Building a Machine Learning Model to Predict the Early Mortality Risk in Pediatric ICU Sepsis Patients

Building a Machine Learning Model to Predict the Early Mortality Risk in Pediatric ICU Sepsis Patients | 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 Building a Machine Learning Model to Predict the Early Mortality Risk in Pediatric ICU Sepsis Patients Lin Yang, Na Zang, Ying Yang, KaiBing Pu, Cong Liu, LiPing Tan This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7842349/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 Objective: This study aims to construct a risk prediction model for early mortality in pediatric intensive care unit (PICU) sepsis patients using machine learning methods. The model is intended to assist clinicians in identifying high-risk patients to enable timely interventions, and to help prevent early deaths in PICU sepsis patients. Methods : A single-center retrospective cohort study design was used; clinical data of sepsis patients admitted to the PICU of Chongqing Medical University Affiliated Children's Hospital from January 2015 to December 2021 were included. The data comprised demographic information, vital signs, complications, laboratory indicators, diagnoses, and treatments. Patients were divided into early mortality and survival groups based on whether death occurred within 14 days of PICU admission. Seventy percent of the data were randomly assigned to the training set, and 30% were assigned to the validation set. Seven machine learning models, including logistic regression (LR), decision tree (DT), random forest (RF), XGBoost (XGB), LightGBM (LGBM), Naive Bayes (NB), and support vector machine (SVM), were used to build the early mortality prediction model for pediatric sepsis patients. The model's predictive performance was evaluated using sensitivity, specificity, receiver operating characteristic curve (ROC curve), and calibration curve. The clinical application value of the model was assessed through decision curve analysis (DCA). Results: A total of 1,559 pediatric sepsis patients were included, with 198 cases of early mortality, resulting in an early mortality rate of 12.7%. After feature selection using LASSO regression, recursive feature elimination (RFE), and Boruta algorithms, six optimal predictive variables were identified: blood transfusion, pediatric Sequential Organ Failure Assessment (pSOFA) score, cardiopulmonary resuscitation, days of mechanical ventilation, secondary infection(s), and septic shock. These variables were used to construct the early mortality prediction model. ROC curve analysis showed that the area under the curve (AUC) values of the seven models ranged from 0.88 to 0.94. The XGBoost and LGBM models performed the best, both achieving an AUC of 0.94. Among all models, XGBoost and LGBM had the highest accuracy, 0.929 and 0.925 respectively; sensitivity was 0.534 and 0.483; specificity was 0.985 and 0.988; and F1 scores were 0.653 and 0.615, respectively. Calibration curves indicated that the XGBoost and LGBM models performed best among the seven machine learning models, showing excellent calibration performance. Decision curve analysis demonstrated that, compared to other models, XGBoost and LGBM exhibited greater net benefits across a wider range of threshold probabilities, indicating good clinical application value and confirming them as the optimal models. Conclusion: Machine learning models are reliable tools for predicting early mortality risk in pediatric patients with sepsis in the PICU. The XGBoost and LGBM models were built using six key variables: blood transfusion, Pediatric Sequential Organ Failure Assessment (pSOFA) score, post-cardiopulmonary resuscitation (CPR), days of mechanical ventilation, secondary infection, and septic shock. These models demonstrate stable performance, high discriminatory ability, and accuracy. Moreover, they aid clinicians in identifying high-risk patients and implementing early interventions. Trial Registration: The trial was registered on the Chinese Clinical Trial Registry with the number ChiCTR2300068260 on February 13, 2023. Pediatric sepsis PICU early mortality risk machine learning prediction models Figures Figure 1 Figure 2 Figure 3 Figure 4 Background Sepsis is one of the leading causes of death among children worldwide, posing a persistent and significant challenge in the field of intensive care (1). According to data from the Global Burden of Disease study, sepsis causes approximately 3 million deaths annually among children under five years old, with low- and middle-income countries bearing over 90% of the disease burden (2). Despite advancements in sepsis diagnosis and treatment, early mortality rates remain high, especially in resource-limited areas, where about 20–30% of pediatric sepsis patients in the PICU die within 72 hours of onset (3). Therefore, predicting the early mortality risk of sepsis patients is of significant clinical importance, as these predictions can help determine the patient's disease status, improve treatment outcomes, and reduce early mortality risk. Sepsis is an acute organ dysfunction syndrome caused by a dysregulated immune response (4). However, we lack precise definitions and diagnostic tests for this dysregulated host response. The heterogeneity of sepsis has long been considered a barrier to translating preclinical research into clinical applications and identifying targeted therapies (5). This heterogeneity leads to significant differentiation in clinical trajectories, specifically manifested as early mortality, rapid recovery, and chronic critical illness (CCI) phenotypes (6), with up to 33% of sepsis survivors not recovering quickly or completely, but rather developing a new syndrome of "chronic critical illness" (CCI) (7,8). This study, based on clinical data from the largest pediatric medical center in Southwest China, found that sepsis patients with underlying respiratory diseases or trauma, as well as those with high SOFA scores, surgical procedures, blood transfusions, or prolonged mechanical ventilation, are prone to developing CCI, which leads to prolonged hospital stays and increased rates of secondary infections caused by CCI (9). The pattern of early mortality sharply contrasts with that of the chronic critical illness (CCI), suggesting the potential existence of different pathophysiological mechanisms and risk factor profiles. Although several studies have explored the treatment and prognostic factors of sepsis, existing research mostly focuses on traditional clinical indicators and empirical treatments. However, existing research lacks risk prediction models that systematically utilize modern technology (10). For example, some studies have identified clinical features associated with poor prognosis, such as higher pediatric sequential organ failure assessment scores and invasive mechanical ventilation; however, there is still insufficient systematic integration of these features for early mortality risk assessment (11). Additionally, existing prediction models often fail to adequately consider the heterogeneity of sepsis, leading to limited applicability across different patient populations (12).Therefore, developing a risk prediction model that can effectively identify high-risk pediatric sepsis patients and support clinical decision-making is particularly important. This retrospective cohort study uses clinical data from the PICU of Chongqing Medical University Affiliated Children's Hospital to explore risk factors for early mortality in pediatric sepsis patients. To achieve this, we employed machine learning techniques to aggregate extensive clinical data, identify key risk factors for early mortality, and develop a prediction model with clinical utility. During data analysis, various machine learning algorithms were applied to evaluate their performance in predicting early mortality risk, aiming to identify the most effective risk prediction model. Subjects and Methods Study population A total of 1735 pediatric patients with sepsis hospitalized in the Pediatric Intensive Care Unit (PICU) of the Chongqing Medical University Affiliated Children's Hospital from January 2015 to December 2024 were included. After excluding patients with repeated admissions, missing clinical or demographic data, or severe cranial injuries based on predefined criteria, 1559 patients remained for analysis. Data on demographics, infection sources, laboratory indicators, treatment measures, and outcomes were collected through the electronic medical record system. Inclusion and Exclusion Criteria Inclusion Criteria: (1) Patients hospitalized in the PICU; (2) Aged 0-18 years; (3) Clinically diagnosed with sepsis or septic shock (defined as SOFA score ≥2 or PSS score ≥2); (4) It was the patient's first episode of sepsis; (5) Patients who received treatment according to the clinical management protocol standards for sepsis in this study. Exclusion Criteria: (1) Patients who were not admitted to the PICU or did not meet the sepsis diagnostic criteria; (2) Patients with severe traumatic brain injury (evidence of neurological injury on CT scan and Glasgow Coma Scale score <8); (3) Patients with incomplete case information. Ethical approval and consent: This study was approved by the Ethics Committee of the Chongqing Medical University Affiliated Children’s Hospital (Approval Number: 534/2022) and informed consent was obtained from the guardians. The study was conducted in accordance with the ethical principles of the Declaration of Helsinki. Clinical trial number: ChiCTR2300068260. Since this study was a retrospective study, Consent to Participate declarations: not applicable. Data Collection and Methods Data Collection Clinical data were collected through hospital electronic medical record systems and follow-up. Demographic information included gender, age, and weight; clinical features included sources of sepsis infection, severity of sepsis (whether septic shock was present), indicators of organ dysfunction, and the presence of underlying diseases. Laboratory tests mainly included white blood cell count, neutrophil count, lymphocyte count, hemoglobin, platelet count, procalcitonin, bilirubin, creatinine, albumin, and D-dimer. Treatment conditions included the use of vasoactive drugs, occurrence of cardiopulmonary resuscitation, duration of mechanical ventilation, surgical procedures, blood transfusions, blood purification, ICU length of stay, and total hospital stay. The occurrence and specific sites of secondary infections during hospitalization were also recorded. Diagnostic Criteria The diagnosis of sepsis was based on the pediatric Sequential Organ Failure Assessment (pSOFA) score to assess the extent of involvement of multiple systems, including the respiratory system, coagulation function, liver, and kidney function. If pSOFA ≥2, the diagnosis of sepsis is established (13). Severe infection leading to cardiovascular dysfunction—including hypotension requiring vasoactive drug treatment or abnormal perfusion—is classified as septic shock. In conjunction with the "2024 International Consensus Standards: Pediatric Sepsis and Septic Shock," the Phoenix Sepsis Score (PSS) was used for assessment, which includes relevant indicators from the cardiovascular, respiratory, neurological, and coagulation systems. If the PSS score ≥2, the child with suspected or confirmed infection is diagnosed with sepsis. Septic shock is defined as sepsis with a PSS cardiovascular score of at least 1 (4). Group classification Based on sepsis definitions, patients diagnosed with sepsis or septic shock in our PICU were divided into two groups to facilitate outcome analysis. Early Mortality Group: Patients who died within 14 days of PICU admission (198 cases). Non-Early Mortality Group: Patients not meeting the criteria for the early mortality group were classified as the non-early mortality group (1361 cases), including the rapid recovery group (RAP, 1058 cases) and the chronic critical illness (CCI) group (303 cases). The CCI group is defined by an ICU length of stay (LOS) ≥14 days and evidence of persistent organ dysfunction, assessed using the Sequential Organ Failure Assessment (SOFA) score at 14 days (i.e., cardiovascular SOFA score ≥1 or any other organ system score ≥2). Statistical Methods Statistical analyses were performed using SPSS 26.0, R 4.4.1, and Python 3.12 software. Normally distributed continuous data were expressed as mean ± standard deviation and analyzed using t-tests; non-normally distributed continuous data were expressed as median with interquartile ranges M[IQR] and analyzed using Mann-Whitney U tests; categorical data were expressed as counts and percentages, and analyzed using χ2 tests or Fisher's exact tests. A P-value < 0.05 was considered statistically significant. The dataset was randomly divided into training and validation sets in a 7:3 ratio. Feature variable selection was performed using LASSO regression, recursive feature elimination (RFE), and the Boruta algorithm, with variables selected by all three algorithms used as predictive factors to construct the prediction model. Seven machine learning models were employed: logistic regression (LR), decision tree (DT), random forest (RF), XGBoost (XGB), LightGBM (LGBM), naive Bayes (NB), and support vector machine (SVM). The predictive performance of the models was evaluated using sensitivity, specificity, receiver operating characteristic curve (ROC curve), area under the curve (AUC), and calibration curve. The clinical utility of the models was assessed through a clinical decision curve analysis. Results Patient Demographics and Sepsis Characteristics After screening, a total of 1,559 pediatric sepsis patients were included in this study. Among them, 198 cases experienced early mortality, resulting in an early mortality rate of 12.7%. There were no significant differences between the early mortality group and the survivor group in terms of gender, age, and weight (p > 0.05). However, the early mortality group had significantly higher pediatric Sequential Organ Failure Assessment (pSOFA) scores and proportions of septic shock, indicating that the degree of organ dysfunction is closely related to disease outcomes and early mortality. The intensive care unit (ICU) length of stay (median 3 days) and total hospital stay (median 4 days) in the early mortality group were significantly shorter than those in the survivor group (ICU length of stay median 7 days, total hospital stay median 19.5 days), and the rate of secondary infections was significantly lower in the early mortality group. The main infection source in the early mortality group was respiratory infection, followed by bloodstream infection; notably, the proportion of bloodstream infections was significantly higher than in the survivor group (19.7% versus 7.0%), suggesting a close association between bloodstream infections and early mortality. The proportion of patients with underlying diseases in the early mortality group was significantly higher than that in the survivor group, particularly those with underlying immune system diseases, mainly immunodeficiencies, and hematological diseases such as leukemia, lymphoma, and aplastic anemia. Additionally, the proportion of patients who underwent cardiopulmonary resuscitation prior to hospitalization was significantly increased in the early mortality group.The early mortality group had significantly lower hemoglobin, absolute lymphocyte count, platelet count, and albumin levels compared to the non-early mortality group, while PCT, D-dimer, and creatinine levels were significantly higher (p < 0.01) (Table 1 ). Table 1 Comparison of Patient Demographics and Characteristics of the Two Groups Sepsis Characteristics Non EM(1359) Early Mortality(198) P-value Male, n (%) 808 (59.5) 121 ( 61.1) 0.714 Age (median [IQR]) 1.75 [0.42, 6.00] 1.88 [0.50, 6.00] 0.269 Weight (median [IQR]) 11.00 [6.94, 20.00] 11.00 [7.00, 21.00] 0.501 PSOFA Score (median [IQR]) 5.00 [3.00, 8.00] 8.00 [6.00, 10.75] < 0.001 Septic Shock (%) 603 (44.4) 155 ( 78.3) < 0.001 Secondary Infection(%) 521 (38.3) 27 ( 13.6) < 0.001 ICU LOS (median [IQR]) 7.00 [4.00, 12.00] 3.00 [1.00, 5.00] < 0.001 Hospital LOS (median [IQR]) 19.500 [13.00, 33.00] 4.00 [2.00, 11.00] < 0.001 Source of sepsis infection Intestinal Infection (%) 246 (18.1) 31 ( 15.7) 0.459 Pneumonia (%) 918 (67.5) 121 ( 61.1) 0.086 NSTI (%) 83 ( 6.1) 9 ( 4.5) 0.478 BloodstreamInfection (%) 95 ( 7.0) 39 ( 19.7) < 0.001 Surgical Site Infection (%) 5 ( 0.4) 0 ( 0.0) 0.855 Urinary TractInfection (%) 14 ( 1.0) 0 ( 0.0) 0.302 CLABSI (%) 13 ( 1.0) 0 ( 0.0) 0.335 Other Infection (%) 11 ( 0.8) 0 ( 0.0) 0.414 Underlying Diseases Cardiovascular Disease (%) 184 (13.5) 21 ( 10.6) 0.304 Immune Disease (%) 71 ( 5.2) 21 ( 10.6) 0.005 Hematologic Disease (%) 123 ( 9.1) 43 ( 21.7) < 0.001 Other Neoplastic Disease (%) 50 ( 3.7) 2 ( 1.0) 0.082 Inherited Metabolic Disease (%) 195 (14.3) 36 ( 18.2) 0.19 Respiratory Disease (%) 27 ( 2.0) 1 ( 0.5) 0.238 Connective Tissue Disease (%) 26 ( 1.9) 7 ( 3.5) 0.224 Neurological Disease (%) 106 ( 7.8) 13 ( 6.6) 0.64 Liver Failure (%) 12 ( 0.9) 2 ( 1.0) 1 Kidney Failure (%) 16 ( 1.2) 3 ( 1.5) 0.954 Post Transplant (%) 30 ( 2.2) 2 ( 1.0) 0.4 Post CPR (%) 30 ( 2.2) 11 ( 5.6) 0.012 Trauma (%) 22 ( 1.6) 4 ( 2.0) 0.908 Drowning (%) 6 ( 0.4) 1 ( 0.5) 1 Laboratory Indicators WBC (median [IQR]) 10.34 [5.81, 16.88] 10.13 [3.93, 16.75] 0.193 Hemoglobin (median [IQR]) 101.00 [88.00, 114.00] 96.00 [82.00, 110.00] 0.002 Lymphocytes (median [IQR]) 2.33 [1.21, 4.44] 1.89 [0.96, 3.62] 0.003 Neutrophils (median [IQR]) 6.49 [3.06, 12.11] 5.94 [1.89, 10.81] 0.064 Platelets (median [IQR]) 267.00 [132.00, 419.50] 106.00 [36.00, 291.00] < 0.001 PCT (median [IQR]) 3.94 [0.69, 14.72] 6.83 [0.70, 33.50] 0.029 Bilirubin (median [IQR]) 6.80 [3.20, 14.00] 7.95 [3.05, 26.30] 0.109 Creatinine (median [IQR]) 32.00 [22.65, 51.85] 42.00 [24.08, 92.03] < 0.001 Albumin (median [IQR]) 32.30 [26.60, 38.70] 28.55 [23.63, 36.50] < 0.001 DD (median [IQR]) 2.76 [1.08, 7.80] 4.27 [1.77, 11.64] 0.001 Treatment measures MechanicalVentilation (%) 968 (71.2) 194 ( 98.0) < 0.001 Surgery (%) 357 (26.3) 23 ( 11.6) < 0.001 BloodTransfusion (%) 427 (31.4) 162 ( 81.8) < 0.001 BloodPurification (%) 162 (11.9) 62 ( 31.3) < 0.001 ECMO (%) 8 ( 0.6) 7 ( 3.5) < 0.001 ROC Curve Analysis of the Predictive Performance of Seven Models for In-Hospital Mortality in Sepsis Patients This study predicted early mortality in pediatric sepsis patients using seven machine learning models: logistic regression (LR), decision tree (DT), random forest (RF), XGBoost (XGB), LightGBM (LGBM), Naive Bayes (NB), and support vector machine (SVM). The ROC curve analysis results showed that the XGBoost and LGBM models had the best predictive performance for early mortality, each achieving an area under the curve (AUC) of 0.94. The LR and RF models followed with AUCs of 0.93, the NB and SVM models both had AUCs of 0.92, and the DT model had the lowest AUC of 0.88(Figure 1 ). These findings indicate that the XGBoost and LGBM models have superior predictive accuracy for early mortality in pediatric sepsis patients, while the DT model’s discriminative ability is lower than that of the other models. Detailed Performance Indicators of the Seven Models Table 2 shows that among the seven models, the XGBoost model had the highest precision, recall, accuracy, and F1 score, demonstrating the best discriminative ability. Calibration curves (Fig. 2 ) indicated that the XGBoost model performed best. Furthermore, decision curve analysis (DCA) (Fig. 3 ) demonstrated that the RF model exhibited greater net benefits over a wider threshold probability range compared to the other models. Taken together, these results suggest that considering all indicators, the XGBoost model is the best model with strong clinical applicability. Model Accuracy Sensitivity Specificity PPV NPV Precision Recall F1 Score ROC AUC LogisticRegression 0.916 (0.890, 0.940) 0.448 (0.342, 0.582) 0.983 (0.971, 0.993) 0.788 (0.596, 0.930) 0.926 (0.902, 0.949) 0.788 (0.656, 0.896) 0.448 (0.342, 0.582) 0.571 (0.438, 0.688) 0.926 (0.880, 0.955) NaiveBayes 0.910 (0.885, 0.931) 0.621 (0.487, 0.764) 0.951 (0.930, 0.970) 0.643 (0.520, 0.793) 0.946 (0.928, 0.968) 0.643 (0.502, 0.750) 0.621 (0.487, 0.764) 0.632 (0.538, 0.743) 0.919 (0.882, 0.952) DecisionTree 0.929 (0.906, 0.948) 0.603 (0.477, 0.729) 0.975 (0.960, 0.988) 0.778 (0.670, 0.875) 0.945 (0.924, 0.964) 0.778 (0.665, 0.884) 0.603 (0.477, 0.729) 0.680 (0.584, 0.763) 0.881 (0.821, 0.921) RandomForest 0.918 (0.893, 0.940) 0.448 (0.318, 0.603) 0.985 (0.975, 0.994) 0.812 (0.678, 0.939) 0.926 (0.903, 0.943) 0.812 (0.672, 0.913) 0.448 (0.318, 0.603) 0.578 (0.435, 0.708) 0.931 (0.889, 0.972) SVM 0.875 (0.852, 0.907) 0.000 (0.000, 0.000) 1.000 (1.000, 1.000) 0.000 (0.000, 0.000) 0.875 (0.849, 0.904) 0.000 (0.000, 0.000) 0.000 (0.000, 0.000) 0.000 (0.000, 0.000) 0.923 (0.880, 0.959) XGBoost 0.929 (0.909, 0.948) 0.534 (0.424, 0.697) 0.985 (0.973, 0.995) 0.838 (0.703, 0.969) 0.937 (0.914, 0.962) 0.838 (0.727, 0.966) 0.534 (0.424, 0.697) 0.653 (0.522, 0.763) 0.936 (0.900, 0.967) LightGBM 0.925 (0.902, 0.945) 0.483 (0.336, 0.604) 0.988 (0.975, 0.995) 0.848 (0.719, 0.969) 0.931 (0.908, 0.951) 0.848 (0.730, 0.945) 0.483 (0.336, 0.604) 0.615 (0.496, 0.732) 0.938 (0.903, 0.968) Table 2 . Performance of Seven Machine Learning Models in Predicting Early Mortality in Pediatric Sepsis Patients Table 2 Performance of Seven Machine Learning Models in Predicting Early Mortality in Pediatric Sepsis Patients Model Accuracy Sensitivity Specificity PPV NPV Precision Recall F1 Score ROC AUC LogisticRegression 0.916 (0.890, 0.940) 0.448 (0.342, 0.582) 0.983 (0.971, 0.993) 0.788 (0.596, 0.930) 0.926 (0.902, 0.949) 0.788 (0.656, 0.896) 0.448 (0.342, 0.582) 0.571 (0.438, 0.688) 0.926 (0.880, 0.955) NaiveBayes 0.910 (0.885, 0.931) 0.621 (0.487, 0.764) 0.951 (0.930, 0.970) 0.643 (0.520, 0.793) 0.946 (0.928, 0.968) 0.643 (0.502, 0.750) 0.621 (0.487, 0.764) 0.632 (0.538, 0.743) 0.919 (0.882, 0.952) DecisionTree 0.929 (0.906, 0.948) 0.603 (0.477, 0.729) 0.975 (0.960, 0.988) 0.778 (0.670, 0.875) 0.945 (0.924, 0.964) 0.778 (0.665, 0.884) 0.603 (0.477, 0.729) 0.680 (0.584, 0.763) 0.881 (0.821, 0.921) RandomForest 0.918 (0.893, 0.940) 0.448 (0.318, 0.603) 0.985 (0.975, 0.994) 0.812 (0.678, 0.939) 0.926 (0.903, 0.943) 0.812 (0.672, 0.913) 0.448 (0.318, 0.603) 0.578 (0.435, 0.708) 0.931 (0.889, 0.972) SVM 0.875 (0.852, 0.907) 0.000 (0.000, 0.000) 1.000 (1.000, 1.000) 0.000 (0.000, 0.000) 0.875 (0.849, 0.904) 0.000 (0.000, 0.000) 0.000 (0.000, 0.000) 0.000 (0.000, 0.000) 0.923 (0.880, 0.959) XGBoost 0.929 (0.909, 0.948) 0.534 (0.424, 0.697) 0.985 (0.973, 0.995) 0.838 (0.703, 0.969) 0.937 (0.914, 0.962) 0.838 (0.727, 0.966) 0.534 (0.424, 0.697) 0.653 (0.522, 0.763) 0.936 (0.900, 0.967) LightGBM 0.925 (0.902, 0.945) 0.483 (0.336, 0.604) 0.988 (0.975, 0.995) 0.848 (0.719, 0.969) 0.931 (0.908, 0.951) 0.848 (0.730, 0.945) 0.483 (0.336, 0.604) 0.615 (0.496, 0.732) 0.938 (0.903, 0.968) Feature Importance in the XGBoost Model In the XGBoost model, the top six features were blood transfusion status, pediatric Sequential Organ Failure Assessment (pSOFA) score, post-cardiopulmonary resuscitation (POST-CPR), number of days of mechanical ventilation, secondary infection, and septic shock (Fig. 4 ). These features indicate key risk factors for early mortality in pediatric sepsis patients in the PICU. Discussion Sepsis, caused by infection, manifests as a systemic inflammatory response syndrome and often progresses to multiple organ dysfunction, making it one of the leading causes of death among children globally. Sepsis patients admitted to the pediatric intensive care unit (PICU) are critically ill with high mortality rates, imposing a heavy socioeconomic burden. Although early identification and intervention have been shown to significantly improve outcomes, existing clinical tools still exhibit notable limitations in precise risk stratification. Therefore, creating an effective prediction model to identify early mortality risk has become a key focus in clinical research. This study, based on the largest single-center retrospective cohort in Southwest China, successfully constructed and validated an early mortality risk prediction model for pediatric sepsis patients in the PICU using various machine learning algorithms. Among the 1,559 children included in this study, the overall mortality rate was 14.4%, which is notably significant. Early mortality (within 14 days of PICU admission) accounted for 12.7%. The median time to death was 3 days. Compared to the mortality rates reported by Tan et al. (14) for children with severe sepsis and septic shock globally (16.4%–22.7%), the data from this center show a lower mortality rate. This difference likely reflects a higher quality of care and regional variations in treatment standards, as well as the critical role of early intervention. These models exhibited excellent predictive performance, with the XGBoost and LightGBM algorithms demonstrating superior discriminative ability (AUC = 0.94). The final model identified six key predictive factors: blood transfusion, pSOFA score, post-cardiopulmonary resuscitation (post-CPR), days of mechanical ventilation, secondary infection, and septic shock. These factors reflect the severity of the disease, complications, and treatment responses from multiple perspectives, encompassing multidimensional information such as organ function, circulatory support, inflammatory burden, and iatrogenic factors, all supported by a solid pathophysiological and clinical foundation. Among these, the pSOFA score is a reliable tool for assessing organ dysfunction related to sepsis, with elevated scores indicating progressive deterioration of multiple organ functions and significantly correlating with mortality risk (15, 16). Studies have shown that pSOFA has better predictive value for the prognosis of pediatric sepsis than traditional scoring systems (17). Septic shock is the most severe manifestation of sepsis, accompanied by circulatory failure and metabolic disturbances, and is one of the core mechanisms leading to death (18); it is also a major cause of hospitalization, morbidity, and mortality among children globally (19). Prolonged mechanical ventilation duration is an independent risk factor for poor outcomes in extremely low birth weight infants (20). In pediatric sepsis, prolonged mechanical ventilation not only reflects primary respiratory failure but also indicates complications such as secondary infections and muscle atrophy, serving as an important marker of disease complexity and prognosis (21). Cardiopulmonary resuscitation (CPR) is a sign of acute deterioration, typically occurring after refractory shock or cardiac arrest; even with aggressive resuscitation, the prognosis for such patients remains poor (22, 23). Therefore, children with sepsis who were admitted after CPR have a significantly higher mortality rate. The need for blood transfusions often indicates active bleeding, coagulopathy, or severe anemia. However, transfusions may exacerbate the condition through immune modulation, leading to extended mechanical ventilation duration and longer ICU stays (24), and they are associated with increased risk of death (25). Secondary infections reflect the immunoparalysis state following sepsis, indicating host defense function failure. This often leads to prolonged illness and elevated mortality risk (26). The XGBoost and LightGBM models constructed in this study exhibited optimal predictive performance (AUC 0.94, specificity > 0.98), significantly outperforming traditional scoring systems. Their high specificity helps minimize false positives, avoiding unnecessary interventions and resource consumption, making them suitable for screening high-risk patients in the PICU. While sensitivity is slightly lower, these models can still serve as auxiliary decision-making tools. They enable earlier and more precise interventions when used alongside clinical assessments. The findings have profound implications for clinical practice and policymaking. By identifying key factors associated with high-risk mortality in pediatric sepsis patients, clinicians can implement interventions earlier, potentially improving patient outcomes and reducing mortality rates. Additionally, the clinical application value of this model lies in providing a basis for the allocation and optimization of medical resources within the PICU, allowing critical care resources to be used more effectively for the patients who need them most.As machine learning technology continues to develop in the medical field, the results of this study provide a foundation for similar applications in other disease areas, promoting the advancement of personalized medicine. This study has some limitations. First, the retrospective cohort design may lead to selection bias; it is necessary to verify whether these results hold up in larger, multicenter studies. Second, the model construction relies on clinical data from the PICU, which may be influenced by practice differences across various hospitals and regions. Additionally, although our model performed well in both the training and validation sets, it has not incorporated dynamic physiological parameters and continuous time-series physiological data, which may also limit the model's predictive performance. Future research could integrate specific real-time vital sign parameters, such as heart rate variability and respiratory rate trends, to further enhance predictive accuracy. Conclusion In summary, the machine learning-based prediction model developed in this study demonstrates promising performance in assessing early mortality risk in pediatric sepsis patients admitted to the PICU. This tool may assist clinicians in early identification of high-risk cases and inform resource allocation decisions. Future work should focus on external validation across diverse settings and investigation of the model's impact on clinical outcomes in real-world applications. This research contributes to the foundation for advancing personalized treatment strategies and improving the management of pediatric sepsis. Abbreviations PICU: Pediatric Intensive Care Unit LR: Logistic Regression DT: Decision Tree RF: Random Forest XGB: XGBoost LGBM: LightGBM NB: Naive Bayes SVM: Support Vector Machine DCA: Decision Curve Analysis LASSO: Least Absolute Shrinkage and Selection Operator RFE: Recursive Feature Elimination pSOFA: Pediatric Sequential Organ Failure Assessment ROC: Receiver Operating Characteristic AUC: Area Under the Curve CPR: Cardiopulmonary Resuscitation CCI: Chronic Critical Illness PSS: Phoenix Sepsis Score CT: Computed Tomography RAP: Rapid Recovery LOS: Length of Stay M[IQR]: Median [Interquartile Range] PCT: Procalcitonin WBC: White Blood Cell Count NSTI: Necrotizing Soft Tissue Infection CLABSI: Central Line-Associated Bloodstream Infection POST-CPR: Post-Cardiopulmonary Resuscitation Declarations Ethical approval and consent This study was approved by the Ethics Committee of the Chongqing Medical University Affiliated Children’s Hospital (Approval Number: 534/2022), and informed consent was obtained from the guardians. The study was conducted in accordance with the ethical principles of the Declaration of Helsinki. Because this study was retrospective, consent to participate declarations are not applicable. The trial was registered on the Chinese Clinical Trial Registry with the number ChiCTR2300068260. Consent for publication Not applicable. Availability of data and materials The datasets generated and analyzed during the current study are not publicly available due to the inclusion of protected health information but are available from the corresponding author on reasonable request. Competing interests The authors declare that they have no competing interests to disclose. Funding This work was financially supported by the 2022 Clinical Medical Research Youth Project of the National Clinical Research Center for Child Health and Diseases, China. Authors' contributions YL and LC were responsible for data collection and drafted the manuscript. ZN performed data curation and screening. YY analyzed and interpreted the data. PKB conducted the literature search and critical review. TLP was responsible for study design, quality control of data, and manuscript revision. All authors read and approved the final manuscript. Acknowledgements We would like to extend our sincere gratitude to all those who contributed to this work but do not meet the criteria for authorship, including individuals who provided professional writing assistance or materials. References Schlapbach LJ, Watson RS, Sorce LR, et al. International consensus criteria for pediatric sepsis and septic shock. JAMA (2024) 331:665–674. doi: 10.1001/jama.2024.0179 Rudd KE, Johnson SC, Agesa KM, et al. Global, regional, and national sepsis incidence and mortality, 1990-2017: analysis for the global burden of disease study. Lancet (Lond Engl) (2020) 395:200–211. doi: 10.1016/S0140-6736(19)32989-7 Weiss SL, Chair C-V, Peters MJ, et al. Surviving sepsis campaign international guidelines for the management of septic shock and sepsis-associated organ dysfunction in children. (2020) Schlapbach LJ, Watson RS, Sorce LR, et al. International Consensus Criteria for Pediatric Sepsis and Septic Shock. JAMA (2024) 331:665–674. doi: 10.1001/jama.2024.0179 Meyer NJ, Prescott HC. Sepsis and septic shock. N Engl J Med (2024) 391:2133–2146. doi: 10.1056/NEJMra2403213 Fenner BP, Darden DB, Kelly LS, et al. Immunological Endotyping of Chronic Critical Illness After Severe Sepsis. Front Med (Lausanne) (2020) 7:616694. doi: 10.3389/fmed.2020.616694 Hawkins RB, Raymond SL, Stortz JA, et al. Chronic Critical Illness and the Persistent Inflammation, Immunosuppression, and Catabolism Syndrome. Front Immunol (2018) 9:1511. doi: 10.3389/fimmu.2018.01511 Gentile LF, Cuenca AG, Efron PA, et al.Persistent inflammation and immunosuppression: a common syndrome and new horizon for surgical intensive care. J Trauma Acute Care Surg (2012) 72:1491–1501. doi: 10.1097/TA.0b013e318256e000 Yang L, Zang N, Liu C, et al. Clinical characteristics and risk factors of chronic critical illness in children with sepsis. Front Pediatr (2025) 13:1561044. doi: 10.3389/fped.2025.1561044 Cruz AT, Lane RD, Balamuth F, et al. Updates on pediatric sepsis. J Am Coll Emerg Physicians Open (2020) 1:981–993. doi: 10.1002/emp2.12173 Liu R, Yu Z, Xiao C, et al. Epidemiology and clinical characteristics of pediatric sepsis in PICUs in southwest China: a prospective multicenter study. Pediatr Crit Care Med: J Soc Crit Care Med World Fed Pediatr Intensive Crit Care Soc (2024) 25:425–433. doi: 10.1097/PCC.0000000000003450 Carrol ED, Ranjit S, Menon K, et al. Operationalizing appropriate sepsis definitions in children worldwide: considerations for the pediatric sepsis definition taskforce. Pediatr Crit Care Med: J Soc Crit Care Med World Fed Pediatr Intensive Crit Care Soc (2023) 24:e263–e271. doi: 10.1097/PCC.0000000000003263 Weiss SL, Peters MJ, Alhazzani W, et al. Surviving Sepsis Campaign International Guidelines for the Management of Septic Shock and Sepsis-Associated Organ Dysfunction in Children. Pediatr Crit Care Med (2020) 21:e52–e106. doi: 10.1097/PCC.0000000000002198 Global case-fatality rates in pediatric severe sepsis and septic shock: a systematic review and meta-analysis - PubMed. https://pubmed.ncbi.nlm.nih.gov/30742207/ [Accessed April 3, 2025] Qiu X, Lei Y-P, Zhou R-X. SIRS, SOFA, qSOFA, and NEWS in the diagnosis of sepsis and prediction of adverse outcomes: a systematic review and meta-analysis. Expert Rev Anti Infect Ther (2023) 21:891–900. doi: 10.1080/14787210.2023.2237192 Aggarwal B, Behera JR, Rup AR, et al. Comparison of the Pediatric Sequential Organ Failure Assessment (p SOFA) Score and Lactate Clearance as Predictors of Morbidity and Mortality in Pediatric Sepsis: A Prospective Observational Study. Cureus (2025) 17:e79172. doi: 10.7759/cureus.79172 Li S, Liu J, Chen F, et al. A risk score based on pediatric sequential organ failure assessment predicts 90-day mortality in children with Klebsiella pneumoniae bloodstream infection. BMC Infect Dis (2020) 20:916. doi: 10.1186/s12879-020-05644-w Weiss SL, Peters MJ, Alhazzani W, et al. Surviving sepsis campaign international guidelines for the management of septic shock and sepsis-associated organ dysfunction in children. Pediatr Crit Care Med: J Soc Crit Care Med World Fed Pediatr Intensive Crit Care Soc (2020) 21:e52–e106. doi: 10.1097/PCC.0000000000002198 Ranjit S, Kissoon N, Argent A, et al. Haemodynamic support for paediatric septic shock: a global perspective. Lancet Child Adolesc Health (2023) 7:588–598. doi: 10.1016/S2352-4642(23)00103-7 Menshykova AO, Dobryanskyy DO. Duration of mechanical ventilation and clinical outcomes in very low birth weight infants: a single center 10-years cohort study. J Neonatal-perinat Med (2023) 16:673–680. doi: 10.3233/NPM-230142 Vaschetto R, Pecere A, Perkins GD, et al. Effects of early extubation followed by noninvasive ventilation versus standard extubation on the duration of invasive mechanical ventilation in hypoxemic non-hypercapnic patients: a systematic review and individual patient data meta-analysis of randomized controlled trials. Crit Care (2021) 25:189. doi: 10.1186/s13054-021-03595-5 Topjian AA, de Caen A, Wainwright MS, et al. Pediatric Post-Cardiac Arrest Care: A Scientific Statement From the American Heart Association. Circulation (2019) 140:e194–e233. doi: 10.1161/CIR.0000000000000697 Hamzah M, Othman HF, Almasri M, et al. Survival outcomes of in-hospital cardiac arrest in pediatric patients in the USA. Eur J Pediatr (2021) 180:2513–2520. doi: 10.1007/s00431-021-04082-3 Yin M, Wang T, Jiang Q, et al.The association of red blood cell transfusion with mortality in pediatric patients with sepsis, severe sepsis, and septic shock: A single-center retrospective cohort study. Transfus Clin Biol (2025) 32:62–68. doi: 10.1016/j.tracli.2024.12.002 Muszynski JA, Banks R, Reeder RW, et al. Outcomes Associated With Early RBC Transfusion in Pediatric Severe Sepsis: A Propensity-Adjusted Multicenter Cohort Study. Shock (2022) 57:88–94. doi: 10.1097/SHK.0000000000001863 Stortz JA, Murphy TJ, Raymond SL, et al. Evidence for persistent immune suppression in patients who develop chronic critical illness after sepsis. Shock (Augusta Ga,) (2018) 49:249–258. doi: 10.1097/SHK.0000000000000981 Additional Declarations No competing interests reported. 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-7842349","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":538791273,"identity":"73c499d4-9356-46fa-b868-47584bd4fb90","order_by":0,"name":"Lin Yang","email":"","orcid":"","institution":"Children’s Hospital of Chongqing Medical University.","correspondingAuthor":false,"prefix":"","firstName":"Lin","middleName":"","lastName":"Yang","suffix":""},{"id":538791274,"identity":"fa15541b-5428-44be-bbcc-4ad0bd0fbcf6","order_by":1,"name":"Na Zang","email":"","orcid":"","institution":"Children's Hospital of Chongqing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Na","middleName":"","lastName":"Zang","suffix":""},{"id":538791275,"identity":"7e648e11-fc1c-4f69-ba46-c5f06cf5dcc9","order_by":2,"name":"Ying Yang","email":"","orcid":"","institution":"Children's Hospital of Chongqing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Ying","middleName":"","lastName":"Yang","suffix":""},{"id":538791276,"identity":"9e6d444f-be05-49d0-b1cd-5d2f6fb09e85","order_by":3,"name":"KaiBing Pu","email":"","orcid":"","institution":"Children’s Hospital of Chongqing Medical University.","correspondingAuthor":false,"prefix":"","firstName":"KaiBing","middleName":"","lastName":"Pu","suffix":""},{"id":538791277,"identity":"4745307c-1e34-431a-a853-b5b6d40f3ad9","order_by":4,"name":"Cong Liu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA60lEQVRIiWNgGAWjYBACPmYIzWN///zHB0BGAkEtbFAtcgw3GIwNoFoYG/BqgdLGQC1mEsRpYecx/PBxR21i4+yGtGqeCrs8Bunm4w/wO4zHWHLmmeOJzTIHjt3mOZNczCBzLBG/Lcw8Zsy8bccS2xgS227ztjEnNkjkGBKnpYchma2Y9189UEv+R2K01BhLSKSxMfM2HAbZQsD7zGzFkjPbDsgZ8Jxhlpxz7Hhim0Sa4Qx8Wvj5D2/88LGtjseAvYfxw5ua6sR+ieQHH/BpgYLDYJKJhwERU4RAHZhk/EGk8lEwCkbBKBhZAACnOUW7KJWiLgAAAABJRU5ErkJggg==","orcid":"","institution":"Children's Hospital of Chongqing Medical University","correspondingAuthor":true,"prefix":"","firstName":"Cong","middleName":"","lastName":"Liu","suffix":""},{"id":538791278,"identity":"bbde7842-c31e-498b-b754-92baea39d724","order_by":5,"name":"LiPing Tan","email":"","orcid":"","institution":"Children’s Hospital of Chongqing Medical University.","correspondingAuthor":false,"prefix":"","firstName":"LiPing","middleName":"","lastName":"Tan","suffix":""}],"badges":[],"createdAt":"2025-10-12 17:53:21","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7842349/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7842349/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":95010435,"identity":"aabde042-4853-4e75-9150-ac4dac13d9ab","added_by":"auto","created_at":"2025-11-03 10:17:43","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":416019,"visible":true,"origin":"","legend":"","description":"","filename":"BuildingaMachineLearningModeltoPredicttheEarlyMortalityRiskinPediatricICUSepsisPatients3.0.docx","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/a8b5e0b5f365864274acb857.docx"},{"id":95010440,"identity":"e175fa5d-595e-4c03-8e5f-4cdbd7938cde","added_by":"auto","created_at":"2025-11-03 10:17:43","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":396008,"visible":true,"origin":"","legend":"","description":"","filename":"Figure14.docx","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/1fb7e7dd6379e2b9854b742f.docx"},{"id":95010436,"identity":"4db0b614-7703-4d81-80d3-69fc238a6f69","added_by":"auto","created_at":"2025-11-03 10:17:43","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":21460,"visible":true,"origin":"","legend":"","description":"","filename":"Table12.docx","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/70c4208ee08bf64e7ddff78b.docx"},{"id":95221050,"identity":"bcddc532-55a8-4418-8c1c-73a899635906","added_by":"auto","created_at":"2025-11-05 16:18:07","extension":"json","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6475,"visible":true,"origin":"","legend":"","description":"","filename":"13e69f6e0f3e475d964934ac665e6cf1.json","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/393434c64c60491a24e88139.json"},{"id":95010438,"identity":"c3c33f72-6b93-4208-affc-716dfa5fca53","added_by":"auto","created_at":"2025-11-03 10:17:43","extension":"docx","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19763,"visible":true,"origin":"","legend":"","description":"","filename":"reference.docx","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/62db7c838e9e5f9bb2ca56f2.docx"},{"id":95221389,"identity":"658737c2-64d0-46d0-af8a-a935404c551e","added_by":"auto","created_at":"2025-11-05 16:18:54","extension":"xml","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":103333,"visible":true,"origin":"","legend":"","description":"","filename":"13e69f6e0f3e475d964934ac665e6cf11enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/58bcfd9ce489d0c9fb05d833.xml"},{"id":95221344,"identity":"960cec65-acd5-4c19-953d-4d7f4dd0445c","added_by":"auto","created_at":"2025-11-05 16:18:50","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":90977,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/aa438860a5b4b503035e19ba.png"},{"id":95221219,"identity":"9cde4c69-4d0a-40b6-964a-1e1969387ecd","added_by":"auto","created_at":"2025-11-05 16:18:37","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":124505,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/33bca1375e8a057be449f6b7.png"},{"id":95010443,"identity":"d92f0b94-6485-44e8-9c73-1316cca4154b","added_by":"auto","created_at":"2025-11-03 10:17:43","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":119894,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/cfc0acd819e7c8d361d3a62c.png"},{"id":95221081,"identity":"62d7f2d9-0a08-4694-8625-9bb852f3634a","added_by":"auto","created_at":"2025-11-05 16:18:12","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":41544,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/53a67cd884bcec684d0ec102.png"},{"id":95221390,"identity":"77ddad83-1b79-421b-9b95-815d03c225d1","added_by":"auto","created_at":"2025-11-05 16:18:55","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":90977,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/b628d8c57ef64d7bf3359340.png"},{"id":95221128,"identity":"e4f70239-1fd7-4e06-9507-6afca33d74e1","added_by":"auto","created_at":"2025-11-05 16:18:24","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":124505,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/0a309439ceaabd85c609f3a1.png"},{"id":95221257,"identity":"5c19e804-db19-4753-8e89-13fa7d45604c","added_by":"auto","created_at":"2025-11-05 16:18:43","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":119894,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/71d6c9ec301633219f3fb6a9.png"},{"id":95220861,"identity":"30197899-2495-4197-8e6d-b5a76d2e5bb7","added_by":"auto","created_at":"2025-11-05 16:16:13","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":41544,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/1f84267f867bc9acbe847ef1.png"},{"id":95221475,"identity":"4113b258-cc67-4cba-b56f-059d5fa00e55","added_by":"auto","created_at":"2025-11-05 16:19:06","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":29432,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/ac6933e0e445205c83e4ff24.png"},{"id":95221933,"identity":"02b45146-7b53-4a11-813a-8656b6ff6da3","added_by":"auto","created_at":"2025-11-05 16:19:55","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":36513,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/bacecc265f12c37e3e995652.png"},{"id":95221058,"identity":"501808dc-73b3-4f6d-8d52-071f31da5dbc","added_by":"auto","created_at":"2025-11-05 16:18:08","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":38702,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/dd5aefafa8a6143868e9ac20.png"},{"id":95010448,"identity":"352e976e-f02b-4886-bd39-1816862ff65c","added_by":"auto","created_at":"2025-11-03 10:17:43","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14280,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/67d72a911cf0bcadecc0a993.png"},{"id":95221318,"identity":"e0ee0447-d75a-426a-99cb-2890697b1a89","added_by":"auto","created_at":"2025-11-05 16:18:49","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":29432,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/524307e7c12f7438ed7f2c8e.png"},{"id":95010452,"identity":"b5957c1b-6e5c-43e0-9cec-c88d0654264b","added_by":"auto","created_at":"2025-11-03 10:17:43","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":36513,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/4001566a6af54f24c9606057.png"},{"id":95221550,"identity":"8a20ac84-6a14-4b4a-bd43-effd8c5da884","added_by":"auto","created_at":"2025-11-05 16:19:17","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":38702,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/a9b948586ebd42f01d91f197.png"},{"id":95220976,"identity":"b9536c2e-1347-47df-8459-7a80a9cfc836","added_by":"auto","created_at":"2025-11-05 16:17:29","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14280,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/5b60cb9abfcf63f0af81dc38.png"},{"id":95010458,"identity":"ace624ee-c3e2-4259-a2ce-8f3db044590b","added_by":"auto","created_at":"2025-11-03 10:17:43","extension":"xml","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":100403,"visible":true,"origin":"","legend":"","description":"","filename":"13e69f6e0f3e475d964934ac665e6cf11structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/825e087f460091b49c32d5ee.xml"},{"id":95010459,"identity":"cb9f6f33-386c-48b8-82bc-2a3ac5a28fa2","added_by":"auto","created_at":"2025-11-03 10:17:43","extension":"html","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":108089,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/34260fb09bfb08a2066fe6c7.html"},{"id":95220863,"identity":"14c818d9-01c1-4494-a693-277931c4dc73","added_by":"auto","created_at":"2025-11-05 16:16:14","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":129648,"visible":true,"origin":"","legend":"\u003cp\u003eROC curves of seven machine learning models for predicting early mortality in children with sepsis.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/fe28eec0362d0cf7b95d9254.png"},{"id":95221242,"identity":"62c479c9-ed89-4633-8d98-c597ccd07661","added_by":"auto","created_at":"2025-11-05 16:18:40","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":146213,"visible":true,"origin":"","legend":"\u003cp\u003eThe calibration curves of seven machine learning models for predicting the early mortality rate among pediatric sepsis patients.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/704d7101e62af5fcf905ea3b.png"},{"id":95221337,"identity":"41fdf785-ba86-4672-af88-f357478f466c","added_by":"auto","created_at":"2025-11-05 16:18:50","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":146392,"visible":true,"origin":"","legend":"\u003cp\u003eDecision Curve Analysis (DCA) of Seven Machine Learning Models for Early Mortality Among Pediatric-Sepsis Patients\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/9f71d1e66501ad8f55cc6303.png"},{"id":95010432,"identity":"7488c279-91aa-46bc-a32a-079560fd08df","added_by":"auto","created_at":"2025-11-03 10:17:42","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":57606,"visible":true,"origin":"","legend":"\u003cp\u003eImportant feature scores from the XGBoost model that predicts early mortality in pediatric sepsis patients.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/6cfcd9d169db781ab94f84f4.png"},{"id":100370371,"identity":"abcd4149-7d06-4303-96dd-511c472c43e8","added_by":"auto","created_at":"2026-01-16 08:05:34","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1368911,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7842349/v1/bc62d6d4-22ed-4d43-a5ef-cb53f5b1fa4e.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Building a Machine Learning Model to Predict the Early Mortality Risk in Pediatric ICU Sepsis Patients","fulltext":[{"header":"Background","content":"\u003cp\u003eSepsis is one of the leading causes of death among children worldwide, posing a persistent and significant challenge in the field of intensive care (1). According to data from the Global Burden of Disease study, sepsis causes approximately 3\u0026nbsp;million deaths annually among children under five years old, with low- and middle-income countries bearing over 90% of the disease burden (2). Despite advancements in sepsis diagnosis and treatment, early mortality rates remain high, especially in resource-limited areas, where about 20\u0026ndash;30% of pediatric sepsis patients in the PICU die within 72 hours of onset (3). Therefore, predicting the early mortality risk of sepsis patients is of significant clinical importance, as these predictions can help determine the patient's disease status, improve treatment outcomes, and reduce early mortality risk.\u003c/p\u003e\u003cp\u003eSepsis is an acute organ dysfunction syndrome caused by a dysregulated immune response (4). However, we lack precise definitions and diagnostic tests for this dysregulated host response. The heterogeneity of sepsis has long been considered a barrier to translating preclinical research into clinical applications and identifying targeted therapies (5). This heterogeneity leads to significant differentiation in clinical trajectories, specifically manifested as early mortality, rapid recovery, and chronic critical illness (CCI) phenotypes (6), with up to 33% of sepsis survivors not recovering quickly or completely, but rather developing a new syndrome of \"chronic critical illness\" (CCI) (7,8). This study, based on clinical data from the largest pediatric medical center in Southwest China, found that sepsis patients with underlying respiratory diseases or trauma, as well as those with high SOFA scores, surgical procedures, blood transfusions, or prolonged mechanical ventilation, are prone to developing CCI, which leads to prolonged hospital stays and increased rates of secondary infections caused by CCI (9). The pattern of early mortality sharply contrasts with that of the chronic critical illness (CCI), suggesting the potential existence of different pathophysiological mechanisms and risk factor profiles.\u003c/p\u003e\u003cp\u003eAlthough several studies have explored the treatment and prognostic factors of sepsis, existing research mostly focuses on traditional clinical indicators and empirical treatments. However, existing research lacks risk prediction models that systematically utilize modern technology (10). For example, some studies have identified clinical features associated with poor prognosis, such as higher pediatric sequential organ failure assessment scores and invasive mechanical ventilation; however, there is still insufficient systematic integration of these features for early mortality risk assessment (11). Additionally, existing prediction models often fail to adequately consider the heterogeneity of sepsis, leading to limited applicability across different patient populations (12).Therefore, developing a risk prediction model that can effectively identify high-risk pediatric sepsis patients and support clinical decision-making is particularly important.\u003c/p\u003e\u003cp\u003eThis retrospective cohort study uses clinical data from the PICU of Chongqing Medical University Affiliated Children's Hospital to explore risk factors for early mortality in pediatric sepsis patients. To achieve this, we employed machine learning techniques to aggregate extensive clinical data, identify key risk factors for early mortality, and develop a prediction model with clinical utility. During data analysis, various machine learning algorithms were applied to evaluate their performance in predicting early mortality risk, aiming to identify the most effective risk prediction model.\u003c/p\u003e"},{"header":"Subjects and Methods","content":"\u003cp\u003e\u003cstrong\u003eStudy population\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA total of 1735 pediatric patients with sepsis hospitalized in the Pediatric Intensive Care Unit (PICU) of the Chongqing Medical University Affiliated Children\u0026apos;s Hospital from January 2015 to December 2024 were included. After excluding patients with repeated admissions, missing clinical or demographic data, or severe cranial injuries based on predefined criteria, 1559 patients remained for analysis. Data on demographics, infection sources, laboratory indicators, treatment measures, and outcomes were collected through the electronic medical record system. \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eInclusion and Exclusion Criteria \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eInclusion Criteria: \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e(1) Patients hospitalized in the PICU; \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e(2) Aged 0-18 years; \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e(3) Clinically diagnosed with sepsis or septic shock (defined as SOFA score \u0026ge;2 or PSS score \u0026ge;2); \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e(4) It was the patient\u0026apos;s first episode of sepsis; \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e(5) Patients who received treatment according to the clinical management protocol standards for sepsis in this study. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eExclusion Criteria: \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e(1) Patients who were not admitted to the PICU or did not meet the sepsis diagnostic criteria; \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e(2) Patients with severe traumatic brain injury (evidence of neurological injury on CT scan and Glasgow Coma Scale score \u0026lt;8); \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e(3) Patients with incomplete case information. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eEthical approval and consent: This study was approved by the Ethics Committee of the Chongqing Medical University Affiliated Children\u0026rsquo;s Hospital (Approval Number: 534/2022) and informed consent was obtained from the guardians. The study was conducted in accordance with the ethical principles of the Declaration of Helsinki.\u0026nbsp;Clinical trial number:\u0026nbsp;ChiCTR2300068260.\u0026nbsp;Since this study was a retrospective study, Consent to Participate declarations: not applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Collection and Methods \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Collection \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eClinical data were collected through hospital electronic medical record systems and follow-up. Demographic information included gender, age, and weight; clinical features included sources of sepsis infection, severity of sepsis (whether septic shock was present), indicators of organ dysfunction, and the presence of underlying diseases. Laboratory tests mainly included white blood cell count, neutrophil count, lymphocyte count, hemoglobin, platelet count, procalcitonin, bilirubin, creatinine, albumin, and D-dimer. Treatment conditions included the use of vasoactive drugs, occurrence of cardiopulmonary resuscitation, duration of mechanical ventilation, surgical procedures, blood transfusions, blood purification, ICU length of stay, and total hospital stay. The occurrence and specific sites of secondary infections during hospitalization were also recorded. \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDiagnostic Criteria\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe diagnosis of sepsis was based on the pediatric Sequential Organ Failure Assessment (pSOFA) score to assess the extent of involvement of multiple systems, including the respiratory system, coagulation function, liver, and kidney function. If pSOFA \u0026ge;2, the diagnosis of sepsis is established (13). Severe infection leading to cardiovascular dysfunction\u0026mdash;including hypotension requiring vasoactive drug treatment or abnormal perfusion\u0026mdash;is classified as septic shock. In conjunction with the \u0026quot;2024 International Consensus Standards: Pediatric Sepsis and Septic Shock,\u0026quot; the Phoenix Sepsis Score (PSS) was used for assessment, which includes relevant indicators from the cardiovascular, respiratory, neurological, and coagulation systems. If the PSS score \u0026ge;2, the child with suspected or confirmed infection is diagnosed with sepsis. Septic shock is defined as sepsis with a PSS cardiovascular score of at least 1 (4). \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGroup\u003c/strong\u003e \u003cstrong\u003eclassification\u003c/strong\u003e \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBased on sepsis definitions, patients diagnosed with sepsis or septic shock in our PICU were divided into two groups to facilitate outcome analysis. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eEarly Mortality Group: Patients who died within 14 days of PICU admission (198 cases). \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eNon-Early Mortality Group: Patients not meeting the criteria for the early mortality group were classified as the non-early mortality group (1361 cases), including the rapid recovery group (RAP, 1058 cases) and the chronic critical illness (CCI) group (303 cases). The CCI group is defined by an ICU length of stay (LOS) \u0026ge;14 days and evidence of persistent organ dysfunction, assessed using the Sequential Organ Failure Assessment (SOFA) score at 14 days (i.e., cardiovascular SOFA score \u0026ge;1 or any other organ system score \u0026ge;2).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical Methods \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eStatistical analyses were performed using SPSS 26.0, R 4.4.1, and Python 3.12 software. Normally distributed continuous data were expressed as mean \u0026plusmn; standard deviation and analyzed using t-tests; non-normally distributed continuous data were expressed as median with interquartile ranges M[IQR] and analyzed using Mann-Whitney U tests; categorical data were expressed as counts and percentages, and analyzed using \u0026chi;2 tests or Fisher\u0026apos;s exact tests. A P-value \u0026lt; 0.05 was considered statistically significant. The dataset was randomly divided into training and validation sets in a 7:3 ratio. Feature variable selection was performed using LASSO regression, recursive feature elimination (RFE), and the Boruta algorithm, with variables selected by all three algorithms used as predictive factors to construct the prediction model. Seven machine learning models were employed: logistic regression (LR), decision tree (DT), random forest (RF), XGBoost (XGB), LightGBM (LGBM), naive Bayes (NB), and support vector machine (SVM). The predictive performance of the models was evaluated using sensitivity, specificity, receiver operating characteristic curve (ROC curve), area under the curve (AUC), and calibration curve. The clinical utility of the models was assessed through a clinical decision curve analysis. \u0026nbsp;\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003ePatient Demographics and Sepsis Characteristics\u003c/h2\u003e\u003cp\u003eAfter screening, a total of 1,559 pediatric sepsis patients were included in this study. Among them, 198 cases experienced early mortality, resulting in an early mortality rate of 12.7%. There were no significant differences between the early mortality group and the survivor group in terms of gender, age, and weight (p\u0026thinsp;\u0026gt;\u0026thinsp;0.05). However, the early mortality group had significantly higher pediatric Sequential Organ Failure Assessment (pSOFA) scores and proportions of septic shock, indicating that the degree of organ dysfunction is closely related to disease outcomes and early mortality. The intensive care unit (ICU) length of stay (median 3 days) and total hospital stay (median 4 days) in the early mortality group were significantly shorter than those in the survivor group (ICU length of stay median 7 days, total hospital stay median 19.5 days), and the rate of secondary infections was significantly lower in the early mortality group. The main infection source in the early mortality group was respiratory infection, followed by bloodstream infection; notably, the proportion of bloodstream infections was significantly higher than in the survivor group (19.7% versus 7.0%), suggesting a close association between bloodstream infections and early mortality. The proportion of patients with underlying diseases in the early mortality group was significantly higher than that in the survivor group, particularly those with underlying immune system diseases, mainly immunodeficiencies, and hematological diseases such as leukemia, lymphoma, and aplastic anemia. Additionally, the proportion of patients who underwent cardiopulmonary resuscitation prior to hospitalization was significantly increased in the early mortality group.The early mortality group had significantly lower hemoglobin, absolute lymphocyte count, platelet count, and albumin levels compared to the non-early mortality group, while PCT, D-dimer, and creatinine levels were significantly higher (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eComparison of Patient Demographics and Characteristics of the Two Groups\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSepsis Characteristics\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNon EM(1359)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eEarly Mortality(198)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eP-value\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMale, n (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e808 (59.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e121 ( 61.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.714\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge (median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.75 [0.42, 6.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1.88 [0.50, 6.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.269\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWeight (median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e11.00 [6.94, 20.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e11.00 [7.00, 21.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.501\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePSOFA Score (median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e5.00 [3.00, 8.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e8.00 [6.00, 10.75]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSeptic Shock (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e603 (44.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e155 ( 78.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSecondary Infection(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e521 (38.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e27 ( 13.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eICU LOS (median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e7.00 [4.00, 12.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e3.00 [1.00, 5.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHospital LOS (median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e19.500 [13.00, 33.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4.00 [2.00, 11.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eSource of sepsis infection\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIntestinal Infection (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e246 (18.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e31 ( 15.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.459\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePneumonia (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e918 (67.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e121 ( 61.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.086\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNSTI (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e83 ( 6.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e9 ( 4.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.478\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBloodstreamInfection (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e95 ( 7.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e39 ( 19.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSurgical Site Infection (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e5 ( 0.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0 ( 0.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.855\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eUrinary TractInfection (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e14 ( 1.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0 ( 0.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.302\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCLABSI (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e13 ( 1.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0 ( 0.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.335\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOther Infection (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e11 ( 0.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0 ( 0.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.414\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eUnderlying Diseases\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCardiovascular Disease (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e184 (13.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e21 ( 10.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.304\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eImmune Disease (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e71 ( 5.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e21 ( 10.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.005\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHematologic Disease (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e123 ( 9.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e43 ( 21.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOther Neoplastic Disease (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e50 ( 3.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2 ( 1.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.082\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eInherited Metabolic Disease (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e195 (14.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e36 ( 18.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.19\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRespiratory Disease (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e27 ( 2.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1 ( 0.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.238\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eConnective Tissue Disease (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e26 ( 1.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e7 ( 3.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.224\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNeurological Disease (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e106 ( 7.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e13 ( 6.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.64\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLiver Failure (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e12 ( 0.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2 ( 1.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eKidney Failure (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e16 ( 1.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e3 ( 1.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.954\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePost Transplant (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e30 ( 2.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2 ( 1.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.4\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePost CPR (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e30 ( 2.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e11 ( 5.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.012\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTrauma (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e22 ( 1.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4 ( 2.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.908\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDrowning (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e6 ( 0.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1 ( 0.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eLaboratory Indicators\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWBC (median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e10.34 [5.81, 16.88]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e10.13 [3.93, 16.75]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.193\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHemoglobin (median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e101.00 [88.00, 114.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e96.00 [82.00, 110.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.002\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLymphocytes (median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e2.33 [1.21, 4.44]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1.89 [0.96, 3.62]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.003\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNeutrophils (median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e6.49 [3.06, 12.11]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e5.94 [1.89, 10.81]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.064\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePlatelets (median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e267.00 [132.00, 419.50]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e106.00 [36.00, 291.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePCT (median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e3.94 [0.69, 14.72]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e6.83 [0.70, 33.50]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.029\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBilirubin (median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e6.80 [3.20, 14.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e7.95 [3.05, 26.30]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.109\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCreatinine (median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e32.00 [22.65, 51.85]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e42.00 [24.08, 92.03]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAlbumin (median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e32.30 [26.60, 38.70]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e28.55 [23.63, 36.50]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDD (median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e2.76 [1.08, 7.80]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4.27 [1.77, 11.64]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eTreatment measures\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMechanicalVentilation (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e968 (71.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e194 ( 98.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSurgery (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e357 (26.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e23 ( 11.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBloodTransfusion (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e427 (31.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e162 ( 81.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBloodPurification (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e162 (11.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e62 ( 31.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eECMO (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e8 ( 0.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e7 ( 3.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eROC Curve Analysis of the Predictive Performance of Seven Models for In-Hospital Mortality in Sepsis Patients\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThis study predicted early mortality in pediatric sepsis patients using seven machine learning models: logistic regression (LR), decision tree (DT), random forest (RF), XGBoost (XGB), LightGBM (LGBM), Naive Bayes (NB), and support vector machine (SVM). The ROC curve analysis results showed that the XGBoost and LGBM models had the best predictive performance for early mortality, each achieving an area under the curve (AUC) of 0.94. The LR and RF models followed with AUCs of 0.93, the NB and SVM models both had AUCs of 0.92, and the DT model had the lowest AUC of 0.88(Figure \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). These findings indicate that the XGBoost and LGBM models have superior predictive accuracy for early mortality in pediatric sepsis patients, while the DT model\u0026rsquo;s discriminative ability is lower than that of the other models.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eDetailed Performance Indicators of the Seven Models\u003c/h2\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eshows that among the seven models, the XGBoost model had the highest precision, recall, accuracy, and F1 score, demonstrating the best discriminative ability. Calibration curves (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e) indicated that the XGBoost model performed best. Furthermore, decision curve analysis (DCA) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) demonstrated that the RF model exhibited greater net benefits over a wider threshold probability range compared to the other models. Taken together, these results suggest that considering all indicators, the XGBoost model is the best model with strong clinical applicability.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"10\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSensitivity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSpecificity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003ePPV\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eNPV\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003eRecall\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c9\"\u003e\u003cp\u003eF1 Score\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c10\"\u003e\u003cp\u003eROC AUC\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLogisticRegression\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.916 (0.890, 0.940)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.448 (0.342, 0.582)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.983 (0.971, 0.993)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.788 (0.596, 0.930)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.926 (0.902, 0.949)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.788 (0.656, 0.896)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.448 (0.342, 0.582)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.571 (0.438, 0.688)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.926 (0.880, 0.955)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNaiveBayes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.910 (0.885, 0.931)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.621 (0.487, 0.764)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.951 (0.930, 0.970)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.643 (0.520, 0.793)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.946 (0.928, 0.968)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.643 (0.502, 0.750)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.621 (0.487, 0.764)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.632 (0.538, 0.743)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.919 (0.882, 0.952)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDecisionTree\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.929 (0.906, 0.948)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.603 (0.477, 0.729)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.975 (0.960, 0.988)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.778 (0.670, 0.875)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.945 (0.924, 0.964)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.778 (0.665, 0.884)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.603 (0.477, 0.729)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.680 (0.584, 0.763)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.881 (0.821, 0.921)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRandomForest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.918 (0.893, 0.940)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.448 (0.318, 0.603)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.985 (0.975, 0.994)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.812 (0.678, 0.939)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.926 (0.903, 0.943)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.812 (0.672, 0.913)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.448 (0.318, 0.603)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.578 (0.435, 0.708)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.931 (0.889, 0.972)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSVM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.875 (0.852, 0.907)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.000 (0.000, 0.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.000 (1.000, 1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.000 (0.000, 0.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.875 (0.849, 0.904)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.000 (0.000, 0.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.000 (0.000, 0.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.000 (0.000, 0.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.923 (0.880, 0.959)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eXGBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.929 (0.909, 0.948)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.534 (0.424, 0.697)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.985 (0.973, 0.995)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.838 (0.703, 0.969)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.937 (0.914, 0.962)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.838 (0.727, 0.966)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.534 (0.424, 0.697)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.653 (0.522, 0.763)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.936 (0.900, 0.967)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLightGBM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.925 (0.902, 0.945)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.483 (0.336, 0.604)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.988 (0.975, 0.995)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.848 (0.719, 0.969)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.931 (0.908, 0.951)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.848 (0.730, 0.945)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.483 (0.336, 0.604)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.615 (0.496, 0.732)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.938 (0.903, 0.968)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e2\u003c/span\u003e. Performance of Seven Machine Learning Models in Predicting Early Mortality in Pediatric Sepsis Patients\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003ePerformance of Seven Machine Learning Models in Predicting Early Mortality in Pediatric Sepsis Patients\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"10\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSensitivity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSpecificity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003ePPV\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eNPV\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003eRecall\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c9\"\u003e\u003cp\u003eF1 Score\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c10\"\u003e\u003cp\u003eROC AUC\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLogisticRegression\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.916 (0.890, 0.940)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.448 (0.342, 0.582)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.983 (0.971, 0.993)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.788 (0.596, 0.930)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.926 (0.902, 0.949)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.788 (0.656, 0.896)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.448 (0.342, 0.582)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.571 (0.438, 0.688)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.926 (0.880, 0.955)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNaiveBayes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.910 (0.885, 0.931)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.621 (0.487, 0.764)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.951 (0.930, 0.970)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.643 (0.520, 0.793)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.946 (0.928, 0.968)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.643 (0.502, 0.750)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.621 (0.487, 0.764)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.632 (0.538, 0.743)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.919 (0.882, 0.952)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDecisionTree\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.929 (0.906, 0.948)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.603 (0.477, 0.729)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.975 (0.960, 0.988)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.778 (0.670, 0.875)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.945 (0.924, 0.964)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.778 (0.665, 0.884)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.603 (0.477, 0.729)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.680 (0.584, 0.763)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.881 (0.821, 0.921)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRandomForest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.918 (0.893, 0.940)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.448 (0.318, 0.603)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.985 (0.975, 0.994)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.812 (0.678, 0.939)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.926 (0.903, 0.943)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.812 (0.672, 0.913)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.448 (0.318, 0.603)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.578 (0.435, 0.708)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.931 (0.889, 0.972)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSVM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.875 (0.852, 0.907)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.000 (0.000, 0.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.000 (1.000, 1.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.000 (0.000, 0.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.875 (0.849, 0.904)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.000 (0.000, 0.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.000 (0.000, 0.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.000 (0.000, 0.000)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.923 (0.880, 0.959)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eXGBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.929 (0.909, 0.948)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.534 (0.424, 0.697)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.985 (0.973, 0.995)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.838 (0.703, 0.969)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.937 (0.914, 0.962)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.838 (0.727, 0.966)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.534 (0.424, 0.697)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.653 (0.522, 0.763)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.936 (0.900, 0.967)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLightGBM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.925 (0.902, 0.945)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.483 (0.336, 0.604)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.988 (0.975, 0.995)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.848 (0.719, 0.969)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.931 (0.908, 0.951)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.848 (0.730, 0.945)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e0.483 (0.336, 0.604)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e0.615 (0.496, 0.732)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e\u003cp\u003e0.938 (0.903, 0.968)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003eFeature Importance in the XGBoost Model\u003c/h2\u003e\u003cp\u003eIn the XGBoost model, the top six features were blood transfusion status, pediatric Sequential Organ Failure Assessment (pSOFA) score, post-cardiopulmonary resuscitation (POST-CPR), number of days of mechanical ventilation, secondary infection, and septic shock (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). These features indicate key risk factors for early mortality in pediatric sepsis patients in the PICU.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eSepsis, caused by infection, manifests as a systemic inflammatory response syndrome and often progresses to multiple organ dysfunction, making it one of the leading causes of death among children globally. Sepsis patients admitted to the pediatric intensive care unit (PICU) are critically ill with high mortality rates, imposing a heavy socioeconomic burden. Although early identification and intervention have been shown to significantly improve outcomes, existing clinical tools still exhibit notable limitations in precise risk stratification. Therefore, creating an effective prediction model to identify early mortality risk has become a key focus in clinical research.\u003c/p\u003e\u003cp\u003eThis study, based on the largest single-center retrospective cohort in Southwest China, successfully constructed and validated an early mortality risk prediction model for pediatric sepsis patients in the PICU using various machine learning algorithms. Among the 1,559 children included in this study, the overall mortality rate was 14.4%, which is notably significant. Early mortality (within 14 days of PICU admission) accounted for 12.7%. The median time to death was 3 days. Compared to the mortality rates reported by Tan et al. (14) for children with severe sepsis and septic shock globally (16.4%\u0026ndash;22.7%), the data from this center show a lower mortality rate. This difference likely reflects a higher quality of care and regional variations in treatment standards, as well as the critical role of early intervention. These models exhibited excellent predictive performance, with the XGBoost and LightGBM algorithms demonstrating superior discriminative ability (AUC\u0026thinsp;=\u0026thinsp;0.94). The final model identified six key predictive factors: blood transfusion, pSOFA score, post-cardiopulmonary resuscitation (post-CPR), days of mechanical ventilation, secondary infection, and septic shock. These factors reflect the severity of the disease, complications, and treatment responses from multiple perspectives, encompassing multidimensional information such as organ function, circulatory support, inflammatory burden, and iatrogenic factors, all supported by a solid pathophysiological and clinical foundation.\u003c/p\u003e\u003cp\u003eAmong these, the pSOFA score is a reliable tool for assessing organ dysfunction related to sepsis, with elevated scores indicating progressive deterioration of multiple organ functions and significantly correlating with mortality risk (15, 16). Studies have shown that pSOFA has better predictive value for the prognosis of pediatric sepsis than traditional scoring systems (17). Septic shock is the most severe manifestation of sepsis, accompanied by circulatory failure and metabolic disturbances, and is one of the core mechanisms leading to death (18); it is also a major cause of hospitalization, morbidity, and mortality among children globally (19). Prolonged mechanical ventilation duration is an independent risk factor for poor outcomes in extremely low birth weight infants (20). In pediatric sepsis, prolonged mechanical ventilation not only reflects primary respiratory failure but also indicates complications such as secondary infections and muscle atrophy, serving as an important marker of disease complexity and prognosis (21). Cardiopulmonary resuscitation (CPR) is a sign of acute deterioration, typically occurring after refractory shock or cardiac arrest; even with aggressive resuscitation, the prognosis for such patients remains poor (22, 23). Therefore, children with sepsis who were admitted after CPR have a significantly higher mortality rate. The need for blood transfusions often indicates active bleeding, coagulopathy, or severe anemia. However, transfusions may exacerbate the condition through immune modulation, leading to extended mechanical ventilation duration and longer ICU stays (24), and they are associated with increased risk of death (25). Secondary infections reflect the immunoparalysis state following sepsis, indicating host defense function failure. This often leads to prolonged illness and elevated mortality risk (26).\u003c/p\u003e\u003cp\u003eThe XGBoost and LightGBM models constructed in this study exhibited optimal predictive performance (AUC 0.94, specificity\u0026thinsp;\u0026gt;\u0026thinsp;0.98), significantly outperforming traditional scoring systems. Their high specificity helps minimize false positives, avoiding unnecessary interventions and resource consumption, making them suitable for screening high-risk patients in the PICU. While sensitivity is slightly lower, these models can still serve as auxiliary decision-making tools. They enable earlier and more precise interventions when used alongside clinical assessments. The findings have profound implications for clinical practice and policymaking. By identifying key factors associated with high-risk mortality in pediatric sepsis patients, clinicians can implement interventions earlier, potentially improving patient outcomes and reducing mortality rates. Additionally, the clinical application value of this model lies in providing a basis for the allocation and optimization of medical resources within the PICU, allowing critical care resources to be used more effectively for the patients who need them most.As machine learning technology continues to develop in the medical field, the results of this study provide a foundation for similar applications in other disease areas, promoting the advancement of personalized medicine.\u003c/p\u003e\u003cp\u003eThis study has some limitations. First, the retrospective cohort design may lead to selection bias; it is necessary to verify whether these results hold up in larger, multicenter studies. Second, the model construction relies on clinical data from the PICU, which may be influenced by practice differences across various hospitals and regions. Additionally, although our model performed well in both the training and validation sets, it has not incorporated dynamic physiological parameters and continuous time-series physiological data, which may also limit the model's predictive performance. Future research could integrate specific real-time vital sign parameters, such as heart rate variability and respiratory rate trends, to further enhance predictive accuracy.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn summary, the machine learning-based prediction model developed in this study demonstrates promising performance in assessing early mortality risk in pediatric sepsis patients admitted to the PICU. This tool may assist clinicians in early identification of high-risk cases and inform resource allocation decisions. Future work should focus on external validation across diverse settings and investigation of the model's impact on clinical outcomes in real-world applications. This research contributes to the foundation for advancing personalized treatment strategies and improving the management of pediatric sepsis.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003ePICU: Pediatric Intensive Care Unit\u003c/p\u003e\n\u003cp\u003eLR: Logistic Regression\u003c/p\u003e\n\u003cp\u003eDT: Decision Tree\u003c/p\u003e\n\u003cp\u003eRF: Random Forest\u003c/p\u003e\n\u003cp\u003eXGB: XGBoost\u003c/p\u003e\n\u003cp\u003eLGBM: LightGBM\u003c/p\u003e\n\u003cp\u003eNB: Naive Bayes\u003c/p\u003e\n\u003cp\u003eSVM: Support Vector Machine\u003c/p\u003e\n\u003cp\u003eDCA: Decision Curve Analysis\u003c/p\u003e\n\u003cp\u003eLASSO: Least Absolute Shrinkage and Selection Operator\u003c/p\u003e\n\u003cp\u003eRFE: Recursive Feature Elimination\u003c/p\u003e\n\u003cp\u003epSOFA: Pediatric Sequential Organ Failure Assessment\u003c/p\u003e\n\u003cp\u003eROC: Receiver Operating Characteristic\u003c/p\u003e\n\u003cp\u003eAUC: Area Under the Curve\u003c/p\u003e\n\u003cp\u003eCPR: Cardiopulmonary Resuscitation\u003c/p\u003e\n\u003cp\u003eCCI: Chronic Critical Illness\u003c/p\u003e\n\u003cp\u003ePSS: Phoenix Sepsis Score\u003c/p\u003e\n\u003cp\u003eCT: Computed Tomography\u003c/p\u003e\n\u003cp\u003eRAP: Rapid Recovery\u003c/p\u003e\n\u003cp\u003eLOS: Length of Stay\u003c/p\u003e\n\u003cp\u003eM[IQR]: Median [Interquartile Range]\u003c/p\u003e\n\u003cp\u003ePCT: Procalcitonin\u003c/p\u003e\n\u003cp\u003eWBC: White Blood Cell Count\u003c/p\u003e\n\u003cp\u003eNSTI: Necrotizing Soft Tissue Infection\u003c/p\u003e\n\u003cp\u003eCLABSI: Central Line-Associated Bloodstream Infection\u003c/p\u003e\n\u003cp\u003ePOST-CPR: Post-Cardiopulmonary Resuscitation\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthical approval and consent \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was approved by the Ethics Committee of the Chongqing Medical University Affiliated Children\u0026rsquo;s Hospital (Approval Number: 534/2022), and informed consent was obtained from the guardians. The study was conducted in accordance with the ethical principles of the Declaration of Helsinki. Because this study was retrospective, consent to participate declarations are not applicable. The trial was registered on the Chinese Clinical Trial Registry with the number ChiCTR2300068260.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated and analyzed during the current study are not publicly available due to the inclusion of protected health information but are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests to disclose.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was financially supported by the 2022 Clinical Medical Research Youth Project of the National Clinical Research Center for Child Health and Diseases, China.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026apos; contributions \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eYL and LC were responsible for data collection and drafted the manuscript. ZN performed data curation and screening. YY analyzed and interpreted the data. PKB conducted the literature search and critical review. TLP was responsible for study design, quality control of data, and manuscript revision. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe would like to extend our sincere gratitude to all those who contributed to this work but do not meet the criteria for authorship, including individuals who provided professional writing assistance or materials.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSchlapbach LJ, Watson RS, Sorce LR, et al. International consensus criteria for pediatric sepsis and septic shock. \u003cem\u003eJAMA\u003c/em\u003e (2024) 331:665\u0026ndash;674. doi: 10.1001/jama.2024.0179\u003c/li\u003e\n\u003cli\u003eRudd KE, Johnson SC, Agesa KM, et al. Global, regional, and national sepsis incidence and mortality, 1990-2017: analysis for the global burden of disease study. \u003cem\u003eLancet (Lond Engl)\u003c/em\u003e (2020) 395:200\u0026ndash;211. doi: 10.1016/S0140-6736(19)32989-7\u003c/li\u003e\n\u003cli\u003eWeiss SL, Chair C-V, Peters MJ, et al. Surviving sepsis campaign international guidelines for the management of septic shock and sepsis-associated organ dysfunction in children. (2020)\u003c/li\u003e\n\u003cli\u003eSchlapbach LJ, Watson RS, Sorce LR, et al. International Consensus Criteria for Pediatric Sepsis and Septic Shock. \u003cem\u003eJAMA\u003c/em\u003e (2024) 331:665\u0026ndash;674. doi: 10.1001/jama.2024.0179\u003c/li\u003e\n\u003cli\u003eMeyer NJ, Prescott HC. Sepsis and septic shock. \u003cem\u003eN Engl J Med\u003c/em\u003e (2024) 391:2133\u0026ndash;2146. doi: 10.1056/NEJMra2403213\u003c/li\u003e\n\u003cli\u003eFenner BP, Darden DB, Kelly LS, et al. Immunological Endotyping of Chronic Critical Illness After Severe Sepsis. \u003cem\u003eFront Med (Lausanne)\u003c/em\u003e (2020) 7:616694. doi: 10.3389/fmed.2020.616694\u003c/li\u003e\n\u003cli\u003eHawkins RB, Raymond SL, Stortz JA, et al. Chronic Critical Illness and the Persistent Inflammation, Immunosuppression, and Catabolism Syndrome. \u003cem\u003eFront Immunol\u003c/em\u003e (2018) 9:1511. doi: 10.3389/fimmu.2018.01511\u003c/li\u003e\n\u003cli\u003eGentile LF, Cuenca AG, Efron PA, et al.Persistent inflammation and immunosuppression: a common syndrome and new horizon for surgical intensive care. \u003cem\u003eJ Trauma Acute Care Surg\u003c/em\u003e (2012) 72:1491\u0026ndash;1501. doi: 10.1097/TA.0b013e318256e000\u003c/li\u003e\n\u003cli\u003eYang L, Zang N, Liu C, et al. Clinical characteristics and risk factors of chronic critical illness in children with sepsis. \u003cem\u003eFront Pediatr\u003c/em\u003e (2025) 13:1561044. doi: 10.3389/fped.2025.1561044\u003c/li\u003e\n\u003cli\u003eCruz AT, Lane RD, Balamuth F, et al. Updates on pediatric sepsis. \u003cem\u003eJ Am Coll Emerg Physicians Open\u003c/em\u003e (2020) 1:981\u0026ndash;993. doi: 10.1002/emp2.12173\u003c/li\u003e\n\u003cli\u003eLiu R, Yu Z, Xiao C, et al. Epidemiology and clinical characteristics of pediatric sepsis in PICUs in southwest China: a prospective multicenter study. \u003cem\u003ePediatr Crit Care Med: J Soc Crit Care Med World Fed Pediatr Intensive Crit Care Soc\u003c/em\u003e (2024) 25:425\u0026ndash;433. doi: 10.1097/PCC.0000000000003450\u003c/li\u003e\n\u003cli\u003eCarrol ED, Ranjit S, Menon K, et al. Operationalizing appropriate sepsis definitions in children worldwide: considerations for the pediatric sepsis definition taskforce. \u003cem\u003ePediatr Crit Care Med: J Soc Crit Care Med World Fed Pediatr Intensive Crit Care Soc\u003c/em\u003e (2023) 24:e263\u0026ndash;e271. doi: 10.1097/PCC.0000000000003263\u003c/li\u003e\n\u003cli\u003eWeiss SL, Peters MJ, Alhazzani W, et al. Surviving Sepsis Campaign International Guidelines for the Management of Septic Shock and Sepsis-Associated Organ Dysfunction in Children. \u003cem\u003ePediatr Crit Care Med\u003c/em\u003e (2020) 21:e52\u0026ndash;e106. doi: 10.1097/PCC.0000000000002198\u003c/li\u003e\n\u003cli\u003eGlobal case-fatality rates in pediatric severe sepsis and septic shock: a systematic review and meta-analysis - PubMed. https://pubmed.ncbi.nlm.nih.gov/30742207/ [Accessed April 3, 2025]\u003c/li\u003e\n\u003cli\u003eQiu X, Lei Y-P, Zhou R-X. SIRS, SOFA, qSOFA, and NEWS in the diagnosis of sepsis and prediction of adverse outcomes: a systematic review and meta-analysis. \u003cem\u003eExpert Rev Anti Infect Ther\u003c/em\u003e (2023) 21:891\u0026ndash;900. doi: 10.1080/14787210.2023.2237192\u003c/li\u003e\n\u003cli\u003eAggarwal B, Behera JR, Rup AR, et al. Comparison of the Pediatric Sequential Organ Failure Assessment (p SOFA) Score and Lactate Clearance as Predictors of Morbidity and Mortality in Pediatric Sepsis: A Prospective Observational Study. \u003cem\u003eCureus\u003c/em\u003e (2025) 17:e79172. doi: 10.7759/cureus.79172\u003c/li\u003e\n\u003cli\u003eLi S, Liu J, Chen F, et al. A risk score based on pediatric sequential organ failure assessment predicts 90-day mortality in children with Klebsiella pneumoniae bloodstream infection. \u003cem\u003eBMC Infect Dis\u003c/em\u003e (2020) 20:916. doi: 10.1186/s12879-020-05644-w\u003c/li\u003e\n\u003cli\u003eWeiss SL, Peters MJ, Alhazzani W, et al. Surviving sepsis campaign international guidelines for the management of septic shock and sepsis-associated organ dysfunction in children. \u003cem\u003ePediatr Crit Care Med: J Soc Crit Care Med World Fed Pediatr Intensive Crit Care Soc\u003c/em\u003e (2020) 21:e52\u0026ndash;e106. doi: 10.1097/PCC.0000000000002198\u003c/li\u003e\n\u003cli\u003eRanjit S, Kissoon N, Argent A, et al. Haemodynamic support for paediatric septic shock: a global perspective. \u003cem\u003eLancet Child Adolesc Health\u003c/em\u003e (2023) 7:588\u0026ndash;598. doi: 10.1016/S2352-4642(23)00103-7\u003c/li\u003e\n\u003cli\u003eMenshykova AO, Dobryanskyy DO. Duration of mechanical ventilation and clinical outcomes in very low birth weight infants: a single center 10-years cohort study. \u003cem\u003eJ Neonatal-perinat Med\u003c/em\u003e (2023) 16:673\u0026ndash;680. doi: 10.3233/NPM-230142\u003c/li\u003e\n\u003cli\u003eVaschetto R, Pecere A, Perkins GD, et al. Effects of early extubation followed by noninvasive ventilation versus standard extubation on the duration of invasive mechanical ventilation in hypoxemic non-hypercapnic patients: a systematic review and individual patient data meta-analysis of randomized controlled trials. \u003cem\u003eCrit Care\u003c/em\u003e (2021) 25:189. doi: 10.1186/s13054-021-03595-5\u003c/li\u003e\n\u003cli\u003eTopjian AA, de Caen A, Wainwright MS, et al. Pediatric Post-Cardiac Arrest Care: A Scientific Statement From the American Heart Association. \u003cem\u003eCirculation\u003c/em\u003e (2019) 140:e194\u0026ndash;e233. doi: 10.1161/CIR.0000000000000697\u003c/li\u003e\n\u003cli\u003eHamzah M, Othman HF, Almasri M, et al. Survival outcomes of in-hospital cardiac arrest in pediatric patients in the USA. \u003cem\u003eEur J Pediatr\u003c/em\u003e (2021) 180:2513\u0026ndash;2520. doi: 10.1007/s00431-021-04082-3\u003c/li\u003e\n\u003cli\u003eYin M, Wang T, Jiang Q, et al.The association of red blood cell transfusion with mortality in pediatric patients with sepsis, severe sepsis, and septic shock: A single-center retrospective cohort study. \u003cem\u003eTransfus Clin Biol\u003c/em\u003e (2025) 32:62\u0026ndash;68. doi: 10.1016/j.tracli.2024.12.002\u003c/li\u003e\n\u003cli\u003eMuszynski JA, Banks R, Reeder RW, et al. Outcomes Associated With Early RBC Transfusion in Pediatric Severe Sepsis: A Propensity-Adjusted Multicenter Cohort Study. \u003cem\u003eShock\u003c/em\u003e (2022) 57:88\u0026ndash;94. doi: 10.1097/SHK.0000000000001863\u003c/li\u003e\n\u003cli\u003eStortz JA, Murphy TJ, Raymond SL, et al. Evidence for persistent immune suppression in patients who develop chronic critical illness after sepsis. \u003cem\u003eShock (Augusta Ga,)\u003c/em\u003e (2018) 49:249\u0026ndash;258. doi: 10.1097/SHK.0000000000000981\u003c/li\u003e\n\u003c/ol\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":"Pediatric sepsis, PICU, early mortality risk, machine learning, prediction models","lastPublishedDoi":"10.21203/rs.3.rs-7842349/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7842349/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eObjective: \u003c/strong\u003eThis study aims to construct a risk prediction model for early mortality in pediatric intensive care unit (PICU) sepsis patients using machine learning methods. The model is intended to assist clinicians in identifying high-risk patients to enable timely interventions, and to help prevent early deaths in PICU sepsis patients.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e: A single-center retrospective cohort study design was used; clinical data of sepsis patients admitted to the PICU of Chongqing Medical University Affiliated Children's Hospital from January 2015 to December 2021 were included. The data comprised demographic information, vital signs, complications, laboratory indicators, diagnoses, and treatments. Patients were divided into early mortality and survival groups based on whether death occurred within 14 days of PICU admission. Seventy percent of the data were randomly assigned to the training set, and 30% were assigned to the validation set. Seven machine learning models, including logistic regression (LR), decision tree (DT), random forest (RF), XGBoost (XGB), LightGBM (LGBM), Naive Bayes (NB), and support vector machine (SVM), were used to build the early mortality prediction model for pediatric sepsis patients. The model's predictive performance was evaluated using sensitivity, specificity, receiver operating characteristic curve (ROC curve), and calibration curve. The clinical application value of the model was assessed through decision curve analysis (DCA).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults: \u003c/strong\u003eA total of 1,559 pediatric sepsis patients were included, with 198 cases of early mortality, resulting in an early mortality rate of 12.7%. After feature selection using LASSO regression, recursive feature elimination (RFE), and Boruta algorithms, six optimal predictive variables were identified: blood transfusion, pediatric Sequential Organ Failure Assessment (pSOFA) score, cardiopulmonary resuscitation, days of mechanical ventilation, secondary infection(s), and septic shock. These variables were used to construct the early mortality prediction model. ROC curve analysis showed that the area under the curve (AUC) values of the seven models ranged from 0.88 to 0.94. The XGBoost and LGBM models performed the best, both achieving an AUC of 0.94. Among all models, XGBoost and LGBM had the highest accuracy, 0.929 and 0.925 respectively; sensitivity was 0.534 and 0.483; specificity was 0.985 and 0.988; and F1 scores were 0.653 and 0.615, respectively. Calibration curves indicated that the XGBoost and LGBM models performed best among the seven machine learning models, showing excellent calibration performance. Decision curve analysis demonstrated that, compared to other models, XGBoost and LGBM exhibited greater net benefits across a wider range of threshold probabilities, indicating good clinical application value and confirming them as the optimal models.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion: \u003c/strong\u003eMachine learning models are reliable tools for predicting early mortality risk in pediatric patients with sepsis in the PICU. The XGBoost and LGBM models were built using six key variables: blood transfusion, Pediatric Sequential Organ Failure Assessment (pSOFA) score, post-cardiopulmonary resuscitation (CPR), days of mechanical ventilation, secondary infection, and septic shock. These models demonstrate stable performance, high discriminatory ability, and accuracy. Moreover, they aid clinicians in identifying high-risk patients and implementing early interventions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTrial Registration: \u003c/strong\u003eThe trial was registered on the Chinese Clinical Trial Registry with the number ChiCTR2300068260 on February 13, 2023.\u003c/p\u003e","manuscriptTitle":"Building a Machine Learning Model to Predict the Early Mortality Risk in Pediatric ICU Sepsis Patients","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-03 10:17:38","doi":"10.21203/rs.3.rs-7842349/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":"a497157b-6c9e-45b0-8009-018d1754121e","owner":[],"postedDate":"November 3rd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-01-14T06:25:08+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-03 10:17:38","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7842349","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7842349","identity":"rs-7842349","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}