Association between systemic immune-inflammation index and in-hospital mortality in ICU patients with aspiration pneumonia: A retrospective cohort study based on MIMIC-IV database

Abstract

Background

Aspiration pneumonia is a common and life-threatening condition among critically ill patients, with mortality reaching 20–50%. The Systemic Immune-Inflammation Index (SII), derived from the routine complete blood count, has shown prognostic value in various inflammatory and infectious diseases, but its association with outcomes in ICU patients with aspiration pneumonia is unknown.

Objectives

To investigate the association between SII and in-hospital mortality in ICU patients with aspiration pneumonia and to characterize the dose-response relationship.

Design

Retrospective cohort study using a large publicly available critical care database.

Methods

Adult patients (≥18 years) admitted to the ICU with aspiration pneumonia and an ICU stay ≥24 hours were identified in the MIMIC-IV database (version 3.1). SII (neutrophil × platelet/lymphocyte count) was derived from laboratory values obtained within the first 24 hours of admission and stratified into quartiles. Multivariable logistic regression with sequential adjustment evaluated the association with in-hospital mortality; restricted cubic splines characterized the dose-response relationship, and subgroup and sensitivity analyses assessed robustness.

Results

A total of 2,132 patients were included (median age 66 years, IQR 54–78; 62.2% male); overall in-hospital mortality was 25.5%. Mortality across SII quartiles was 24.6% (Q1), 20.5% (Q2), 26.1% (Q3) and 31.0% (Q4), exhibiting a U-shaped pattern. Using Q2 as reference, the highest quartile was independently associated with increased mortality (adjusted OR 1.53, 95% CI 1.14–2.06; P=0.005), and restricted cubic spline analysis confirmed a non-linear U-shaped relationship (P for trend = 0.003). Results were consistent after additional adjustment for vasopressor use and mechanical ventilation (adjusted OR 1.70, 95% CI 1.11–2.60; P=0.014) and across age, sex and comorbidity subgroups, with no significant interactions.

Conclusion

In ICU patients with aspiration pneumonia, SII showed a U-shaped association with in-hospital mortality. Pending prospective and external validation, SII may represent a promising candidate biomarker for risk stratification.

Plain language summary

Why we did this study Aspiration pneumonia is a lung infection that happens when food, drink, saliva or stomach contents are breathed into the lungs. People who are very ill and treated in an intensive care unit (ICU) can become seriously unwell from it, and some do not survive. Doctors need quick, low-cost ways to tell which patients are most at risk so they can watch them more closely. What we looked at We studied a score that can be worked out from an ordinary blood test. The score combines three types of blood cells that change when the body is inflamed. We wanted to find out whether this score is linked to the chance of dying in hospital. What we did We used a large, anonymous database of ICU patients in the United States. We included 2,132 adults who were treated in intensive care for this type of pneumonia. We compared the blood-test score on arrival with whether patients survived their hospital stay. What we found • Patients with a middle-range score had the lowest risk of dying. • Patients with either a very low or a very high score were more likely to die. • When we drew this on a graph, the risk formed a U-shape: lowest in the middle, higher at both ends. • This pattern stayed the same across different groups of patients, such as older people and those with other illnesses. What this means A score from a routine blood test, which costs nothing extra, may help doctors spot patients who need closer attention. Because both low and high scores signalled higher risk, doctors should not assume that only a high score is a warning sign. More studies are needed before this score is used to guide treatment.

Introduction

Aspiration pneumonia is a common and potentially life-threatening pulmonary condition that occurs when oropharyngeal or gastric contents are inhaled into the lower respiratory tract. ¹ It accounts for approximately 5-15% of community-acquired pneumonia cases and represents a significant proportion of pneumonia cases in intensive care unit (ICU) settings. ² Patients with impaired consciousness, dysphagia, neurological disorders, and those receiving mechanical ventilation are particularly susceptible to aspiration events ³ The mortality rate of aspiration pneumonia in critically ill patients ranges from 20% to 50%, imposing a substantial burden on healthcare systems. ⁴ Early identification of high-risk patients is crucial for timely intervention and improved outcomes.

The systemic inflammatory response plays a pivotal role in the pathophysiology and prognosis of pneumonia. ⁵ Excessive inflammation can lead to tissue damage, organ dysfunction, and ultimately death, while inadequate immune response may result in failure to control infection. ⁶ Traditional inflammatory markers such as C-reactive protein (CRP), procalcitonin, and white blood cell count have been widely used for diagnosis and prognosis assessment in pneumonia. ⁷ However, these single markers may not fully capture the complex interplay between inflammation and immune status.

The Systemic Immune-Inflammation Index (SII), calculated as platelet count × neutrophil count/lymphocyte count, is a novel biomarker that integrates three key components of the peripheral blood reflecting both inflammatory and immune status. ⁸ Originally developed in oncological research, SII has demonstrated robust prognostic value in various malignancies including hepatocellular carcinoma, gastric cancer, and lung cancer.^9–11 More recently, its utility has been extended to cardiovascular diseases, sepsis, and respiratory conditions.^12–14 Studies have shown that elevated SII is associated with adverse outcomes in patients with community-acquired pneumonia and chronic obstructive pulmonary disease (COPD).^15,16 Notably, SII offers several advantages over other inflammatory indices: it is easily calculated from routine complete blood count, requires no additional cost, and provides information about platelet activation in addition to the neutrophil-to-lymphocyte ratio.

Despite the growing evidence supporting the prognostic value of SII in various inflammatory conditions, its role in predicting outcomes among ICU patients with aspiration pneumonia has not been investigated. Given the high mortality rate and the need for reliable prognostic tools in this population, we hypothesized that SII might serve as a useful biomarker for risk stratification.

Therefore, we conducted this retrospective cohort study using the Medical Information Mart for Intensive Care IV (MIMIC-IV) database to investigate the association between SII and in-hospital mortality in ICU patients with aspiration pneumonia, and to explore the dose-response relationship between SII levels and mortality risk.

Methods

Study design and data source

This was a retrospective cohort study utilizing data from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database, version 3.1. MIMIC-IV is a large, freely available single-center database comprising de-identified electronic health records of patients admitted to the ICU at Beth Israel Deaconess Medical Center, a tertiary academic medical center located in Boston, Massachusetts, United States, between 2008 and 2019. ¹⁷ The database contains comprehensive clinical information including demographics, vital signs, laboratory measurements, medications, diagnoses, and outcomes. The reporting of this study conforms to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement. ¹⁸

Study population

We included adult patients (≥18 years) who were admitted to the ICU with a diagnosis of aspiration pneumonia. Aspiration pneumonia was identified using the International Classification of Diseases (ICD) codes: ICD-9-CM codes 507.0, 507.1, and 997.32; ICD-10-CM codes J69.0, J69.1, and J69.8. For patients with multiple ICU admissions, only the first ICU stay was included. Patients were excluded if they met any of the following criteria: (1) ICU length of stay less than 24 hours; (2) missing data on all components required for SII calculation; or (3) diagnosis of hematologic malignancy.

Calculation of systemic immune-inflammation index

SII was calculated using laboratory values obtained within the time window of 6 hours before to 24 hours after ICU admission, using the following formula:

$$\text{SII}=\text{neutrophil}\hspace{0.17em}\text{count}\hspace{0.17em}(\times {10}^9/\mathrm{L})\times \text{platelet}\hspace{0.17em}\text{count}\hspace{0.17em}(\times {10}^9/\mathrm{L})/\text{lymphocyte}\hspace{0.17em}\text{count}\hspace{0.17em}(\times {10}^9/\mathrm{L})$$

All cell counts were expressed as ×10⁹/L. Where absolute neutrophil and lymphocyte counts were not directly available, they were derived from the total white blood cell count multiplied by the corresponding percentages. When multiple measurements were available within the time window, the first recorded value was used. SII values were then categorized into quartiles (Q1, Q2, Q3, and Q4) for analysis.

Outcome

The primary outcome was in-hospital mortality, defined as death occurring during the index hospitalization.

Covariates

Baseline characteristics collected at ICU admission included: (1) demographics: age and sex; (2) comorbidities: hypertension, diabetes mellitus, congestive heart failure, COPD, chronic kidney disease, stroke, dementia, liver disease, solid malignancy, myocardial infarction, and atrial fibrillation; (3) laboratory parameters: hemoglobin, serum creatinine, blood urea nitrogen, glucose, lactate, and albumin; and (4) simplified Charlson Comorbidity Index.

Statistical analysis

Continuous variables were expressed as median with interquartile range (IQR) and compared using the Kruskal-Wallis test. Categorical variables were presented as frequencies and percentages and compared using the chi-square test. The association between SII quartiles and in-hospital mortality was evaluated using multivariable logistic regression with three sequential models: Model 1 was unadjusted; Model 2 was adjusted for age and sex; Model 3 was fully adjusted for age, sex, comorbidities, Charlson index, and laboratory parameters. The second quartile (Q2) was selected as the reference group. To explore the dose-response relationship, restricted cubic splines were used. Subgroup analyses were conducted with interaction terms tested. To address potential residual confounding by acute illness severity, we performed a sensitivity analysis (Model 4) by additionally adjusting for vasopressor use (norepinephrine, epinephrine, dopamine, vasopressin, or phenylephrine recorded in the inputevents table) and mechanical ventilation status (invasive or non-invasive, recorded in the procedureevents table) within the first 24 hours of ICU admission. To assess potential selection bias arising from missing SII components, we further compared baseline characteristics between included and excluded patients using the Mann-Whitney U test for continuous variables and the chi-square test for categorical variables (Supplementary Table S2). All statistical analyses were performed using Python (version 3.14) with the following packages: pandas (data management), NumPy and SciPy (numerical computation and statistical testing), and statsmodels (logistic regression and restricted cubic spline analysis). Visualizations were generated using Matplotlib. A two-sided P value < 0.05 was considered statistically significant.

Results

Patient selection and characteristics

The patient selection process is illustrated in Figure 1. From the MIMIC-IV database, we identified 9,491 hospital admissions with aspiration pneumonia. After restricting to ICU admissions (n=7,424) and selecting only the first ICU stay per patient (n=5,478), we excluded 649 patients with ICU length of stay less than 24 hours. Among the remaining 4,829 patients, 2,598 were excluded due to missing SII components. Finally, 99 patients with hematologic malignancies were excluded, resulting in a final cohort of 2,132 patients for analysis.

Figure 1.

Patient selection flowchart. From 9,491 hospitalizations with aspiration pneumonia diagnosis in the MIMIC-IV database, 2,132 patients were included in the final analysis after applying exclusion criteria.

Baseline characteristics by SII quartiles

The baseline characteristics of the study population stratified by SII quartiles are presented in Table 1. The median age was 66 years (IQR: 54-78), and 1,327 patients (62.2%) were male. The median SII was 2,125 (IQR: 993-4,360). The overall in-hospital mortality was 25.5% (544/2,132). Patients in higher SII quartiles tended to be older (median age: 63, 65, 67, and 68 years for Q1 through Q4, respectively; P<0.001). Significant differences were observed across quartiles for several laboratory parameters. Patients in Q4 had higher white blood cell counts, higher neutrophil percentages, and lower lymphocyte percentages compared to Q1 (all P<0.001). Platelet counts also increased across quartiles (P<0.001). Regarding comorbidities, patients with higher SII were more likely to have congestive heart failure (25.9% in Q1 vs. 34.0% in Q4; P=0.002), COPD (17.3% vs. 29.3%; P<0.001), and atrial fibrillation (27.0% vs. 37.0%; P=0.002). Conversely, liver disease was more prevalent in the lowest SII quartile (28.0% in Q1 vs. 12.6% in Q4; P<0.001).

Table 1.

Baseline characteristics of ICU patients with aspiration pneumonia stratified by SII quartiles.

Variable	Total (n=2132)	Q1 (n=533)	Q2 (n=533)	Q3 (n=533)	Q4 (n=533)	P value
SII range	0–147,136	0–986	995–2,125	2,125–4,353	4,380–147,136	​
Age, years	66.0 (53.8–78.0)	63.0 (51.0–76.0)	65.0 (51.0–77.0)	67.0 (56.0–79.0)	68.0 (56.0–80.0)	<0.001
SII, ×109/L	2,125 (993–4,360)	556 (342–775)	1,507 (1,257–1,778)	2,958 (2,490–3,566)	7,302 (5,482–11,097)	<0.001
Platelet, ×109/L	207 (143–284)	131 (87–183)	193 (143–249)	235 (180–299)	286 (220–387)	<0.001
WBC, ×109/L	12.1 (8.2–17.5)	7.8 (5.3–11.2)	10.9 (8.1–14.1)	14.0 (10.3–18.2)	17.3 (12.7–22.1)	<0.001
Neutrophils, %	83.1 (75.0–88.6)	71.5 (60.3–77.8)	81.6 (76.5–85.0)	86.0 (81.6–88.7)	90.3 (87.0–92.8)	<0.001
Lymphocytes, %	8.0 (4.5–13.0)	17.1 (12.6–23.0)	10.1 (7.6–13.0)	6.8 (5.0–8.6)	3.2 (2.0–4.5)	<0.001
Hemoglobin, g/dL	10.8 (9.0–12.6)	10.5 (8.7–12.4)	10.9 (9.2–12.7)	10.9 (9.3–12.8)	10.6 (8.9–12.3)	0.008
Creatinine, mg/dL	1.1 (0.8–1.7)	1.1 (0.8–1.7)	1.1 (0.8–1.8)	1.1 (0.8–1.7)	1.0 (0.7–1.7)	0.309
BUN, mg/dL	23.0 (15.0–38.0)	23.0 (14.0–37.0)	22.0 (14.0–38.2)	24.0 (16.0–36.0)	24.0 (16.0–40.0)	0.141
Glucose, mg/dL	131 (106–176)	120 (99–163)	129 (106–172)	138 (111–185)	141 (110–181)	<0.001
Lactate, mmol/L	1.9 (1.3–3.1)	2.1 (1.4–3.1)	1.8 (1.2–2.7)	2.0 (1.3–3.3)	1.8 (1.3–3.0)	0.008
Albumin, g/dL	3.1 (2.6–3.5)	3.1 (2.6–3.5)	3.1 (2.7–3.6)	3.1 (2.6–3.5)	3.0 (2.5–3.4)	0.055
ICU LOS, hours	109.2 (55.3–220.3)	110.5 (52.5–239.7)	110.4 (59.5–212.0)	113.8 (57.1–220.6)	101.7 (51.8–210.9)	0.514
Charlson index	2.0 (1.0–3.0)	2.0 (1.0–3.0)	2.0 (1.0–3.0)	2.0 (1.0–4.0)	2.0 (1.0–3.0)	0.009
Male, n (%)	1,327 (62.2)	349 (65.5)	332 (62.3)	330 (61.9)	316 (59.3)	0.223
Hypertension, n (%)	1,322 (62.0)	309 (58.0)	327 (61.4)	349 (65.5)	337 (63.2)	0.077
Diabetes, n (%)	617 (28.9)	142 (26.6)	171 (32.1)	158 (29.6)	146 (27.4)	0.197
Heart failure, n (%)	693 (32.5)	138 (25.9)	180 (33.8)	194 (36.4)	181 (34.0)	0.002
COPD, n (%)	518 (24.3)	92 (17.3)	120 (22.5)	150 (28.1)	156 (29.3)	<0.001
CKD, n (%)	462 (21.7)	95 (17.8)	125 (23.5)	128 (24.0)	114 (21.4)	0.060
Stroke, n (%)	388 (18.2)	103 (19.3)	113 (21.2)	103 (19.3)	69 (12.9)	0.003
Dementia, n (%)	191 (9.0)	40 (7.5)	45 (8.4)	62 (11.6)	44 (8.3)	0.088
Liver disease, n (%)	364 (17.1)	149 (28.0)	82 (15.4)	66 (12.4)	67 (12.6)	<0.001
Malignancy, n (%)	297 (13.9)	67 (12.6)	51 (9.6)	65 (12.2)	114 (21.4)	<0.001
Myocardial infarction, n (%)	538 (25.2)	115 (21.6)	135 (25.3)	150 (28.1)	138 (25.9)	0.098
Atrial fibrillation, n (%)	701 (32.9)	144 (27.0)	168 (31.5)	192 (36.0)	197 (37.0)	0.002
In-hospital mortality, n (%)	544 (25.5)	131 (24.6)	109 (20.5)	139 (26.1)	165 (31.0)	0.001

Continuous variables are presented as median (interquartile range) and compared using the Kruskal-Wallis test. Categorical variables are presented as n (%) and compared using the chi-square test. SII, Systemic Immune-Inflammation Index; WBC, white blood cell count; BUN, blood urea nitrogen; ICU LOS, intensive care unit length of stay; COPD, chronic obstructive pulmonary disease; CKD, chronic kidney disease.

Association between SII and in-hospital mortality

The mortality rates across SII quartiles demonstrated a U-shaped pattern: 24.6% (Q1), 20.5% (Q2), 26.1% (Q3), and 31.0% (Q4) (P=0.001 for overall comparison) (Figure 2). The second quartile (Q2; SII range: 995-2,125) had the lowest mortality rate and was therefore used as the reference group. The results of multivariable logistic regression analyses are shown in Table 2. In the unadjusted model (Model 1), compared with Q2, patients in Q3 (OR: 1.37, 95% CI: 1.03-1.83, P=0.030) and Q4 (OR: 1.74, 95% CI: 1.32-2.31, P<0.001) had significantly higher odds of in-hospital mortality. The association for Q1 did not reach statistical significance (OR: 1.27, 95% CI: 0.95-1.69, P=0.107). After adjusting for age and sex (Model 2), Q4 remained significantly associated with increased mortality (OR: 1.65, 95% CI: 1.24-2.18, P<0.001). In the fully adjusted model (Model 3), patients in Q4 had 53% higher odds of in-hospital mortality compared to Q2 (adjusted OR: 1.53, 95% CI: 1.14-2.06, P=0.005).

Figure 2.

Bar chart showing in-hospital mortality risk by SII quartiles. The green bar (Q2) represents the reference group with the lowest mortality rate.

Table 2.

Association between SII quartiles and in-hospital mortality.

SII quartile	SII range	N	Deaths, n (%)	Model 1 OR (95% CI)	Model 1 P	Model 2 OR (95% CI)	Model 2 P	Model 3 OR (95% CI)	Model 3 P
Q1	0–986	533	131 (24.6)	1.27 (0.95–1.69)	0.107	1.30 (0.97–1.74)	0.074	1.14 (0.84–1.55)	0.411
Q2	995–2,125	533	109 (20.5)	1.00 (Reference)	—	1.00 (Reference)	—	1.00 (Reference)	—
Q3	2,125–4,353	533	139 (26.1)	1.37 (1.03–1.83)	0.030	1.32 (0.99–1.76)	0.061	1.22 (0.91–1.64)	0.192
Q4	≥4,380	533	165 (31.0)	1.74 (1.32–2.31)	<0.001	1.65 (1.24–2.18)	<0.001	1.53 (1.14–2.06)	0.005

Model 1: unadjusted. Model 2: adjusted for age and sex. Model 3: fully adjusted for age, sex, comorbidities (hypertension, diabetes, heart failure, COPD, CKD, stroke, malignancy), Charlson index, creatinine, lactate, and albumin. The second quartile (Q2) served as the reference. SII, Systemic Immune-Inflammation Index; OR, odds ratio; CI, confidence interval.

Dose-response relationship

The dose-response relationship between SII and in-hospital mortality is illustrated in Figure 3(a). The restricted cubic spline analysis revealed a non-linear, U-shaped association between SII levels and mortality risk. The lowest mortality risk was observed at SII values around 1,500-2,000, corresponding to the second quartile. The trend test confirmed a significant positive association (P for trend = 0.003).

Figure 3.

Association between Systemic Immune-Inflammation Index (SII) and in-hospital mortality. (A) Restricted cubic spline curve showing the U-shaped relationship between SII and odds ratio for mortality. The second quartile (Q2, SII 995-2,125) was used as the reference. The shaded area represents the 95% confidence interval. (B) Forest plot of subgroup analyses showing the association between high SII (Q4 vs. Q1-Q3) and in-hospital mortality across different patient subgroups. All P values for interaction were >0.05.

Subgroup analyses

Subgroup analyses were performed to assess the consistency of the association between high SII (Q4) and in-hospital mortality across different patient subgroups (Figure 3(b) and Table 3). The overall association showed that high SII was significantly associated with increased mortality (OR: 1.36, 95% CI: 1.08-1.71, P=0.008). Significant associations were observed in patients aged ≥65 years (OR: 1.37, 95% CI: 1.02-1.85, P=0.037), males (OR: 1.46, 95% CI: 1.08-1.97, P=0.013), patients with hypertension (OR: 1.47, 95% CI: 1.11-1.97, P=0.008), diabetes (OR: 1.60, 95% CI: 1.04-2.45, P=0.032), and CKD (OR: 1.63, 95% CI: 1.02-2.61, P=0.043). Importantly, no significant interactions were detected (all P for interaction > 0.05).

Table 3.

Subgroup analysis of the association between high SII (Q4 vs Q1–Q3) and in-hospital mortality.

Subgroup	N	Events	OR (95% CI)	P value	P interaction
Overall	2,132	544	1.36 (1.08–1.71)	0.008	—
Age					0.940
<65 years	1,044	221	1.34 (0.92–1.94)	0.123	​
≥65 years	1,088	323	1.37 (1.02–1.85)	0.037	​
Sex					0.764
Male	1,327	324	1.46 (1.08–1.97)	0.013	​
Female	805	220	1.25 (0.87–1.80)	0.221	​
Hypertension					0.488
Yes	1,322	336	1.47 (1.11–1.97)	0.008	​
No	810	208	1.28 (0.86–1.89)	0.222	​
Diabetes					0.494
Yes	617	164	1.60 (1.04–2.45)	0.032	​
No	1,515	380	1.31 (0.99–1.72)	0.056	​
Heart failure					0.504
Yes	693	195	1.22 (0.82–1.80)	0.328	​
No	1,439	349	1.45 (1.09–1.94)	0.010	​
COPD					0.950
Yes	518	142	1.33 (0.85–2.07)	0.211	​
No	1,614	402	1.36 (1.04–1.79)	0.024	​
CKD					0.579
Yes	462	134	1.63 (1.02–2.61)	0.043	​
No	1,670	410	1.29 (0.99–1.68)	0.063	​
Malignancy					0.488
Yes	297	105	1.35 (0.80–2.28)	0.259	​
No	1,835	439	1.30 (1.00–1.68)	0.049	​

High SII was defined as the highest quartile (Q4, ≥4,380). All models were adjusted for age, sex, comorbidities (excluding the stratification variable), Charlson index, creatinine, lactate, and albumin. SII, Systemic Immune-Inflammation Index; OR, odds ratio; CI, confidence interval; COPD, chronic obstructive pulmonary disease; CKD, chronic kidney disease.

Sensitivity analysis including vasopressor use and mechanical ventilation

Among the 2,132 patients, 823 (38.6%) received vasopressors and 1,059 (49.7%) received mechanical ventilation during the first 24 hours of ICU admission. The distribution of these severity markers across SII quartiles is presented in Supplementary Table S1. The proportion of patients receiving vasopressors did not differ significantly across SII quartiles (37.0%, 37.5%, 42.4%, and 37.5% for Q1–Q4, respectively; P = 0.224). Mechanical ventilation use varied modestly across quartiles (48.4%, 55.7%, 50.3%, and 44.3% for Q1–Q4; P = 0.002), but did not show a monotonic relationship with SII level. When these two severity markers were added to the fully adjusted model (Model 4; complete-case N = 1,051), the association between Q4 SII and in-hospital mortality remained significant (adjusted OR: 1.70, 95% CI: 1.11–2.60, P = 0.014). For direct comparability, Model 3 was also re-estimated in this same subset and yielded an adjusted OR of 1.64 (95% CI: 1.08–2.51, P = 0.021). The fact that the OR for Q4 was not attenuated—and was in fact slightly higher—after additional adjustment for vasopressor use and mechanical ventilation suggests that these markers of acute illness severity do not substantially confound the observed SII–mortality association (Table 4). In addition, comparison of baseline characteristics between included (n = 2,132) and excluded patients (n = 2,697) showed broadly comparable populations: in-hospital mortality differed only modestly (25.5% vs. 23.0%, P = 0.045), and most demographic and laboratory parameters were similar between groups (Supplementary Table S2).

Table 4.

Sensitivity analysis: Association between SII quartiles and in-hospital mortality with additional adjustment for vasopressor use and mechanical ventilation.

SII quartile (vs Q2)	Model 3 (recomputed) OR (95% CI), P	Model 4 OR (95% CI), P
Q1	1.50 (0.98–2.28), 0.060	1.53 (1.00–2.33), 0.048
Q3	1.25 (0.82–1.92), 0.302	1.25 (0.82–1.93), 0.302
Q4	1.64 (1.08–2.51), 0.021	1.70 (1.11–2.60), 0.014
N (events)	1,051 (297)	1,051 (297)

Model 3 (recomputed): adjusted for age, sex, comorbidities, Charlson index, creatinine, lactate, and albumin. Model 4: Model 3 covariates plus vasopressor use and mechanical ventilation within the first 24 hours of ICU admission. Q2 served as the reference. The smaller sample size compared with the main analysis (N = 2,132) reflects complete-case analysis after inclusion of lactate and albumin. SII, Systemic Immune-Inflammation Index; OR, odds ratio; CI, confidence interval.

Discussion

In this large retrospective cohort study of 2,132 ICU patients with aspiration pneumonia from the MIMIC-IV database, we comprehensively investigated the association between the Systemic Immune-Inflammation Index and in-hospital mortality. Our study yielded several important findings. First, we observed that the relationship between SII and mortality followed a distinct U-shaped pattern, with patients in the second quartile (SII range 995-2,125) experiencing the lowest mortality rate of 20.5%. Second, patients with the highest SII values (Q4, ≥4,380) demonstrated a 53% increased risk of in-hospital death compared to those with moderate SII levels, even after comprehensive adjustment for demographic factors, comorbidities, and laboratory parameters. Third, this prognostic association remained remarkably consistent across all examined subgroups, including stratifications by age, sex, and various comorbid conditions, with no evidence of significant effect modification. To our knowledge, this represents the first study to systematically examine the relationship between SII and clinical outcomes specifically in ICU patients with aspiration pneumonia.

Our findings are largely consistent with and extend the rapidly expanding body of literature supporting SII as a valuable prognostic biomarker across diverse inflammatory and infectious conditions. The concept of SII was first introduced by Hu and colleagues in 2014, who demonstrated its prognostic utility in patients with hepatocellular carcinoma undergoing curative resection. ⁸ Since then, numerous studies have validated the prognostic value of SII in various malignancies, including gastric cancer, colorectal cancer, and non-small cell lung cancer.^9–11 More recently, the application of SII has expanded beyond oncology to encompass cardiovascular diseases, where elevated SII has been associated with adverse outcomes in patients with acute coronary syndrome and heart failure. ¹² Recent ICU-based studies have firmly established the prognostic value of SII in critically ill populations. Emgin et al. (2025), in a tertiary-level ICU cohort, demonstrated that elevated admission SII was independently associated with 28-day mortality. ¹⁹ Gao et al. (2025), in a large MIMIC-IV cohort of over 17,000 ICU admissions, showed that early dynamic changes in SII within the first 24 hours predicted both acute kidney injury and 1-year mortality, with a nomogram integrating SII alongside SOFA and APS III scores achieving superior discrimination compared with severity scores alone. ²⁰ In the specific context of respiratory disease, recent MIMIC-IV-based evidence has shown SII to be independently associated with mortality in critically ill COPD patients, ¹⁶ and Chen et al. (2025) demonstrated that elevated SII levels (>4,689) were predictive of 28-day mortality in patients with severe community-acquired pneumonia. ²¹ In sepsis, Xu et al. (2025) identified distinct SII trajectory subgroups with markedly different mortality risks and observed a U-shaped relationship between admission SII and in-hospital mortality—closely mirroring the pattern observed in our study. ²² Notably, in stroke-associated pneumonia—a population that overlaps substantially with aspiration pneumonia given the high prevalence of dysphagia in stroke patients—Duan et al. (2025) reported that elevated SII independently predicted pneumonia development. ²³ Together, these recent studies provide robust contemporary evidence that SII captures clinically meaningful prognostic information across the spectrum of critical illness.

However, an important distinction of our study lies in the identification of a U-shaped rather than linear relationship between SII and mortality. Most previous investigations have focused primarily on elevated SII as a risk factor, potentially overlooking the clinical significance of low SII values. Our restricted cubic spline analysis clearly demonstrated that both extremes of the SII distribution were associated with increased mortality risk, with the nadir of the curve occurring at SII values around 1,500-2,000. This non-linear pattern has profound implications for clinical interpretation and suggests that the prognostic value of SII cannot be captured by simple dichotomization or linear modeling approaches. Importantly, an emerging body of evidence supports our observation: Xu et al. (2025), in a contemporary cohort of patients with sepsis, demonstrated a U-shaped association between admission SII and in-hospital mortality using restricted cubic spline analysis—a finding remarkably parallel to ours. ²² Similar non-linear patterns have been reported across diverse patient populations, including a recent meta-analysis of all-cause mortality data spanning multiple cohorts. ²⁴ A comparable U-shaped relationship has also been observed for other inflammatory markers in critical illness, including white blood cell count and body temperature, supporting the broader concept that optimal clinical outcomes require a finely balanced inflammatory and immune response—neither suppressed nor excessive. ²⁵

The biological mechanisms underlying the U-shaped association between SII and mortality can be understood through the lens of immune homeostasis and the dual roles of inflammation in infectious disease. SII integrates information from three distinct cell populations, each with specific functions in the inflammatory and immune response. Neutrophils serve as the primary effector cells in acute bacterial infection, providing rapid antimicrobial defense through phagocytosis, degranulation, and the formation of neutrophil extracellular traps. Platelets, beyond their hemostatic function, actively participate in inflammatory processes through the release of chemokines, interaction with leukocytes, and modulation of vascular permeability. Lymphocytes, particularly T cells, orchestrate adaptive immunity and regulate the magnitude and duration of inflammatory responses. ²⁶

Elevated SII, characterized by high neutrophil and platelet counts with relative lymphopenia, likely reflects an exaggerated systemic inflammatory response that may paradoxically contribute to tissue injury and organ dysfunction. In severe pneumonia, excessive neutrophil activation can lead to the release of proteolytic enzymes and reactive oxygen species, causing alveolar epithelial damage, increased vascular permeability, and impaired gas exchange. ²⁷ The neutrophil-mediated injury may be further amplified by activated platelets, which promote neutrophil recruitment and enhance their pro-inflammatory functions. Concurrently, lymphopenia in the setting of severe infection often reflects stress-induced redistribution, apoptosis, or functional exhaustion of lymphocytes, compromising the host’s ability to mount an effective adaptive immune response and potentially predisposing to secondary infections. This state of immune dysregulation, sometimes termed immunoparalysis, has been well-documented in sepsis and severe pneumonia and is associated with poor outcomes. ²⁸

Conversely, low SII values may indicate inadequate inflammatory response or pre-existing immunocompromise, both of which can impair pathogen clearance and tissue repair. In our cohort, we observed that liver disease was significantly more prevalent among patients in the lowest SII quartile (28.0% vs. 12.6% in Q4). Chronic liver disease is associated with impaired hepatic synthesis of acute phase reactants, reduced thrombopoietin production leading to thrombocytopenia, and functional defects in both innate and adaptive immunity. ²⁹ Additionally, thrombocytopenia in critically ill patients may reflect consumptive coagulopathy, bone marrow suppression, or sequestration, all of which portend poor prognosis independent of infection. The finding that low SII carries prognostic significance underscores the importance of adequate inflammatory response for successful resolution of pneumonia and suggests that immunomodulatory strategies in this population must be carefully calibrated to avoid both excessive inflammation and immune suppression.

The clinical implications of our findings are substantial and multifaceted. From a practical standpoint, SII can be readily calculated from the complete blood count, a laboratory test that is universally obtained in ICU patients upon admission and frequently monitored thereafter. Unlike specialized biomarkers such as procalcitonin, interleukin-6, or presepsin, which require dedicated assays and incur additional costs, SII utilizes existing laboratory data without any incremental expense. This characteristic makes SII particularly attractive for resource-limited settings where access to specialized inflammatory markers may be constrained. Furthermore, the components of SII—platelet count, neutrophil count, and lymphocyte count—are objective measurements with well-established quality control procedures, minimizing inter-laboratory variability and enhancing reproducibility across different healthcare settings. ³⁰

The identification of a U-shaped relationship provides clinicians with more nuanced guidance for risk stratification than would a simple linear model. Based on our data, patients presenting with SII values substantially below 1,000 or above 4,000 should be recognized as high-risk individuals who may benefit from more intensive monitoring, earlier escalation of care, or consideration for advanced supportive therapies. The optimal SII range of approximately 1,000-2,000 may serve as a therapeutic target, although this hypothesis requires validation in prospective interventional studies. Importantly, our subgroup analyses demonstrated that the prognostic value of SII was consistent across diverse patient populations, including different age groups, both sexes, and patients with or without specific comorbidities. The absence of significant interactions suggests that SII can be applied broadly without the need for subgroup-specific cutoffs, simplifying clinical implementation and enhancing generalizability. ³¹

SII in the context of established ICU severity scores

SII may also be considered in relation to established ICU severity scoring systems such as APACHE II and SOFA. Although our study did not perform a direct head-to-head comparison, recent evidence suggests that SII complements rather than replaces these scores. Gao et al. (2025), in a MIMIC-IV cohort of over 17,000 ICU patients, demonstrated that incorporating dynamic SII metrics alongside SOFA and APS III in a prognostic nomogram improved discrimination for 1-year mortality, with each log-transformed SII parameter contributing independently to risk stratification. ²⁰ Similarly, Chen et al. (2026) showed that SII-based pneumonia severity scores provided complementary prognostic information when used alongside conventional severity assessments in elderly patients with community-acquired pneumonia. ¹⁵ Unlike APACHE II and SOFA, which require multiple physiological and laboratory parameters and time-consuming calculations, SII can be derived instantly from a routine complete blood count, making it particularly attractive in resource-limited settings or as an early triage tool prior to availability of full physiological data. Future studies directly comparing SII with established severity scores in patients with aspiration pneumonia would help clarify its incremental prognostic value beyond conventional risk stratification tools.

Our study possesses several methodological strengths that enhance the validity and reliability of our findings. First, we utilized a large, well-established critical care database with comprehensive clinical information, enabling robust statistical analysis and multivariable adjustment for a wide range of potential confounders. The MIMIC-IV database has been extensively validated and widely used in critical care research, facilitating comparison with other studies and supporting reproducibility. Second, we employed multiple complementary analytical approaches to characterize the SII-mortality relationship. The quartile-based analysis provided clinically interpretable risk estimates, while the restricted cubic spline analysis allowed visualization of the continuous dose-response relationship without imposing assumptions of linearity. The subgroup analyses and interaction testing provided evidence for the consistency and robustness of our findings across different patient populations. Third, the sequential modeling approach, progressing from unadjusted to fully adjusted models, allowed assessment of the independent prognostic value of SII while accounting for potential confounding and mediating pathways. ¹⁸

Several important limitations of our study warrant acknowledgment and should be considered when interpreting our results. First, the diagnosis of aspiration pneumonia was identified solely using ICD-9-CM and ICD-10-CM codes, which is an inherent limitation of database-based research. ICD-coded aspiration pneumonia may overlap with aspiration pneumonitis—the chemical injury caused by inhalation of sterile gastric contents—as the two conditions share clinical and radiographic features and may not always be reliably distinguished by treating clinicians at the point of coding. This potential misclassification is a recognized limitation of MIMIC-IV-based pneumonia studies and may have introduced some heterogeneity into our cohort. Second, the retrospective observational design inherently precludes causal inference. Although we adjusted for multiple potential confounders including demographics, comorbidities, and laboratory parameters, residual confounding by unmeasured or incompletely measured variables cannot be excluded. Factors such as nutritional status, functional capacity, goals of care discussions, and specific treatment decisions may influence both SII and mortality but were not captured in our analysis. Third, SII was calculated from laboratory values obtained at a single time point near ICU admission; we did not examine dynamic changes in SII during the ICU stay. Serial measurements may provide additional prognostic information, as the trajectory of inflammatory markers often carries greater predictive value than baseline values alone. Fourth, a substantial proportion of otherwise eligible patients were excluded from the primary analysis due to missing SII components (n = 2,697; 55.8% of the eligible cohort), reflecting the well-documented limitation that differential white blood cell counts are not routinely available in the MIMIC-IV database. To assess potential selection bias, we compared baseline characteristics between included and excluded patients (Supplementary Table S2). Patients in the two groups were broadly comparable in terms of sex distribution (62.2% vs. 63.6% male, P = 0.337), albumin levels, ICU length of stay, and hemoglobin. Importantly, in-hospital mortality was only modestly higher in the included group (25.5% vs. 23.0%, P = 0.045), with an absolute difference of 2.5 percentage points, suggesting that the excluded patients did not represent a systematically lower- or higher-risk population. While these small differences cannot be entirely ruled out as a source of residual selection bias, the overall similarity between groups supports the generalizability of our findings. Fifth, MIMIC-IV represents a single tertiary academic medical center in the United States, and external validation in other populations, healthcare systems, and geographic regions is essential before widespread clinical implementation. Sixth, we did not perform a direct head-to-head comparison with established severity scores such as APACHE II, SOFA, or pneumonia-specific scores like CURB-65, although recent literature suggests SII complements rather than replaces these tools. Finally, detailed information on specific causative pathogens, antibiotic selection and timing, and oxygenation parameters was not included in our analysis. ³²

Future research should address several important questions raised by our study. Prospective validation of the U-shaped SII-mortality relationship in independent cohorts is essential to confirm our findings and establish appropriate cutoff values for clinical application. Investigation of serial SII measurements and their trajectory may enhance prognostic accuracy and potentially identify patients who are responding to treatment versus those who are deteriorating. Head-to-head comparison with existing severity scores and inflammatory biomarkers would help establish the incremental value of SII and potentially support its integration into existing risk prediction models. Additionally, exploring whether SII can guide therapeutic decisions, such as intensity of monitoring, timing of interventions, or selection of patients for specific therapies, would be valuable for translating our findings into improved patient care.

Conclusions

In this retrospective cohort study of 2,132 ICU patients with aspiration pneumonia from the MIMIC-IV database, we found that the Systemic Immune-Inflammation Index exhibited a U-shaped association with in-hospital mortality. Patients with moderate SII levels (approximately 1,000-2,000) had the lowest mortality risk, while both low and high SII values were associated with increased mortality. After comprehensive multivariable adjustment, patients with the highest SII quartile (≥4,380) had a 53% increased odds of in-hospital death compared to those with moderate SII levels, and this association remained robust after additional adjustment for vasopressor use and mechanical ventilation. The relationship was consistent across different patient subgroups defined by age, sex, and comorbidities, with no evidence of significant effect modification. Given the retrospective single-center design and the absence of external validation, our findings should be interpreted as evidence of an association that requires further prospective validation rather than a definitive causal relationship. Nevertheless, SII—as a readily available and cost-free biomarker derived from routine complete blood counts—may represent a promising candidate for inclusion in future risk stratification frameworks in critically ill patients with aspiration pneumonia. Multicenter prospective studies and external validation cohorts are warranted to confirm these findings and to clarify SII’s incremental prognostic value beyond established severity scores before routine clinical implementation.

Consent for publication

This study did not report any individual person’s data, images, videos or other identifiable information requiring consent for publication.

ORCID iD

Hong Zhou https://orcid.org/0000-0003-3086-6692

Ethical considerations

The Medical Information Mart for Intensive Care IV (MIMIC-IV) database was approved by the Institutional Review Board of the Beth Israel Deaconess Medical Center, Boston, MA, USA (approval number 2001-P-001699/14), and by the Institutional Review Board of the Massachusetts Institute of Technology, Cambridge, MA, USA (approval number 0403000206). Because all records in the database are de-identified in accordance with the Health Insurance Portability and Accountability Act (HIPAA) Safe Harbor provision, the requirement for individual informed consent to participate was waived by these institutional review boards; no written or verbal consent was therefore obtained from individual participants. The first author (Na Liu) completed the Collaborative Institutional Training Initiative (CITI) programme examination (certification number 74975582) and was granted authorised access to the database under the PhysioNet data use agreement. As the present work is a secondary analysis of an existing, fully de-identified dataset, no additional ethical approval was required from the institutional review board of Jiangxi Provincial People’s Hospital, Nanchang, China.

Data Availability Statement

The data that support the findings of this study are publicly available from the PhysioNet repository at https://physionet.org/content/mimiciv/. Access requires completion of the Collaborative Institutional Training Initiative (CITI) programme and execution of the standard PhysioNet data use agreement. The analysis code used in this study is available from the corresponding author upon reasonable request.