Neutrophil-to-lymphocyte ratio predicts mortality for sepsis-induced coagulopathy: a retrospective study

Abstract

Background

Sepsis often leads to coagulation dysfunction, potentially progressing to disseminated intravascular coagulation (DIC), leading to higher mortality. Sepsis-induced coagulopathy (SIC), an early stage of DIC, benefits from early detection and intervention. Neutrophil-to-lymphocyte ratio (NLR) predicts mortality in inflammatory diseases, but its link to SIC mortality is unclear. This study explored NLR’s relevance to clinical outcomes in SIC patients.

Methods

SIC patients from the MIMIC-IV database were divided into four groups by NLR quartiles. The primary outcome was 28-day mortality, while secondary outcomes included 90-day and 1-year mortality, intensive care unit (ICU) and hospital mortality. The association between NLR and clinical outcomes in SIC patients was analyzed using Kaplan–Meier, Cox proportional hazards regression analysis, and restricted cubic spline (RCS) analysis.

Results

Among 5323 SIC patients (60.49% male), the median NLR was 7.64 (IQR: 4.30–14.82). The 28-day mortality rate was 16.96% (903), with a higher NLR in non-survivors (12.01 vs. 7.02, P < 0.001), and with a difference of 4.99 (95% CI 3.38–4.62; Z = 13.5). Kaplan–Meier curves showed that higher NLR was linked to increased short- and long-term mortality. Adjusted Cox proportional hazards regression analysis showed that elevated NLR was significantly associated with increased 28-day mortality (HR: 1.36; 95% CI 1.16–1.59; P < 0.001). RCS analysis indicated a J-shaped relationship between NLR and mortality, with the lowest risk at the second quartile (Q2). ROC analysis showed that NLR had modest predictive value for mortality risk, with an AUC of 0.642 (95% CI 0.622–0.663) for 28-day mortality.

Conclusion

NLR is significantly associated with mortality in SIC patients and can predict short-and long-term mortality. This finding suggests that the NLR may be useful in identifying patients with SIC at high risk of mortality. Elevated NLR is significantly associated with increased mortality in SIC patients and can predict short-and long-term mortality. The relationship between NLR and mortality risk is J-shaped, with the lowest risk observed in the second quartile. While NLR demonstrates modest predictive value, its accessibility and ease of use make it a potentially valuable tool for early risk stratification and informing clinical decision-making in SIC patients, especially in emergencies or resource-limited settings.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40001-025-03428-z.

Introduction

Sepsis is a complex syndrome of life-threatening organ dysfunction caused by a dysregulated host response to infection [1], and remains a major cause of health damage worldwide [2]. About 70% of sepsis causes coagulopathy [3], and of these patients, about 35% are likely to develop disseminated intravascular coagulation (DIC) [4, 5], exacerbating organ failure and mortality [6]. Although therapeutic treatments for sepsis-induced DIC have been explored in recent years, its curative effect is not satisfactory [7]. However, early detection and appropriate management of DIC will help to improve sepsis outcomes [8]. The concept of coagulopathy induced by sepsis has therefore been proposed.

Sepsis-induced coagulopathy (SIC) is vascular endothelial cell damage and coagulation disorders designed to identify patients with early coagulation dysfunction whose coagulation status is still in a reversible stage. The proposal of this concept greatly improves the diagnostic efficacy of DIC [9]. Current guidelines encourage early detection and treatment with anticoagulants based on this, improving prognosis [10, 11]. However, although the current scoring system greatly improves the diagnosis of SIC, there is still no effective judgment on the mortality of SIC. Compared with the blood routine examination, the timeliness of coagulation-related tests has some shortcomings. Meanwhile, currently widely used indicators such as D-dimer and fibrinogen level have some limitations. Therefore, it is important to find other potential biomarkers for early prediction of SIC prognosis to facilitate the exploration of more personalized management and treatment options [12]. However, no indicators are specifically designed to predict the prognosis of SIC so far [13].

The neutrophil-to-lymphocyte ratio (NLR), derived from neutrophil and lymphocyte counts, is considered a marker of systemic inflammation and stress in critical illnesses [14]. Additionally, an elevated NLR value is associated with increased levels of cytokines and C-reactive protein. Studies reported that NLR is more reliable than predicting neutrophil count or lymphocyte count alone in predicting patient survival [15]. Moreover, NLR is a cost-effective parameter in emergencies.

Critical reviews of existing literature highlight the association between elevated NLR and adverse sepsis outcomes. However, heterogeneity in study designs and endpoints can complicate the direct application of these findings to SIC. For instance, variations in patient populations and the lack of SIC-specific analysis limit the understanding of NLR’s predictive value in this specific context [16, 17].

Here, we retrospectively analyzed the relationship between NLR and mortality in patients with SIC from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database, to explore the association between NLR level and mortality in SIC patients. By focusing on this subset of sepsis patients, we may identify a valuable tool for early risk stratification, potentially leading to improved management and outcomes.

Methods

Study population

MIMIC-IV-2.2 is a freely accessible database of more than 50,000 ICU admissions in Beth Israel Deaconess Medical Center in Boston, Massachusetts between 2008 and 2019. The current survey utilizes health-related data from the MIMIC-IV (version 2.2) database, a widely used database developed and supervised by the MIT Computational Physiology Laboratory [18]. One author complied with all access requirements and performed data extraction (Certification ID: 58045733). The study flow chart is shown in Supplementary Figure S1.

The International Society on Thrombosis and Haemostasis (ISTH) has established a simplified diagnostic criterion for SIC, comprising three parameters: the Sequential Organ Failure Assessment (SOFA) score, platelet count (PLT), and International Normalized Ratio (INR) [19]. To avoid redundant scoring, the SOFA score in this context includes only four organ systems: cardiovascular, respiratory, hepatic, and renal systems, and is used to confirm the presence of sepsis as defined by the Sepsis-3.0 criteria. Notably, if the SOFA score is ≥ 2, it is capped at 2 points for the purpose of SIC diagnosis. A diagnosis of SIC is considered when the cumulative score from the platelet count, INR, and SOFA score reaches or exceeds 4 points.

The inclusion criteria for this study were: (1) patients with SIC diagnostic criteria in ICU within 24 h: conform to sepsis 3.0 and SIC total score ≥ 4, as shown in Table 1; (2) admission to ICU for the first time.

Table 1.

Scoring for sepsis-induced coagulopathy

Category	Parameter	0 point (n)	1 point (n)	2 points (n)
Coagulation	Platelet count (109/L)	≥ 150 (n = 2651)	≥ 100, < 150 (n = 1586)	< 100 (n = 1086)
Prothrombin time	INR	≤ 1.2 (n = 1218)	> 1.2, ≤ 1.4 (n = 1515)	> 1.4 (n = 2590)
Total SOFA	SOFA score	0 (n = 296)	1 (n = 838)	≥ 2 (n = 4189)
Total score for SIC		≥ 4

INR international normalized ratio; SIC sepsis-induced coagulopathy

Total SIC score is 4 or more with the sum of total SOFA score, coagulation and prothrombin time criteria. Total SOFA score is the sum of four items (respiratory SOFA, cardiovascular SOFA, hepatic SOFA, and renal SOFA). The score of total SOFA is defined as 2 if the total score exceeds 2. The n in parentheses represents the distribution of patients.

The exclusion criteria were as follows: (1) less than 18 years old; (2) ICU stay < 24 h; (3) with malignant tumor; (4) severe hepatic and renal dysfunction; (5) use of anticoagulants (refers to patients either on chronic oral anticoagulants prior to admission or those newly treated during hospitalization); (6) hypercoagulable state (excluding patients with known hereditary thrombophilia such as factor V Leiden mutation, or a history of venous thromboembolism to differentiate non-sepsis-related hypercoagulable conditions); (7) pregnant; (8) missing neutrophil or lymphocyte count.

Data extraction

The information extraction process utilizes the PostgreSQL (version 16) software and the Navicat Premium (version 16), where the Structured Query Language (SQL) is the extraction tool.

Potential variables are classified as:

1. Demographic variables, including age, race, gender, and body mass index (BMI).

2. Complications, including hypertension, diabetes, hyperlipidemia, myocardial infarction (MI), asthma, and chronic obstructive pulmonary disease (COPD).

3. Vital signs and indicators, including heart rate (HR), respiratory rate (RR), mean arterial pressure (MAP), oxygen saturation (SpO₂) and temperature (T).

4. Laboratory indicators, including platelets, international normalized ratio (INR), red blood cells (RBC), hemoglobin (Hb), white blood cells (WBC), neutrophils, lymphocytes, blood glucose, serum creatinine (Cr), blood urea nitrogen (BUN), lactate, serum sodium, serum potassium, serum calcium, etc.

5. Treatment measures, including the use of vasoactive drugs, mechanical ventilation, mechanical ventilation duration, continuous renal replacement therapy (CRRT), and CRRT duration.

All laboratory variables and vital signs indicators were derived from data generated for the first time within 24 h after patients entered the ICU. To avoid possible bias, variables were excluded if their missing values exceeded 20%. Variables with less than 20% missing data were processed by multiple interpolation by the SPSS 25.0 software. The original dataset was subjected to random interpolation. The interpolated variables primarily included patient height and weight (used to calculate BMI), vital signs, and laboratory indicators. Five complete datasets were generated through this interpolation process, and reliability analysis was performed on each of these datasets. The dataset exhibiting the highest retention coefficient was selected for subsequent analysis. Ultimately, a final complete dataset (n = 5323) was obtained and used as the basis for further research and statistical analysis.

Clinical outcomes

The primary outcome was mortality within 28 days of admission. Secondary outcomes were 90-day mortality, 1-year mortality, ICU mortality and hospital mortality.

Statistical analysis

Continuous variables were expressed as medians (interquartile distance, IQR) and analyzed by the Mann–Whitney U test. Categorical variables were expressed as frequencies and percentages, and their differences were analyzed by chi-square tests. The incidence of endpoints was evaluated between groups according to different levels of NLR using Kaplan–Meier, survival analysis. To assess the influencing factors associated with the risk of death, Cox regression analysis was performed. The Cox proportional hazards model was used to calculate the hazard ratio (HR) and 95% confidence interval (CI) between NLR and clinical outcomes, and was adjusted for some models. To mitigate the possibility of overadjustment for potential confounders, a stepwise approach with three adjustment levels was applied. Model 1: unadjusted; Model 2: partially adjusted for demographic variables; Model 3: fully adjusted for included latent variables. The covariates included in the model were selected based on existing literature reports and clinical expertise. We included all variables that were considered clinically important and whose missing values met the standard. These variables have been widely recognized in previous studies as important prognostic factors affecting patient survival. After conducting the correlation analysis, the correlations among the variables were not significant.

In addition, we also analyzed the association between ICU baseline NLR and 28-day mortality of admission using a restricted cubic spline (RCS) regression model with four nodes. Receiver operating characteristic (ROC) curve was used to determine the cutoff value of the NLR. The NLR was entered into the model as a continuous or ordinal variable (with the first quartile of the NLR as the reference group). P values for trends were calculated using the quartile level.

Subgroup analysis was performed with further stratified analysis exploring the potential association of NLR as a continuous variable by sex, age (≤ 60 and > 60 years), race (the White and others), BMI (< 30 and 30 kg/m²), hypertension, diabetes mellitus, hyperlipidemia, myocardial infarction, asthma, and COPD in different subgroups.

All statistical analyses were performed using the software IBM SPSS Statistics 25 and R 4.3.0. Two-sided P < 0.05 was considered statistically significant.

Results

Baseline characteristics

This study included 5323 patients with SIC, with a median age of 67.36 years (IQR: 56.76–77.43). The cohort was predominantly male (60.49%) and White (36.78%), with a median BMI of 27.77 kg/m² (IQR: 24.17–32.68). The 28-day mortality rate was 16.96% (903 patients).

The median NLR for all participants was 7.64 (IQR: 4.30–14.82). Baseline characteristics stratified by NLR quartile are detailed in Table 2. Patients were categorized into quartiles based on their NLR level at ICU admission: Q1 (0–4.295), Q2 (4.295–7.642), Q3 (7.642–14.824), and Q4 (14.824–243.75). The median NLR for each quartile was 3.03 (IQR: 2.14, 3.70), 5.87 (IQR: 5.03, 6.77), 10.32 (IQR: 8.83, 12.20), and 25.00 (IQR: 18.75, 41.11), respectively.

Table 2.

Characteristics and outcomes of participants categorized by NLR

Variables	Total (n = 5323)	Q1 (n = 1331)	Q2 (n = 1331)	Q3 (n = 1331)	Q4 (n = 1330)	χ2/F	P
Demographics
Age (years)	67.36 (56.76, 77.43)	67.42 (57.87, 77.25)	67.27 (56.71, 76.55)	67.27 (56.30, 77.64)	67.44 (55.77, 78.10)	χ2 = 2.20#	0.533
Male (%)	3220 (60.49)	793 (59.58)	878 (65.97)	782 (58.75)	767 (57.67)	χ2 = 23.27	< 0.001
White (%)	1958 (36.78)	505 (37.94)	456 (34.26)	506 (38.02)	491 (36.92)	χ2 = 5.29	0.152
BMI (kg/m2)	27.77 (24.17, 32.68)	27.64 (24.30, 32.03)	27.99 (24.43, 32.86)	27.90 (23.91, 32.68)	27.67 (23.67, 32.98)	χ2 = 3.84#	0.279
Complications
Hypertension, n (%)	3483 (65.43)	909 (68.29)	882 (66.27)	876 (65.82)	816 (61.35)	χ2 = 15.10	0.002
Diabetes, n (%)	1706 (32.05)	439 (32.98)	420 (31.56)	443 (33.28)	404 (30.38)	χ2 = 3.32	0.345
Myocardial Infarction, n (%)	870 (16.34)	208 (15.63)	203 (15.25)	248 (18.63)	211 (15.86)	χ2 = 6.98	0.072
Asthma, n (%)	455 (8.55)	106 (7.96)	116 (8.72)	120 (9.02)	113 (8.50)	χ2 = 1.01	0.800
COPD, n (%)	285 (5.35)	59 (4.43)	50 (3.76)	73 (5.48)	103 (7.74)	χ2 = 23.97	< 0.001
Hyperlipidemia, n (%)	2133 (40.07)	620 (46.58)	601 (45.15)	492 (36.96)	420 (31.58)	χ2 = 83.10	< 0.001
Vital signs
HR (beats/min)	86 (77, 102)	81 (75, 93)	82 (75, 96)	88 (79, 103)	96 (81, 111)	χ2 = 287.97#	< 0.001
RR (beats/min)	18 (15, 22.5)	16 (14, 20)	16 (14, 21)	18 (16, 23)	21 (17, 25.5)	χ2 = 435.61#	< 0.001
MBP (mmHg)	78 (68, 880)	78 (69, 87)	78 (68, 88)	77 (68, 89)	77 (66, 89)	χ2 = 1.28#	0.734
SpO2 (%)	99 (95, 100)	100 (97, 100)	100 (97, 100)	98 (95, 100)	97 (94, 99)	χ2 = 421.89#	< 0.001
T (℃)	36.61 (36.17, 37.06)	36.44 (35.79, 36.89)	36.56 (36.10, 37.00)	36.67 (36.28, 37.17)	36.83 (36.44, 37.28)	χ2 = 215.58#	< 0.001
Laboratory tests
SOFA	4 (3, 5)	4 (3, 5)	4 (3, 5)	4 (3, 5)	4 (3, 6)	χ2 = 34.96#	< 0.001
WBC (K/uL)	12.00 (8.00, 16.90)	8.60 (5.50, 12.20)	10.90 (8.00, 14.80)	13.20 (9.50, 17.50)	16.75 (11.62, 22.90)	χ2 = 948.34#	< 0.001
RBC (m/uL)	3.33 (2.84, 3.90)	3.13 (2.71, 3.64)	3.26 (2.78, 3.78)	3.44 (2.91, 4.00)	3.53 (3.02, 4.13)	χ2 = 186.82#	< 0.001
Neutrophils (K/uL)	9.72 (6.20, 14.23)	5.85 (3.26, 8.36)	8.98 (6.27, 12.02)	11.30 (8.02, 14.86)	15.12 (10.50, 20.43)	χ2 = 1632.24#	< 0.001
Lymphocytes (K/uL)	1.17 (0.67, 1.85)	2.02 (1.30, 2.82)	1.52 (1.09, 2.04)	1.09 (0.75, 1.43)	0.55 (0.30, 0.81)	χ2 = 2104.80#	< 0.001
INR	1.4 (1.3, 1.7)	1.4 (1.3, 1.6)	1.4 (1.3, 1.6)	1.4 (1.2, 1.7)	1.5 (1.3, 1.9)	χ2 = 28.12#	< 0.001
Platelet (K/uL)	149 (107, 219)	130 (99, 176)	143 (109, 202.5)	167 (118, 240)	173 (111, 257.75)	χ2 = 216.21#	< 0.001
Lactate (mg/dL)	1.70 (1.20, 2.70)	1.50 (1.10, 2.20)	1.50 (1.10, 2.29)	1.70 (1.20, 2.90)	2.20 (1.40, 3.50)	χ2 = 235.77#	< 0.001
Glucose (mg/dL)	126 (105, 161)	120 (102, 146)	120 (103, 148)	130 (108, 174.5)	138 (109, 186)	χ2 = 143.12#	< 0.001
Cr (mg/dL)	1.1 (0.8, 1.7)	0.9 (0.7, 1.2)	1.0 (0.7, 1.4)	1.1 (0.8, 1.8)	1.4 (0.9, 2.4)	χ2 = 384.44#	< 0.001
BUN (mg/dL)	21 (14, 36)	18 (13, 26)	18 (13, 28)	23 (16, 38)	30 (18, 54)	χ2 = 413.86#	< 0.001
Calcium (mmol/L)	8.10 (7.60, 8.60)	8.20 (7.70, 8.70)	8.20 (7.70, 8.63)	8.10 (7.50, 8.60)	8.00 (7.40, 8.50)	χ2 = 57.25#	< 0.001
Potassium (mmol/L)	4.2 (3.8, 4.6)	4.1 (3.8, 4.5)	4.2 (3.8, 4.6)	4.1 (3.7, 4.6)	4.2 (3.7, 4.7)	χ2 = 3.61#	0.307
Sodium (mmol/L)	139 (136, 141)	139 (137, 141)	139 (137, 141)	139 (136, 141)	138 (134, 141)	χ2 = 49.82#	< 0.001
NLR	7.64 (4.30, 14.82)	3.03 (2.14, 3.70)	5.87 (5.03, 6.77)	10.32 (8.83, 12.20)	25.00 (18.75, 41.11)	χ2 = 4989.38#	< 0.001
Treatments
Vasoactive, n (%)	3610 (68.46)	926 (70.26)	913 (69.27)	874 (66.31)	897 (68.01)	χ2 = 5.32	0.150
Ventilator, n (%)	4861 (91.32)	1232 (92.56)	1239 (93.09)	1206 (90.61)	1184 (89.02)	χ2 = 17.55	< 0.001
Ventilation Hours (h)	48.00 (24.00, 105.00)	36.53 (21.00, 77.00)	42.75 (22.00, 91.19)	56.58 (26.16, 125.50)	63.76 (32.00, 140.85)	χ2 = 149.20#	< 0.001
CRRT, n (%)	477 (8.96)	80 (6.01)	82 (6.16)	122 (9.17)	193 (14.51)	χ2 = 77.29	< 0.001
CRRT Days (d)	5.00 (2.00, 9.00)	4.00 (2.00, 7.25)	4.50 (2.25, 8.00)	5.00 (2.25, 8.00)	5.00 (3.00, 9.00)	χ2 = 1.45#	0.694
Outcomes
ICU LOS (d)	3.11 (1.83, 6.32)	2.34 (1.38, 4.52)	2.69 (1.52, 5.38)	3.62 (2.06, 7.16)	4.12 (2.25, 8.18)	χ2 = 261.40#	< 0.001
Hospital LOS (d)	8.82 (5.49, 15.70)	7.54 (5.22, 13.01)	8.02 (5.26, 14.01)	9.59 (5.87, 16.91)	10.81 (6.12, 18.77)	χ2 = 92.71#	< 0.001
28-day death, n (%)	903 (16.96)	141 (10.59)	148 (11.12)	233 (17.51)	381 (28.65)	χ2 = 199.76	< 0.001
90-day death, n (%)	1202 (22.58)	192 (14.43)	201 (15.10)	320 (24.04)	489 (36.77)	χ2 = 247.96	< 0.001
One-year death, n (%)	1551 (29.14)	268 (20.14)	277 (20.81)	424 (31.86)	582 (43.76)	χ2 = 239.41	< 0.001
ICU death, n (%)	594 (11.16)	98 (7.36)	96 (7.21)	164 (12.32)	236 (17.74)	χ2 = 100.25	< 0.001
Hospital death, n (%)	837 (15.72)	137 (10.29)	129 (9.69)	224 (16.83)	347 (26.09)	χ2 = 175.25	< 0.001

^#: Kruskal-waills test, χ²: Chi-square test

M: Median, Q₁: 1 st Quartile, Q₃: 3 st Quartile

BMI: Body Mass Index; COPD: Chronic Obstructive Pulmonary Disease; HR: Heart Rate; RR: Respiratory Rate; MBP: Mean Blood Pressure; SpO₂: Oxygen Saturation; T: Temperature; SOFA: Sequential Organ Failure Assessment; WBC: White Blood Cell; RBC: Red Blood Cell; Cr: Creatinine; BUN: Blood Urea Nitrogen; NLR: Neutrophil-to-Lymphocyte Ratio; ICU: Intensive Care Unit; CRRT: Continuous Renal Replacement Therapy; LOS: Length of Stay

In general, there were fewer male patients in higher NLR quartiles who exhibited lower prevalence of hypertension and hyperlipidemia, higher prevalence of COPD, more unstable vital signs, elevated platelet counts and INR, and worse renal function compared to those in lower quartiles. Although patients with higher NLR were less likely to receive mechanical ventilation during hospitalization, they experienced longer ventilation durations and a greater need for CRRT.

The Q1 group (lowest NLR) presented a distinct clinical profile. This group had a higher prevalence of hypertension (68.29%) and hyperlipidemia (46.58%) compared to the other quartiles. Despite requiring substantial vasopressor (70.26%) and ventilator support (92.56%), the Q1 group showed more stable vital signs, with lower HR, RR, and temperature, and higher SpO₂ levels. Consistent with this, the Q1 group had significantly lower levels of lactate, Cr, BUN, and platelet count compared to the other groups (P < 0.001).

In contrast, patients in the highest NLR quartile (Q4) had the longest ICU stay (4.12 days, IQR: 2.25, 8.18) and total hospital stay (10.81 days, IQR: 6.12, 18.77). Moreover, Q4 had significantly higher 28-day mortality (28.65%), 90-day mortality (36.77%), 1-year mortality (43.76%), ICU mortality (17.74%), and hospital mortality (26.09%) compared to the others (P < 0.001 for all). Notably, the Q2 group (NLR: 4.295–7.642) had the lowest ICU mortality (7.21%) and hospital mortality (9.69%).

Patients were divided into survival and non-survival groups based on their 28-day outcome. Baseline characteristics of the two groups are presented in Supplementary Table S1. Compared to survivors, non-survivors were older, had a lower proportion of White individuals, a higher prevalence of MI and COPD, but a lower prevalence of hyperlipidemia. Non-survivors also exhibited more unstable vital signs, higher SOFA scores and INR, and worse renal function. Consistent with findings in the higher NLR group, non-survivors had a lower rate of mechanical ventilation but a longer ventilation duration, and a higher rate of CRRT but a shorter CRRT duration. Importantly, NLR levels were significantly higher in non-survivors (12.01) than in survivors (7.02, P < 0.001), with a mean difference of 4.99 (95% CI 3.38–4.62; Z = 13.5).

Main outcome

Kaplan–Meier survival curves were generated to visualize differences in the incidence of primary outcomes across NLR quartiles (Fig. 1). While the cumulative death rates between Q1 and Q2 were similar, patients in Q4 exhibited a significantly higher mortality risk at 28 days, 90 days, 1 year, and for both ICU and hospital mortality (log-rank P < 0.001). Thus, elevated NLR levels were associated with significantly increased mortality.

Fig. 1.

Kaplan–Meier survival curves for mortality stratified by NLR quartiles

Patients with SIC were categorized into four groups (Q1–Q4) based on NLR quartiles at ICU admission. The survival probability over time is shown for (A) 28-day, (B) 90-day, and (C) 1-year all-cause mortality. (D) ICU mortality and (E) in-hospital mortality are also presented. The differences between the groups were assessed using the log-rank test, and all comparisons were statistically significant (P < 0.001). Higher NLR quartiles (especially Q4) are associated with progressively lower survival probabilities across all time points.

Receiver operating characteristic (ROC) curve analysis (Fig. 2) revealed an area under the curve (AUC) of 0.642 (95% CI 0.622–0.663) for NLR in predicting 28-day mortality. Moreover, NLR demonstrated significantly greater predictive value for SIC mortality compared to neutrophils or lymphocytes considered independently (Supplementary Figure S2). Using ROC analysis, the optimal NLR cutoff value for predicting 28-day mortality was 10.898, yielding a sensitivity of 0.551 and a specificity of 0.685. The AUC values for NLR in predicting 90-day, 1-year, ICU, and hospital mortality were 0.645, 0.631, 0.620, and 0.637, respectively. In conclusion, NLR provides modest predictive value for mortality risk at various time points in patients with SIC.

Fig. 2.

ROC curves of the NLR for predicting mortality in patients with SIC

The predictive performance of the baseline NLR for (A) 28-day, (B) 90-day, and (C) 1-year mortality, as well as for (D) ICU mortality and (E) in-hospital mortality, is shown. The AUC values with 95% confidence intervals are provided for each outcome, demonstrating the discriminative ability of NLR. The diagonal line represents the reference line of no discriminative value (AUC = 0.5).

Cox proportional hazards analysis was performed to assess the association between NLR and 28-day mortality, adjusting for potential confounders as shown in Table 3. Three models were constructed: Model 1 (unadjusted), Model 2 (adjusted for age, sex, race, and BMI), and Model 3 (adjusted for age, sex, race, BMI, hypertension, diabetes, hyperlipidemia, MI, asthma, COPD, RBC, INR, platelet, Cr, BUN, lactate, vital signs, blood glucose, serum sodium, serum potassium, and serum calcium).

Table 3.

Cox proportional hazard ratios (HR) for mortality

Variables	Model1		Model2		Model3
	HR (95%CI)	P	HR (95%CI)	P	HR (95%CI)	P
Hospital mortality
NLR as continuous	1.90 (1.66–2.17)	< 0.001	1.86 (1.62–2.13)	< 0.001	1.36 (1.16–1.59)	< 0.001
Quartile
Q1	1.00 (Reference)		1.00 (Reference)		1.00 (Reference)
Q2	1.05 (0.83–1.32)	0.689	1.07 (0.85–1.35)	0.579	1.02 (0.81–1.29)	0.849
Q3	1.72 (1.40–2.13)	< 0.001	1.70 (1.38–2.10)	< 0.001	1.26 (1.01–1.57)	0.023
Q4	2.97 (2.45–3.60)	< 0.001	2.81 (2.28–3.46)	< 0.001	1.52 (1.22–1.91)	< 0.001

HR: Hazard Ratio, CI Confidence Interval

Model1: Crude

Model2: Adjust: gender, age, race, BMI

Model3: Adjust: gender, age, race, BMI, INR, platelet, RBC, lactate, calcium, BUN, potassium, sodium, glucose, Cr, HR, RR, temperature, SpO₂, MBP, hypertension, diabetes, MI, asthma, COPD, hyperlipidemia

When treated as a continuous variable, NLR was a significant predictor of 28-day mortality in all models (Model 1: P < 0.001; Model 2: P < 0.001; Model 3: P < 0.001). Furthermore, when categorized by quartiles, patients in Q4 demonstrated a significantly increased risk of 28-day mortality compared to Q1 and Q2 across all three models. Specifically, the hazard ratios (HR) for Q4 were: Model 1 [HR, 2.97 (95% CI, 2.45–3.60) P < 0.001], Model 2 [HR, 2.81 (95% CI, 2.28–3.46) P < 0.001], and Model 3 [HR, 1.52 (95% CI, 1.22–1.91) P < 0.001].

Restricted cubic spline (RCS) analysis was employed to further investigate the relationship between NLR and 28-day mortality (Fig. 3). This analysis revealed a non-linear, J-shaped association between NLR and mortality risk (non-linear P < 0.001). The inflection points of the RCS curve, indicating a shift in the relationship between NLR and mortality risk, was estimated to be at NLR = 7.64. Consistent with this finding, Cox proportional hazards regression identified the lowest risk of mortality within quartile 2 (Q2, NLR: 4.295–7.642), where the RCS turning point was located. Similar J-shaped associations with increased NLR were observed for 90-day, 1-year, ICU, and hospital mortality.

Fig. 3.

Dose–response relationship between NLR and mortality risk using RCS analysis

The solid curves represent the adjusted HRs for (A) 28-day, (B) 90-day, (C) 1-year, (D) ICU, and (E) in-hospital mortality, with the shaded areas indicating the 95% confidence intervals. The analyses were adjusted for gender, hypertension, COPD, hyperlipidemia, vital signs, SOFA score, lactate, Cr, BUN, etc. The horizontal dotted lines represent the hazard ratio of 1.0. The non-linear relationship was statistically significant (P for non-linearity < 0.001), showing a J-shaped association where the risk is lowest around the second quartile (Q2) and increases sharply at higher NLR levels.

Subgroup analysis

To further examine the association between NLR and mortality at different time points (28-day, 90-day, 1-year, ICU, and hospital mortality), subgroup analyses were conducted, stratifying patients by gender, race, age, BMI, hypertension, diabetes, MI, asthma, COPD, and hyperlipidemia (Fig. 4). These analyses generally demonstrated a consistent positive association between NLR and mortality across various subgroups. Specifically, the interaction terms for race, age, BMI, hypertension, diabetes, MI, asthma, COPD, and hyperlipidemia were not statistically significant (P > 0.05), indicating a consistent effect of NLR on mortality across these strata. Notably, the predictive value of NLR for 28-day mortality appeared more pronounced in male patients (HR: 2.18, 95% CI 1.85–2.58, interaction P = 0.017), with statistically significant interaction p-values (P < 0.05) observed for mortality risk at different time points. Overall, the stratified analyses consistently suggested that NLR had similar associations with mortality across most of the subpopulations examined, regardless of the mortality endpoint.

Fig. 4.

Subgroup analysis of the association between NLR and mortality

Forest plots display the HRs and 95% confidence intervals for the association between NLR (as a continuous variable) and (A) 28-day, (B) 90-day, (C) 1-year, (D) ICU, and (E) in-hospital mortality across various patient subgroups. The solid vertical line represents the null effect (HR = 1.0). A HR to the right of this line indicates increased risk, while a value to the left indicates decreased risk. P for interaction was calculated to test for effect modification across subgroups.

Discussion

In this retrospective study, we found that the NLR can serve as a predictor of short-term and long-term mortality in patients with SIC and is an important indicator for prognostic assessment. Our study revealed that patients with elevated NLR levels had a significantly increased mortality rate, and there was a non-linear J-shaped relationship between NLR and the risk of death. Further Cox regression and subgroup analyses also validated the prognostic value of NLR across different patient subgroups. These findings suggest that NLR may be a practical parameter for early risk stratification and individualized management of SIC patients, and could be an independent risk factor for SIC.

Sepsis is a clinically common critical syndrome characterized by high morbidity and mortality rates. In recent years, the NLR has become a popular indicator for assessing sepsis prognosis due to its simple and feasible detection methods. Numerous studies have explored the prognostic value of NLR in the overall population of sepsis patients, confirming its correlation with the degree of inflammation, organ dysfunction, and risk of mortality. However, the vast majority of current research focuses on generalized sepsis populations, and studies on the specific prognostic value and mechanisms of NLR in sepsis-related subpopulations, especially patients with SIC, remain very limited.

SIC is a critical stage in the progression of sepsis to DIC, with significant differences in clinical features and pathophysiological mechanisms compared to general sepsis patients. Patients with SIC often exhibit more pronounced coagulopathy and microcirculatory disturbances, leading to disease deterioration and poor prognosis. Notably, SIC has a short diagnostic window with rapid progression, and traditional inflammatory markers such as CRP and PCT have limited specificity in predicting SIC. Therefore, exploring novel prognostic indicators that are applicable to SIC patients and can reflect the interaction between inflammation and coagulation is of significant clinical importance.

Unlike previous studies related to NLR in general sepsis patients, this study strictly incorporates and classifies patients according to the defined criteria for SIC. It systematically analyzes the performance patterns and predictive efficacy of NLR in the SIC subgroup, clarifying its unique value in the inflammatory-coagulation interaction process. Additionally, based on NLR and related parameters, a prognostic model specifically for the SIC population is constructed, aiming to achieve precise identification of high-risk SIC patients and providing new tools for early clinical intervention and resource optimization. Therefore, this study enhances the scientific and clinical value of NLR prognostic research in specific populations and fills the research gap regarding the specific clinical application of NLR in SIC patients. The findings not only contribute to optimizing follow-up and intervention decisions for SIC patients but also provide solid evidence for understanding the mechanisms of inflammation-coagulation interaction in sepsis and for stratified management.

SIC is associated with a poor prognosis in sepsis patients due to its high incidence. Early intervention with anticoagulants is currently favored for its significant potential to improve outcomes in individuals with SIC [13]. However, there is ongoing debate regarding the optimal dose and timing of anticoagulation. Current clinical strategies are guided by patient severity, allowing for proactive treatment initiation, yet objective indicators to accurately determine SIC severity are lacking. Although machine learning models have been developed to dynamically predict SIC occurrence [20], they still fall short of predicting mortality in SIC patients.

Unlike other studies that use D-dimers or fibrinogen to predict SIC mortality, our study demonstrates that NLR can quickly and effectively predict mortality in SIC patients [21]. We found that elevated NLR levels were strongly associated with increased 28-day mortality and showed similar trends for 90-day, 1-year, ICU, and hospital mortality. Notably, patients in the highest NLR quartile (Q4) had a significantly higher risk of death compared to other groups, aligning with the understanding that elevated NLR reflects excessive inflammatory activation [22, 23].

NLR, derived from routine blood tests, reflects both neutrophil-mediated inflammation and lymphopenia, explaining its close association with poor prognosis. Furthermore, restrictive cubic spline regression analysis revealed a nonlinear J-shaped relationship between NLR and mortality risk. Patients with NLR values between 4.295 and 7.642 (Group Q2) exhibited the lowest mortality risk. NLR beyond this range suggests a severe inflammatory response and immune dysfunction, heightening mortality risk.

During sepsis, neutrophils, and lymphocytes rapidly respond to infection. Initially, counts rise; however, neutrophil migration to infection sites and anti-inflammatory cytokine-induced immunosuppression result in lymphocyte apoptosis, reducing lymphocyte counts [24]. Lymphopenia is linked to higher mortality and shorter survival in patients [25]. NLR serves as a systemic inflammatory indicator, with its changes reflecting neutrophil and lymphocyte balance. Notably, in RCS results, NLR below a certain point was inversely related to mortality, while higher values increased mortality risk, particularly at high inflammation levels like Q4. This suggests adverse outcomes from both inadequate and excessive immune responses in SIC, similar to other NLR studies [26, 27]. This nonlinear relationship provides a crucial basis for clinical NLR risk stratification.

Although the ROC analysis showed a modest AUC of approximately 0.64, it is important to consider the clinical context of SIC. In the absence of highly accurate predictive tools, even a modestly discriminating biomarker like NLR can be valuable for risk stratification, especially when combined with other clinical parameters. An AUC of 0.64 suggests that NLR is better than random chance at distinguishing between patients at higher and lower risk of mortality. Its clinical usefulness lies in its ease of accessibility, low cost, and ability to provide rapid information. In resource-limited settings, or in situations requiring rapid decision-making, NLR can serve as an initial screening tool to identify patients who may benefit from closer monitoring and more aggressive interventions. Furthermore, NLR can be easily integrated into existing clinical workflows and used in conjunction with other clinical data, such as platelet count to improve the overall predictive accuracy and clinical decision-making.

Our findings support the identification of critically ill SIC patients for earlier intervention opportunities, potentially enhancing clinical prognosis. In anticoagulation decision-making, NLR may offer valuable supplementary information. For SIC patients with high NLR, clinicians might consider proactive monitoring and earlier anticoagulation, especially with established parameters like platelet count and D-dimer. Conversely, patients with lower NLR may avoid unnecessary aggressive anticoagulation, minimizing bleeding complications.

Previous studies fully confirm the role of NLR as a marker of homeostasis of the immune system, although the precise and unique cut-off values require further investigation. In our study, we determined by the ROC curve analysis that the cut-off value of NLR predicted 28-day mortality was 10.898, with certain sensitivity and specificity. The finding of this threshold provides a reference for clinicians to identify high-risk SIC patients at an early stage. It is worth noting that NLR is a convenient, economical and easily accessible indicator, especially in resource-limited medical settings. In addition, compared with conventional coagulation-related indicators (such as D-dimer and fibrinogen), NLR has higher timeliness and flexibility and can be used as a complementary tool for early prognosis assessment of SIC.

The SIC is the result of the interaction between inflammation and the coagulation system, and the NLR may play an important role in this process. On the one hand, neutrophils play a key role in coagulation activation and thrombosis by the release of NETs, proinflammatory cytokines and TF. On the other hand, the reduction of lymphocytes reflects the damage to the host immune defense, which may lead to an uncontrolled inflammatory response and aggravate the pathological process of SIC. Therefore, elevated NLR may be a combined marker of inflammation and coagulation disorders that can sensitively reflect deterioration in patients with SIC.

Subgroup analysis showed a consistent association between NLR and mortality regardless of patient sex, age, ethnicity, BMI, or comorbidities status. This suggests a general prognostic value of NLR in different patient subsets. However, we noted that the predictive value of the NLR was more significant in male patients (interaction P = 0.017). This sex difference may be related to the more pronounced inflammatory response in sepsis in male patients and deserves further investigation. Moreover, the predictive value of NLR is equally prominent in COPD patients and MI patients, which may be related to the basal inflammatory status in these patients.

Despite the value of the NLR as a predictive tool highlighted in this study, we also recognize its limitations. First, this was a single-center observational study based on the MIMIC-IV database. The results may lack generalizability and require external validation. Furthermore, there may be unmeasured confounding factors, such as stroke, autoimmune diseases, and unrecorded medical histories. Second, our focus was primarily on the baseline NLR value at admission, overlooking the impact of dynamic changes in the NLR during hospitalization. An improvement in the NLR often reflects the recovery of immune function and may influence long-term prognosis. Therefore, based solely on the association between admission NLR and poor long-term outcomes, it is difficult to distinguish whether this is due to a decline in Activities of Daily Living (ADL) at discharge or other factors unrelated to sepsis. Clearly distinguishing the effects of post-discharge functional status and factors independent of sepsis on long-term survival is crucial. Finally, to mitigate the confounding effect of anticoagulant therapy on mortality in SIC patients, we excluded patients receiving anticoagulation, including those on long-term anticoagulation prior to admission and those who newly started it during their hospital stay. While this ensured sample homogeneity, it also limited the applicability of our findings to patients who might benefit from anticoagulant therapy. The influence of anticoagulant therapy on the relationship between the NLR and mortality in SIC patients warrants further investigation. Future research should consider these patient populations to broaden the predictive value of the NLR in different therapeutic contexts, for instance, by exploring the relationship between NLR values and outcomes in SIC patients receiving various anticoagulant regimens, thereby providing insights for optimizing treatment strategies. It is noteworthy that patients in the Q1 group received more vasopressors and CRRT, which is inconsistent with conventional understanding. We believe the possible reasons include: 1. The early inflammatory response was not yet fully activated, resulting in a non-elevated NLR; although the inflammation was insufficient to raise the NLR, it was enough to trigger coagulation dysfunction and SIC. 2. Patients had immunosuppression; the higher proportion of patients with hypertension and hyperlipidemia in the Q1 group might lead to reduced lymphocyte reactivity, exacerbating the infection and ultimately causing clinical deterioration [28, 29]. 3. Other non-inflammatory factors led to clinical deterioration, such as cardiac dysfunction, hypovolemia, or severe metabolic disorders. Combined with the higher use of vasopressors in the Q1 group, this may suggest the presence of cardiac dysfunction or hypovolemia. 4. The NLR, as a single indicator, has limitations and cannot fully reflect the overall condition of the body.

Conclusion

This study provides evidence supporting the potential of NLR as a readily available and easily calculated predictor of mortality in SIC patients. Elevated NLR levels are associated with increased short-term and long-term mortality, exhibiting a J-shaped relationship. While NLR demonstrates modest predictive value, its accessibility and ease of use suggest its potential as a valuable tool for early risk stratification and informing clinical decision-making in SIC patients, especially in emergencies or resource-limited settings. Further research is warranted to validate these findings in external cohorts and to investigate the optimal integration of NLR into clinical algorithms for the management of SIC.

Author contributions

SP and HS contributed to the study design. HZ and GY performed the data analyses and wrote the first draft of the manuscript. SP, HS and CH revised the manuscript. The authors read and approved the final manuscript.

Funding

Not applicable.

Data availability

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://physionet.org/content/mimiciv/2.2/.

Declarations

Ethics approval and consent to participate

The study from MIMIC-IV database was approved by the review committee of Massachusetts Institute of Technology and Beth Israel Deaconess Medical Center. In the MIMIC-IV database, the data is publicly available. Therefore, the ethical approval statement and the requirement for informed consent were waived for this study.

Consent for publication

All authors have reviewed the final version of the manuscript and approved it for publication.

Competing interests

The authors declare no competing interests.