. A study on the relationship between neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR) in neurosurgery and the occurrence and prognosis of progressive hemorrhagic brain injury (PHI) in patients with traumatic brain injury. BMC Neurol 24:484. 2023;

15. Liu YW, Li S, Dai SS. Neutrophils in traumatic brain injury (TBI): friend or foe? J Neuroinflammation 15:146. 2018;

16. McDonald SJ, Shultz SR. Blood-based biomarkers and health-related quality of life outcomes in mild TBI: new insights and unexpected associations. Neurology 104e210238. 2023;

17. Miao HT, Wang J, Shao JJ, Song RX, Li WG, Sun JK, et al. Astrocytic NLRP3 cKO mitigates depression-like behaviors induced by mild TBI in mice. Neurobiol Dis 205:106785. 2023;

18. Mikolić A, van Klaveren D, Jost M, Maas AI, Shi S, Silverberg ND, et al. Prognostic models for depression and post-traumatic stress disorder symptoms following traumatic brain injury: a CENTER-TBI study. BMJ Ment Health 28e301181. 2023;

19. Qian F, Zhong Q, Chen Z. Role of mitochondrial dysfunction in acute traumatic brain injury: evidence from bioinformatics analysis. Heliyon 10e31121. 2023;

20. Rajakumar HK, Coimbatore Sathyabal V, Palaniyandi A, Balakrishnan D. Exploring mean platelet volume and neutrophil-to-albumin ratio as surrogate markers for monitoring tuberculosis treatment: a prospective longitudinal study. BMC Pulm Med 24:406. 2023;

21. Richter S, Winzeck S, Correia MM, Czeiter E, Whitehouse D, Kornaropoulos EN, et al. Predicting recovery in patients with mild traumatic brain injury and a normal CT using serum biomarkers and diffusion tensor imaging (CENTERTBI): an observational cohort study. EClinicalMedicine 75:102751. 2024;

22. Samson K. Two-year TBI outcomes worse with albumin versus saline resuscitation. Neurology Today 7:5. 2007;

23. Strogulski NR, Portela LV, Polster BM, Loane DJ. Fundamental neurochemistry review: microglial immunometabolism in traumatic brain injury. J Neurochem 167:129–153. 2023;

24. van Wessem KJP, Benders KEM, Leenen LPH, Hietbrink F. TBI related death has become the new epidemic in polytrauma: a 10-year prospective cohort analysis in severely injured patients. Eur J Trauma Emerg Surg 50:3083–3094. 2023;

25. Wang C, Xian L, Zheng S, Li J, Chen X, Wang S. Cranial venous-outflow obstruction promotes neuroinflammation via ADAM17/solTNF-α/NF-κB pathway following experimental TBI. Brain Res Bull 204:110804. 2023;

26. Wang H, Liu Y, Yuan J, Wang Y, Yuan Y, Liu Y, et al. Development and validation of a nomogram for predicting mortality in patients with acute severe traumatic brain injury: a retrospective analysis. Neurol Sci 45:4931–4956. 2024;

27. Whitehouse DP, Wilson L, Czeiter E, Buki A, Wang KKW, von Steinbüchel N, et al. Association of blood-based biomarkers and 6-month patient-reported outcomes in patients with mild TBI: a CENTER-TBI analysis. Neurology 104e210040. 2025;

28. Wiedermann CJ. Albumin in normovolemic fluid management for severe traumatic brain injury: controversies and research gaps. J Clin Med 13:5452. 2024;

29. Wiedermann CJ. Use of hyperoncotic human albumin solution in severe traumatic brain injury revisited-a narrative review and meta-analysis. J Clin Med 11:2662. 2022;

30. Xiao L, Li F, Sheng Y, Hou X, Liao X, Zhou P, et al. Predictive value analysis of albumin-related inflammatory markers for short-term outcomes in patients with In-hospital cardiac arrest. Expert Rev Clin Immunol 21:249–257. 2025;

31. Zhang Y, Shen J, Yang J, Huang G, Niu H, Zhang S, et al. Brain-targeted M2 macrophage membrane-hybrid biomimetic liposomes for treatment of traumatic brain injury. Int J Pharm 672:125316. 2023;

Article information Continued

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Variable	Survivors (n=90)	Non-survivors (n=22)	Total (n=112)	p-value
Demographic
Age (years)	67.1±12.6	63.4±21.5	64.8±12.3	0.338^*
Gender				0.523^†
Female	43 (47.8)	8 (36.4)	51
Male	47 (52.2)	14 (63.6)	61
Weight (kg)	61.2 (55.3–75.6)	63.4 (57.5–73.2)	62 (54.6–77.3)	0.448^‡
Vital signs
HR (beats/min)	104 (81.3–123.4)	112 (83–124)	109 (84–121)	0.001^‡
SBP (mmHg)	93.2 (81–101)	86 (71–93)	91 (85–103)	<0.001^‡
DBP (mmHg)	54 (41–63)	47 (39–55)	50 (43–57)	<0.001^‡
RR (beats/min)	23 (22–31)	27 (24–31)	25.0 (19.0–44.0)	0.546^‡
T (℃)	37.4 (37–37.8)	37.8 (37.9–39.1)	37.5 (37.2–37.9)	0.007^‡
Laboratory tests
Neutrophils (K/µL)	8.49 (5.3–11.7)	9.8 (7.2–13.2)	9.13 (5.08–11.27)	<0.001^‡
Lactate (mmol/L)	1.3 (1–2.2)	2.6 (1.5–6.3)	1.9 (1.3–4.4)	<0.001^‡
Albumin (g/L)	3.6±0.8	3.5±0.7	3.2±0.7	<0.001^*
NAR (×10^-1)	2.3 (1.4–3.2)	3.2 (2.3–4.6)	2.7 (1.5–3.3)	<0.001^‡
Severity score
GCS	7 (5–8)	6 (3–9)	5 (4–7)	<0.001^‡
SOFA	3 (2–5)	6 (3–9)	5 (3–8)	<0.001^‡
Comorbidities
Vasopressin	14 (15.6)	7 (31.8)	21	<0.001^†
HBP	38 (42.2)	5 (22.7)	43	0.023^†
Diabetes	16 (17.8)	4 (18.2)	20	0.438^†
COPD	5 (5.6)	2 (9.1)	7	0.774^†
Fracture				0.369^†
Pelvic	21 (23.3)	6 (27.3)	13 (6.1)
Spinal	20 (22.2)	9 (40.9)	26 (12.2)
Femur	14 (15.6)	3 (13.6)	11 (9.8)
Injury type				0.447^†
EDH	23 (25.6)	7 (31.8)	30 (26.8)
SDH	25 (27.8)	7 (31.8)	32 (28.6)
SAH	31 (34.4)	4 (18.2)	35 (31.3)
Other	11 (12.2)	4 (18.2)	15 (13.4)
Mechanism of injury				0.662^†
Traffic accident	33 (36.7)	10 (45.5)	43 (38.4)
High fall	14 (15.6)	5 (22.7)	19 (17.0)
Stumble	27 (30.0)	5 (22.7)	32 (28.6)
Others	16 (17.8)	2 (9.1)	18 (16.1)
Surgical intervention	56 (62.2)	15 (68.2)	71 (63.4)	0.639^†

Variable	28-day survivors (n=180)	28-day non-survivors (n=33)	Total (n=213)	p-value
Demographic
Age (years)	64.1±18.6	61.8±23.3	63.8±19.3	0.528^*
Gender				0.047^†
Female	58 (32.2)	10 (30.3)	68
Male	122(67.8)	23 (69.7)	145
Race				<0.001^†
White	104 (57.8)	10 (30.3)	114 (53.4)
Other	76 (42.2)	23 (69.7)	99 (46.6)
Weight (kg)	73.7 (63.2–86.6)	74 (66.5–84)	74.1 (63.3–86.4)	0.627^‡
Vital signs
HR (beats/min)	102 (89.5–116.5)	116 (87–134)	103 (89–120)	0.004^‡
SBP (mmHg)	95.5 (88–106)	87 (75–97)	95 (85–104)	<0.001^‡
DBP (mmHg)	51 (44–59)	43 (36–50)	50 (43–58)	<0.001^‡
RR (beats/min)	27 (24–32)	26 (25–32)	26.0 (12.0–59.0)	0.896^‡
T (℃)	37.3 (37–37.7)	37.8 (37.2–39)	37.4 (37–37.9)	0.003^‡
Laboratory tests
Sodium (K/µL)	139 (135.5–142)	140 (138–145)	139 (136–142)	0.009^‡
Potassium (K/µL)	4.1 (3.7–4.5)	4 (3.8–4.2)	4.1 (3.7–4.5)	0.585^‡
RBC (m/µL)	3.8 (3.3–4.4)	3.9 (3.3–4.5)	3.8 (3.3–4.4)	0.623^‡
WBC (K/µL)	10.5 (7.4–13.6)	11.6 (9.9–14.5)	10.7 (7.6–13.7)	0.034^‡
Platelet (K/µL)	186 (134–220.5)	154 (130–202)	183 (133–219)	0.307^‡
Hemoglobin (g/dL)	11.3±2.10	11.0±1.99	11.3±2.11	0.383^*
Neutrophils (K/µL)	7.8 (4.5–10.7)	9.7 (6.7–12.1)	8.0 (5.1–11.1)	<0.001^‡
Lymphocyte (10⁹ /L)	2.8 (2.2–3.7)	2.2 (1.7–2.8)	2.5 (1.8–3.4)	<0.001^‡
Creatinine (mg/dL)	0.9 (0.7–1.2)	1.0 (0.8–1.1)	0.9 (0.7–1.2)	0.347^‡
BUN (K/µL)	14.5 (10–23)	14.0 (13–23)	14.0 (11–23)	0.353^‡
SPO₂ (%)	93 (91–95)	95 (92–98)	81 (65–113)	0.437^‡
GLU (mmol/L)	119.0 (103–144)	146.0 (118–172)	120.4 (103–149)	0.031^‡
Lactate (mmol/L)	1.6 (1.0–2.3)	2.6 (1.5–5.2)	1.9 (1.3–3.3)	<0.001^‡
Albumin(g/L)	3.3±0.6	3.0±0.7	3.2±0.7	<0.001^*
NAR (×10^-1)	2.2 (1.4–3.2)	3.0 (2.1–4.2)	2.4 (1.5–3.3)	<0.001^‡
NLR (×10^-1)	2.9 (2.1–4.4)	4.2 (3.4–6.6)	3.6 (1.2–5.9)	<0.001^‡
Severity score
GCS	6 (5–8)	5 (3–6)	5 (4–7)	<0.001^‡
SOFA	3 (2–5)	6 (4–8)	4 (2–7)	<0.001^‡
Comorbidities
Vasopressin	21 (11.7)	14 (42.4)	35 (16.4)	<0.001^†
HBP	69 (38.3)	11 (33.3)	80 (37.6)	0.586^†
Diabetes	44 (24.4)	9 (27.3)	53 (24.9)	0.730^†
COPD	19 (10.6)	2 (6.1)	21 (9.9)	0.426^†
Fracture				0.108^†
Pelvic	12 (6.7)	1 (3.0)	13 (6.1)
Spinal	17 (9.4)	9 (27.3)	26 (12.2)
Femur	9 (5.0)	2 (6.1)	11 (5.2)
Sepsis	73 (40.4)	20 (60.6)	93 (43.7)	0.033^†
Injury type				0.393^†
EDH	55 (30.6)	10 (30.3)	65 (30.5)
SDH	51 (28.3)	10 (30.3)	61 (28.6)
SAH	43 (23.9)	8 (24.2)	51 (23.9)
Other	31 (17.2)	5 (15.2)	36 (16.9)
Surgical intervention	88 (48.9)	23 (69.7)	111 (52.1)	0.384^†

Variable	Univariable Cox
Variable	HR	95% CI	p-value Wald’s test
Age	0.994	0.9751–1.0129	0.526
Gender, male	1.093	0.4886–2.4470	0.828
Weight	1.005	0.9860–1.0236	0.625
HR	1.025	1.0071–1.0425	0.006
SBP	0.952	0.9290–0.9755	<0.001
DBP	0.943	0.9096–0.9769	0.001
RR	0.987	0.9330–1.0451	0.663
T	2.304	1.4689–3.6133	<0.001
Sodium	1.116	1.0290–1.2110	0.008
Potassium	0.862	0.5067–1.4658	0.583
chloride ions	0.805	0.4886–1.3258	0.394
WBC	1.066	1.0030–1.1331	0.040
Platelet	0.997	0.9918–1.0025	0.306
Hemoglobin	0.925	0.7777–1.1009	0.381
Neutrophils	1.233	1.0723–1.3988	0.009
creatinine	0.805	0.4886–1.3258	0.394
BUN	0.989	0.9653–1.0129	0.358
SPO₂	1.041	0.9407–1.1526	0.435
GLU	1.008	1.0001–1.0154	0.045
Lactate	1.586	1.1823–2.1270	0.002
Albumin	0.414	0.2046–0.8387	0.014
NAR	1.809	1.1855–2.4839	0.006
GCS	1.071	1.0373–1.1347	<0.001
SOFA	1.349	1.1967–1.5210	<0.001
Vasopressin	10.276	4.4948–23.4906	<0.001
HBP	0.804	0.3673–1.7610	0.586
Diabetes	1.159	0.5013–2.6798	0.730
COPD	1.547	0.1211–2.4670	0.432
Surgical intervention	1.229	1.1239–2.1136	0.741

Variable	Univariable Cox
Variable	HR	95% CI	p-value Wald’s test
HR	1.019	1.0071–1.0425	0.006
SBP	1.106	0.9290–1.1755	<0.001
DBP	0.817	0.7096–0.9769	0.001
T	2.138	1.4689–3.6133	<0.001
WBC	0.832	0.7003–1.1331	0.040
GLU	0.998	0.9001–1.0154	0.045
Lactate	1.700	1.1823–2.1270	0.002
Neutrophils	1.157	0.9088–1.2234	0.003
Albumin	0.523	0.3345–0.7817	0.014
NAR	3.224	1.3211–4.5943	<0.001
GCS	1.103	1.0373–1.1347	<0.001
SOFA	1.336	1.1967–1.5210	<0.001
Vasopressin	3.661	3.4948–13.4906	<0.001
Surgical intervention	1.154	1.9822–4.33291	0.642