Abstract
Background
The prognostic value of the neutrophil-to-lymphocyte ratio (NLR) in elderly patients with sepsis remains uncertain. This retrospective cohort study investigated the association between NLR at hospital admission and 28-day mortality in this population.
Methods
Data were extracted from the MIMIC-IV database (2008–2019), including 7,522 patients aged ≥65 years diagnosed with sepsis. The primary exposure was NLR measured upon intensive care unit (ICU) admission, and the outcome was 28-day all-cause mortality. Multivariable Cox proportional hazards models and Kaplan–Meier survival analyses were employed to assess the relationship between NLR and mortality risk. Sensitivity analyses were performed to validate the robustness of the results.
Results
Among the 7,522 patients analyzed, each one–standard-deviation increase in NLR (16.29 units) was associated with a 6% higher risk of 28-day mortality (HR: 1.06; 95% CI: 1.03–1.10). When stratified by quartiles, patients in the highest NLR group (>16.0) had a 44% greater mortality risk compared with those in the lowest quartile (HR: 1.44; 95% CI: 1.24–1.68). Survival distributions differed significantly across quartiles (log-rank p < 0.001). However, restricted cubic spline analysis did not demonstrate a statistically significant overall or nonlinear association (P = 0.353 and P = 0.798), and ROC analysis showed modest discrimination (AUC = 0.609; 95% CI: 0.594–0.624).
Conclusions
NLR has a modest association with 28-day mortality in elderly patients with sepsis. Although its predictive value is limited, NLR may serve as an accessible adjunct to established clinical assessments and should be interpreted in conjunction with other clinical indicators.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12879-026-12625-y.
Keywords: Neutrophil-to-lymphocyte ratio, Mortality, Sepsis, Intensive care unit, MIMIC-IV
Introduction
Sepsis remains a major challenge in global critical care, with a disproportionately high incidence and mortality among older adults [1]. In this population, 28-day mortality rates have been reported to range from 30% to 60%, largely due to dysregulated systemic inflammation and age-related immune dysfunction [1, 2].
The NLR, a readily available and cost-effective biomarker of systemic inflammation, has attracted increasing interest for its prognostic value in sepsis. By reflecting both neutrophil-mediated pro-inflammatory responses and lymphocyte-driven immunosuppression, NLR offers an integrative measure of immune-inflammatory dysregulation [3, 4].
Although elevated NLR has been linked to worse outcomes in septic patients, its prognostic significance in elderly populations remains unclear. Some studies have shown that early increases in NLR (within 48–72 h post-diagnosis) are independently associated with higher 28-day mortality [5, 6]. However, other large-scale cohort studies suggest that the predictive performance of NLR may be influenced by underlying comorbidities such as diabetes or chronic kidney disease, as well as by the source of infection [7, 8]. Moreover, in specific subgroups—such as septic patients with lymphopenia—a retrospective cohort study involving 172 such patients reported that an NLR ≥ 18.93 (optimal cut-off determined via receiver operating characteristic [ROC] curve analysis) demonstrated strong predictive accuracy for in-hospital mortality (AUC = 0.750, 95% CI: 0.634–0.788, p < 0.001), thereby highlighting the potential of NLR to inform individualized prognostication [9].
Nevertheless, most prior research has either examined mixed-age cohorts or failed to adequately control for sepsis severity. Given the unique pathophysiological characteristics of elderly patients—including immunosenescence, frailty, and a higher burden of comorbidities—the prognostic utility of NLR in this demographic remains inadequately defined. Therefore, this study aims to investigate the association between admission NLR and 28-day mortality in elderly patients with sepsis, using data from a large, publicly available database. Clarifying this relationship may enhance early risk stratification and clinical decision-making in this vulnerable population.
Methods
Data source
This retrospective cohort analysis utilized data extracted from the Medical Information Mart for Intensive Care IV (MIMIC-IV, version 3.1; https://mimic.mit.edu/iv), a publicly accessible critical care database established by the Massachusetts Institute of Technology. MIMIC-IV includes detailed clinical records of individuals admitted to intensive care units (ICUs) at Beth Israel Deaconess Medical Center between 2008 and 2019, encompassing demographics, physiological measurements, laboratory findings, medical procedures, and patient outcomes [10].
To safeguard privacy, all data in MIMIC-IV were thoroughly de-identified in alignment with the Health Insurance Portability and Accountability Act (HIPAA) standards, eliminating the need for institutional review board approval or informed consent. One author (Zhan-peng Lu) was granted access to the dataset after completing a mandatory training course on human subject research protection (Certification ID: 51480606). This study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) recommendations.
Study population
The study cohort consisted of elderly patients (≥ 65 years) diagnosed with sepsis and admitted to the ICU, as recorded in the MIMIC-IV database. Sepsis was defined according to the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3), which characterizes sepsis as a suspected or documented infection accompanied by an acute increase of ≥ 2 points in the Sequential Organ Failure Assessment (SOFA) score [11]. The following inclusion and exclusion criteria were applied. Inclusion Criteria: First ICU admission during the hospital stay(only the first admission was analyzed in cases of multiple ICU admissions); Age ≥ 65 years at the time of ICU admission. Exclusion Criteria: ICU length of stay < 24 h (n = 1,652); Diagnoses of malignancy, metastatic cancer, hematologic disorders, autoimmune diseases, or human immunodeficiency virus (HIV) infection (n = 2,041); Missing baseline NLR data (n = 4,871). Based on the inclusion criteria, the final cohort consisted of 7,522 older adults diagnosed with sepsis (Fig. 1).
Fig. 1.
Flowchart of patient selection
Data collection
Clinical data were obtained from the MIMIC-IV database using Navicat Premium 15 in conjunction with Structured Query Language (SQL). The extracted variables were subsequently classified as follows: Demographics: Included patient age, sex, race, and body mass index (BMI). Vital Signs: Recorded values comprised body temperature, heart rate (HR), respiratory rate (RR), systolic and diastolic blood pressure (SBP and DBP), as well as peripheral oxygen saturation (SpO₂). Laboratory Findings: Data encompassed white blood cell (WBC) count, hemoglobin levels, platelet count, serum lactate, alanine aminotransferase (ALT), serum creatinine, blood urea nitrogen (BUN), international normalized ratio (INR), blood pH, glucose, and serum concentrations of sodium, potassium, and calcium. Comorbid Conditions: Documented pre-existing conditions included congestive heart failure (CHF), cerebrovascular disease (CVD), chronic obstructive pulmonary disease (COPD), myocardial infarction (MI), liver disease, renal impairment, and diabetes. Organ Support Interventions: Information on the application of mechanical ventilation (MV), administration of vasopressors, use of renal replacement therapy (RRT). Disease severity indicators: Simplified Acute Physiology Score II [SAPS II], Sequential Organ Failure Assessment [SOFA] was obtained.
Exposure variable
The NLR was determined by dividing the absolute neutrophil count by the absolute lymphocyte count measured within the first 24 h after ICU admission. The NLR exhibited a markedly right-skewed distribution. To facilitate comparison of effect sizes across different continuous predictors and to express hazard ratios (HRs) in a scale-independent manner, NLR was standardized using Z-scores (i.e., per SD increase), resulting in the variable NLR.z. For categorical analysis, patients were divided into quartiles based on their NLR levels: Q1, < 4.9; Q2, 4.9–8.8; Q3, 8.8–16.0; and Q4, > 16.0.
Outcome variables
The primary outcome was 28-day all-cause mortality, defined as death from any cause occurring within 28 days after ICU admission, whether during hospitalization or following discharge, as recorded in the MIMIC-IV database.
Statistical analysis
The Shapiro–Wilk test was used to assess the normality of continuous variables. Variables with a normal distribution were expressed as mean ± SD, whereas those with a skewed distribution were reported as median and interquartile range (IQR). Categorical variables were presented as counts and percentages. Differences among groups were evaluated using one-way analysis of variance (ANOVA) or the Kruskal–Wallis rank-sum test for continuous variables, and the χ² test or Fisher’s exact test for categorical variables, as appropriate.
The NLR was analyzed both as a continuous variable (standardized using the Z-score method, denoted as NLRz) and as a categorical variable based on quartiles. The association between NLR and 28-day mortality was examined using multivariable Cox proportional hazards regression models. We first validated the proportional hazards assumption via the Schoenfeld residuals test, which showed no significant deviation (χ² = 0.277, p = 0.599), confirming the assumption holds and NLR’s effect on mortality remains constant during follow-up. Three models were constructed to evaluate the robustness of the findings. Model 1: Adjusted for age and sex. Model 2: Further adjusted for race, BMI, SBP and DBP, RR, HR, temperature, SpO₂, hemoglobin, platelet count, ALT, INR, serum creatinine, BUN, lactate, pH, blood glucose, serum sodium, potassium, calcium, and comorbidities including CHF, CVD, COPD, MI, liver disease, renal disease, and diabetes. Model 3: Additionally adjusted for disease severity indicators (SAPS II and SOFA) and organ support therapies, including MV, RRT, and vasopressor use.
Kaplan–Meier survival curves, accompanied by log-rank tests, were employed to compare cumulative mortality across quartiles of the NLR. Subgroup analyses were performed to evaluate the robustness of these associations across different strata, including age, sex, BMI, comorbid conditions, and severity scores. To elucidate the dose–response relationship between the NLR and 28-day mortality, we employed a restricted cubic spline (RCS) analysis embedded within a fully adjusted multivariable Cox regression model (with NLR truncated at the 0.5th–99.5th percentiles; covariates specified according to Model 3). In addition, ROC curve analysis was performed to evaluate the discriminative ability of NLR and to determine its clinically relevant threshold.
We summarized and assessed missing data for all variables (Supplementary material Table S2). Most variables had no or minimal missingness, while a few laboratory indicators (ALT, BMI, lactate, and pH) showed higher proportions. To reduce bias and efficiency loss from complete-case analysis or single imputation, all variables with missing data were processed using multiple imputation by chained equations (MICE) under the missing at random (MAR) assumption. The imputation model included all covariates from the multivariable Cox regression. Five imputed datasets were generated, analyzed separately using Cox proportional hazards models, and combined with Rubin’s rules to obtain pooled HRs and 95% CIs. All statistical procedures were executed using R software (version 4.1.3) and Free Statistics software (version 2.1). Statistical significance was defined by a two-sided P-value of less than 0.05.
Results
Baseline characteristics of patients
A total of 7,522 elderly patients with sepsis were included in the final analysis (Fig. 1; Table 1). Participants were categorized into four NLR quartiles: Q1 (< 4.9, n = 1,872), Q2 (4.9–8.8, n = 1,879), Q3 (8.8–16.0, n = 1,885), and Q4 (> 16.0, n = 1,886). The mean age was 78.0 ± 8.1 years, and 55.1% (n = 4,145) were male. Patients in Q4 had higher laboratory values than those in Q1, including WBC [18.4 (13.8–24.6) vs. 11.5 (8.6–15.3) ×10⁹/L, p < 0.001], absolute neutrophil count [15.4 (11.2–20.6) vs. 6.3 (4.3–8.7) ×10⁹/L, p < 0.001], serum creatinine (1.5 vs. 1.1 mg/dL, p < 0.001), BUN (34.0 vs. 21.0 mg/dL, p < 0.001), and ALT (30 vs. 24 U/L, p < 0.001). Comorbidities were more prevalent in Q4, including CHF (44.8% vs. 33.4%), COPD (34.2% vs. 25.5%), and renal disease (32.3% vs. 24.5%) (all p < 0.001). Regarding clinical outcomes, patients in Q4 had a higher 28-day mortality rate than those in Q1 (31.9% vs. 15.0%, p < 0.001) and a longer ICU length of stay [4.0 (2.0–7.0) vs. 3.0 (2.0–5.0) days, p < 0.001]. SAPS II (47.4 ± 12.9 vs. 41.0 ± 12.3) and SOFA scores (3.8 ± 2.0 vs. 3.5 ± 1.9) were also higher in Q4 (both p < 0.001), as was the use of renal replacement therapy (12.1% vs. 6.2%, p < 0.001).
Table 1.
Baseline characteristics of the study population according to neutrophil-to-lymphocyte ratio
| Variables | Total (n = 7522) | Neutrophil-to-Lymphocyte Ratio level | P value | ||||
|---|---|---|---|---|---|---|---|
| < 4.9 | 4.9–8.8 | 8.8–16.0 | > 16.0 | ||||
|
Quartile 1
(n = 1872) |
Quartile 2 (n = 1879) |
Quartile 3 (n = 1885) |
Quartile 4 (n = 1886) |
||||
| NLR | 8.9 (5.0, 16.1) | 3.3 (2.5, 4.1) | 6.7 (5.8, 7.7) | 11.7 (10.1, 13.6) | 25.0 (19.4, 36.1) | < 0.001 | |
| Age (years) | 78.0 ± 8.1 | 77.4 ± 8.0 | 77.4 ± 7.9 | 78.5 ± 8.3 | 78.7 ± 8.2 | < 0.001 | |
| Sex Male (n, %) | 0.503 | ||||||
| Female | 3377 (44.9) | 858 (45.8) | 819 (43.6) | 841 (44.6) | 859 (45.5) | ||
| Male | 4145 (55.1) | 1014 (54.2) | 1060 (56.4) | 1044 (55.4) | 1027 (54.5) | ||
| Race(n, %) | < 0.001 | ||||||
| White | 5073 (67.4) | 1211 (64.7) | 1290 (68.7) | 1251 (66.4) | 1321 (70) | ||
| Black | 571 ( 7.6) | 203 (10.8) | 144 (7.7) | 143 (7.6) | 81 (4.3) | ||
| Others | 1878 (25.0) | 458 (24.5) | 445 (23.7) | 491 (26) | 484 (25.7) | ||
| BMI (kg/m2) | 28.3 ± 7.1 | 28.1 ± 6.4 | 28.6 ± 7.2 | 28.2 ± 7.2 | 28.1 ± 7.7 | 0.064 | |
| SBP (mmHg) | 116.1 ± 15.0 | 116.6 ± 13.9 | 116.6 ± 14.9 | 116.9 ± 16.2 | 114.3 ± 14.7 | < 0.001 | |
| DBP (mmHg) | 59.5 ± 9.7 | 59.0 ± 9.5 | 59.4 ± 9.4 | 59.6 ± 10.2 | 59.7 ± 9.8 | 0.125 | |
| RR (breaths/min) | 19.8 ± 3.9 | 18.8 ± 3.6 | 19.4 ± 3.6 | 20.2 ± 4.0 | 20.9 ± 4.0 | < 0.001 | |
| HR (beats/min) | 83.9 ± 15.6 | 80.9 ± 14.1 | 83.1 ± 14.5 | 84.0 ± 16.0 | 87.6 ± 16.9 | < 0.001 | |
| Temperature (◦C) | 36.8 ± 0.6 | 36.7 ± 0.5 | 36.8 ± 0.5 | 36.8 ± 0.6 | 36.8 ± 0.6 | < 0.001 | |
| SpO2(%) | 96.9 ± 2.2 | 97.3 ± 2.0 | 97.1 ± 2.0 | 96.8 ± 2.3 | 96.5 ± 2.2 | < 0.001 | |
| WBC(×109/L) | 14.4 (10.5, 19.2) | 11.5 (8.6, 15.3) | 13.4 (10.0, 17.6) | 14.9 (11.2, 19.1) | 18.4 (13.8, 24.6) | < 0.001 | |
| Hemoglobin(g/dL) | 9.9 ± 2.1 | 9.6 ± 2.0 | 9.8 ± 2.1 | 10.0 ± 2.2 | 10.2 ± 2.2 | < 0.001 | |
| Platelets(×109/L) | 182.0 (133.0, 247.0) | 161.0 (121.5, 220.0) | 178.0 (133.0, 239.0) | 195.0 (146.0, 265.0) | 194.0 (139.0, 270.0) | < 0.001 | |
| ALT (U/L) | 26.0 (16.0, 55.8) | 24.0 (15.0, 46.0) | 24.0 (15.0, 47.0) | 27.0 (16.0, 65.0) | 30.0 (17.0, 77.0) | < 0.001 | |
| INR | 1.4 (1.2, 1.7) | 1.4 (1.2, 1.6) | 1.4 (1.2, 1.7) | 1.4 (1.2, 1.7) | 1.4 (1.2, 1.9) | < 0.001 | |
| Creatinine (mg/dL) | 1.3 (0.9, 2.0) | 1.1 (0.8, 1.6) | 1.2 (0.9, 1.8) | 1.4 (0.9, 2.2) | 1.5 (1.0, 2.4) | < 0.001 | |
| BUN (mg/dL) | 27.0 (18.0, 44.0) | 21.0 (15.0, 32.0) | 25.0 (17.0, 41.0) | 30.0 (20.0, 48.0) | 34.0 (22.0, 54.0) | < 0.001 | |
| Lactate (mmol/L) | 2.1 (1.4, 3.3) | 2.3 (1.6, 3.4) | 2.1 (1.4, 3.2) | 2.0 (1.4, 3.3) | 2.1 (1.4, 3.3) | < 0.001 | |
| pH | 7.3 ± 0.1 | 7.3 ± 0.1 | 7.3 ± 0.1 | 7.3 ± 0.1 | 7.3 ± 0.1 | 0.013 | |
| Glucose (mg/dL) | 134.8 (117.2, 165.7) | 129.5 (117.0, 151.8) | 133.3 (118.0, 157.3) | 140.3 (117.8, 175.0) | 140.4 (115.8, 176.8) | < 0.001 | |
| Sodium (mmol/L) | 140.5 ± 5.6 | 140.6 ± 5.0 | 140.6 ± 5.7 | 140.6 ± 5.9 | 140.2 ± 5.7 | 0.152 | |
| Potassium (mmol/L) | 4.7 ± 0.9 | 4.7 ± 0.9 | 4.7 ± 0.9 | 4.8 ± 0.9 | 4.8 ± 0.9 | 0.026 | |
| Calcium (mg/dL) | 8.0 ± 0.9 | 8.1 ± 0.8 | 8.1 ± 0.8 | 8.0 ± 0.9 | 7.9 ± 0.8 | < 0.001 | |
| Lymphocytes (×109/L) | 1.1 (0.7, 1.7) | 2.0 (1.4, 2.7) | 1.4 (1.0, 1.8) | 1.0 (0.7, 1.3) | 0.6 (0.4, 0.8) | < 0.001 | |
| Neutrophils (×109/L) | 10.1 (6.9, 14.5) | 6.3 (4.3, 8.7) | 9.2 (6.9, 12.2) | 11.4 (8.6, 14.8) | 15.4 (11.2, 20.6) | < 0.001 | |
| CHF | 3031 (40.3) | 625 (33.4) | 738 (39.3) | 824 (43.7) | 844 (44.8) | < 0.001 | |
| CVD | 1407 (18.7) | 370 (19.8) | 382 (20.3) | 370 (19.6) | 285 (15.1) | < 0.001 | |
| COPD | 2204 (29.3) | 478 (25.5) | 532 (28.3) | 549 (29.1) | 645 (34.2) | < 0.001 | |
| MI | 1777 (23.6) | 407 (21.7) | 424 (22.6) | 484 (25.7) | 462 (24.5) | 0.018 | |
| Liver disease | 258 ( 3.4) | 62 (3.3) | 60 (3.2) | 58 (3.1) | 78 (4.1) | 0.268 | |
| Renal disease | 2196 (29.2) | 459 (24.5) | 540 (28.7) | 587 (31.1) | 610 (32.3) | < 0.001 | |
| Diabetes | 2754 (36.6) | 673 (36) | 704 (37.5) | 713 (37.8) | 664 (35.2) | 0.29 | |
| SAPS II | 43.9 ± 12.8 | 41.0 ± 12.3 | 42.3 ± 12.0 | 44.8 ± 12.8 | 47.4 ± 12.9 | < 0.001 | |
| SOFA | 3.7 ± 1.9 | 3.5 ± 1.9 | 3.6 ± 1.9 | 3.6 ± 2.0 | 3.8 ± 2.0 | < 0.001 | |
| MV | 4107 (54.6) | 1065 (56.9) | 1061 (56.5) | 1030 (54.6) | 951 (50.4) | < 0.001 | |
| RRT | 705 ( 9.4) | 116 (6.2) | 164 (8.7) | 196 (10.4) | 229 (12.1) | < 0.001 | |
| Vasopressor use | 3899 (51.8) | 1037 (55.4) | 1001 (53.3) | 904 (48) | 957 (50.7) | < 0.001 | |
| LOS hospital, day | 9.0 (5.0, 15.0) | 8.0 (5.0, 12.0) | 9.0 (5.0, 15.0) | 9.0 (6.0, 15.0) | 9.0 (6.0, 16.0) | < 0.001 | |
| LOS ICU, day | 3.0 (2.0, 7.0) | 3.0 (2.0, 5.0) | 3.0 (2.0, 7.0) | 4.0 (2.0, 7.0) | 4.0 (2.0, 7.0) | < 0.001 | |
| 28-day mortality, n (%) | 1706 (22.7) | 281 (15) | 343 (18.3) | 481 (25.5) | 601 (31.9) | < 0.001 | |
Abbreviations: NLR, Neutrophil-to-Lymphocyte ratio; BMI, body mass Index; SBP, systolic blood pressure; DBP, diastolic blood pressure; RR, respiratory rate; HR, heart rate; Spo2, peripheral capillary oxygen Saturation; WBC, white blood cells; ALT, Alanine transaminase; INR, international normalized ratio; BUN, blood Urea nitrogen; pH, potential of Hydrogen; CHF, congestive heart failure; CVD, cerebrovascular disease; COPD, chronic obstructive pulmonary disease; MI, myocardial infarction; SAPS II, simplified acute physiology score II; SOFA, sequential organ failure assessment; MV, mechanical ventilation; RRT, renal replacement therapy; LOS, length of stay
Multivariable Cox regression analysis
Three Cox proportional hazards models were constructed to examine the association between NLR and 28-day mortality in elderly patients with sepsis (Table 2). In the baseline model adjusted for age and sex (Model 1), each one–SD increase in NLR (16.29 units) was associated with a higher risk of 28-day mortality (HR = 1.14, 95% CI: 1.11–1.18, p < 0.001). After additional adjustment for laboratory variables and comorbidities (Model 2), the association remained statistically significant (HR = 1.08, 95% CI: 1.04–1.12, p < 0.001). In the fully adjusted model incorporating disease-severity indicators and organ support therapies (Model 3), the hazard ratio was 1.06 (95% CI: 1.03–1.10, p = 0.001). When NLR was categorized into quartiles, Model 3 showed no statistically significant difference in mortality between Q2 (4.9–8.8) and Q1 (< 4.9) (HR = 1.12, 95% CI: 0.95–1.31, p = 0.176). Higher mortality risks were observed in Q3 (8.8–16.0) (HR = 1.32, 95% CI: 1.13–1.54, p < 0.001) and Q4 (> 16.0) (HR = 1.44, 95% CI: 1.24–1.68, p < 0.001). The trend test indicated a statistically significant increase in 28-day mortality across NLR quartiles (p < 0.001).
Table 2.
Multivariate cox regression for neutrophil-to-lymphocyte ratio on all-cause mortality of the elderly patients with sepsis
| Variable | Model 1 | Model 2 | Model 3 | |||
|---|---|---|---|---|---|---|
| HR (95% CI) | P value | HR (95% CI) | P value | HR (95% CI) | P value | |
| Continuous variable | ||||||
| NLR.z | 1.14 (1.11 ~ 1.18) | < 0.001 | 1.08 (1.04 ~ 1.12) | < 0.001 | 1.06 (1.03 ~ 1.1) | 0.001 |
| Binary variable | ||||||
| Quartile 1(<4.9) | 1(Ref) | 1(Ref) | 1(Ref) | |||
| Quartile 2(4.9–8.8) | 1.23 (1.05 ~ 1.44) | 0.009 | 1.14 (0.97 ~ 1.33) | 0.12 | 1.12 (0.95 ~ 1.31) | 0.176 |
| Quartile 3(8.8–16.0) | 1.81 (1.56 ~ 2.1) | < 0.001 | 1.4 (1.2 ~ 1.63) | < 0.001 | 1.32 (1.13 ~ 1.54) | < 0.001 |
| Quartile 4(>16.0) | 2.34 (2.03 ~ 2.7) | < 0.001 | 1.56 (1.34 ~ 1.81) | < 0.001 | 1.44 (1.24 ~ 1.68) | < 0.001 |
| Trend.test | < 0.001 | < 0.001 | < 0.001 | |||
Notes: Model 1: Adjusted for age and gender Model 2: Model 1 + race, BMI, SBP, DBP, RR, HR, temperature, SpO2, hemoglobin, platelets, ALT, INR, creatinine, BUN, lactate, pH, glucose, sodium, potassium, calcium, CHF, CVD, COPD, MI, liver disease, renal disease, diabetes Model 3: Model 2 + SAPS II, SOFA, MV, RRT, vasopressor use
Abbreviations: NLR.z: Z-standardized value of the Neutrophil-to-Lymphocyte Ratio (NLR). BMI, Body Mass Index; SBP, systolic blood pressure; DBP, diastolic blood pressure; RR, respiratory rate; HR, heart rate; SpO2, Peripheral Capillary Oxygen Saturation; ALT, alanine transaminase; INR, international normalized ratio; BUN, blood urea nitrogen; pH, Potential of Hydrogen; CHF, congestive heart failure; CVD, cerebrovascular disease; COPD, chronic obstructive pulmonary disease; MI, myocardial infarction; SAPS II, simplified acute physiology score II; SOFA, sequential organ failure assessment; MV, mechanical ventilation; RRT, renal replacement therapy
Kaplan–meier survival curve analysis
Figure 2 shows the Kaplan–Meier survival curves stratified by NLR quartiles. The survival distributions differed significantly among the quartile groups according to the log-rank test (p < 0.0001). Patients in the highest quartile (Q4, NLR > 16.0) had the lowest survival probability, whereas those in the lowest quartile (Q1) showed the highest survival probability. A progressively lower survival probability was observed across increasing NLR quartiles.
Fig. 2.
Kaplan–Meier survival curve for the cumulative hazard of 28-day mortality
Subgroup analyses
Subgroup analyses were performed to examine the stability of the association between NLR and 28-day mortality across different clinical strata (Fig. 3). Although the hazard ratios varied slightly by age, sex, disease severity, and diabetes status, none of the interaction terms reached statistical significance (all p for interaction > 0.05). These findings indicate that the association between elevated NLR and increased mortality risk was broadly consistent across subgroups, suggesting that age, comorbidity, or severity did not materially modify this relationship.
Fig. 3.
Subgroup analysis of the association between NLR and 28-day mortality in elderly septic patients
NLR cutoffs and linear associations
In the fully adjusted multivariable Cox model, restricted cubic spline (RCS) analysis showed the estimated relationship between NLR and 28-day mortality (Supplementary Figure S1). Restricted cubic spline analysis demonstrated a visually linear increasing trend of NLR with 28-day mortality risk; however, neither the overall association (P = 0.353) nor the nonlinear component (P = 0.798) reached statistical significance, indicating a modest incremental effect. A ROC curve analysis was conducted to assess discrimination (Supplementary Figure S2). The area under the curve was 0.609 (95% CI: 0.594–0.624). The optimal cutoff value based on the Youden index was approximately 0.217, with a sensitivity of 62.5% and specificity of 55.5%.
Sensitivity analysis
To assess the robustness of the association between NLR and 28-day mortality, sensitivity analyses were conducted using different NLR transformations. In addition to the primary analysis based on NLR.z, we evaluated log-transformed NLR [log (NLR + 1)] and the original-scale NLR. As summarized in Supplementary material Table S1, all three approaches yielded consistent results, with elevated NLR associated with increased 28-day mortality. These findings confirm that the association was robust across alternative variable transformations. To assess the impact of missing data, we performed Cox regression analyses using both the complete-case dataset and the multiple-imputation dataset (Supplementary material Table S3). The HR from the complete-case analysis was 1.03 (95% CI: 0.97–1.09, p = 0.333), while the adjusted HR after multiple imputation was 1.06 (95% CI: 1.03–1.10, p = 0.001). In both analyses, the direction of the association between the NLR.z and 28-day mortality remained consistent, which supports the robustness of our findings.
Discussion
In this retrospective cohort of 7,522 elderly patients with sepsis, NLR measured within 24 h of ICU admission showed a statistically significant yet modest association with 28-day all-cause mortality. After adjusting for demographics, comorbidities, and validated severity scores, patients in the highest NLR quartile (> 16.0) had a 44% higher risk of death compared with those in the lowest quartile, while each one–standard-deviation increase in NLR corresponded to only a 6% elevation in mortality risk. The attenuation of the hazard ratio from 1.14 in the crude model to 1.06 in the fully adjusted model suggests that illness severity and comorbidity burden account for much of the association. These findings therefore indicate a limited, rather than strong, independent prognostic contribution of NLR. Kaplan–Meier curves also demonstrated a graded but modest increase in mortality across NLR quartiles.
Our findings are directionally consistent with prior studies reporting associations between elevated NLR and adverse outcomes in sepsis, while providing additional insights into elderly patients specifically. Lorente et al. observed that persistently elevated NLR during the first week of ICU stay was associated with higher 30-day mortality [3]. and other MIMIC-IV analyses have identified NLR as a risk indicator in mixed-age septic cohorts [12]. Although our results align with these observations [13, 14], the smaller effect sizes in this study highlight the limited independent prognostic value of NLR in older adults. Prior investigations in elderly subgroups—such as those with diabetes [15] or malignancies [16]—have likewise reported associations but were constrained by smaller sample sizes. Our large, elderly-specific cohort extends this literature by showing that the association persists but is modest. Although NLR categorized by quartiles revealed a stepwise increase in mortality risk, the limited hazard ratios and absence of statistical significance in spline analyses indicate that this pattern does not represent a strong dose–response relationship. Accordingly, NLR may be better conceptualized as an adjunct marker that reflects underlying inflammatory burden and complements established severity assessments. Alternative inflammatory ratios such as the neutrophil-to-monocyte ratio [17] or neutrophil-to-platelet ratio [12] have been explored, but our results suggest that NLR’s main contribution in elderly sepsis lies in its simplicity and contextual value within multidimensional clinical evaluations.
An elevated NLR reflects the combined effects of neutrophil-driven inflammation and lymphocyte-mediated immunosuppression, both of which play central roles in sepsis pathophysiology. Neutrophilia suggests heightened innate immune activation, whereas lymphopenia indicates impaired adaptive immunity and immune exhaustion [18, 19]. In elderly individuals, immunosenescence further disrupts these responses by diminishing lymphocyte function, promoting chronic low-grade inflammation, and increasing susceptibility to secondary infections and organ dysfunction [20–22]. This immunologic milieu—often described as “inflammaging”—is characterized by pro-inflammatory yet functionally compromised immune cell populations, including monocytes, dendritic cells, neutrophils, and lymphocyte subsets [23–25]. While these mechanisms help explain why NLR is frequently abnormal in elderly septic patients, they also underscore why its prognostic performance is modest: when multiple immune pathways are simultaneously dysregulated, no single biomarker can fully capture the complexity of host responses. Thus, although NLR reflects important aspects of inflammation and immunosuppression, its predictive capacity is inherently constrained in the context of widespread immune exhaustion. More integrative approaches—such as multi-omics profiling, immune phenotyping, or dynamic biomarker trajectories—may be necessary to improve risk stratification in this population.
The primary strength of this study is the use of a large, real-world ICU cohort, which enabled comprehensive multivariable and stratified analyses. By focusing specifically on elderly patients, the study addresses a key evidence gap for this high-risk population. Although higher NLR levels were associated with increased 28-day mortality, the modest effect size indicates limited prognostic utility when applied in isolation. NLR is therefore most appropriately interpreted as a supplementary indicator of inflammatory burden rather than as a standalone prognostic tool.
RCS analysis did not demonstrate a statistically significant overall or nonlinear association between NLR and 28-day mortality, indicating that the incremental prognostic contribution of NLR in this elderly population is modest. This finding suggests that, although higher NLR values reflect a heightened inflammatory burden, their independent predictive value is limited when comprehensive adjustment for disease severity and comorbidities is applied. Age-related immunosenescence and multimorbidity likely introduce substantial heterogeneity in immune responses, which may attenuate the prognostic performance of single inflammatory biomarkers such as NLR. Consistent with this interpretation, ROC analysis showed weak discriminatory ability, further supporting the view that NLR should be interpreted as a complementary rather than a standalone prognostic indicator. Nevertheless, the accessibility and biological relevance of NLR underscore its potential utility within more comprehensive predictive frameworks. Future studies may improve risk stratification by integrating NLR with established severity scores and by incorporating dynamic (serial) measurements, rather than relying on a single baseline value. Several limitations warrant consideration. First, the retrospective design may introduce selection bias and residual confounding. Factors such as frailty, nutritional status, and prior immunomodulatory or corticosteroid therapy—each of which could influence NLR and outcomes—were not available. Second, reliance on a single-center database (MIMIC-IV) may limit generalizability. Third, NLR was measured only once within the first 24 h of ICU admission; given that NLR is a dynamic biomarker, its temporal profile may offer additional prognostic insight. Future prospective, multicenter studies incorporating serial NLR measurements and more comprehensive immune and nutritional profiling are needed to further clarify the prognostic relevance of NLR in elderly patients with sepsis.
Conclusions
NLR shows a modest association with 28-day mortality in elderly patients with sepsis. Although its predictive value is limited, its accessibility and reflection of inflammatory burden make it a useful adjunct to established clinical assessments. NLR should be interpreted alongside comprehensive clinical indicators and other biomarkers to improve risk stratification in this vulnerable population.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
We deeply appreciate the contributions of all participants in this cohort study. We extend our heartfelt thanks to the Laboratory for Computational Physiology at the Massachusetts Institute of Technology (LCP-MIT) for sharing the raw data. We are also grateful to the Free Statistics team for their technical support and for supplying valuable tools for data analysis and visualization.
Author contributions
LZ and HL contributed to the study design as well as the initial data analysis and interpretation. They were responsible for the management, data retrieval, and initial manuscript draft. WY and HF conducted the data analysis, FL reviewed the manuscript. All authors contributed to the manuscript and approved the submitted version.
Funding
This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.
Data availability
All the data used to support this study are available from the corresponding author upon request.
Declarations
Compliance with ethics guidelines
This study was performed in accordance with the Helsinki Declaration of 1964 and its later amendments. MIMIC-IV is an anonymized public database. The project was approved by the institutional review boards of the Massachusetts Institute of Technology (MIT) and Beth Israel Deaconess Medical Center (BIDMC) and was given a waiver of informed consent.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Zhan-peng Lu and Yu-Chen Wang equally to this work and share first authorship.
References
- 1.Ibarz M, Haas LE, Ceccato A, Artigas A. The critically ill older patient with sepsis: a narrative review. Ann Intensiv Care. 2024;14(1):6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Pérez-González A, Almudí-Ceinos D, Del López O, Martín-Alfonso S, Rico-Feijoo J, Del López J et al. Is mortality in elderly septic patients as high as expected? Long-term mortality in a surgical sample cohort. (2173–5727 (Electronic)). [DOI] [PubMed]
- 3.Lorente L, Martín MM, Ortiz-López R, Alvarez-Castillo A, Ruiz C, Uribe L et al. Association between neutrophil-to-lymphocyte ratio in the first seven days of sepsis and mortality. LID - S0213-005X(20)30405-5 [pii] LID – 10.1016/j.eimc.2020.11.004IF: 2.5 Q2 B4 [doi]. (2529-993X (Electronic)).
- 4.Kourilovitch M, Galarza-Maldonado C. Could a simple biomarker as neutrophil-to-lymphocyte ratio reflect complex processes orchestrated by neutrophils? (2589–9090 (Electronic)). [DOI] [PMC free article] [PubMed]
- 5.Hwang SY, Shin TG, Jo IJ, Jeon K, Suh GY, Lee TR et al. Neutrophil-to-lymphocyte ratio as a prognostic marker in critically-ill septic patients. (1532–8171 (Electronic)). [DOI] [PubMed]
- 6.Singhal A, Dubey S, Khan S, Tiwari R, Das S, Ahmad R. Neutrophil-to-lymphocyte ratio and procalcitonin in sepsis patients: do they have any prognostic significance? (2168–8184 (Print)). [DOI] [PMC free article] [PubMed]
- 7.Gao JL, Shen J, Yang LP, Liu L, Zhao K, Pan XR et al. Neutrophil-to-lymphocyte ratio associated with renal function in type 2 diabetic patients. (2307–8960 (Print)). [DOI] [PMC free article] [PubMed]
- 8.Xiao W, Lu Z, Liu Y, Hua T, Zhang J, Hu J et al. Influence of the initial neutrophils to lymphocytes and platelets ratio on the incidence and severity of sepsis-associated acute kidney injury: A double robust Estimation based on a large public database. (1664–3224 (Electronic)). [DOI] [PMC free article] [PubMed]
- 9.Qiu X, Wang Q, Zhang Y, Zhao Q, Jiang Z, Zhou L. Prognostic value of neutrophils-to-lymphocytes ratio and platelets-to-lymphocytes ratio in sepsis patients with lymphopenia. Biomark Insights. 2024;19:11772719231223156. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Johnson AEW, Bulgarelli LA-O, Shen L, Gayles A, Shammout A, Horng SA-O et al. MIMIC-IV, a freely accessible electronic health record dataset. (2052–4463 (Electronic)). [DOI] [PMC free article] [PubMed]
- 11.Evans L, Rhodes A, Alhazzani W, Antonelli M, Coopersmith CM, French C et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021. (1530 – 0293 (Electronic)). [DOI] [PubMed]
- 12.Zhang Y, Peng W, Zheng X. The prognostic value of the combined neutrophil-to-lymphocyte ratio (NLR) and neutrophil-to-platelet ratio (NPR) in sepsis. (2045–2322 (Electronic)). [DOI] [PMC free article] [PubMed]
- 13.Li W, Ai X, Ni Y, Ye Z, Liang Z. The association between the Neutrophil-to-Lymphocyte ratio and mortality in patients with acute respiratory distress syndrome: A retrospective cohort study. (1540 – 0514 (Electronic)). [DOI] [PubMed]
- 14.Krishnan A, Hsu J, Ha JS, Broderick SR, Shah PD, Higgins RS et al. Elevated neutrophil to lymphocyte ratio is associated with poor long-term survival and graft failure after lung transplantation. (1879–83 (Electronic)). [DOI] [PubMed]
- 15.Jin S, Yin JB, Li W, Zang LL. Effect of neutrophil to lymphocyte ratio on prognosis of elderly patients with severe sepsis combined with diabetes mellitus. (1471–2318 (Electronic)). [DOI] [PMC free article] [PubMed]
- 16.Wang C-Y, Li X-L, Ma X-L, Yang X-F, Liu Y-Y, Yu Y-J. Preoperative neutrophil-to-lymphocyte ratio predicts symptomatic anastomotic leakage in elderly colon cancer patients: multicenter propensity score-matched analysis. World J Gastrointest Surg. 2024;16(2):438. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Xia Y, Xu H, Xie J, Niu H, Cai X, Zhan F et al. Prognostic value of neutrophil-to-monocyte/lymphocyte ratio for 28-day mortality in ICU sepsis patients: a retrospective cohort study. (2296-858X (Print)). [DOI] [PMC free article] [PubMed]
- 18.Zhang H, Wang Y, Qu M, Li W, Wu D, Cata JP et al. Neutrophil, neutrophil extracellular traps and endothelial cell dysfunction in sepsis. (2001 – 1326 (Electronic)). [DOI] [PMC free article] [PubMed]
- 19.Yang Q, Langston JC, Prosniak R, Pettigrew S, Zhao H, Perez E et al. Distinct functional neutrophil phenotypes in sepsis patients correlate with disease severity. (1664–3224 (Electronic)). [DOI] [PMC free article] [PubMed]
- 20.Margalin B, Arfijanto MV, Hadi U. Effector function and neutrophil cell death in the severity of sepsis with diabetes mellitus. (2807 – 2618 (Electronic)). [DOI] [PMC free article] [PubMed]
- 21.Zhu CL, Wang Y, Ren SC, Fau - Sun YC, Sun Xy Fau - Liu X-Y. Z-L, Liu Zl Fau - Li Q-Q, The delivery of PD-l1 sirna by neutrophil-targeted lipid nanoparticles effectively ameliorates sepsis. (1540 – 0514 (Electronic)). [DOI] [PubMed]
- 22.Pearce J, Hadcocks L, Mansour S, van Zanten M, Jeffery S, Gordon K, et al. Profound and selective lymphopaenia in primary lymphatic anomaly patients demonstrates the significance of lymphatic-lymphocyte interactions. Front Immunol. 2023;14:1279077. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.He W, Yao C, Wang K, Duan Z, Wang S, Xie L. Single-cell landscape of immunological responses in elderly patients with sepsis. Immun Ageing. 2024;21(1):40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.He W, Xiao K, Fang M, Xie L. Immune cell Number, Phenotype, and function in the elderly with sepsis. Aging Dis. 2021;12(1):277–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Wang Y, Zhang H, Miao C. Unraveling Immunosenescence in sepsis: from cellular mechanisms to therapeutics. Cell Death Dis. 2025;16(1):393. [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.
Supplementary Materials
Data Availability Statement
All the data used to support this study are available from the corresponding author upon request.