Association Between Whole Blood Cell-Derived Inflammatory Markers and All-Cause Mortality in Patients with Sepsis-Related Acute Kidney Injury and Development of a Machine Learning Prognostic Model

Association Between Whole Blood Cell-Derived Inflammatory Markers and All-Cause Mortality in Patients with Sepsis-Related Acute Kidney Injury and Development of a Machine Learning Prognostic Model | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Association Between Whole Blood Cell-Derived Inflammatory Markers and All-Cause Mortality in Patients with Sepsis-Related Acute Kidney Injury and Development of a Machine Learning Prognostic Model Xinghe Shangguan, Yongjie Luo, Xiayoumei Wu, Jianning Xu, Yuanqi Gong, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7986649/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Sepsis-associated acute kidney injury (SA-AKI) is a common and severe complication of sepsis, associated with significantly increased patient mortality. Whole blood cell-derived inflammatory markers, such as the neutrophil-to-lymphocyte ratio (NLR) and the systemic immune-inflammation index (SII), are valuable tools for assessing inflammatory status and prognosis due to their ready availability and cost-effectiveness. However, the association between these markers and prognosis in patients with SA-AKI requires further investigation. This study aims to clarify the relationship between these inflammatory markers and all-cause mortality in SA-AKI patients and to develop a prognostic model using machine learning methods, thereby providing new evidence to support clinical decision-making.. Methods This study included patients with SA-AKI, defined by Sepsis-3 and KDIGO criteria, from the MIMIC-IV database (development cohort) and an external cohort from The Second Affiliated Hospital of Nanchang University (validation cohort). We collected demographic data, vital signs, laboratory parameters (including NLR, NPAR, PNR, and SII), comorbidities, and treatment information. The primary outcome was 14-day all-cause mortality following ICU admission.Kaplan-Meier survival analysis and multivariate Cox regression models were used to evaluate the association between inflammatory markers and mortality. For model development, the MIMIC-IV cohort was randomly split into training (70%) and testing (30%) sets. Feature selection was performed using LASSO regression, the Boruta algorithm, and univariate logistic regression. Six machine learning models (Random Forest, Logistic Regression, XGBoost, GAMboost, CatBoost, and GBM) were subsequently developed. Five-fold cross-validation was employed for model tuning, and the SMOTE (Synthetic Minority Over-sampling Technique) algorithm was applied to address class imbalance.Model performance was evaluated using the Area Under the Receiver Operating Characteristic Curve (AUC) and Brier score. Model interpretability was assessed using SHAP (SHapley Additive exPlanations), and clinical utility was determined via Decision Curve Analysis (DCA). The best-performing model was then validated using the external cohort. Results A total of 4,311 patients with SA-AKI were included in the development cohort, of whom 696 (16.1%) died within 14 days. Survival analysis demonstrated that elevated NLR, SII, and NPAR, as well as a low PNR, were significantly associated with 14-day all-cause mortality. Following feature selection, 17 predictors were used to develop six machine learning models. The CatBoost model achieved the best discrimination, with an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.861 on the internal test set and 0.755 on the external validation cohort. Decision curve analysis confirmed its potential clinical utility across a wide range of threshold probabilities. Furthermore, SHAP analysis identified lactate and blood urea nitrogen as the most significant contributors to the model's predictions. Conclusion The predictive model developed using the CatBoost algorithm effectively assesses the 14-day mortality risk in patients with SA-AKI. The model demonstrated robust predictive performance and clinical applicability in both internal and external validation cohorts. This tool holds promise for assisting clinicians with risk stratification and guiding personalized treatment strategies for SA-AKI patients. Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Introduction Sepsis, defined as life-threatening organ dysfunction caused by a dysregulated host response to infection, ranks among the leading causes of death globally[ 1 ]. Acute kidney injury (AKI) is a frequent complication, occurring in 40–50% of sepsis patients and increasing mortality 6- to 8-fold, representing one of the most severe complications of this syndrome [ 1 , 2 ].This condition, sepsis-associated acute kidney injury (SA-AKI), is common in critically ill patients and strongly associated with poor outcomes. These adverse outcomes include increased risks of chronic kidney disease, cardiovascular events, and mortality, in addition to prolonged hospital stays and higher healthcare costs [ 1 , 3 – 5 ]. Consequently, improving the prognostic assessment of patients with SA-AKI holds significant clinical importance. Whole blood cell-derived inflammatory markers, including the neutrophil-to-lymphocyte ratio (NLR), neutrophil percentage to albumin ratio (NPAR), platelet to neutrophil ratio (PNR), and systemic immune inflammation index (SII), are increasingly recognized for their utility in disease diagnosis, prognostic assessment, and severity grading [ 6 ]. As these markers are derived from components (e.g., neutrophils, lymphocytes, platelets, and albumin) measured in routine complete blood counts and biochemical panels, they are cost-effective, reproducible, and widely accessible.These indices reflect the body's inflammatory state and immune balance [ 7 , 8 ] and have shown utility in diverse fields such as cardiovascular disease, oncology, and infectious diseases. Specifically in sepsis, both NLR and SII have been established as predictors of 90-day mortality, with elevated values indicating a poor prognosis [ 9 – 11 ]. However, research on the association between NPAR and PNR with sepsis prognosis remains limited. Moreover, the prognostic value of these specific markers (NLR, SII, NPAR, and PNR) in the SA-AKI subpopulation is not well established. Therefore, investigating the relationship between these inflammatory indices and the prognosis of SA-AKI patients may deepen our understanding of the pathophysiology driving poor outcomes and offer novel targets for clinical management. While previous research has predominantly focused on sepsis diagnosis and treatment, and some studies have examined the correlation between inflammatory markers and sepsis or AKI, the specific relationship between whole-blood cell-derived inflammatory markers and all-cause mortality in SA-AKI patients remains understudied. Concurrently, machine learning (ML) methods are increasingly utilized for predicting clinical events across various medical fields [ 12 , 13 ]. Although several predictive models have been developed for sepsis[ 14 – 16 ], few have specifically incorporated whole blood cell-derived inflammatory markers to predict outcomes in the SA-AKI subpopulation. Therefore, this study has two primary objectives. First, we aim to thoroughly analyze the association between these inflammatory markers and all-cause mortality in SA-AKI patients. Second, we aim to develop and validate a machine learning prognostic model incorporating these markers, providing robust evidence for clinical risk stratification and decision-making. Methods and Data Data Sources and Study Cohort s This retrospective study utilized two distinct cohorts. The primary development and testing cohort was extracted from the Medical Information Mart for Intensive Care (MIMIC-IV, v3.1) database. An independent external validation cohort was assembled from patients diagnosed with SA-AKI at The Second Affiliated Hospital of Nanchang University between 2017 and 2025. Ethical Considerations The MIMIC-IV (v3.1) database is a large, single-center, open-access repository containing de-identified data for 730,141 intensive care unit admissions at the Beth Israel Deaconess Medical Center (BIDMC) in the United States between 2008 and 2019 (Johnson et al., 2022). The establishment of this database was approved by the Institutional Review Boards (IRBs) of the Massachusetts Institute of Technology (MIT) and BIDMC. As the data are publicly available and de-identified, the requirement for individual patient consent for the present study was waived. One of the authors (Certification No. 13822018) completed the required training and obtained access to the database. The study protocol for the external validation cohort was approved by the Medical Ethics Committee of The Second Affiliated Hospital of Nanchang University (Approval No.: IIT-O-2025-319). All data from this cohort were anonymized prior to analysis. This study conforms to the principles of the Declaration of Helsinki. 1.3 Inclusion and Exclusion Criteria Patient Selection We identified patients from the databases who met the diagnostic criteria for sepsis according to the Sepsis-3 guidelines[ 17 ]. From this cohort, we selected adult patients (aged ≥ 18 years) who also developed acute kidney injury (AKI), defined in accordance with the Kidney Disease: Improving Global Outcomes (KDIGO) guidelines. For patients with multiple hospitalizations, only data from the first ICU admission of their first hospitalization were included.Patients were excluded if they met any of the following criteria:ICU length of stay < 24 hours (due to early discharge or death).Missing data for lymphocyte, neutrophil, or platelet counts, or serum albumin levels, which are necessary to calculate the inflammatory markers.A diagnosis of active malignancy. Data Extraction and Variables For the development cohort, data were extracted from the MIMIC-IV (v3.1) database using Structured Query Language (SQL) in Navicat Premium (version 17). For the external validation cohort, data were retrieved from the electronic medical and record systems of The Second Affiliated Hospital of Nanchang University.To ensure all variables represented the baseline status, we collected data recorded within the first 24 hours of each patient's ICU admission. The extracted variables included the following categories:Demographics: Age and sex.Comorbidities: A history of hypertension (HTN), pneumonia (PNA), cerebrovascular accident (CVA), chronic kidney disease (CKD), type 2 diabetes mellitus (T2DM), heart failure (HF), myocardial infarction (MI), ischemic heart disease (IHD), chronic obstructive pulmonary disease (COPD), and COVID-19 infection (COV).Vital Signs: The worst values within the first 24 hours for systolic blood pressure (SBP), heart rate (HR), respiratory rate (RR), body temperature, and peripheral oxygen saturation (SpO₂).Laboratory Results: The first-measured values for lactate, blood glucose, arterial blood pH, partial pressure of oxygen (PO₂), hemoglobin (Hb), creatinine, blood urea nitrogen (BUN), alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin, potassium, sodium, prothrombin time (PT), and activated partial thromboplastin time (aPTT). The components for the inflammatory indices (neutrophil, lymphocyte, and platelet counts; albumin) were also extracted from these initial tests.Fluid Balance: Total 24-hour fluid balance.Treatments: Use of continuous renal replacement therapy (CRRT), mechanical ventilation, neuromuscular blockers, sedation and analgesia, vasoactive drugs (e.g., vasopressors), immunosuppressants, nephrotoxic drugs, glucocorticoids, and antihypertensive therapy. NLR was calculated as neutrophil count / lymphocyte count, NPAR as neutrophil percentage / serum albumin, PNR as platelet count / neutrophil count, and SII as platelet count × neutrophil count / lymphocyte count. Primary Outcome The primary outcome was defined as 14-day all-cause mortality following ICU admission. Statistical Analysis Variables with more than 20% missing values were excluded from the analysis. For the remaining variables, missing data were imputed using Multiple Imputation by Chained Equations (MICE). Baseline characteristics were summarized and compared between 14-day survivors and non-survivors. Continuous variables were presented as mean (standard deviation, SD) or median (interquartile range, IQR) based on data distribution (e.g., assessed by normality tests). Intergroup differences were compared using Student's t-test or the Mann-Whitney U test, as appropriate. Categorical variables were presented as frequencies and percentages (%). These were compared using the Pearson's chi-square test or Fisher's exact test, particularly when expected cell counts were low (e.g., < 5).All analyses were performed using R software (version 4.5.1) and DecisionLine (version 1.1.7.3)[ 18 ]. Survival and Association AnalysisTo assess the association between inflammation markers and 14-day mortality, we first generated Kaplan-Meier survival curves, stratifying patients by baseline inflammation levels (quartiles), and compared groups using the log-rank test.We then employed multivariate Cox proportional hazards models to calculate hazard ratios (HR) and 95% confidence intervals (CI) for each inflammatory marker (NLR, SII, PNR, and NPAR). These markers were analyzed as both continuous variables and as categorical variables (quartiles), with the first quartile (Q1) serving as the reference group.Three sequential models were constructed:Model 1: Unadjusted.Model 2: Adjusted for age and sex.Model 3: Adjusted for the covariates in Model 2 plus all baseline demographic, comorbidity, vital sign, laboratory, and treatment variables listed in Section 2.4.Restricted cubic splines (RCS) were used to explore potential non-linear associations between each continuous marker and 14-day mortality. Finally, we conducted subgroup analyses to evaluate the prognostic consistency of these markers across different strata, including age (< 65 vs. ≥65 years), sex, history of chronic kidney disease (CKD), and heart failure (HF).2.8 Machine Learning Model DevelopmentData Splitting and PreprocessingThe MIMIC-IV cohort (development cohort) was randomly partitioned into a training set (70%) and an internal test set (30%). Within the training set, all selected features were standardized using the Z-score method. To address the class imbalance between survivors and non-survivors, the Synthetic Minority Over-sampling Technique (SMOTE) was applied to the training data.Feature SelectionWe performed feature selection on the training set to identify the most relevant predictors. Three methods were employed:LASSO (Least Absolute Shrinkage and Selection Operator) regression.The Boruta algorithm (a random forest-based wrapper method).Univariate logistic regression (retaining variables with p < 0.05).Only features identified as significant by all three methods were retained for the final model building.Model TrainingUsing the selected features and the preprocessed training data, we developed six machine learning models: Random Forest, Logistic Regression, XGBoost, GAMboost, CatBoost, and Gradient Boosting Machine (GBM). Model training and hyperparameter optimization (via grid search) were performed using 5-fold cross-validation on the training set. Model Evaluation and InterpretationModel Performance and SelectionThe performance of the six models was evaluated on the internal test set using the Area Under the Receiver Operating Characteristic Curve (AUC), Brier score, precision, recall, F-beta score, and specificity. The model demonstrating the best overall performance was selected as the final model.External Validation and Clinical UtilityThe final model was then validated using the independent external cohort from The Second Affiliated Hospital of Nanchang University. Model calibration (the agreement between predicted probabilities and observed outcomes) was assessed using the Hosmer-Lemeshow goodness-of-fit test. The clinical utility and net benefit of the model were visualized using Decision Curve Analysis (DCA).Model InterpretabilityTo explain the model's predictions, we employed SHAP (SHapley Additive exPlanations). Feature contributions and overall importance were visualized using SHAP summary plots. Results Cohort Characteristics Baseline Patient CharacteristicsThis study ultimately included 4,311 patients with SA-AKI admitted to the ICU. Of this cohort, 3,615 patients survived (survivor group) and 696 died within 14 days (non-survivor group). The patient selection process is detailed in Fig. 1. The median age of the overall cohort was 65.0 years (IQR 54.0–75.0), and 40.6% were female.Baseline characteristics, compared between the 14-day survivor and non-survivor groups, are presented in Table 1. The non-survivor group was significantly older than the survivor group (median 70.0 vs. 64.0 years, p < 0.001), although there was no significant difference in sex distribution (p = 0.727). Regarding comorbidities, the non-survivor group had significantly higher prevalences of heart failure, myocardial infarction, ischemic heart disease, and chronic obstructive pulmonary disease (all p < 0.002).Comparisons of vital signs and laboratory indicators (within the first 24 hours of ICU admission) showed that the non-survivor group had significantly higher heart rates, respiratory rates, and lactate levels, as well as lower arterial pH and PaO₂ (all p < 0.001). Furthermore, non-survivors presented with significantly elevated creatinine, blood urea nitrogen, total bilirubin, prothrombin time (PT), and activated partial thromboplastin time (aPTT) (all p < 0.001), and lower hemoglobin levels (9.8 vs. 10.4 g/dL, p = 0.009).For the primary inflammatory markers, the non-survivor group exhibited significantly higher NLR (13.09 vs. 9.06, p < 0.001), NPAR (29.34 vs. 27.23, p = 0.002), and SII (2018.98 vs. 1555.91, p < 0.001). Conversely, PNR was significantly lower in non-survivors (14.47 vs. 18.11, p = 0.046). The non-survivor group also had significantly higher rates of receiving CRRT, vasoactive drugs, glucocorticoids, and mechanical ventilation (all p < 0.05). Association of Inflammatory Markers with 14-Day All-Cause Mortality Cox proportional hazards models were used to assess the association between exposure factors and 14-day mortality. Three sequential models were constructed: Model 1 was unadjusted; Model 2 adjusted for age and sex; and Model 3 was fully adjusted for over 30 confounding variables, including comorbidities (e.g., hypertension, diabetes, heart failure, CKD, COPD), treatments (e.g., CRRT, vasoactive drugs), and key laboratory values (e.g., lactate, creatinine, hemoglobin, blood glucose). The hazard ratios (HR) and 95% confidence intervals (95% CI) for 14-day mortality risk, stratified by inflammatory marker groups (NLR, SII, PNR, NPAR), are presented in Table 2. Kaplan-Meier analysis (Fig. 3) demonstrated a significant difference in survival based on inflammatory marker levels. Patients in the highest tertile (T3) of NLR and SII, as well as those in the highest quartile (Q4) of NPAR, had significantly higher 14-day all-cause mortality than those in the lowest reference groups. Conversely, PNR showed an inverse association: patients in the lowest quartile (Q1) had significantly higher mortality than those in the highest quartile (Q4) (all log-rank p < 0.05).Univariate Cox regression analysis (Table 2) confirmed that all four markers, when analyzed as continuous variables, were significantly associated with 14-day mortality risk in patients with SA-AKI (all trend p < 0.05).In the fully adjusted multivariate model (Model 3), these associations remained significant. For this analysis, NLR and SII were grouped by tertiles, while NPAR and PNR were grouped by quartiles.High NLR (T3 vs. T1) yielded an HR of 1.50 (95% CI: 1.23–1.83).High NPAR (Q4 vs. Q1) yielded an HR of 1.23 (95% CI: 1.02–1.49).For SII, the combined upper tertiles (T2/T3 vs. T1) yielded an HR of 1.19 (95% CI: 1.10–1.39).In contrast, high PNR (Q4 vs. Q1) was protective, indicating a negative correlation with mortality (HR: 0.73, 95% CI: 0.58–0.93). Restricted cubic spline (RCS) analysis was employed to evaluate potential non-linear associations between the inflammatory markers and 14-day mortality (Fig. 2). The analysis revealed significant non-linear relationships for NLR, NPAR, and PNR (all p for non-linearity < 0.05), whereas SII demonstrated a linear relationship with mortality risk (p for non-linearity = 0.642). Specifically, the mortality risk began to increase significantly when NLR exceeded 9.54 and when NPAR exceeded 27.56, with the risk for both markers plateauing at high values (approx. 50–100). Conversely, PNR exhibited a protective effect; the risk was highest at low values and began to decrease significantly once PNR rose above 17.48, stabilizing at levels between 100 and 200. For SII, the risk showed a continuous linear increase for all values above 1610.77. Subgroup Analysis Subgroup analyses were conducted based on numerous baseline comorbidities and treatments, as detailed in Fig. 4. Significant interactions were detected: NLR interacted with CRRT, AHT, and MI; NPAR interacted with IMMUNOS, VP, and HF; SII showed interactions with IHD, AHT, and HTN; and PNR interacted with CKD, CRRT, and HF. Despite these interactions, the overall prognostic trends were consistent: a higher NLR correlated with increased mortality risk, and higher NPAR was associated with increased mortality risk. For SII, mortality risk was also higher with SA, NEPHTOX, AHT, combined IHD and HF, and a high SII. In contrast, high PNR was associated with a lower mortality risk overall. Construction of Machine Learning Predictive Models Feature selection was performed using three methods (LASSO regression, the Boruta algorithm, and univariate logistic regression), resulting in 17 clinical features being identified by all three techniques (Fig. 5). These 17 predictors—AHT, NEPHTOX, NLR, SA, SII, VENTILATION, VP, MI, Age, First bilirubin total, First hemoglobin, First lactate, First pH, First pO2, First PT, First PTT, and First Urea Nitrogen (Fig. 6)—were subsequently standardized using Z-scores prior to model building. The comparative performance metrics for the six machine learning models on the training and internal test sets are presented in Table 3 and Fig. 7. The CatBoost model demonstrated the best and most consistent performance, achieving the highest Area Under the Receiver Operating Characteristic Curve (AUC) on both the training set (0.861) and the internal test set (0.861). Across all models, the recall and specificity ranges were 0.7103–0.8215 and 0.6148–0.7288, respectively, on the training set. On the test set, the ranges were 0.6547–0.7265 for recall and 0.6275–0.7333 for specificity. Establishment and Evaluation of the Optimal Model Based on its superior performance on the internal test set—achieving the highest AUC (0.861) and an optimal balance between recall and specificity—the CatBoost model was selected as the final model. The model was then optimized using grid search hyperparameter tuning . The final CatBoost model was subsequently validated on the independent, external cohort (Table S2 ). In this external validation, the model demonstrated robust discrimination, achieving an AUC of 0.755 (Fig. 8). The calibration curve indicated good agreement between predicted probabilities and observed outcomes. Furthermore, Decision Curve Analysis (DCA) showed a net benefit across the full range of risk thresholds (0 to 1), reflecting its potential applicability and practical utility in supporting clinical decision-making. SHAP Interpretation of the Catboost Model To interpret the final CatBoost model, SHAP (SHapley Additive exPlanations) values were used to illustrate how each feature influences the prediction of 14-day mortality. As shown in the SHAP summary plot (Fig. 9), feature contributions are represented by colored dots, where red indicates high feature values (correlating with higher risk) and blue indicates low feature values. The analysis revealed that metabolic indicators—specifically lactate, blood urea nitrogen, and pH—were the most critical factors driving the model's predictions. The use of vasoactive drugs (VP) was also ranked as a highly important contributor, correlating strongly with a high predicted risk of mortality. Discussion To our knowledge, this study is the first to systematically investigate the risk association between four whole-blood cell-derived inflammatory markers (NLR, SII, PNR, NPAR) and 14-day mortality in patients with sepsis-associated acute kidney injury (SA-AKI), confirming the critical role of inflammatory imbalance in this population. Through multidimensional clinical feature analysis, integrating these inflammatory markers with other key clinical indicators (e.g., lactate levels, mechanical ventilation use), we developed an optimal mortality risk prediction model using machine learning and interpreted the model's predictions using SHAP analysis. This approach provides an efficient and interpretable tool for clinical prognostic assessment in patients with SA-AKI. This study demonstrates that NLR, SII, and NPAR are significantly positively associated with 14-day all-cause mortality, while PNR is negatively associated. Our findings for NLR and SII are largely consistent with recent meta-analyses in general sepsis populations: Hongsheng Wu et al. demonstrated that elevated NLR correlates with poorer short-term outcomes[19] (PMID: 38562922), and Lingbo Liang et al. indicated that high SII values at admission significantly increase the risk of short-term all-cause mortality in sepsis patients [20]. Our finding that elevated NPAR independently correlates with higher mortality risk is supported by Yuqiang Gong et al., who analyzed 2,166 patients with severe sepsis or septic shock from the MIMIC-III database, revealing that higher NPAR levels are associated with increased risk of 30-day, 90-day, and 365-day mortality[21]. A retrospective study by Dan Wu et al., however, analyzed clinical data from 203 sepsis patients, revealing that significantly elevated PNR in sepsis patients was positively correlated with mortality, suggesting PNR serves as an independent prognostic indicator to identify high-risk patients [22]. Sepsis-associated acute kidney injury (SA-AKI) represents a complex physiological and immunological response, with systemic inflammation and endothelial dysfunction as key pathophysiological features [23].Inflammatory markers such as NLR, SII, NPAR, and PNR may reflect distinct inflammatory-coagulation-tissue injury dynamics within the context of SA-AKI. In sepsis, neutrophils participate in inflammatory responses through multiple mechanisms, including releasing granules containing MPO, NE, and metalloproteinases, and forming neutrophil extracellular traps (NETs) [24].Neutrophil-associated matrix metalloproteinase-9 (MMP-9) activity is upregulated and can directly degrade the glycocalyx [25]. Disruption of the glomerular endothelial glycocalyx leads to microvascular barrier impairment [26]and exposure of adhesion molecules (e.g., ICAM-1) [27], potentially exacerbating tissue injury. Loss of the endothelial anticoagulant barrier activates the coagulation cascade (e.g., vWF release), exacerbating intrarenal thrombotic microangiopathy (TMA) [28-30] .Concurrently, delayed neutrophil apoptosis and impaired migration prolong the inflammatory response, creating a vicious cycle [31, 32]. Neutrophil elastase and cathepsin G activate platelets via the receptor for activated serine protease (RAPS) [33]. As Braedon McDonald et al. demonstrated in septic mice, the NET-platelet-thrombin axis promotes intravascular coagulation and microvascular dysfunction in sepsis (PMID: 28073784)[34]. Conversely, lymphocytes undergo widespread apoptosis and functional failure in sepsis, particularly T-cell exhaustion[35] and PD-1 pathway upregulation, blunting the immune response to innate immune overactivation. This dysfunction also impairs the immune regulation of renal tubular injury repair and antifibrosis, hindering the resolution of the inflammation-coagulation axis [36, 37]. Albumin functions as both a carrier/antioxidant molecule and a maintainer of endothelial glycocalyx integrity and colloid osmotic pressure; hypoalbuminemia exacerbates capillary leakage, interstitial edema, and elevated renal venous pressure, thereby amplifying renal ischemia-reperfusion vulnerability and drug/endotoxin burden[38]. Collectively, neutrophil-platelet-coagulation network-driven microthrombosis and endothelial dysfunction, compounded by lymphocyte-mediated immunosuppression and barrier disruption due to hypoalbuminemia, constitute the key pathophysiological framework of SA-AKI. Subgroup analyses revealed significant interactions between the inflammatory markers and distinct clinical characteristics: NLR interacted with CRRT, AHT, and MI; NPAR interacted with immunosuppressive drugs (IMMUNOS), vasoactive drug use (VP), and heart failure (HF); SII interacted with IHD, AHT, and HTN; and PNR interacted with CKD, CRRT, and HF. The interaction between NLR and CRRT may arise because CRRT removes metabolic byproducts and filters/adsorbs certain inflammatory mediators, thereby altering the patient's inflammatory burden and immune microenvironment [39].This suggests CRRT may partially "offset" the adverse risk associated with an elevated NLR. Furthermore, patients exhibiting severe inflammation/stress states (high NLR) may derive greater benefit from CRRT. The interaction with antihypertensive therapy (AHT) may also be influential. Physiological stress from infection or inflammation increases endogenous catecholamine production, elevating neutrophil counts and causing lymphopenia[40] .Antihypertensive drugs, particularly angiotensin-converting enzyme inhibitors (ACEIs), effectively protect the endothelial glycocalyx and mitigate capillary ultrastructural abnormalities [41], potentially influencing outcomes. Finally, MI itself induces robust neutrophil infiltration and lymphocyte apoptosis, amplifying the inflammatory-immune response [42], which may diminish the sensitivity and specificity of NLR as a prognostic marker. NPAR's interaction with IMMUNOS may be twofold: first, immunosuppressants directly inhibit neutrophil activation and inflammatory cytokine release, thereby affecting NPAR's predictive capacity. Second, immunosuppressant users are often organ transplant recipients or autoimmune disease patients, and these underlying conditions themselves influence inflammation and nutritional status, introducing heterogeneity to NPAR risk stratification. Regarding vasoactive drugs (VP), while they improve microcirculation by maintaining blood pressure, they may simultaneously affect endothelial function. For instance, vasoactive amines can alter polymorphonuclear leukocyte expression in vitro via direct endothelial cell action, thereby influencing permeability and subsequently affecting albumin levels [43].Lastly, heart failure patients often exhibit chronic low-grade inflammation and protein-energy wasting, leading to elevated baseline neutrophils and decreased albumin. This existing pathophysiology may enhance NPAR's prognostic ability in the heart failure population[44, 45]. SII's interaction with ischemic heart disease (IHD) is notable, as studies indicate that elevated SII is independently associated with an increased risk of adverse cardiovascular events in patients undergoing percutaneous coronary intervention (PCI) [46] .In the pathogenesis of hypertension (HTN), classically activated macrophages (M1), neutrophils, and dendritic cells secrete inflammatory mediators, potentially enhancing SII's sensitivity to adverse outcomes [47] .Furthermore, the long-term use of antihypertensive drugs (AHT), such as ACEi/ARBs or β-blockers, exerts anti-inflammatory and immunomodulatory effects that can alter baseline neutrophil, lymphocyte, and platelet levels, thereby influencing the observed relationship between SII and outcomes. PNR interacts with CKD, possibly because uremic solutes in CKD can induce endothelial dysfunction, inflammation, and oxidative stress, thereby altering PNR and its prognostic capacity [48] .Its interaction with CRRT may be explained by research from Jian-Biao Meng et al., which indicates that early initiation of continuous veno-venous hemofiltration improves endothelial function and hemodynamic stability, potentially influencing the PNR-prognosis relationship[49]. Finally, the interaction with HF may be linked to shared chronic inflammation pathways, where inflammatory responses stimulate NLRP3 inflammasome activation and IL-1ß release, promoting the migration of multiple immune cells[22, 50]. This retrospective cohort study, while representing the first systematic investigation of the association between these whole-blood-derived inflammatory markers and SA-AKI prognosis, has several limitations. First, despite enrolling 4,311 patients and conducting external validation, selection bias and information bias are inherent to its retrospective design. Moreover, causal relationships between the inflammatory markers and mortality risk cannot be confirmed. Second, the study relied solely on baseline measurements of inflammatory components (neutrophil, lymphocyte, platelet counts, and serum albumin) at ICU admission, failing to assess the dynamic changes of these indicators throughout the disease course or their relationship with clinical progression. Therefore, future research should focus on the relationship between the dynamic changes in these inflammatory markers and sepsis-associated kidney injury. Concurrently, leveraging prospective designs and multi-omics integration holds promise for systematically elucidating the molecular mechanisms by which systemic inflammatory responses mediate kidney injury and revealing their biological basis. This would advance clinical practice in precision risk stratification, metabolically targeted interventions, and personalized treatment for high-risk patients. Conclusion This study demonstrates that the whole-blood cell-derived inflammatory markers NLR, SII, NPAR, and PNR are significantly associated with 14-day all-cause mortality in patients with SA-AKI. Specifically, NLR, SII, and NPAR were identified as independent risk factors, whereas PNR was found to be a protective factor. Furthermore, in the machine learning prognostic model developed from these findings, the CatBoost algorithm exhibited optimal predictive performance and generalization capability. This model not only enables precise risk stratification but also offers high clinical interpretability, as demonstrated by SHAP analysis. Collectively, this study confirms the significant value of these inflammatory markers in SA-AKI prognosis assessment and provides an effective tool that may assist clinical decision-making and facilitate personalized patient management. Declarations Ethics approval and consent to participate Not applicable Consent for publication Not applicable. Availability of data and materials The datasets analysed during the current study are available in the MIMIC-IV 3.1 and Medical Ethics Committee of The Second Affiliated Hospital of Nanchang University(Approval No.: IIT-O-2025-319) Competing Interests The authors declare that they have no competing interests. Funding This work was supported by the Regional Science Foundation Program of National Natural Science Foundation of China (82260374). Author Contributions All authors contributed to the study conception and design. Writing - original draft preparation:Xinghe Shangguan; Writing - review and editing: Xinghe Shangguan,Yongjie Luo,Yang Zhou; Conceptualization:Xinghe Shangguan,Xiayoumei Wu,Xiaofan Zou; Methodology:Xinghe Shangguan,Qianyu Yuan; Formal analysis and investigation: Xinghe Shangguan; Funding acquisition:Xinghe Shangguan; Resources:Xinghe Shangguan,Yuanqi Gong,Jianning Xu; Supervision: Yuanqi Gong,Jianning Xu,and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. Acknowledgments Not applicable. References Cohen J, Vincent J L, Adhikari N K J, et al. Sepsis: A Roadmap for Future Research. The Lancet Infectious Diseases, 2015, 15(5): 581–614. Gómez H, Kellum J A. Sepsis-Induced Acute Kidney Injury. Current Opinion in Critical Care, 2016, 22(6): 546–553. Angus D C, Linde-Zwirble W T, Lidicker J, et al. Epidemiology of Severe Sepsis in the United States: Analysis of Incidence, Outcome, and Associated Costs of Care:. Critical Care Medicine, 2001, 29(7): 1303–1310. Chalfin D B, Fein A M, Carlon G C, et al. Cost-Effectiveness of Monoclonal Antibodies to Gram-Negative Endotoxin in the Treatment of Gram-Negative Sepsis in ICU Patients. Peerapornratana S, Manrique-Caballero C L, Gómez H, et al. Acute Kidney Injury from Sepsis: Current Concepts, Epidemiology, Pathophysiology, Prevention and Treatment. Kidney International, 2019, 96(5): 1083–1099. Haybar H, Pezeshki S M S, Saki N. Evaluation of Complete Blood Count Parameters in Cardiovascular Diseases: An Early Indicator of Prognosis?. Experimental and Molecular Pathology, 2019, 110: 104267. Dong G, Gan M, Xu S, et al. The Neutrophil–Lymphocyte Ratio as a Risk Factor for All-Cause and Cardiovascular Mortality among Individuals with Diabetes: Evidence from the NHANES 2003–2016. Cardiovascular Diabetology, 2023, 22(1): 267. Tudurachi B S, Anghel L, Tudurachi A, et al. Assessment of Inflammatory Hematological Ratios (NLR, PLR, MLR, LMR and Monocyte/HDL–Cholesterol Ratio) in Acute Myocardial Infarction and Particularities in Young Patients. International Journal of Molecular Sciences, 2023, 24(18): 14378. Chen J, Yasrebinia S, Ghaedi A, et al. Meta-Analysis of the Role of Neutrophil to Lymphocyte Ratio in Neonatal Sepsis. BMC Infectious Diseases, 2023, 23(1): 837. Drăgoescu A N, Pădureanu V, Stănculescu A D, et al. Neutrophil to Lymphocyte Ratio (NLR)—a Useful Tool for the Prognosis of Sepsis in the ICU. Biomedicines, 2021, 10(1): 75. Zhang Y, Peng W, Zheng X. The Prognostic Value of the Combined Neutrophil-to-Lymphocyte Ratio (NLR) and Neutrophil-to-Platelet Ratio (NPR) in Sepsis. Scientific Reports, 2024, 14(1): 15075. Ren Y, Loftus T J, Datta S, et al. Performance of a Machine Learning Algorithm Using Electronic Health Record Data to Predict Postoperative Complications and Report on a Mobile Platform. JAMA Network Open, 2022, 5(5): e2211973. Song Y, Yang X, Luo Y, et al. Comparison of Logistic Regression and Machine Learning Methods for Predicting Postoperative Delirium in Elderly Patients: A Retrospective Study. CNS Neuroscience & Therapeutics, 2023, 29(1): 158–167. Liu Z, Shu W, Li T, et al. Interpretable Machine Learning for Predicting Sepsis Risk in Emergency Triage Patients. Scientific Reports, 2025, 15(1): 887. Hou N. Predicting 30-Days Mortality for MIMIC-III Patients with Sepsis-3: A Machine Learning Approach Using XGboost.2020. Yue S, Li S, Huang X, et al. Machine Learning for the Prediction of Acute Kidney Injury in Patients with Sepsis. Journal of Translational Medicine, 2022, 20(1): 215. Singer M, Deutschman C S, Seymour C W, et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA, 2016, 315(8): 801. Shen D, Sha L, Yang L, et al. Identification of Multiple Complications as Independent Risk Factors Associated with 1-, 3-, and 5-Year Mortality in Hepatitis B-Associated Cirrhosis Patients. BMC Infectious Diseases, 2025, 25(1): 151. Wu H, Cao T, Ji T, et al. Predictive Value of the Neutrophil-to-Lymphocyte Ratio in the Prognosis and Risk of Death for Adult Sepsis Patients: A Meta-Analysis. Frontiers in Immunology, 2024, 15: 1336456. Liang L, Su Q. Systemic Immune-Inflammation Index and the Short-Term Mortality of Patients with Sepsis: A Meta-Analysis. Biomolecules and Biomedicine, 2024, 25(4): 798–809. Gong Y, Li D, Cheng B, et al. Increased Neutrophil Percentage-to-Albumin Ratio Is Associated with All-Cause Mortality in Patients with Severe Sepsis or Septic Shock. Epidemiology and Infection, 2020, 148: e87. Wu D, Qin H. Diagnostic and Prognostic Values of Immunocyte Ratios in Patients with Sepsis in the Intensive Care Unit. The Journal of Infection in Developing Countries, 2023, 17(10): 1362–1372. Li J C, Wang L J, Feng F, et al. Role of Heparanase in Sepsis‑related Acute Kidney Injury (Review). Experimental and Therapeutic Medicine, 2023, 26(2): 379. Zhang H, Wang Y, Qu M, et al. Neutrophil, Neutrophil Extracellular Traps and Endothelial Cell Dysfunction in Sepsis. Clinical and Translational Medicine, 2023, 13(1): e1170. Lin L, Niu M, Gao W, et al. Predictive Role of Glycocalyx Components and MMP-9 in Cardiopulmonary Bypass Patients for ICU Stay. Heliyon, 2024, 10(1): e23299. Gamez M, Elhegni H E, Fawaz S, et al. Heparanase Inhibition as a Systemic Approach to Protect the Endothelial Glycocalyx and Prevent Microvascular Complications in Diabetes. Cardiovascular Diabetology, 2024, 23(1): 50. Zhang W, Guan Y, Bayliss G, et al. Class IIa HDAC Inhibitor TMP195 Alleviates Lipopolysaccharide-Induced Acute Kidney Injury. American Journal of Physiology Renal Physiology, 2020, 319(6): F1015–F1026. Kei C Y, Singh K, Dautov R F, et al. Coronary “Microvascular Dysfunction”: Evolving Understanding of Pathophysiology, Clinical Implications, and Potential Therapeutics. International Journal of Molecular Sciences, 2023, 24(14): 11287. Peter B, Kanyo N, Kovacs K D, et al. Glycocalyx Components Detune the Cellular Uptake of Gold Nanoparticles in a Size- and Charge-Dependent Manner. ACS Applied Bio Materials, 2023, 6(1): 64–73. Medica D, Quercia A D, Marengo M, et al. High-Volume Hemofiltration Does Not Protect Human Kidney Endothelial and Tubular Epithelial Cells from Septic Plasma-Induced Injury. Scientific Reports, 2024, 14(1): 18323. Sônego F, Castanheira F V E S, Ferreira R G, et al. Paradoxical Roles of the Neutrophil in Sepsis: Protective and Deleterious. Frontiers in Immunology, 2016, 7. Mantovani A, Bonecchi R, Locati M. Tuning Inflammation and Immunity by Chemokine Sequestration: Decoys and More. Nature Reviews Immunology, 2006, 6(12): 907–918. Sambrano G R, Huang W, Faruqi T, et al. Cathepsin G Activates Protease-Activated Receptor-4 in Human Platelets. Journal of Biological Chemistry, 2000, 275(10): 6819–6823. McDonald B, Davis R P, Kim S J, et al. Platelets and Neutrophil Extracellular Traps Collaborate to Promote Intravascular Coagulation during Sepsis in Mice. Blood, 2017, 129(10): 1357–1367. Van Der Poll T, Van De Veerdonk F L, Scicluna B P, et al. The Immunopathology of Sepsis and Potential Therapeutic Targets. Nature Reviews Immunology, 2017, 17(7): 407–420. Cao C, Yao Y, Zeng R. Lymphocytes: Versatile Participants in Acute Kidney Injury and Progression to Chronic Kidney Disease. Frontiers in Physiology, 2021, 12: 729084. Hotchkiss R S, Monneret G, Payen D. Sepsis-Induced Immunosuppression: From Cellular Dysfunctions to Immunotherapy. Nature Reviews Immunology, 2013, 13(12): 862–874. Aldecoa C, Llau J V, Nuvials X, et al. Role of Albumin in the Preservation of Endothelial Glycocalyx Integrity and the Microcirculation: A Review. Annals of Intensive Care, 2020, 10(1): 85. Chen J J, Lai P C, Lee T H, et al. Blood Purification for Adult Patients with Severe Infection or Sepsis/Septic Shock: A Network Meta-Analysis of Randomized Controlled Trials. Critical Care Medicine, 2023, 51(12): 1777–1789. Chen J J, Kuo G, Fan P C, et al. Neutrophil-to-Lymphocyte Ratio Is a Marker for Acute Kidney Injury Progression and Mortality in Critically Ill Populations: A Population-Based, Multi-Institutional Study. Journal of Nephrology, 2022, 35(3): 911–920. Locatelli M, Rottoli D, Mahmoud R, et al. Endothelial Glycocalyx of Peritubular Capillaries in Experimental Diabetic Nephropathy: A Target of ACE Inhibitor-Induced Kidney Microvascular Protection. International Journal of Molecular Sciences, 2023, 24(22): 16543. Forteza M J, Trapero I, Hervas A, et al. Apoptosis and Mobilization of Lymphocytes to Cardiac Tissue Is Associated with Myocardial Infarction in a Reperfused Porcine Model and Infarct Size in post‐PCI Patients. Madamanchi N. Oxidative Medicine and Cellular Longevity, 2018, 2018(1): 1975167. Doukas J, Shepro D, Hechtman H B. Vasoactive Amines Directly Modify Endothelial Cells to Affect. Amara M, Stoler O, Birati E Y. The Role of Inflammation in the Pathophysiology of Heart Failure. Cells, 2025, 14(14): 1117. Nguyen A P, Kawi J, Meraz R, et al. Hidden Malnutrition in Overweight and Obese Individuals with Chronic Heart Failure: Insights from the pro-HEART Trial. Nutrients, 2025, 17(16): 2694. Zhang C, Li M, Liu L, et al. Systemic Immune-Inflammation Index as a Novel Predictor of Major Adverse Cardiovascular Events in Patients Undergoing Percutaneous Coronary Intervention: A Meta-Analysis of Cohort Studies. BMC Cardiovascular Disorders, 2024, 24(1): 189. Zhang Z, Zhao L, Zhou X, et al. Role of Inflammation, Immunity, and Oxidative Stress in Hypertension: New Insights and Potential Therapeutic Targets. Frontiers in Immunology, 2023, 13: 1098725. Dou L, Sallée M, Cerini C, et al. The Cardiovascular Effect of the Uremic Solute Indole-3 Acetic Acid. Journal of the American Society of Nephrology, 2015, 26(4): 876–887. Meng J biao, Lai Z zhen, Xu X juan, et al. Effects of Early Continuous Venovenous Hemofiltration on E-Selectin, Hemodynamic Stability, and Ventilatory Function in Patients with Septic-Shock-Induced Acute Respiratory Distress Syndrome. Biomed Research International, 2016, 2016: 1–9. Sandanger Ø, Ranheim T, Vinge L E, et al. The NLRP3 Inflammasome Is Up-Regulated in Cardiac Fibroblasts and Mediates Myocardial Ischaemia–Reperfusion Injury. Cardiovascular Research, 2013, 99(1): 164–174. Tables Tables 1 to 3 are available in the Supplementary Files section Additional Declarations No competing interests reported. Supplementary Files TableS1.UnivariateLogisticRegressionAnalysisofFactors.docx TableS2.PerformanceoftheFinalGatboostModelontheExternalValidationCohort.docx TableS3.HyperparameterSettingsforCatBoostModel.docx TableS4CharacteristicsofparticipantsintheExternalValidationCohort.docx Table1.CharacteristicsofparticipantsincludedinstudyfromtheMIMIC.docx Table2.CoxRegressionAnalysisoftheAssociationBetweenNLRNPARPNRSIIand14DayAllCauseMortality.docx Table3.MachineLearningModelEvaluationMetricsfortheTrainingSetandTestSet.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7986649","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":542335999,"identity":"a5838873-f209-4825-bf37-833401903cb1","order_by":0,"name":"Xinghe Shangguan","email":"","orcid":"","institution":"Nanchang University","correspondingAuthor":false,"prefix":"","firstName":"Xinghe","middleName":"","lastName":"Shangguan","suffix":""},{"id":542336001,"identity":"39de1970-c586-4b9f-b41c-20307352cc1d","order_by":1,"name":"Yongjie Luo","email":"","orcid":"","institution":"Nanchang University","correspondingAuthor":false,"prefix":"","firstName":"Yongjie","middleName":"","lastName":"Luo","suffix":""},{"id":542336003,"identity":"1b2e0b05-8757-4563-951a-a2ae3f77c62c","order_by":2,"name":"Xiayoumei Wu","email":"","orcid":"","institution":"Nanchang University","correspondingAuthor":false,"prefix":"","firstName":"Xiayoumei","middleName":"","lastName":"Wu","suffix":""},{"id":542336005,"identity":"1607a47c-3bbe-4b1f-8b3c-d035dd7ea371","order_by":3,"name":"Jianning Xu","email":"","orcid":"","institution":"Nanchang University","correspondingAuthor":false,"prefix":"","firstName":"Jianning","middleName":"","lastName":"Xu","suffix":""},{"id":542336007,"identity":"91db5138-8bc6-463e-ad1b-29f8c8f93f13","order_by":4,"name":"Yuanqi Gong","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzklEQVRIiWNgGAWjYBAC++ONjQ8+/vkvZ3+8gVg9Zw4fNpzZwGzMcOYAsVpuuKVJ8zYwJzbcSCBSB+MMHgMJ3h1sxowzH2+8wVBjE01QC7N0j4GB5BkeOWbptGILhmNpuQ2EtLDJnDFIMGCTMGaTzjGTYGw4TFgLj0SOwYEENoPEHskzRGqRkEhLbDjYlpA4Q4KHSC0GPIcPMzacOWBswAP0SwIxfjFgb2z//afigJwB++GNNz7U2BDWgqJdIoEU5RAtpOoYBaNgFIyCkQEAg7lBDOKv1OcAAAAASUVORK5CYII=","orcid":"","institution":"Nanchang University","correspondingAuthor":true,"prefix":"","firstName":"Yuanqi","middleName":"","lastName":"Gong","suffix":""},{"id":542336008,"identity":"3847626c-c23b-48c8-98ea-27a612877c53","order_by":5,"name":"Xiaofan Zou","email":"","orcid":"","institution":"Nanchang University","correspondingAuthor":false,"prefix":"","firstName":"Xiaofan","middleName":"","lastName":"Zou","suffix":""},{"id":542336010,"identity":"df86f79e-4a10-4a94-ab14-2b390911abac","order_by":6,"name":"Qianyu Yuan","email":"","orcid":"","institution":"Nanchang University","correspondingAuthor":false,"prefix":"","firstName":"Qianyu","middleName":"","lastName":"Yuan","suffix":""},{"id":542336012,"identity":"76770217-d9a4-4a7b-9680-496a5fbe826e","order_by":7,"name":"Yang Zhou","email":"","orcid":"","institution":"Nanchang University","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Zhou","suffix":""}],"badges":[],"createdAt":"2025-10-30 08:08:23","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7986649/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7986649/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":95664977,"identity":"8f37bac7-eb22-461e-ae27-91b7e84c97a5","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"tif","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":50607164,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.Thepatientselectionprocess.tif.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/50eb046b2f264ccd5d7db159.tif"},{"id":95664956,"identity":"ce068851-a87e-4590-80c9-06dfa1f977e4","added_by":"auto","created_at":"2025-11-11 16:43:34","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":94130,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.docx","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/090ce1776cb45f2ebf7f80f9.docx"},{"id":95664968,"identity":"66ca0279-2f03-4c7e-96c2-c4bd809d444a","added_by":"auto","created_at":"2025-11-11 16:43:34","extension":"tif","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":28523084,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2.AssociationBetweenInflammatoryMarkersand14DayAllCauseMortalityUsingRestrictedCubicSplines.tif.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/d5daf5864af14564f9a89744.tif"},{"id":95664971,"identity":"ddfcc2b1-6402-4188-ba52-d00de3c6a413","added_by":"auto","created_at":"2025-11-11 16:43:34","extension":"tif","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17347368,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3.KaplanMeierSurvivalCurvesfor14DayAllCauseMortalityStratifiedbyInflammatoryMarkerLevels.tif.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/9cf8533daff71b6e74fb574e.tif"},{"id":95664998,"identity":"cb41070a-ffd0-465f-ad8d-ee8629133f61","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"tif","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14685580,"visible":true,"origin":"","legend":"","description":"","filename":"Figure4.SubgroupAnalysisoftheAssociationBetweenInflammatoryMarkersand14DayMortality.tif.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/2f3b1ff81cdc8ea9f6b56b37.tif"},{"id":95797794,"identity":"6cdbb586-83da-4d50-bed0-54c940146428","added_by":"auto","created_at":"2025-11-13 08:10:55","extension":"tif","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":29008008,"visible":true,"origin":"","legend":"","description":"","filename":"Figure5.FeatureSelectionProcessforPredictiveModelDevelopment.tif.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/322783c61e99ae33e68ed373.tif"},{"id":95664982,"identity":"5c9b0cba-6821-4062-9c32-23ef08a47032","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"tif","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":27933388,"visible":true,"origin":"","legend":"","description":"","filename":"Figure6.FinalSetof17ClinicalFeaturesSelectedforMachineLearningModelConstruction.tif.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/51fcebd627a88d8314ecbd38.tif"},{"id":95798240,"identity":"3ac5b5c2-117f-48a1-b73c-01239a227f42","added_by":"auto","created_at":"2025-11-13 08:16:08","extension":"tif","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10202992,"visible":true,"origin":"","legend":"","description":"","filename":"Figure7.ReceiverOperatingCharacteristicROCCurvesoftheSixMachineLearningintheTrainingandInternalValidationSets.tif.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/6d2340d55b543012c9f7afa7.tif"},{"id":95664975,"identity":"7caacfe8-8fe0-4374-bd2e-607b469ae434","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"tif","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9137972,"visible":true,"origin":"","legend":"","description":"","filename":"Figure8.PerformanceoftheGatboostModelintheExternalValidationCohort.tif.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/431b2b1a9cecc73373644811.tif"},{"id":95665003,"identity":"a6733d19-cb42-478f-adad-f62be546c909","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"tif","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9591516,"visible":true,"origin":"","legend":"","description":"","filename":"Figure9.SHAPSummaryPlotIllustratingFeatureImportanceandImpactintheFinalCatBoostModel.tif2.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/ae42c693e9f4b07cfac49323.tif"},{"id":95797194,"identity":"346663cf-d1da-460a-8f33-cdbbc132483d","added_by":"auto","created_at":"2025-11-13 08:01:45","extension":"json","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11028,"visible":true,"origin":"","legend":"","description":"","filename":"10db8d4ab1b843f7a3601bd5627af395.json","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/86a66ad7a5d546be53aa3a07.json"},{"id":95664990,"identity":"c9b9fbe7-f801-4b88-96d6-ee4175bc099f","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"docx","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":30615,"visible":true,"origin":"","legend":"","description":"","filename":"TableS4CharacteristicsofparticipantsintheExternalValidationCohort.docx","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/fa6b68e36d0e23d90599001e.docx"},{"id":95797500,"identity":"9a8ad79a-1e47-45d6-9c99-aa0f70f18be3","added_by":"auto","created_at":"2025-11-13 08:06:01","extension":"docx","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":29678,"visible":true,"origin":"","legend":"","description":"","filename":"TableS1.UnivariateLogisticRegressionAnalysisofFactors.docx","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/e3ef2d253e9c551e46d6070e.docx"},{"id":95664973,"identity":"deb33225-d72a-4871-a7cc-be7ae2f8f93c","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"docx","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15956,"visible":true,"origin":"","legend":"","description":"","filename":"TableS2.PerformanceoftheFinalGatboostModelontheExternalValidationCohort.docx","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/80056876530ec1ca07169a9a.docx"},{"id":95664963,"identity":"705b7ed6-a3e2-4300-b8b4-833fdffbd8ef","added_by":"auto","created_at":"2025-11-11 16:43:34","extension":"docx","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14160,"visible":true,"origin":"","legend":"","description":"","filename":"TableS3.HyperparameterSettingsforCatBoostModel.docx","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/e718f7d4bfc24289d4b90da5.docx"},{"id":95797620,"identity":"a7f04d5b-6ae5-4138-8eb2-c68b8494ebae","added_by":"auto","created_at":"2025-11-13 08:08:09","extension":"xml","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":171025,"visible":true,"origin":"","legend":"","description":"","filename":"10db8d4ab1b843f7a3601bd5627af3951enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/29b009b1bea1e3827d4cd741.xml"},{"id":95664984,"identity":"ad42b069-d412-4bd1-bb00-a14cdf4c318c","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"tif","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":50607164,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.Thepatientselectionprocess.tif.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/4833648856af90111e4fb318.tif"},{"id":95797163,"identity":"cc4f4cd3-9382-42f9-abb5-c0a72128c50a","added_by":"auto","created_at":"2025-11-13 08:01:32","extension":"tif","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":28523084,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2.AssociationBetweenInflammatoryMarkersand14DayAllCauseMortalityUsingRestrictedCubicSplines.tif.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/b370b448e61b55f5d9ff7bce.tif"},{"id":95797125,"identity":"01d5cfe4-354b-4279-abf5-99be539d4fad","added_by":"auto","created_at":"2025-11-13 08:01:10","extension":"tif","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17347368,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3.KaplanMeierSurvivalCurvesfor14DayAllCauseMortalityStratifiedbyInflammatoryMarkerLevels.tif.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/4291228c1179eb9c1cc38b8b.tif"},{"id":95664985,"identity":"44b5cd6c-ebdd-4830-9e26-3addbc3f0517","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"tif","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14685580,"visible":true,"origin":"","legend":"","description":"","filename":"Figure4.SubgroupAnalysisoftheAssociationBetweenInflammatoryMarkersand14DayMortality.tif.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/80bf018e124e95fae142e883.tif"},{"id":95664996,"identity":"b25bd9b6-a7b1-494f-9f19-23b5f939a7cd","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"tif","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":29008008,"visible":true,"origin":"","legend":"","description":"","filename":"Figure5.FeatureSelectionProcessforPredictiveModelDevelopment.tif.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/f3e13895e1d27af309aa7bee.tif"},{"id":95665002,"identity":"5bbcbc3c-69cf-4085-ad23-aa4bee7d2aef","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"tif","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":27933388,"visible":true,"origin":"","legend":"","description":"","filename":"Figure6.FinalSetof17ClinicalFeaturesSelectedforMachineLearningModelConstruction.tif.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/3d0228c8d55637cd28750447.tif"},{"id":95665004,"identity":"066c2055-a3fd-458f-9c78-d5126da63bf5","added_by":"auto","created_at":"2025-11-11 16:43:36","extension":"tif","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10202992,"visible":true,"origin":"","legend":"","description":"","filename":"Figure7.ReceiverOperatingCharacteristicROCCurvesoftheSixMachineLearningintheTrainingandInternalValidationSets.tif.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/c338d28b01e2dced3957295e.tif"},{"id":95797191,"identity":"ea6be271-0f8c-4112-bd3c-0a04595195b9","added_by":"auto","created_at":"2025-11-13 08:01:43","extension":"tif","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9137972,"visible":true,"origin":"","legend":"","description":"","filename":"Figure8.PerformanceoftheGatboostModelintheExternalValidationCohort.tif.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/c1db876bc7345adb33d3bc23.tif"},{"id":95664979,"identity":"e70c98ca-edce-489c-b662-76553c5f352d","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"tif","order_by":27,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9591516,"visible":true,"origin":"","legend":"","description":"","filename":"Figure9.SHAPSummaryPlotIllustratingFeatureImportanceandImpactintheFinalCatBoostModel.tif2.tif","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/56e4136d3e2c0bd6f9f80900.tif"},{"id":95664987,"identity":"47f8f9e6-de1f-4f17-a836-2bc4e504aa5c","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"jpeg","order_by":28,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1074,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/92e3e8961cf30194f1518b7d.jpeg"},{"id":95798622,"identity":"3a08e75d-8a89-4c23-835e-e012d6634e10","added_by":"auto","created_at":"2025-11-13 08:17:21","extension":"png","order_by":29,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":499305,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure1.Thepatientselectionprocess.tif.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/115e3c55a4c21b455fe3375d.png"},{"id":95664999,"identity":"3428db64-7217-4885-a8a1-eb9563a753a9","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"png","order_by":30,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":657475,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure2.AssociationBetweenInflammatoryMarkersand14DayAllCauseMortalityUsingRestrictedCubicSplines.tif.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/c3f8f8fb041575a38bacfb35.png"},{"id":95664983,"identity":"6968ee78-5f0d-47c8-9419-0d10e116dfb5","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"png","order_by":31,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":69597,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure3.KaplanMeierSurvivalCurvesfor14DayAllCauseMortalityStratifiedbyInflammatoryMarkerLevels.tif.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/986c189c8a7913b86850585e.png"},{"id":95665008,"identity":"3a1fc738-d748-4a33-939e-7cd5db4c03a8","added_by":"auto","created_at":"2025-11-11 16:43:36","extension":"png","order_by":32,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":178885,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure4.SubgroupAnalysisoftheAssociationBetweenInflammatoryMarkersand14DayMortality.tif.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/1be7a9eed4cfec359ca248b1.png"},{"id":95665007,"identity":"30bbe3e3-fb5a-42ba-b6bb-13f3c1e9faec","added_by":"auto","created_at":"2025-11-11 16:43:36","extension":"png","order_by":33,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":939673,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure5.FeatureSelectionProcessforPredictiveModelDevelopment.tif.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/28ca32f369d2476b5766e488.png"},{"id":95664978,"identity":"86647edc-c0db-4fb7-8bcf-e2821c6320ec","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"png","order_by":34,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":229556,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure6.FinalSetof17ClinicalFeaturesSelectedforMachineLearningModelConstruction.tif.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/2daa7fe1b14f17d060369476.png"},{"id":95797233,"identity":"09c8bc67-6152-42c7-80a9-3e3a3ba08c34","added_by":"auto","created_at":"2025-11-13 08:02:05","extension":"png","order_by":35,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":88543,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure7.ReceiverOperatingCharacteristicROCCurvesoftheSixMachineLearningintheTrainingandInternalValidationSets.tif.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/46919f1ac67a6af84472ea82.png"},{"id":95797956,"identity":"109596c1-0c4a-4b98-a8b5-5cb859566f58","added_by":"auto","created_at":"2025-11-13 08:12:57","extension":"png","order_by":36,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":45414,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure8.PerformanceoftheGatboostModelintheExternalValidationCohort.tif.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/051f3bd73a6623abc3bdb5bb.png"},{"id":95664989,"identity":"d6919ce8-2a61-4c2e-b673-d4a2895bff9a","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"png","order_by":37,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":44775,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure9.SHAPSummaryPlotIllustratingFeatureImportanceandImpactintheFinalCatBoostModel.tif2.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/f915d01990035d1d14f4886f.png"},{"id":95797129,"identity":"c0a695c6-9ca2-4749-9dd2-e857b230983c","added_by":"auto","created_at":"2025-11-13 08:01:13","extension":"png","order_by":38,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":935,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/35c59e98ee92bac84b08036a.png"},{"id":95665000,"identity":"0ebd98ef-fcfc-476d-825c-46cb1b646b0d","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"xml","order_by":39,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":170050,"visible":true,"origin":"","legend":"","description":"","filename":"10db8d4ab1b843f7a3601bd5627af3951structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/cd53abad23ca61edd056a4df.xml"},{"id":95797494,"identity":"4021c43b-103a-4ced-b8fc-72f0b2b14e73","added_by":"auto","created_at":"2025-11-13 08:05:55","extension":"html","order_by":40,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":184488,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/1cb99fd19d45b826a0ccbfb3.html"},{"id":95798834,"identity":"bec355a2-76e5-4174-8241-199ecf832250","added_by":"auto","created_at":"2025-11-13 08:17:56","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":5739793,"visible":true,"origin":"","legend":"\u003cp\u003eFigure legend not provided with this version\u003c/p\u003e","description":"","filename":"Figure1.Thepatientselectionprocess.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/d3ddb42c0d0e46772f2c0316.png"},{"id":95664955,"identity":"c83af985-5460-4529-ac48-83d616cc6124","added_by":"auto","created_at":"2025-11-11 16:43:34","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1756460,"visible":true,"origin":"","legend":"\u003cp\u003eFigure legend not provided with this version\u003c/p\u003e","description":"","filename":"Figure2.AssociationBetweenInflammatoryMarkersand14DayAllCauseMortalityUsingRestrictedCubicSplines.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/f197ab410e3570dfd56cebec.png"},{"id":95797452,"identity":"5782680a-7eeb-46ca-8016-e636f6c9360a","added_by":"auto","created_at":"2025-11-13 08:05:26","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1172036,"visible":true,"origin":"","legend":"\u003cp\u003eFigure legend not provided with this version\u003c/p\u003e","description":"","filename":"Figure3.KaplanMeierSurvivalCurvesfor14DayAllCauseMortalityStratifiedbyInflammatoryMarkerLevels.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/e370a9d7742872f9ddaf5fb2.png"},{"id":95797464,"identity":"7fb03e74-df95-4671-b4bb-5714f75c991d","added_by":"auto","created_at":"2025-11-13 08:05:37","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":4234441,"visible":true,"origin":"","legend":"\u003cp\u003eFigure legend not provided with this version\u003c/p\u003e","description":"","filename":"Figure4.SubgroupAnalysisoftheAssociationBetweenInflammatoryMarkersand14DayMortality.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/106c5802f708bdbae475b3b4.png"},{"id":95798035,"identity":"663001f2-49ec-4c71-8f32-173baaed7d77","added_by":"auto","created_at":"2025-11-13 08:14:45","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":3276493,"visible":true,"origin":"","legend":"\u003cp\u003eFigure legend not provided with this version\u003c/p\u003e","description":"","filename":"Figure5.FeatureSelectionProcessforPredictiveModelDevelopment.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/bad9dd9d751e52a39498e9fc.png"},{"id":95797935,"identity":"ea0a344e-0da8-4bca-a3a9-03841e0a49db","added_by":"auto","created_at":"2025-11-13 08:12:52","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":971517,"visible":true,"origin":"","legend":"\u003cp\u003eFigure legend not provided with this version\u003c/p\u003e","description":"","filename":"Figure6.FinalSetof17ClinicalFeaturesSelectedforMachineLearningModelConstruction.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/3aeb98946448df97a72bbfaa.png"},{"id":95664994,"identity":"93bd0193-da3d-4d88-8b26-2b4178bfb996","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":1414611,"visible":true,"origin":"","legend":"\u003cp\u003eFigure legend not provided with this version\u003c/p\u003e","description":"","filename":"Figure7.ReceiverOperatingCharacteristicROCCurvesoftheSixMachineLearningintheTrainingandInternalValidationSets.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/e69a3b40e31700fbb692cab7.png"},{"id":95798899,"identity":"f579a1ca-8ea8-4a29-9a4c-f8cd1d08fd23","added_by":"auto","created_at":"2025-11-13 08:18:09","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":705583,"visible":true,"origin":"","legend":"\u003cp\u003eFigure legend not provided with this version\u003c/p\u003e","description":"","filename":"Figure8.PerformanceoftheGatboostModelintheExternalValidationCohort.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/9ef98df3fb8690cbd5a87b18.png"},{"id":95797659,"identity":"50d43a61-5847-41c3-ae80-660909ecfedc","added_by":"auto","created_at":"2025-11-13 08:08:32","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":796401,"visible":true,"origin":"","legend":"\u003cp\u003eFigure legend not provided with this version\u003c/p\u003e","description":"","filename":"Figure9.SHAPSummaryPlotIllustratingFeatureImportanceandImpactintheFinalCatBoostModel.png","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/b08b4df43a69e38e55507802.png"},{"id":96245717,"identity":"21861637-9e17-45e0-b656-4df4e9b0b0d3","added_by":"auto","created_at":"2025-11-19 07:22:05","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":18893309,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/791486d4-f684-4569-952a-ace5cc6a7839.pdf"},{"id":95664959,"identity":"e665a774-a48d-405a-8616-55e72ee52720","added_by":"auto","created_at":"2025-11-11 16:43:34","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":29678,"visible":true,"origin":"","legend":"","description":"","filename":"TableS1.UnivariateLogisticRegressionAnalysisofFactors.docx","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/2a3847c29737e6b4aa62a32c.docx"},{"id":95664961,"identity":"dc34624f-1880-4d07-851e-46ef522c6808","added_by":"auto","created_at":"2025-11-11 16:43:34","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":15956,"visible":true,"origin":"","legend":"","description":"","filename":"TableS2.PerformanceoftheFinalGatboostModelontheExternalValidationCohort.docx","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/249769352c81c326b48e46a1.docx"},{"id":95664964,"identity":"c181b10f-8e3a-47bb-99df-69e2db3e24a4","added_by":"auto","created_at":"2025-11-11 16:43:34","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":14160,"visible":true,"origin":"","legend":"","description":"","filename":"TableS3.HyperparameterSettingsforCatBoostModel.docx","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/986ec7bb272a5a78a8184147.docx"},{"id":95664974,"identity":"c41e84cc-fa0c-4e8f-a76a-2d7dd246eeca","added_by":"auto","created_at":"2025-11-11 16:43:35","extension":"docx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":30615,"visible":true,"origin":"","legend":"","description":"","filename":"TableS4CharacteristicsofparticipantsintheExternalValidationCohort.docx","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/8521ee46467de68305ea60f0.docx"},{"id":95664967,"identity":"d8963d02-9771-4b7e-9929-ec04ef0872a4","added_by":"auto","created_at":"2025-11-11 16:43:34","extension":"docx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":27674,"visible":true,"origin":"","legend":"","description":"","filename":"Table1.CharacteristicsofparticipantsincludedinstudyfromtheMIMIC.docx","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/3cfc8ba13896259471c6f5a6.docx"},{"id":95664960,"identity":"bc485903-2d22-42a7-a0f6-02e85d425945","added_by":"auto","created_at":"2025-11-11 16:43:34","extension":"docx","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":23980,"visible":true,"origin":"","legend":"","description":"","filename":"Table2.CoxRegressionAnalysisoftheAssociationBetweenNLRNPARPNRSIIand14DayAllCauseMortality.docx","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/a674b2ed8bd958331a77c453.docx"},{"id":95664972,"identity":"ac62ad55-48bc-47d2-a20d-2876a72768ce","added_by":"auto","created_at":"2025-11-11 16:43:34","extension":"docx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":17963,"visible":true,"origin":"","legend":"","description":"","filename":"Table3.MachineLearningModelEvaluationMetricsfortheTrainingSetandTestSet.docx","url":"https://assets-eu.researchsquare.com/files/rs-7986649/v1/ce0535948152aadc469da464.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eAssociation Between Whole Blood Cell-Derived Inflammatory Markers and All-Cause Mortality in Patients with Sepsis-Related Acute Kidney Injury and Development of a Machine Learning Prognostic Model\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eSepsis, defined as life-threatening organ dysfunction caused by a dysregulated host response to infection, ranks among the leading causes of death globally[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Acute kidney injury (AKI) is a frequent complication, occurring in 40\u0026ndash;50% of sepsis patients and increasing mortality 6- to 8-fold, representing one of the most severe complications of this syndrome [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e].This condition, sepsis-associated acute kidney injury (SA-AKI), is common in critically ill patients and strongly associated with poor outcomes. These adverse outcomes include increased risks of chronic kidney disease, cardiovascular events, and mortality, in addition to prolonged hospital stays and higher healthcare costs [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan additionalcitationids=\"CR4\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Consequently, improving the prognostic assessment of patients with SA-AKI holds significant clinical importance.\u003c/p\u003e\u003cp\u003eWhole blood cell-derived inflammatory markers, including the neutrophil-to-lymphocyte ratio (NLR), neutrophil percentage to albumin ratio (NPAR), platelet to neutrophil ratio (PNR), and systemic immune inflammation index (SII), are increasingly recognized for their utility in disease diagnosis, prognostic assessment, and severity grading [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. As these markers are derived from components (e.g., neutrophils, lymphocytes, platelets, and albumin) measured in routine complete blood counts and biochemical panels, they are cost-effective, reproducible, and widely accessible.These indices reflect the body's inflammatory state and immune balance [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e] and have shown utility in diverse fields such as cardiovascular disease, oncology, and infectious diseases. Specifically in sepsis, both NLR and SII have been established as predictors of 90-day mortality, with elevated values indicating a poor prognosis [\u003cspan additionalcitationids=\"CR10\" citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. However, research on the association between NPAR and PNR with sepsis prognosis remains limited. Moreover, the prognostic value of these specific markers (NLR, SII, NPAR, and PNR) in the SA-AKI subpopulation is not well established. Therefore, investigating the relationship between these inflammatory indices and the prognosis of SA-AKI patients may deepen our understanding of the pathophysiology driving poor outcomes and offer novel targets for clinical management.\u003c/p\u003e\u003cp\u003eWhile previous research has predominantly focused on sepsis diagnosis and treatment, and some studies have examined the correlation between inflammatory markers and sepsis or AKI, the specific relationship between whole-blood cell-derived inflammatory markers and all-cause mortality in SA-AKI patients remains understudied. Concurrently, machine learning (ML) methods are increasingly utilized for predicting clinical events across various medical fields [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Although several predictive models have been developed for sepsis[\u003cspan additionalcitationids=\"CR15\" citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e], few have specifically incorporated whole blood cell-derived inflammatory markers to predict outcomes in the SA-AKI subpopulation.\u003c/p\u003e\u003cp\u003eTherefore, this study has two primary objectives. First, we aim to thoroughly analyze the association between these inflammatory markers and all-cause mortality in SA-AKI patients. Second, we aim to develop and validate a machine learning prognostic model incorporating these markers, providing robust evidence for clinical risk stratification and decision-making.\u003c/p\u003e"},{"header":"Methods and Data","content":"\u003cp\u003e\u003cb\u003eData Sources and Study Cohort\u003c/b\u003es\u003c/p\u003e\u003cp\u003eThis retrospective study utilized two distinct cohorts. The primary development and testing cohort was extracted from the Medical Information Mart for Intensive Care (MIMIC-IV, v3.1) database. An independent external validation cohort was assembled from patients diagnosed with SA-AKI at The Second Affiliated Hospital of Nanchang University between 2017 and 2025.\u003c/p\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eEthical Considerations\u003c/h2\u003e\u003cp\u003e The MIMIC-IV (v3.1) database is a large, single-center, open-access repository containing de-identified data for 730,141 intensive care unit admissions at the Beth Israel Deaconess Medical Center (BIDMC) in the United States between 2008 and 2019 (Johnson et al., 2022). The establishment of this database was approved by the Institutional Review Boards (IRBs) of the Massachusetts Institute of Technology (MIT) and BIDMC. As the data are publicly available and de-identified, the requirement for individual patient consent for the present study was waived. One of the authors (Certification No. 13822018) completed the required training and obtained access to the database.\u003c/p\u003e\u003cp\u003e The study protocol for the external validation cohort was approved by the Medical Ethics Committee of The Second Affiliated Hospital of Nanchang University (Approval No.: IIT-O-2025-319). All data from this cohort were anonymized prior to analysis. This study conforms to the principles of the Declaration of Helsinki. 1.3 Inclusion and Exclusion Criteria\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003ePatient Selection\u003c/h3\u003e\n\u003cp\u003eWe identified patients from the databases who met the diagnostic criteria for sepsis according to the Sepsis-3 guidelines[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. From this cohort, we selected adult patients (aged\u0026thinsp;\u0026ge;\u0026thinsp;18 years) who also developed acute kidney injury (AKI), defined in accordance with the Kidney Disease: Improving Global Outcomes (KDIGO) guidelines. For patients with multiple hospitalizations, only data from the first ICU admission of their first hospitalization were included.Patients were excluded if they met any of the following criteria:ICU length of stay\u0026thinsp;\u0026lt;\u0026thinsp;24 hours (due to early discharge or death).Missing data for lymphocyte, neutrophil, or platelet counts, or serum albumin levels, which are necessary to calculate the inflammatory markers.A diagnosis of active malignancy.\u003c/p\u003e\n\u003ch3\u003eData Extraction and Variables\u003c/h3\u003e\n\u003cp\u003eFor the development cohort, data were extracted from the MIMIC-IV (v3.1) database using Structured Query Language (SQL) in Navicat Premium (version 17). For the external validation cohort, data were retrieved from the electronic medical and record systems of The Second Affiliated Hospital of Nanchang University.To ensure all variables represented the baseline status, we collected data recorded within the first 24 hours of each patient's ICU admission. The extracted variables included the following categories:Demographics: Age and sex.Comorbidities: A history of hypertension (HTN), pneumonia (PNA), cerebrovascular accident (CVA), chronic kidney disease (CKD), type 2 diabetes mellitus (T2DM), heart failure (HF), myocardial infarction (MI), ischemic heart disease (IHD), chronic obstructive pulmonary disease (COPD), and COVID-19 infection (COV).Vital Signs: The worst values within the first 24 hours for systolic blood pressure (SBP), heart rate (HR), respiratory rate (RR), body temperature, and peripheral oxygen saturation (SpO₂).Laboratory Results: The first-measured values for lactate, blood glucose, arterial blood pH, partial pressure of oxygen (PO₂), hemoglobin (Hb), creatinine, blood urea nitrogen (BUN), alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin, potassium, sodium, prothrombin time (PT), and activated partial thromboplastin time (aPTT). The components for the inflammatory indices (neutrophil, lymphocyte, and platelet counts; albumin) were also extracted from these initial tests.Fluid Balance: Total 24-hour fluid balance.Treatments: Use of continuous renal replacement therapy (CRRT), mechanical ventilation, neuromuscular blockers, sedation and analgesia, vasoactive drugs (e.g., vasopressors), immunosuppressants, nephrotoxic drugs, glucocorticoids, and antihypertensive therapy.\u003c/p\u003e\u003cp\u003eNLR was calculated as neutrophil count / lymphocyte count, NPAR as neutrophil percentage / serum albumin, PNR as platelet count / neutrophil count, and SII as platelet count \u0026times; neutrophil count / lymphocyte count.\u003c/p\u003e\n\u003ch3\u003ePrimary Outcome\u003c/h3\u003e\n\u003cp\u003eThe primary outcome was defined as 14-day all-cause mortality following ICU admission.\u003c/p\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003eStatistical Analysis\u003c/h2\u003e\u003cp\u003eVariables with more than 20% missing values were excluded from the analysis. For the remaining variables, missing data were imputed using Multiple Imputation by Chained Equations (MICE).\u003c/p\u003e\u003cp\u003eBaseline characteristics were summarized and compared between 14-day survivors and non-survivors. Continuous variables were presented as mean (standard deviation, SD) or median (interquartile range, IQR) based on data distribution (e.g., assessed by normality tests). Intergroup differences were compared using Student's t-test or the Mann-Whitney U test, as appropriate. Categorical variables were presented as frequencies and percentages (%). These were compared using the Pearson's chi-square test or Fisher's exact test, particularly when expected cell counts were low (e.g., \u0026lt;\u0026thinsp;5).All analyses were performed using R software (version 4.5.1) and DecisionLine (version 1.1.7.3)[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eSurvival and Association AnalysisTo assess the association between inflammation markers and 14-day mortality, we first generated Kaplan-Meier survival curves, stratifying patients by baseline inflammation levels (quartiles), and compared groups using the log-rank test.We then employed multivariate Cox proportional hazards models to calculate hazard ratios (HR) and 95% confidence intervals (CI) for each inflammatory marker (NLR, SII, PNR, and NPAR). These markers were analyzed as both continuous variables and as categorical variables (quartiles), with the first quartile (Q1) serving as the reference group.Three sequential models were constructed:Model 1: Unadjusted.Model 2: Adjusted for age and sex.Model 3: Adjusted for the covariates in Model 2 plus all baseline demographic, comorbidity, vital sign, laboratory, and treatment variables listed in Section 2.4.Restricted cubic splines (RCS) were used to explore potential non-linear associations between each continuous marker and 14-day mortality. Finally, we conducted subgroup analyses to evaluate the prognostic consistency of these markers across different strata, including age (\u0026lt;\u0026thinsp;65 vs. \u0026ge;65 years), sex, history of chronic kidney disease (CKD), and heart failure (HF).2.8 Machine Learning Model DevelopmentData Splitting and PreprocessingThe MIMIC-IV cohort (development cohort) was randomly partitioned into a training set (70%) and an internal test set (30%). Within the training set, all selected features were standardized using the Z-score method. To address the class imbalance between survivors and non-survivors, the Synthetic Minority Over-sampling Technique (SMOTE) was applied to the training data.Feature SelectionWe performed feature selection on the training set to identify the most relevant predictors. Three methods were employed:LASSO (Least Absolute Shrinkage and Selection Operator) regression.The Boruta algorithm (a random forest-based wrapper method).Univariate logistic regression (retaining variables with p\u0026thinsp;\u0026lt;\u0026thinsp;0.05).Only features identified as significant by all three methods were retained for the final model building.Model TrainingUsing the selected features and the preprocessed training data, we developed six machine learning models: Random Forest, Logistic Regression, XGBoost, GAMboost, CatBoost, and Gradient Boosting Machine (GBM). Model training and hyperparameter optimization (via grid search) were performed using 5-fold cross-validation on the training set. Model Evaluation and InterpretationModel Performance and SelectionThe performance of the six models was evaluated on the internal test set using the Area Under the Receiver Operating Characteristic Curve (AUC), Brier score, precision, recall, F-beta score, and specificity. The model demonstrating the best overall performance was selected as the final model.External Validation and Clinical UtilityThe final model was then validated using the independent external cohort from The Second Affiliated Hospital of Nanchang University. Model calibration (the agreement between predicted probabilities and observed outcomes) was assessed using the Hosmer-Lemeshow goodness-of-fit test. The clinical utility and net benefit of the model were visualized using Decision Curve Analysis (DCA).Model InterpretabilityTo explain the model's predictions, we employed SHAP (SHapley Additive exPlanations). Feature contributions and overall importance were visualized using SHAP summary plots.\u003c/p\u003e\u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003eCohort Characteristics\u003c/h2\u003e\u003cp\u003eBaseline Patient CharacteristicsThis study ultimately included 4,311 patients with SA-AKI admitted to the ICU. Of this cohort, 3,615 patients survived (survivor group) and 696 died within 14 days (non-survivor group). The patient selection process is detailed in Fig.\u0026nbsp;1. The median age of the overall cohort was 65.0 years (IQR 54.0\u0026ndash;75.0), and 40.6% were female.Baseline characteristics, compared between the 14-day survivor and non-survivor groups, are presented in Table\u0026nbsp;1. The non-survivor group was significantly older than the survivor group (median 70.0 vs. 64.0 years, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), although there was no significant difference in sex distribution (p\u0026thinsp;=\u0026thinsp;0.727). Regarding comorbidities, the non-survivor group had significantly higher prevalences of heart failure, myocardial infarction, ischemic heart disease, and chronic obstructive pulmonary disease (all p\u0026thinsp;\u0026lt;\u0026thinsp;0.002).Comparisons of vital signs and laboratory indicators (within the first 24 hours of ICU admission) showed that the non-survivor group had significantly higher heart rates, respiratory rates, and lactate levels, as well as lower arterial pH and PaO₂ (all p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Furthermore, non-survivors presented with significantly elevated creatinine, blood urea nitrogen, total bilirubin, prothrombin time (PT), and activated partial thromboplastin time (aPTT) (all p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and lower hemoglobin levels (9.8 vs. 10.4 g/dL, p\u0026thinsp;=\u0026thinsp;0.009).For the primary inflammatory markers, the non-survivor group exhibited significantly higher NLR (13.09 vs. 9.06, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), NPAR (29.34 vs. 27.23, p\u0026thinsp;=\u0026thinsp;0.002), and SII (2018.98 vs. 1555.91, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Conversely, PNR was significantly lower in non-survivors (14.47 vs. 18.11, p\u0026thinsp;=\u0026thinsp;0.046). The non-survivor group also had significantly higher rates of receiving CRRT, vasoactive drugs, glucocorticoids, and mechanical ventilation (all p\u0026thinsp;\u0026lt;\u0026thinsp;0.05).\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eAssociation of Inflammatory Markers with 14-Day All-Cause Mortality\u003c/h3\u003e\n\u003cp\u003eCox proportional hazards models were used to assess the association between exposure factors and 14-day mortality. Three sequential models were constructed: Model 1 was unadjusted; Model 2 adjusted for age and sex; and Model 3 was fully adjusted for over 30 confounding variables, including comorbidities (e.g., hypertension, diabetes, heart failure, CKD, COPD), treatments (e.g., CRRT, vasoactive drugs), and key laboratory values (e.g., lactate, creatinine, hemoglobin, blood glucose). The hazard ratios (HR) and 95% confidence intervals (95% CI) for 14-day mortality risk, stratified by inflammatory marker groups (NLR, SII, PNR, NPAR), are presented in Table\u0026nbsp;2.\u003c/p\u003e\u003cp\u003eKaplan-Meier analysis (Fig.\u0026nbsp;3) demonstrated a significant difference in survival based on inflammatory marker levels. Patients in the highest tertile (T3) of NLR and SII, as well as those in the highest quartile (Q4) of NPAR, had significantly higher 14-day all-cause mortality than those in the lowest reference groups. Conversely, PNR showed an inverse association: patients in the lowest quartile (Q1) had significantly higher mortality than those in the highest quartile (Q4) (all log-rank p\u0026thinsp;\u0026lt;\u0026thinsp;0.05).Univariate Cox regression analysis (Table\u0026nbsp;2) confirmed that all four markers, when analyzed as continuous variables, were significantly associated with 14-day mortality risk in patients with SA-AKI (all trend p\u0026thinsp;\u0026lt;\u0026thinsp;0.05).In the fully adjusted multivariate model (Model 3), these associations remained significant. For this analysis, NLR and SII were grouped by tertiles, while NPAR and PNR were grouped by quartiles.High NLR (T3 vs. T1) yielded an HR of 1.50 (95% CI: 1.23\u0026ndash;1.83).High NPAR (Q4 vs. Q1) yielded an HR of 1.23 (95% CI: 1.02\u0026ndash;1.49).For SII, the combined upper tertiles (T2/T3 vs. T1) yielded an HR of 1.19 (95% CI: 1.10\u0026ndash;1.39).In contrast, high PNR (Q4 vs. Q1) was protective, indicating a negative correlation with mortality (HR: 0.73, 95% CI: 0.58\u0026ndash;0.93).\u003c/p\u003e\u003cp\u003eRestricted cubic spline (RCS) analysis was employed to evaluate potential non-linear associations between the inflammatory markers and 14-day mortality (Fig.\u0026nbsp;2). The analysis revealed significant non-linear relationships for NLR, NPAR, and PNR (all p for non-linearity\u0026thinsp;\u0026lt;\u0026thinsp;0.05), whereas SII demonstrated a linear relationship with mortality risk (p for non-linearity\u0026thinsp;=\u0026thinsp;0.642). Specifically, the mortality risk began to increase significantly when NLR exceeded 9.54 and when NPAR exceeded 27.56, with the risk for both markers plateauing at high values (approx. 50\u0026ndash;100). Conversely, PNR exhibited a protective effect; the risk was highest at low values and began to decrease significantly once PNR rose above 17.48, stabilizing at levels between 100 and 200. For SII, the risk showed a continuous linear increase for all values above 1610.77.\u003c/p\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003eSubgroup Analysis\u003c/h2\u003e\u003cp\u003eSubgroup analyses were conducted based on numerous baseline comorbidities and treatments, as detailed in Fig.\u0026nbsp;4. Significant interactions were detected: NLR interacted with CRRT, AHT, and MI; NPAR interacted with IMMUNOS, VP, and HF; SII showed interactions with IHD, AHT, and HTN; and PNR interacted with CKD, CRRT, and HF. Despite these interactions, the overall prognostic trends were consistent: a higher NLR correlated with increased mortality risk, and higher NPAR was associated with increased mortality risk. For SII, mortality risk was also higher with SA, NEPHTOX, AHT, combined IHD and HF, and a high SII. In contrast, high PNR was associated with a lower mortality risk overall.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eConstruction of Machine Learning Predictive Models\u003c/h2\u003e\u003cp\u003eFeature selection was performed using three methods (LASSO regression, the Boruta algorithm, and univariate logistic regression), resulting in 17 clinical features being identified by all three techniques (Fig.\u0026nbsp;5). These 17 predictors\u0026mdash;AHT, NEPHTOX, NLR, SA, SII, VENTILATION, VP, MI, Age, First bilirubin total, First hemoglobin, First lactate, First pH, First pO2, First PT, First PTT, and First Urea Nitrogen (Fig.\u0026nbsp;6)\u0026mdash;were subsequently standardized using Z-scores prior to model building.\u003c/p\u003e\u003cp\u003eThe comparative performance metrics for the six machine learning models on the training and internal test sets are presented in Table\u0026nbsp;3 and Fig.\u0026nbsp;7. The CatBoost model demonstrated the best and most consistent performance, achieving the highest Area Under the Receiver Operating Characteristic Curve (AUC) on both the training set (0.861) and the internal test set (0.861). Across all models, the recall and specificity ranges were 0.7103\u0026ndash;0.8215 and 0.6148\u0026ndash;0.7288, respectively, on the training set. On the test set, the ranges were 0.6547\u0026ndash;0.7265 for recall and 0.6275\u0026ndash;0.7333 for specificity.\u003c/p\u003e\u003cp\u003eEstablishment and Evaluation of the Optimal Model\u003c/p\u003e\u003cp\u003eBased on its superior performance on the internal test set\u0026mdash;achieving the highest AUC (0.861) and an optimal balance between recall and specificity\u0026mdash;the CatBoost model was selected as the final model. The model was then optimized using grid search hyperparameter tuning .\u003c/p\u003e\u003cp\u003eThe final CatBoost model was subsequently validated on the independent, external cohort (Table \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e). In this external validation, the model demonstrated robust discrimination, achieving an AUC of 0.755 (Fig.\u0026nbsp;8). The calibration curve indicated good agreement between predicted probabilities and observed outcomes. Furthermore, Decision Curve Analysis (DCA) showed a net benefit across the full range of risk thresholds (0 to 1), reflecting its potential applicability and practical utility in supporting clinical decision-making.\u003c/p\u003e\u003cp\u003eSHAP Interpretation of the Catboost Model\u003c/p\u003e\u003cp\u003eTo interpret the final CatBoost model, SHAP (SHapley Additive exPlanations) values were used to illustrate how each feature influences the prediction of 14-day mortality. As shown in the SHAP summary plot (Fig.\u0026nbsp;9), feature contributions are represented by colored dots, where red indicates high feature values (correlating with higher risk) and blue indicates low feature values. The analysis revealed that metabolic indicators\u0026mdash;specifically lactate, blood urea nitrogen, and pH\u0026mdash;were the most critical factors driving the model's predictions. The use of vasoactive drugs (VP) was also ranked as a highly important contributor, correlating strongly with a high predicted risk of mortality.\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eTo our knowledge, this study is the first to systematically investigate the risk association between four whole-blood cell-derived inflammatory markers (NLR, SII, PNR, NPAR) and 14-day mortality in patients with sepsis-associated acute kidney injury (SA-AKI), confirming the critical role of inflammatory imbalance in this population. Through multidimensional clinical feature analysis, integrating these inflammatory markers with other key clinical indicators (e.g., lactate levels, mechanical ventilation use), we developed an optimal mortality risk prediction model using machine learning and interpreted the model's predictions using SHAP analysis. This approach provides an efficient and interpretable tool for clinical prognostic assessment in patients with SA-AKI.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;This study demonstrates that NLR, SII, and NPAR are significantly positively associated with 14-day all-cause mortality, while PNR is negatively associated. Our findings for NLR and SII are largely consistent with recent meta-analyses in general sepsis populations: Hongsheng Wu et al. demonstrated that elevated NLR correlates with poorer short-term outcomes[19] (PMID: 38562922), and Lingbo Liang et al. indicated that high SII values at admission significantly increase the risk of short-term all-cause mortality in sepsis patients [20]. Our finding that elevated NPAR independently correlates with higher mortality risk is supported by Yuqiang Gong et al., who analyzed 2,166 patients with severe sepsis or septic shock from the MIMIC-III database, revealing that higher NPAR levels are associated with increased risk of 30-day, 90-day, and 365-day mortality[21]. A retrospective study by Dan Wu et al., however, analyzed clinical data from 203 sepsis patients, revealing that significantly elevated PNR in sepsis patients was positively correlated with mortality, suggesting PNR serves as an independent prognostic indicator to identify high-risk patients \u0026nbsp;[22].\u003c/p\u003e\n\u003cp\u003eSepsis-associated acute kidney injury (SA-AKI) represents a complex physiological and immunological response, with systemic inflammation and endothelial dysfunction as key pathophysiological features [23].Inflammatory markers such as NLR, SII, NPAR, and PNR may reflect distinct inflammatory-coagulation-tissue injury dynamics within the context of SA-AKI.\u003c/p\u003e\n\u003cp\u003eIn sepsis, neutrophils participate in inflammatory responses through multiple mechanisms, including releasing granules containing MPO, NE, and metalloproteinases, and forming neutrophil extracellular traps (NETs)\u0026nbsp;[24].Neutrophil-associated matrix metalloproteinase-9 (MMP-9) activity is upregulated and can directly degrade the glycocalyx [25]. Disruption of the glomerular endothelial glycocalyx leads to microvascular barrier impairment [26]and exposure of adhesion molecules (e.g., ICAM-1) [27], potentially exacerbating tissue injury. Loss of the endothelial anticoagulant barrier activates the coagulation cascade (e.g., vWF release), exacerbating intrarenal thrombotic microangiopathy (TMA) [28-30] .Concurrently, delayed neutrophil apoptosis and impaired migration prolong the inflammatory response, creating a vicious cycle [31, 32]. Neutrophil elastase and cathepsin G activate platelets via the receptor for activated serine protease (RAPS) [33]. As Braedon McDonald et al. demonstrated in septic mice, the NET-platelet-thrombin axis promotes intravascular coagulation and microvascular dysfunction in sepsis (PMID: 28073784)[34]. Conversely, lymphocytes undergo widespread apoptosis and functional failure in sepsis, particularly T-cell exhaustion[35] and PD-1 pathway upregulation, blunting the immune response to innate immune overactivation. This dysfunction also impairs the immune regulation of renal tubular injury repair and antifibrosis, hindering the resolution of the inflammation-coagulation axis [36, 37]. Albumin functions as both a carrier/antioxidant molecule and a maintainer of endothelial glycocalyx integrity and colloid osmotic pressure; hypoalbuminemia exacerbates capillary leakage, interstitial edema, and elevated renal venous pressure, thereby amplifying renal ischemia-reperfusion vulnerability and drug/endotoxin burden[38]. Collectively, neutrophil-platelet-coagulation network-driven microthrombosis and endothelial dysfunction, compounded by lymphocyte-mediated immunosuppression and barrier disruption due to hypoalbuminemia, constitute the key pathophysiological framework of SA-AKI.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;Subgroup analyses revealed significant interactions between the inflammatory markers and distinct clinical characteristics: NLR interacted with CRRT, AHT, and MI; NPAR interacted with immunosuppressive drugs (IMMUNOS), vasoactive drug use (VP), and heart failure (HF); SII interacted with IHD, AHT, and HTN; and PNR interacted with CKD, CRRT, and HF.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;The interaction between NLR and CRRT may arise because CRRT removes metabolic byproducts and filters/adsorbs certain inflammatory mediators, thereby altering the patient's inflammatory burden and immune microenvironment \u0026nbsp;\u0026nbsp;[39].This suggests CRRT may partially \"offset\" the adverse risk associated with an elevated NLR. Furthermore, patients exhibiting severe inflammation/stress states (high NLR) may derive greater benefit from CRRT. The interaction with antihypertensive therapy (AHT) may also be influential. Physiological stress from infection or inflammation increases endogenous catecholamine production, elevating neutrophil counts and causing lymphopenia[40]\u0026nbsp;.Antihypertensive drugs, particularly angiotensin-converting enzyme inhibitors (ACEIs), effectively protect the endothelial glycocalyx and mitigate capillary ultrastructural abnormalities\u0026nbsp;[41], potentially influencing outcomes. Finally, MI itself induces robust neutrophil infiltration and lymphocyte apoptosis, amplifying the inflammatory-immune response\u0026nbsp;[42], which may diminish the sensitivity and specificity of NLR as a prognostic marker.\u003c/p\u003e\n\u003cp\u003eNPAR's interaction with IMMUNOS may be twofold: first, immunosuppressants directly inhibit neutrophil activation and inflammatory cytokine release, thereby affecting NPAR's predictive capacity. Second, immunosuppressant users are often organ transplant recipients or autoimmune disease patients, and these underlying conditions themselves influence inflammation and nutritional status, introducing heterogeneity to NPAR risk stratification. Regarding vasoactive drugs (VP), while they improve microcirculation by maintaining blood pressure, they may simultaneously affect endothelial function. For instance, vasoactive amines can alter polymorphonuclear leukocyte expression in vitro via direct endothelial cell action, thereby influencing permeability and subsequently affecting albumin levels\u0026nbsp;[43].Lastly, heart failure patients often exhibit chronic low-grade inflammation and protein-energy wasting, leading to elevated baseline neutrophils and decreased albumin. This existing pathophysiology may enhance NPAR's prognostic ability in the heart failure population[44, 45].\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;SII's interaction with ischemic heart disease (IHD) is notable, as studies indicate that elevated SII is independently associated with an increased risk of adverse cardiovascular events in patients undergoing percutaneous coronary intervention (PCI)\u0026nbsp;[46]\u0026nbsp;.In the pathogenesis of hypertension (HTN), classically activated macrophages (M1), neutrophils, and dendritic cells secrete inflammatory mediators, potentially enhancing SII's sensitivity to adverse outcomes\u0026nbsp;[47]\u0026nbsp;.Furthermore, the long-term use of antihypertensive drugs (AHT), such as ACEi/ARBs or β-blockers, exerts anti-inflammatory and immunomodulatory effects that can alter baseline neutrophil, lymphocyte, and platelet levels, thereby influencing the observed relationship between SII and outcomes.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;PNR interacts with CKD, possibly because uremic solutes in CKD can induce endothelial dysfunction, inflammation, and oxidative stress, thereby altering PNR and its prognostic capacity [48] .Its interaction with CRRT may be explained by research from Jian-Biao Meng et al., which indicates that early initiation of continuous veno-venous hemofiltration improves endothelial function and hemodynamic stability, potentially influencing the PNR-prognosis relationship[49]. Finally, the interaction with HF may be linked to shared chronic inflammation pathways, where inflammatory responses stimulate NLRP3 inflammasome activation and IL-1ß release, promoting the migration of multiple immune cells[22, 50].\u003c/p\u003e\n\u003cp\u003eThis retrospective cohort study, while representing the first systematic investigation of the association between these whole-blood-derived inflammatory markers and SA-AKI prognosis, has several limitations. First, despite enrolling 4,311 patients and conducting external validation, selection bias and information bias are inherent to its retrospective design. Moreover, causal relationships between the inflammatory markers and mortality risk cannot be confirmed. Second, the study relied solely on baseline measurements of inflammatory components (neutrophil, lymphocyte, platelet counts, and serum albumin) at ICU admission, failing to assess the dynamic changes of these indicators throughout the disease course or their relationship with clinical progression.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTherefore, future research should focus on the relationship between the dynamic changes in these inflammatory markers and sepsis-associated kidney injury. Concurrently, leveraging prospective designs and multi-omics integration holds promise for systematically elucidating the molecular mechanisms by which systemic inflammatory responses mediate kidney injury and revealing their biological basis. This would advance clinical practice in precision risk stratification, metabolically targeted interventions, and personalized treatment for high-risk patients.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study demonstrates that the whole-blood cell-derived inflammatory markers NLR, SII, NPAR, and PNR are significantly associated with 14-day all-cause mortality in patients with SA-AKI. Specifically, NLR, SII, and NPAR were identified as independent risk factors, whereas PNR was found to be a protective factor. Furthermore, in the machine learning prognostic model developed from these findings, the CatBoost algorithm exhibited optimal predictive performance and generalization capability. This model not only enables precise risk stratification but also offers high clinical interpretability, as demonstrated by SHAP analysis. Collectively, this study confirms the significant value of these inflammatory markers in SA-AKI prognosis assessment and provides an effective tool that may assist clinical decision-making and facilitate personalized patient management.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets analysed during the current study are available in the\u0026nbsp;MIMIC-IV 3.1 and\u003c/p\u003e\n\u003cp\u003eMedical Ethics Committee of The Second Affiliated Hospital of Nanchang University(Approval No.: IIT-O-2025-319)\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the Regional Science Foundation Program of National Natural Science Foundation of China (82260374).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors contributed to the study conception and design. Writing - original draft preparation:Xinghe Shangguan; Writing - review and editing: Xinghe Shangguan,Yongjie Luo,Yang Zhou; Conceptualization:Xinghe Shangguan,Xiayoumei Wu,Xiaofan Zou; Methodology:Xinghe Shangguan,Qianyu Yuan; Formal analysis and investigation: Xinghe Shangguan; Funding acquisition:Xinghe Shangguan; Resources:Xinghe Shangguan,Yuanqi Gong,Jianning Xu; Supervision: Yuanqi Gong,Jianning Xu,and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eCohen J, Vincent J L, Adhikari N K J, et al. Sepsis: A Roadmap for Future Research. The Lancet Infectious Diseases, 2015, 15(5): 581\u0026ndash;614.\u003c/li\u003e\n \u003cli\u003eG\u0026oacute;mez H, Kellum J A. Sepsis-Induced Acute Kidney Injury. Current Opinion in Critical Care, 2016, 22(6): 546\u0026ndash;553.\u003c/li\u003e\n \u003cli\u003eAngus D C, Linde-Zwirble W T, Lidicker J, et al. Epidemiology of Severe Sepsis in the United States: Analysis of Incidence, Outcome, and Associated Costs of Care:. Critical Care Medicine, 2001, 29(7): 1303\u0026ndash;1310.\u003c/li\u003e\n \u003cli\u003eChalfin D B, Fein A M, Carlon G C, et al. Cost-Effectiveness of Monoclonal Antibodies to Gram-Negative Endotoxin in the Treatment of Gram-Negative Sepsis in ICU Patients.\u003c/li\u003e\n \u003cli\u003ePeerapornratana S, Manrique-Caballero C L, G\u0026oacute;mez H, et al. Acute Kidney Injury from Sepsis: Current Concepts, Epidemiology, Pathophysiology, Prevention and Treatment. Kidney International, 2019, 96(5): 1083\u0026ndash;1099.\u003c/li\u003e\n \u003cli\u003eHaybar H, Pezeshki S M S, Saki N. Evaluation of Complete Blood Count Parameters in Cardiovascular Diseases: An Early Indicator of Prognosis?. Experimental and Molecular Pathology, 2019, 110: 104267.\u003c/li\u003e\n \u003cli\u003eDong G, Gan M, Xu S, et al. The Neutrophil\u0026ndash;Lymphocyte Ratio as a Risk Factor for All-Cause and Cardiovascular Mortality among Individuals with Diabetes: Evidence from the NHANES 2003\u0026ndash;2016. Cardiovascular Diabetology, 2023, 22(1): 267.\u003c/li\u003e\n \u003cli\u003eTudurachi B S, Anghel L, Tudurachi A, et al. Assessment of Inflammatory Hematological Ratios (NLR, PLR, MLR, LMR and Monocyte/HDL\u0026ndash;Cholesterol Ratio) in Acute Myocardial Infarction and Particularities in Young Patients. International Journal of Molecular Sciences, 2023, 24(18): 14378.\u003c/li\u003e\n \u003cli\u003eChen J, Yasrebinia S, Ghaedi A, et al. Meta-Analysis of the Role of Neutrophil to Lymphocyte Ratio in Neonatal Sepsis. BMC Infectious Diseases, 2023, 23(1): 837.\u003c/li\u003e\n \u003cli\u003eDrăgoescu A N, Pădureanu V, Stănculescu A D, et al. Neutrophil to Lymphocyte Ratio (NLR)\u0026mdash;a Useful Tool for the Prognosis of Sepsis in the ICU. Biomedicines, 2021, 10(1): 75.\u003c/li\u003e\n \u003cli\u003eZhang Y, Peng W, Zheng X. The Prognostic Value of the Combined Neutrophil-to-Lymphocyte Ratio (NLR) and Neutrophil-to-Platelet Ratio (NPR) in Sepsis. Scientific Reports, 2024, 14(1): 15075.\u003c/li\u003e\n \u003cli\u003eRen Y, Loftus T J, Datta S, et al. Performance of a Machine Learning Algorithm Using Electronic Health Record Data to Predict Postoperative Complications and Report on a Mobile Platform. JAMA Network Open, 2022, 5(5): e2211973.\u003c/li\u003e\n \u003cli\u003eSong Y, Yang X, Luo Y, et al. Comparison of Logistic Regression and Machine Learning Methods for Predicting Postoperative Delirium in Elderly Patients: A Retrospective Study. CNS Neuroscience \u0026amp; Therapeutics, 2023, 29(1): 158\u0026ndash;167.\u003c/li\u003e\n \u003cli\u003eLiu Z, Shu W, Li T, et al. Interpretable Machine Learning for Predicting Sepsis Risk in Emergency Triage Patients. Scientific Reports, 2025, 15(1): 887.\u003c/li\u003e\n \u003cli\u003eHou N. Predicting 30-Days Mortality for MIMIC-III Patients with Sepsis-3: A Machine Learning Approach Using XGboost.2020.\u003c/li\u003e\n \u003cli\u003eYue S, Li S, Huang X, et al. Machine Learning for the Prediction of Acute Kidney Injury in Patients with Sepsis. Journal of Translational Medicine, 2022, 20(1): 215.\u003c/li\u003e\n \u003cli\u003eSinger M, Deutschman C S, Seymour C W, et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA, 2016, 315(8): 801.\u003c/li\u003e\n \u003cli\u003eShen D, Sha L, Yang L, et al. Identification of Multiple Complications as Independent Risk Factors Associated with 1-, 3-, and 5-Year Mortality in Hepatitis B-Associated Cirrhosis Patients. BMC Infectious Diseases, 2025, 25(1): 151.\u003c/li\u003e\n \u003cli\u003eWu H, Cao T, Ji T, et al. Predictive Value of the Neutrophil-to-Lymphocyte Ratio in the Prognosis and Risk of Death for Adult Sepsis Patients: A Meta-Analysis. Frontiers in Immunology, 2024, 15: 1336456.\u003c/li\u003e\n \u003cli\u003eLiang L, Su Q. Systemic Immune-Inflammation Index and the Short-Term Mortality of Patients with Sepsis: A Meta-Analysis. Biomolecules and Biomedicine, 2024, 25(4): 798\u0026ndash;809.\u003c/li\u003e\n \u003cli\u003eGong Y, Li D, Cheng B, et al. Increased Neutrophil Percentage-to-Albumin Ratio Is Associated with All-Cause Mortality in Patients with Severe Sepsis or Septic Shock. Epidemiology and Infection, 2020, 148: e87.\u003c/li\u003e\n \u003cli\u003eWu D, Qin H. Diagnostic and Prognostic Values of Immunocyte Ratios in Patients with Sepsis in the Intensive Care Unit. The Journal of Infection in Developing Countries, 2023, 17(10): 1362\u0026ndash;1372.\u003c/li\u003e\n \u003cli\u003eLi J C, Wang L J, Feng F, et al. Role of Heparanase in Sepsis‑related Acute Kidney Injury (Review). Experimental and Therapeutic Medicine, 2023, 26(2): 379.\u003c/li\u003e\n \u003cli\u003eZhang H, Wang Y, Qu M, et al. Neutrophil, Neutrophil Extracellular Traps and Endothelial Cell Dysfunction in Sepsis. Clinical and Translational Medicine, 2023, 13(1): e1170.\u003c/li\u003e\n \u003cli\u003eLin L, Niu M, Gao W, et al. Predictive Role of Glycocalyx Components and MMP-9 in Cardiopulmonary Bypass Patients for ICU Stay. Heliyon, 2024, 10(1): e23299.\u003c/li\u003e\n \u003cli\u003eGamez M, Elhegni H E, Fawaz S, et al. Heparanase Inhibition as a Systemic Approach to Protect the Endothelial Glycocalyx and Prevent Microvascular Complications in Diabetes. Cardiovascular Diabetology, 2024, 23(1): 50.\u003c/li\u003e\n \u003cli\u003eZhang W, Guan Y, Bayliss G, et al. Class IIa HDAC Inhibitor TMP195 Alleviates Lipopolysaccharide-Induced Acute Kidney Injury. American Journal of Physiology Renal Physiology, 2020, 319(6): F1015\u0026ndash;F1026.\u003c/li\u003e\n \u003cli\u003eKei C Y, Singh K, Dautov R F, et al. Coronary \u0026ldquo;Microvascular Dysfunction\u0026rdquo;: Evolving Understanding of Pathophysiology, Clinical Implications, and Potential Therapeutics. International Journal of Molecular Sciences, 2023, 24(14): 11287.\u003c/li\u003e\n \u003cli\u003ePeter B, Kanyo N, Kovacs K D, et al. Glycocalyx Components Detune the Cellular Uptake of Gold Nanoparticles in a Size- and Charge-Dependent Manner. ACS Applied Bio Materials, 2023, 6(1): 64\u0026ndash;73.\u003c/li\u003e\n \u003cli\u003eMedica D, Quercia A D, Marengo M, et al. High-Volume Hemofiltration Does Not Protect Human Kidney Endothelial and Tubular Epithelial Cells from Septic Plasma-Induced Injury. Scientific Reports, 2024, 14(1): 18323.\u003c/li\u003e\n \u003cli\u003eS\u0026ocirc;nego F, Castanheira F V E S, Ferreira R G, et al. Paradoxical Roles of the Neutrophil in Sepsis: Protective and Deleterious. Frontiers in Immunology, 2016, 7.\u003c/li\u003e\n \u003cli\u003eMantovani A, Bonecchi R, Locati M. Tuning Inflammation and Immunity by Chemokine Sequestration: Decoys and More. Nature Reviews Immunology, 2006, 6(12): 907\u0026ndash;918.\u003c/li\u003e\n \u003cli\u003eSambrano G R, Huang W, Faruqi T, et al. Cathepsin G Activates Protease-Activated Receptor-4 in Human Platelets. Journal of Biological Chemistry, 2000, 275(10): 6819\u0026ndash;6823.\u003c/li\u003e\n \u003cli\u003eMcDonald B, Davis R P, Kim S J, et al. Platelets and Neutrophil Extracellular Traps Collaborate to Promote Intravascular Coagulation during Sepsis in Mice. Blood, 2017, 129(10): 1357\u0026ndash;1367.\u003c/li\u003e\n \u003cli\u003eVan Der Poll T, Van De Veerdonk F L, Scicluna B P, et al. The Immunopathology of Sepsis and Potential Therapeutic Targets. Nature Reviews Immunology, 2017, 17(7): 407\u0026ndash;420.\u003c/li\u003e\n \u003cli\u003eCao C, Yao Y, Zeng R. Lymphocytes: Versatile Participants in Acute Kidney Injury and Progression to Chronic Kidney Disease. Frontiers in Physiology, 2021, 12: 729084.\u003c/li\u003e\n \u003cli\u003eHotchkiss R S, Monneret G, Payen D. Sepsis-Induced Immunosuppression: From Cellular Dysfunctions to Immunotherapy. Nature Reviews Immunology, 2013, 13(12): 862\u0026ndash;874.\u003c/li\u003e\n \u003cli\u003eAldecoa C, Llau J V, Nuvials X, et al. Role of Albumin in the Preservation of Endothelial Glycocalyx Integrity and the Microcirculation: A Review. Annals of Intensive Care, 2020, 10(1): 85.\u003c/li\u003e\n \u003cli\u003eChen J J, Lai P C, Lee T H, et al. Blood Purification for Adult Patients with Severe Infection or Sepsis/Septic Shock: A Network Meta-Analysis of Randomized Controlled Trials. Critical Care Medicine, 2023, 51(12): 1777\u0026ndash;1789.\u003c/li\u003e\n \u003cli\u003eChen J J, Kuo G, Fan P C, et al. Neutrophil-to-Lymphocyte Ratio Is a Marker for Acute Kidney Injury Progression and Mortality in Critically Ill Populations: A Population-Based, Multi-Institutional Study. Journal of Nephrology, 2022, 35(3): 911\u0026ndash;920.\u003c/li\u003e\n \u003cli\u003eLocatelli M, Rottoli D, Mahmoud R, et al. Endothelial Glycocalyx of Peritubular Capillaries in Experimental Diabetic Nephropathy: A Target of ACE Inhibitor-Induced Kidney Microvascular Protection. International Journal of Molecular Sciences, 2023, 24(22): 16543.\u003c/li\u003e\n \u003cli\u003eForteza M J, Trapero I, Hervas A, et al. Apoptosis and Mobilization of Lymphocytes to Cardiac Tissue Is Associated with Myocardial Infarction in a Reperfused Porcine Model and Infarct Size in post‐PCI Patients. Madamanchi N. Oxidative Medicine and Cellular Longevity, 2018, 2018(1): 1975167.\u003c/li\u003e\n \u003cli\u003eDoukas J, Shepro D, Hechtman H B. Vasoactive Amines Directly Modify Endothelial Cells to Affect.\u003c/li\u003e\n \u003cli\u003eAmara M, Stoler O, Birati E Y. The Role of Inflammation in the Pathophysiology of Heart Failure. Cells, 2025, 14(14): 1117.\u003c/li\u003e\n \u003cli\u003eNguyen A P, Kawi J, Meraz R, et al. Hidden Malnutrition in Overweight and Obese Individuals with Chronic Heart Failure: Insights from the pro-HEART Trial. Nutrients, 2025, 17(16): 2694.\u003c/li\u003e\n \u003cli\u003eZhang C, Li M, Liu L, et al. Systemic Immune-Inflammation Index as a Novel Predictor of Major Adverse Cardiovascular Events in Patients Undergoing Percutaneous Coronary Intervention: A Meta-Analysis of Cohort Studies. BMC Cardiovascular Disorders, 2024, 24(1): 189.\u003c/li\u003e\n \u003cli\u003eZhang Z, Zhao L, Zhou X, et al. Role of Inflammation, Immunity, and Oxidative Stress in Hypertension: New Insights and Potential Therapeutic Targets. Frontiers in Immunology, 2023, 13: 1098725.\u003c/li\u003e\n \u003cli\u003eDou L, Sall\u0026eacute;e M, Cerini C, et al. The Cardiovascular Effect of the Uremic Solute Indole-3 Acetic Acid. Journal of the American Society of Nephrology, 2015, 26(4): 876\u0026ndash;887.\u003c/li\u003e\n \u003cli\u003eMeng J biao, Lai Z zhen, Xu X juan, et al. Effects of Early Continuous Venovenous Hemofiltration on E-Selectin, Hemodynamic Stability, and Ventilatory Function in Patients with Septic-Shock-Induced Acute Respiratory Distress Syndrome. Biomed Research International, 2016, 2016: 1\u0026ndash;9.\u003c/li\u003e\n \u003cli\u003eSandanger \u0026Oslash;, Ranheim T, Vinge L E, et al. The NLRP3 Inflammasome Is Up-Regulated in Cardiac Fibroblasts and Mediates Myocardial Ischaemia\u0026ndash;Reperfusion Injury. Cardiovascular Research, 2013, 99(1): 164\u0026ndash;174.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTables 1 to 3 are available in the Supplementary Files section\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7986649/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7986649/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSepsis-associated acute kidney injury (SA-AKI) is a common and severe complication of sepsis, associated with significantly increased patient mortality. Whole blood cell-derived inflammatory markers, such as the neutrophil-to-lymphocyte ratio (NLR) and the systemic immune-inflammation index (SII), are valuable tools for assessing inflammatory status and prognosis due to their ready availability and cost-effectiveness. However, the association between these markers and prognosis in patients with SA-AKI requires further investigation. This study aims to clarify the relationship between these inflammatory markers and all-cause mortality in SA-AKI patients and to develop a prognostic model using machine learning methods, thereby providing new evidence to support clinical decision-making..\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study included patients with SA-AKI, defined by Sepsis-3 and KDIGO criteria, from the MIMIC-IV database (development cohort) and an external cohort from The Second Affiliated Hospital of Nanchang University (validation cohort). We collected demographic data, vital signs, laboratory parameters (including NLR, NPAR, PNR, and SII), comorbidities, and treatment information. The primary outcome was 14-day all-cause mortality following ICU admission.Kaplan-Meier survival analysis and multivariate Cox regression models were used to evaluate the association between inflammatory markers and mortality. For model development, the MIMIC-IV cohort was randomly split into training (70%) and testing (30%) sets. Feature selection was performed using LASSO regression, the Boruta algorithm, and univariate logistic regression. Six machine learning models (Random Forest, Logistic Regression, XGBoost, GAMboost, CatBoost, and GBM) were subsequently developed. Five-fold cross-validation was employed for model tuning, and the SMOTE (Synthetic Minority Over-sampling Technique) algorithm was applied to address class imbalance.Model performance was evaluated using the Area Under the Receiver Operating Characteristic Curve (AUC) and Brier score. Model interpretability was assessed using SHAP (SHapley Additive exPlanations), and clinical utility was determined via Decision Curve Analysis (DCA). The best-performing model was then validated using the external cohort.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA total of 4,311 patients with SA-AKI were included in the development cohort, of whom 696 (16.1%) died within 14 days. Survival analysis demonstrated that elevated NLR, SII, and NPAR, as well as a low PNR, were significantly associated with 14-day all-cause mortality. Following feature selection, 17 predictors were used to develop six machine learning models. The CatBoost model achieved the best discrimination, with an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.861 on the internal test set and 0.755 on the external validation cohort. Decision curve analysis confirmed its potential clinical utility across a wide range of threshold probabilities. Furthermore, SHAP analysis identified lactate and blood urea nitrogen as the most significant contributors to the model's predictions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe predictive model developed using the CatBoost algorithm effectively assesses the 14-day mortality risk in patients with SA-AKI. The model demonstrated robust predictive performance and clinical applicability in both internal and external validation cohorts. This tool holds promise for assisting clinicians with risk stratification and guiding personalized treatment strategies for SA-AKI patients.\u003c/p\u003e","manuscriptTitle":"Association Between Whole Blood Cell-Derived Inflammatory Markers and All-Cause Mortality in Patients with Sepsis-Related Acute Kidney Injury and Development of a Machine Learning Prognostic Model","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-11 16:43:29","doi":"10.21203/rs.3.rs-7986649/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"71da55c9-8ec1-4f99-85fb-2baf2982c633","owner":[],"postedDate":"November 11th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-11-15T01:08:07+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-11 16:43:29","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7986649","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7986649","identity":"rs-7986649","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}