Systemic Immune-Inflammation Index is Associated with Pressure Injury in Hospitalized Patients with Non-Small Cell Lung Cancer: A Retrospective Cohort Study

Abstract

Objective

To examine the association between admission systemic immune-inflammation index (SII) and incident pressure injury (PI) in hospitalized patients with non-small cell lung cancer (NSCLC) and assess its discriminative performance.

Materials and Methods

This single-center retrospective cohort included 428 hospitalized patients with NSCLC (119 with PI and 309 without PI). Admission SII was calculated from platelet, neutrophil, and lymphocyte counts. Logistic regression and ROC analyses were performed, with DeLong comparisons, bootstrap internal validation, calibration, and decision-curve analysis.

Results

Patients with PI had higher SII than those without PI (median 1223.49 [IQR 963.35–1519.65] vs 868.37 [694.89–1089.19], P < 0.001). In multivariable analysis, SII (OR per 100 units = 1.339, 95% CI: 1.223–1.466), BMI (OR = 0.549, 95% CI: 0.467–0.645), and CRP (OR = 1.254, 95% CI: 1.173–1.341) were independently associated with PI. The AUC was 0.763 for SII and 0.908 for the combined SII-BMI-CRP model; bootstrap optimism-corrected AUC was 0.904 (95% CI: 0.878–0.934).

Conclusion

Higher admission SII was independently associated with incident PI and showed discrimination comparable to BMI, CRP, and NLR. The combined SII-BMI-CRP model showed greater discrimination with stable bootstrap internal validation. These findings support further evaluation of SII as an adjunct to clinical risk assessment, with external validation and baseline Braden Scale comparison warranted before routine use.

Introduction

Pressure injury (PI), formerly known as pressure ulcer, refers to localized damage to the skin and/or subcutaneous tissue, typically over bony prominences, caused by sustained mechanical loading, primarily pressure, shear forces, or their interaction.^1,2 PI remains a significant clinical challenge in hospitalized populations, particularly among individuals with impaired mobility, protein-energy malnutrition, or chronic systemic conditions.^1,2 In oncologic cohorts, especially non-small cell lung cancer (NSCLC) patients, the incidence of PI is notably higher. This increased vulnerability is multifactorial, arising from cancer-related cachexia, prolonged immobility, immunosuppression due to therapies, and frequent ICU exposure.^3,4 PI in these patients not only worsens functional outcomes and quality of life but also increases the risk of nosocomial infections, prolongs hospitalization, raises healthcare costs, and heightens mortality risk.⁴

NSCLC accounts for approximately 85% of lung cancer cases and remains a major contributor to global cancer-related mortality.⁵ Despite significant therapeutic advancements, particularly with targeted therapies and immune checkpoint inhibitors, many patients present with advanced disease and progressive physiological decline.⁶ These patients often require hospitalization due to acute decompensations, such as respiratory insufficiency,⁷ opportunistic infections,⁸ or treatment-induced toxicities,⁹ increasing the risk of secondary complications like pressure injury (PI). Given these risk factors, early patient stratification for PI risk is essential to implement preemptive, multidisciplinary preventive strategies to reduce the impact of this clinical complication.

Traditional PI risk stratification tools, such as the Braden Scale, primarily assess extrinsic factors like mobility, moisture exposure, and nutritional status, but demonstrate limited sensitivity and specificity in medically complex populations, including cancer patients. Recent studies have focused on systemic pathophysiological mechanisms in PI development, particularly inflammatory and immune dysregulation. Increasing evidence suggests that a persistently elevated systemic inflammatory state may contribute to PI pathophysiology, with inflammatory mediators implicated in endothelial dysfunction, microvascular compromise, and tissue degradation.^10–12 Although biomarkers like C-reactive protein (CRP) have been linked to impaired wound healing,¹³ their ability to capture the multifactorial and dynamic nature of systemic inflammation remains limited.

The systemic immune-inflammation index (SII), calculated as platelet count × neutrophil count / lymphocyte count, is a composite biomarker that integrates inflammatory and immune-cell components of the host response.¹⁴ SII has been evaluated in several oncologic, cardiovascular, infectious, and perioperative settings.^15–17 Because it incorporates platelet, neutrophil, and lymphocyte counts, SII captures information that overlaps with but is not identical to the neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR).^18,19 Although comparative prognostic performance has been reported in other populations,^20,21 these data do not establish that SII is superior to NLR or PLR for PI risk assessment in hospitalized patients with NSCLC.

The association between SII and incident PI has not been well characterized in hospitalized patients with NSCLC. Therefore, this study aimed to examine the association between admission SII and incident PI during hospitalization and to evaluate the discriminative performance of SII alone and in combination with BMI and CRP. Secondary analyses compared SII with NLR and PLR. We hypothesized that higher SII would be associated with greater odds of PI and that combining SII with BMI and CRP would improve discrimination.

Methods

Participants and Study Design

This single-center retrospective cohort included patients with NSCLC admitted to the Respiratory Medicine Center at Weifang People’s Hospital between March 2022 and April 2024. A total of 1022 records were screened, of which 594 were excluded. The archived screening log categorized 59 exclusions as coexisting autoimmune disease, hematologic malignancy, or other inflammatory conditions; 89 as systemic corticosteroid or immunosuppressant use within 4 weeks before admission; 148 as skin ulcer or wound at admission; and 297 as incomplete clinical or laboratory data. These categories accounted for 593 exclusions; one excluded record had no retrievable category-level designation. The remaining 428 patients comprised the analytic cohort, including 119 who developed PI during hospitalization and 309 who did not. PI was defined and staged according to National Pressure Injury Advisory Panel (NPIAP) criteria.²²

The patient-selection process is summarized in Supplementary Figure S1.

Inclusion criteria were: (1) histopathologically confirmed NSCLC; (2) age ≥18 years; (3) hospitalization for ≥7 days; and (4) availability of the variables required for the primary analysis. Exclusion criteria were: (1) autoimmune disease, hematologic malignancy, or another acute or chronic inflammatory condition; (2) systemic corticosteroid or immunosuppressant use within 4 weeks before admission; (3) a skin ulcer or wound documented at admission; or (4) incomplete clinical or laboratory records. The ≥7-day requirement was part of the cohort eligibility criteria but has not been validated as a PI-specific threshold. The primary analysis used complete cases. Because patient-level records for individuals excluded because of missing data or shorter hospitalization were not retained in the study database, variable-specific missing-data patterns, included-versus-excluded comparisons, and alternative ≥3-day/≥5-day eligibility analyses could not be performed. Data were de-identified before analysis. The protocol was approved by the Ethics Committee of Weifang People’s Hospital (KYLL20240517-11), the requirement for informed consent was waived because of the retrospective use of de-identified data, and the study was conducted in accordance with the Declaration of Helsinki and relevant institutional regulations.

Clinical Data Collection

Demographic, clinical, treatment, and laboratory data were retrospectively abstracted from electronic health records. Baseline variables included age, sex, BMI, smoking status, histologic subtype, recorded comorbidities, and anticancer treatment. Neutrophil, lymphocyte, and platelet counts (×10⁹/L) and CRP (mg/L) were obtained within 24 hours of admission and before the documented onset of incident PI. Patients with a PI or other wound documented at admission were excluded. For patients who developed PI, the anatomical site, NPIAP stage, Braden Scale score at the documented onset of PI, and PI status at discharge were recorded. Admission Braden scores were not available; therefore, the Braden Scale was used descriptively and was not evaluated as a baseline comparator. Records were independently reviewed by two investigators, with discrepancies resolved with a third senior reviewer; formal inter-rater agreement and assessor blinding to SII were not documented. The binary study endpoint was incident PI during the index hospitalization. Exact dates of PI onset, discharge, and in-hospital death were not available in the retrospective dataset; consequently, time-to-event and competing-risk analyses were not performed.

Calculation of SII

Admission SII was calculated using the formula SII = platelet count × neutrophil count / lymphocyte count. Platelet, neutrophil, and lymphocyte counts were expressed in ×10⁹/L, and SII was reported on the corresponding ×10⁹/L scale. For secondary comparator analyses, NLR was calculated as neutrophil count / lymphocyte count and PLR as platelet count / lymphocyte count.

Statistical Analysis

Statistical analyses were performed using IBM SPSS Statistics version 26.0 and Python 3.13 (statsmodels and scikit-learn). Distributional assumptions were assessed using the Kolmogorov–Smirnov test together with graphical inspection and skewness. Approximately normally distributed continuous variables were summarized as mean ± standard deviation and compared using the independent-samples t test; non-normally distributed variables were summarized as median (interquartile range [IQR]) and compared using the Mann–Whitney U-test. SII was non-normally distributed and was therefore summarized using median (IQR). Categorical variables were summarized as n (%) and compared using Pearson’s χ²-test when expected cell counts were adequate; Fisher’s exact test was used when Univariable logistic regression was used to evaluate demographic, clinical, treatment, and laboratory variables in relation to incident PI. The primary multivariable model evaluated SII while adjusting for BMI and CRP, which were clinically relevant and significantly associated with PI in univariable analyses. Because SII is mathematically derived from platelet, neutrophil, and lymphocyte counts, these component counts were not entered simultaneously with SII. SII effects were additionally reported per 100-unit increase for clinical interpretability. A sensitivity model additionally adjusted for age, sex, smoking status, and histologic subtype. Linearity of continuous predictors on the logit scale was examined by adding quadratic terms, and multicollinearity was assessed using variance inflation factors (VIFs). With 119 PI events, the primary three-variable model had 39.7 events per fitted predictor, and the expanded sensitivity model had approximately 14.9 events per parameter. ROC curves and 95% CIs were calculated for BMI, CRP, SII, NLR, PLR, and the combined SII-BMI-CRP model. Pairwise AUCs were compared using DeLong tests. Data-derived cut points were selected by the maximum Youden index and treated as cohort-specific rather than established clinical thresholds. Internal validation of the combined model used 2000 bootstrap resamples to estimate optimism-corrected AUC and calibration slope. Calibration was further assessed using a calibration plot, the Hosmer–Lemeshow test, and the Brier score. Decision-curve analysis compared net benefit across threshold probabilities of 5–50%. All tests were two-sided, with P < 0.05 considered statistically significant.

Results

Baseline Characteristics of the Control and PI Groups

The baseline characteristics are summarized in Table 1. The cohort included 309 patients without PI and 119 who developed PI. Age, sex, smoking status, histologic subtype, recorded comorbidities, and anticancer treatments showed no statistically significant between-group differences. Patients with PI had lower BMI (21.32 ± 2.07 vs 23.47 ± 1.94 kg/m², P < 0.001), higher CRP (20.73 ± 4.68 vs 15.90 ± 5.10 mg/L, P < 0.001), higher neutrophil counts (8.55 ± 1.55 vs 7.51 ± 1.38 ×10⁹/L, P < 0.001), and lower lymphocyte counts (2.25 ± 0.59 vs 2.77 ± 0.59 ×10⁹/L, P < 0.001). Platelet counts did not differ significantly (322.74 ± 36.61 vs 320.55 ± 35.12 ×10⁹/L, P = 0.568). SII was non-normally distributed and was higher in the PI group (median 1223.49 [IQR 963.35–1519.65]) than in the control group (868.37 [694.89–1089.19], P < 0.001).

Table 1.

Baseline Characteristics of the Control and Pressure Injury Groups

Variable	Control Group (n=309)	PI Group (n=119)	P value
Sex [n (%)]			0.570
Male	220 (71.20)	88 (73.95)
Female	89 (28.80)	31 (26.05)
Age (years)	67.70 ± 7.18	67.51 ± 7.36	0.808
BMI (kg/m2)	23.47 ± 1.94	21.32 ± 2.07	<0.001
Smoking status [n (%)]			0.790
Current	66 (21.36)	22 (18.49)
Former	85 (27.51)	35 (29.41)
Never	158 (51.13)	62 (52.10)
Histologic subtype [n (%)]			0.394
Squamous cell carcinoma	116 (37.54)	50 (42.02)
Adenocarcinoma	193 (62.46)	69 (57.98)
Comorbidities [n (%)]
Coronary artery disease	30 (9.71)	15 (12.61)	0.381
COPD	128 (41.42)	46 (38.66)	0.601
Diabetes mellitus	71 (22.98)	24 (20.17)	0.531
Treatment [n (%)]
Chemotherapy	136 (44.01)	44 (36.97)	0.186
Immunotherapy	67 (21.68)	23 (19.33)	0.592
Radiotherapy	61 (19.74)	21 (17.65)	0.622
Laboratory variables
CRP (mg/L)	15.90 ± 5.10	20.73 ± 4.68	<0.001
Platelet count (×109/L)	320.55 ± 35.12	322.74 ± 36.61	0.568
Neutrophil count (×109/L)	7.51 ± 1.38	8.55 ± 1.55	<0.001
Lymphocyte count (×109/L)	2.77 ± 0.59	2.25 ± 0.59	<0.001
SII (×109/L)	868.37 (694.89–1089.19)	1223.49 (963.35–1519.65)	<0.001

Notes: Data are mean ± SD, median (IQR), or n (%), as appropriate. SII was calculated as platelet count × neutrophil count / lymphocyte count and was compared using the Mann–Whitney U-test; other continuous variables were compared using independent-samples t tests. Categorical P values are from Pearson χ²-tests.

Abbreviations: BMI, body mass index; COPD, chronic obstructive pulmonary disease; CRP, C-reactive protein; IQR, interquartile range; PI, pressure injury; SII, systemic immune-inflammation index.

Characteristics of Pressure Injury Among Patients with PI

Among the 119 patients who developed PI, 88 (73.95%) were male and 31 (26.05%) were female. At the documented onset of PI, 97 (81.51%) had a Braden Scale score ≤15 and 22 (18.49%) had a score >15; these onset scores were descriptive and were not baseline predictors. The coccyx was the most frequent site (n = 76, 63.87%), followed by the buttock (n = 15, 12.61%) and trochanter (n = 14, 11.76%). Stage 2 PI was most common (n = 96, 80.67%), followed by deep tissue pressure injury (n = 11, 9.24%), Stage 1 (n = 7, 5.88%), unstageable PI (n = 3, 2.52%), and Stage 3 (n = 2, 1.68%); no Stage 4 injuries were recorded. At discharge, 26 patients (21.85%) had improved PI, 73 (61.34%) were unchanged, and 20 (16.81%) had aggravated PI (Table 2).

Table 2.

Characteristics of Pressure Injury Among Patients Who Developed Pressure Injury

Variable	n	%
Sex
Male	88	73.95
Female	31	26.05
Braden Scale score at documented PI onset
≤15	97	81.51
>15	22	18.49
Anatomical site
Coccyx	76	63.87
Buttock	15	12.61
Trochanter	14	11.76
Back	5	4.20
Heel/malleolus	4	3.36
Other	5	4.20
Stage of PI (NPIAP)
Stage 1	7	5.88
Stage 2	96	80.67
Stage 3	2	1.68
Stage 4	0	0.00
Deep tissue pressure injury	11	9.24
Unstageable	3	2.52
PI status at discharge
Improved	26	21.85
Unchanged	73	61.34
Aggravated	20	16.81

Notes: Braden Scale scores were recorded at the documented onset of pressure injury and were used descriptively; they were not admission-baseline predictors.

Abbreviations: NPIAP, National Pressure Injury Advisory Panel; PI, pressure injury.

Univariable and Multivariable Logistic Regression Analyses of Factors Associated with Pressure Injury

Complete univariable regression results are presented in Table 3. Lower BMI, higher CRP, higher neutrophil count, lower lymphocyte count, and higher SII were associated with PI, whereas platelet count was not (P = 0.567). For SII, each 100-unit increase was associated with 31.4% higher odds of PI in univariable analysis (OR = 1.314, 95% CI: 1.224–1.411, P < 0.001). In the primary multivariable model (Table 4), SII remained independently associated with PI (OR per 100 units = 1.339, 95% CI: 1.223–1.466, P < 0.001), together with BMI (OR = 0.549, 95% CI: 0.467–0.645, P < 0.001) and CRP (OR = 1.254, 95% CI: 1.173–1.341, P < 0.001). VIFs were 1.061 for SII, 1.043 for BMI, and 1.044 for CRP. There was no evidence of departure from linearity when quadratic terms were jointly added to the model (likelihood-ratio P = 0.755). In the sensitivity model additionally adjusted for age, sex, smoking status, and histologic subtype, the associations remained similar for SII (OR per 100 units = 1.350, 95% CI: 1.231–1.481), BMI (OR = 0.545, 95% CI: 0.463–0.641), and CRP (OR = 1.257, 95% CI: 1.174–1.346); the additional covariates were not statistically significant.

Table 3.

Complete Univariable Logistic Regression Analyses of Factors Associated with Incident Pressure Injury

Variable	β	SE	Wald χ2	P value	OR	95% CI
Age, per year	−0.004	0.015	0.059	0.807	0.996	0.968–1.026
Male sex	0.138	0.244	0.322	0.570	1.148	0.712–1.852
BMI, per kg/m2	−0.523	0.064	66.733	<0.001	0.593	0.523–0.672
Smoking: former vs current	0.211	0.318	0.443	0.506	1.235	0.663–2.303
Smoking: never vs current	0.163	0.288	0.321	0.571	1.177	0.669–2.071
Adenocarcinoma vs squamous	−0.187	0.220	0.724	0.395	0.829	0.539–1.276
Coronary artery disease	0.294	0.336	0.763	0.383	1.341	0.694–2.594
COPD	−0.115	0.221	0.273	0.601	0.891	0.578–1.374
Diabetes mellitus	−0.166	0.265	0.393	0.531	0.847	0.503–1.425
Chemotherapy	−0.293	0.222	1.742	0.187	0.746	0.483–1.153
Immunotherapy	−0.145	0.270	0.287	0.592	0.865	0.510–1.469
Radiotherapy	−0.138	0.280	0.243	0.622	0.871	0.504–1.507
CRP, per mg/L	0.201	0.026	57.666	<0.001	1.222	1.161–1.287
Platelet count, per 109/L	0.002	0.003	0.328	0.567	1.002	0.996–1.008
Neutrophil count, per 109/L	0.517	0.084	37.559	<0.001	1.677	1.421–1.979
Lymphocyte count, per 109/L	−1.544	0.218	50.025	<0.001	0.214	0.139–0.328
SII, per 100 units	0.273	0.036	56.868	<0.001	1.314	1.224–1.411

Notes: Outcome: incident pressure injury during hospitalization. SII was calculated from platelet, neutrophil, and lymphocyte counts and rescaled per 100 units for interpretability.

Abbreviations: BMI, body mass index; CI, confidence interval; COPD, chronic obstructive pulmonary disease; CRP, C-reactive protein; OR, odds ratio; SE, standard error; SII, systemic immune-inflammation index.

Table 4.

Multivariable Logistic Regression Analyses of Factors Associated with Incident Pressure Injury

Variables	β	SE	Wald χ2	P value	OR	95% CI
Primary model
SII, per 100 units	0.292	0.046	39.715	<0.001	1.339	1.223–1.466
BMI, per kg/m2	−0.599	0.082	53.139	<0.001	0.549	0.467–0.645
CRP, per mg/L	0.227	0.034	44.005	<0.001	1.254	1.173–1.341
Sensitivity model: additionally adjusted for age, sex, smoking, and histology
SII, per 100 units	0.300	0.047	40.435	<0.001	1.350	1.231–1.481
BMI, per kg/m2	−0.608	0.083	53.492	<0.001	0.545	0.463–0.641
CRP, per mg/L	0.229	0.035	42.931	<0.001	1.257	1.174–1.346
Age, per year	−0.016	0.020	0.640	0.424	0.984	0.945–1.024
Male sex	0.100	0.334	0.090	0.764	1.105	0.575–2.126
Smoking: former vs current	0.383	0.463	0.686	0.407	1.467	0.592–3.632
Smoking: never vs current	0.350	0.421	0.694	0.405	1.420	0.623–3.237
Adenocarcinoma vs squamous	−0.360	0.306	1.385	0.239	0.698	0.383–1.271

Notes: Primary-model VIFs were 1.061 for SII, 1.043 for BMI, and 1.044 for CRP. Addition of standardized quadratic terms provided no evidence of nonlinearity (joint likelihood-ratio P=0.755). With 119 PI events, the primary model had 39.7 events per fitted predictor; the expanded sensitivity model had approximately 14.9 events per parameter.

Abbreviations: BMI, body mass index; CI, confidence interval; CRP, C-reactive protein; OR, odds ratio; SE, standard error; SII, systemic immune-inflammation index; VIF, variance inflation factor.

Discrimination, Internal Validation, Calibration, and Decision-Curve Analyses

ROC results are shown in Table 5 and Figure 1. The AUC was 0.763 (95% CI: 0.711–0.815) for SII, 0.775 (0.725–0.825) for BMI, and 0.755 (0.705–0.805) for CRP. SII showed discrimination comparable to BMI (DeLong P = 0.755) and CRP (P = 0.837). Secondary comparison with other blood-cell indices showed an AUC of 0.774 (0.724–0.824) for NLR and 0.707 (0.651–0.763) for PLR; SII was comparable to NLR (P = 0.188) and had a higher AUC than PLR (P = 0.002). The data-derived Youden cut points were SII ≥896.28 (sensitivity 85.7%, specificity 55.7%), BMI ≤21.845 kg/m² (sensitivity 62.2%, specificity 80.9%), and CRP ≥20.5 mg/L (sensitivity 57.1%, specificity 81.2%). The combined SII-BMI-CRP model had an apparent AUC of 0.908 (95% CI: 0.880–0.936), which was higher than the AUCs of SII, BMI, and CRP individually (all DeLong P < 0.001). At the corresponding probability cut point of 0.253, sensitivity was 86.6% and specificity was 80.6%. Bootstrap internal validation yielded an optimism-corrected AUC of 0.904 (95% CI: 0.878–0.934) and an optimism-corrected calibration slope of 0.965. Bootstrap resampling also showed variability in data-derived cut points: the median SII cut point was 925.85 (95% bootstrap range 886.11–1273.28), and the median combined-model probability cut point was 0.270 (95% bootstrap range 0.152–0.393), indicating cohort dependence. Calibration was acceptable in the development cohort (Hosmer–Lemeshow χ² = 3.476, P = 0.901; Brier score = 0.110; Figure 2). Decision-curve analysis showed greater net benefit for the combined model than treat-all, treat-none, or SII alone across threshold probabilities of 5%–50% (Figure 3); external validation is needed to confirm clinical utility.

Table 5.

ROC Discrimination, Data-Derived Cut Points, and Internal Validation

Variables	AUC	95% CI	Data-Derived Cut Point	Sensitivity (%)	Specificity (%)	DeLong P vs SII
BMI	0.775	0.725–0.825	≤21.845 kg/m2	62.2	80.9	0.755
CRP	0.755	0.705–0.805	≥20.5 mg/L	57.1	81.2	0.837
SII	0.763	0.711–0.815	≥896.28	85.7	55.7	Reference
NLR	0.774	0.724–0.824	≥2.855	85.7	58.9	0.188
PLR	0.707	0.651–0.763	≥141.406	55.5	80.3	0.002
Combined SII-BMI-CRP model	0.908	0.880–0.936	Probability ≥0.253	86.6	80.6	<0.001

Notes: The directions shown correspond to higher PI risk. Cut points were selected by the maximum Youden index in the development cohort and are cohort-specific rather than established clinical thresholds. For the combined model, 2000 bootstrap resamples yielded an optimism-corrected AUC of 0.904 (95% CI 0.878–0.934), an optimism-corrected calibration slope of 0.965, Hosmer–Lemeshow χ²=3.476 (P=0.901), and Brier score=0.110. AUC comparisons used DeLong tests.

Abbreviations: AUC, area under the curve; BMI, body mass index; CI, confidence interval; CRP, C-reactive protein; NLR, neutrophil-to-lymphocyte ratio; PI, pressure injury; PLR, platelet-to-lymphocyte ratio; ROC, receiver operating characteristic; SII, systemic immune-inflammation index.

Figure 1.

Receiver operating characteristic curves showing discrimination of BMI, CRP, SII, and the combined SII-BMI-CRP model for incident pressure injury in the study cohort.

Figure 2.

Calibration plot for the combined SII-BMI-CRP model in the development cohort. Points show observed pressure-injury proportions across deciles of predicted risk; the dashed line represents ideal calibration.

Figure 3.

Decision-curve analysis comparing the combined SII-BMI-CRP model, SII alone, treat-all, and treat-none strategies across threshold probabilities of 5–50%.

Discussion

In this single-center retrospective cohort, higher admission SII was independently associated with incident PI after adjustment for BMI and CRP, and this association remained stable after additional adjustment for measured demographic and tumor-related covariates. SII showed discrimination comparable to BMI, CRP, and NLR as an individual marker. The combined SII-BMI-CRP model demonstrated substantially higher discrimination with minimal bootstrap-estimated optimism and acceptable calibration, supporting further external evaluation of this multivariable approach.

One of the primary findings of our study is the significant elevation of SII in patients who developed PI. SII, a composite marker derived from neutrophil, platelet, and lymphocyte counts, reflects both inflammatory activation and immune status.^23,24 Prior studies suggest that neutrophil elevations may accompany systemic inflammation and are associated with the release of reactive oxygen species, proteolytic enzymes, and pro-inflammatory cytokines such as TNF-α and IL-6.^25,26 These pathways have been linked to endothelial dysfunction, thrombosis, and altered local perfusion,^25,26 which could plausibly influence tissue tolerance under pressure; however, these mechanisms were not directly assessed in the present cohort. Platelets can participate in clot formation, vascular occlusion, and inflammatory-cell recruitment in other inflammatory and thrombotic settings.^27,28 Reduced lymphocyte counts may reflect altered adaptive immune status and have been linked to impaired cytokine regulation, tissue repair, and epithelial regeneration.²⁹ In NSCLC patients, already prone to chronic inflammation from malignancy, chemotherapy, or cachexia, the elevated SII may be associated with reduced skin resilience and impaired wound repair.^30–32 Taken together, SII may function as a composite marker of inflammatory-immune status rather than a measure of a single causal pathway. Chronic inflammation related to tumor progression, paraneoplastic phenomena, and treatment provides biological context for the observed association.^33,34 Additionally, cancer-related cachexia, characterized by systemic inflammation, muscle wasting, and metabolic imbalance, has been associated with reduced tissue reserves and may increase vulnerability to PI.^34–36 Although inflammatory and thrombotic pathways have been described in NSCLC, the present study did not measure endothelial injury, microvascular thrombosis, or hypoperfusion. These mechanisms therefore remain hypotheses rather than demonstrated mediators of the SII–PI association.

Importantly, platelet count itself neither differed significantly between the PI and non-PI groups nor showed an association with PI in univariable regression. Thus, the SII finding should not be interpreted as evidence that thrombocytosis or platelet-mediated thrombosis caused PI in this cohort.

Our data also show significantly higher CRP levels and lower lymphocyte counts in patients with PI, further supporting the link between systemic inflammation and PI occurrence. CRP, an acute-phase reactant produced by hepatocytes in response to IL-6, IL-1β, and TNF-α, is a sensitive marker of systemic inflammation.^37,38 Its elevation reflects both tumor-related processes, such as necrosis and immune cell infiltration, and non-tumor-related conditions like infection, malnutrition, and physical stress.^39,40 Elevated CRP may accompany both systemic inflammation and hypercatabolic states associated with impaired tissue repair. The relationship between CRP, endothelial function, and microcirculatory changes is complex,⁴¹ and should not be interpreted here as evidence that CRP directly caused ischemia or tissue necrosis. The lower lymphocyte counts observed in patients with PI are consistent with altered adaptive immune status, but immune function was not directly measured. Reduced lymphocyte counts, particularly T-cell depletion, have been associated with impaired cytokine regulation, delayed healing, and infection risk, all of which provide biological context for PI.^29,42,43

Lower BMI was independently associated with PI in hospitalized patients with NSCLC and showed an AUC of 0.775 in this cohort. Malnutrition is an established correlate of impaired skin integrity.^44–46 Inadequate protein and energy intake has been associated with impaired collagen synthesis, fibroblast activity, and keratinocyte proliferation, potentially reducing skin resilience.⁴⁷ Malnutrition is also associated with reduced skin turgor and subcutaneous tissue reserves, which may reduce tolerance to pressure and shear forces.^48,49 Cancer-related cachexia in NSCLC may compound nutritional and inflammatory vulnerability through systemic inflammation, muscle wasting, and progressive weight loss, which nutritional support alone cannot fully reverse.^50,51 Loss of muscle and adipose tissue may increase bony prominence and local pressure concentration.⁵² This anatomical vulnerability, combined with impaired immune function and delayed healing due to nutrient deficiencies (eg, zinc, vitamin C, arginine),⁵³ may increase susceptibility to PI. Because albumin, prealbumin, detailed nutritional intake, and sarcopenia were not measured in the present analysis, BMI should be interpreted as a limited anthropometric marker rather than a complete measure of nutritional status. These unmeasured factors remain potential confounders.

The multivariable analysis identified independent associations of SII, CRP, and BMI with PI, consistent with the multifactorial nature of PI in hospitalized patients with NSCLC. Individual discrimination was comparable for SII (AUC = 0.763), BMI (AUC = 0.775), CRP (AUC = 0.755), and NLR (AUC = 0.774), indicating that SII provided discrimination similar to these individual markers. The combined SII-BMI-CRP model demonstrated substantially higher discrimination (apparent AUC = 0.908; optimism-corrected AUC = 0.904) with limited bootstrap-estimated optimism, supporting further external validation of the combined approach. Regarding PI characteristics, the coccyx was the most common site and Stage 2 was the most frequent category. These findings are consistent with established knowledge linking mechanical loading over bony prominences to PI occurrence.^54,55 At the documented assessment, 16 of 119 patients (13.45%) had Stage 3, Stage 4, deep-tissue pressure injury, or an unstageable injury. At discharge, 93 of 119 (78.15%) were unchanged or aggravated. These figures describe stage at assessment and discharge status, respectively, and should not be interpreted as a 27% progression rate because individual stage transitions were not available.

Given the independent association between SII and PI in this cohort, it is also important to emphasize established preventive strategies for reducing PI risk among hospitalized patients. Preventive measures, including routine repositioning, pressure-redistributing surfaces, and skin care protocols, are foundational in PI management and widely recommended in clinical guidelines.^56–58 Nutritional optimization, especially adequate protein and micronutrient intake, supports tissue tolerance and wound healing.⁵⁹ Early risk assessment using validated tools, along with multidisciplinary nursing interventions, offers a proactive framework to reduce PI incidence.^60–62 Biomarker-based risk stratification is best considered alongside established preventive strategies and clinical risk assessment. In this context, SII may provide complementary biological information, although its incremental value beyond baseline Braden Scale assessment remains to be established prospectively.

Several limitations should be considered. First, the retrospective, single-center design is susceptible to selection and information bias. Of the screened records, 297 were excluded because of incomplete clinical or laboratory data; because patient-level information for these excluded records was unavailable, variable-specific missingness and included-versus-excluded comparisons could not be assessed, and complete-case selection may have influenced the observed associations. Second, the ≥7-day eligibility criterion has not been validated as a PI-specific threshold. Shorter admissions could not be re-evaluated, and the binary logistic model did not account for time to PI or competing discharge/death; longer hospitalization may therefore increase opportunity for PI ascertainment. Third, only admission inflammatory measurements were analyzed, admission Braden scores were unavailable, and several clinically relevant factors—including mobility/performance status, disease severity, infection, detailed nutritional status, sarcopenia, repositioning practices, and pressure-redistributing devices—were not fully measured. Finally, the combined model and Youden cut points were developed in the same cohort. Bootstrap validation indicated limited optimism and acceptable internal calibration, but multicenter external validation and prospective comparison with baseline clinical risk assessment are needed before routine clinical implementation.

Conclusion

In conclusion, higher admission SII was independently associated with incident PI in hospitalized patients with NSCLC. SII showed discrimination comparable to BMI, CRP, and NLR, while a model combining SII, BMI, and CRP demonstrated good discrimination, stable bootstrap internal validation, and acceptable calibration. These findings support further evaluation of SII as an adjunctive biomarker for PI risk stratification in hospitalized patients with NSCLC. External validation and prospective comparison with baseline clinical risk tools, including the Braden Scale, should precede routine clinical implementation.

Funding Statement

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Abbreviations

SII, systemic immune-inflammation index; PI, pressure injury; NSCLC, non-small cell lung cancer; ROC, receiver operating characteristic; NLR, neutrophil-to-lymphocyte ratio; PLR, platelet-to-lymphocyte ratio; CRP, C-reactive protein; BMI, body mass index.

Data Sharing Statement

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Author Contributions

All authors made significant contributions to the work reported, whether in conception, study design, execution, acquisition of data, analysis and interpretation of results. They took part in drafting, revising or critically reviewing the article and gave approval for the final version published and agreed on the journal to which the article has been submitted and agreed to be accountable for all aspects of the work.

Disclosure

The authors declare that they have no conflicts of interest.