Development and Validation of a Machine-Learning Model for Predicting Prognosis in Critically Ill Patients Undergoing Major Surgery

Development and Validation of a Machine-Learning Model for Predicting Prognosis in Critically Ill Patients Undergoing Major Surgery | 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 Development and Validation of a Machine-Learning Model for Predicting Prognosis in Critically Ill Patients Undergoing Major Surgery Xian Zhang, Tao Wang, Zhe Chen, Shuliu Zhang, Jihong Yuan, YingYi Qin, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5411439/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 29 Oct, 2025 Read the published version in BMC Medical Informatics and Decision Making → Version 1 posted 4 You are reading this latest preprint version Abstract Background Major surgery can result in elevated mortality rates, poorer prognoses, and extended hospital stays. This study sought to develop and validate an effective machine-learning model capable of accurately forecasting outcomes in critically ill patients who have undergone major surgery. Methods Using the publicly accessible Medical Information Mart for Intensive Care (MIMIC)-IV database, we developed and validated multiple machine-learning models to predict postoperative outcomes in critically ill patients who had at least an overnight ICU stay after major surgery. Seven predictive models were tested to forecast prognosis, with the highest-performing model selected based on its accuracy and the area under the receiver operating characteristic curve (AUC). An advanced model, eXtremely Gradient Boosting (XGBoost), was created using all variables, followed by a streamlined model built from 10 features chosen for their importance and clinical applicability. The performance of both models was assessed using Decision Curve Analysis (DCA), while survival analyses distinguished high- and low-risk groups within the internal validation sets. Results A cohort of 2,335 critically ill patients who had undergone major surgery were included in the MIMIC-IV cohorts. The full XGBoost model achieved an accuracy of 80.6% and an AUC of 0.828, indicating high predictive power. A more practical selection model with 10 features demonstrated a slightly lower AUC of 0.824 (95% CI: 0.762–0.886) but offered advantages in clinical usability. The ten key features were identified based on their SHapley Additive exPlanations (SHAP) values, which included the Charlson Comorbidity Index (CCI), Simplified Acute Physiology Score (SAPS)-II, Sequential Organ Failure Assessment (SOFA) score, mechanical ventilation, ARDS and sepsis complications, blood urea nitrogen (BUN) levels, estimated glomerular filtration rate (eGFR), respiratory rate, and marital status. Additionally, survival curves showed a clear distinction between high- and low-risk groups based on predictions from both the full and selection XGBoost models. Conclusions This study developed two XGBoost model variants that outperform other ICU predictive methods for forecasting the prognosis of patients undergoing major surgery. These models have the potential to assist healthcare providers in making more informed decisions, thereby improving clinical outcomes in the ICU setting. Major surgery Dynamic prediction Machine learning eXtremely Gradient Boosting Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 INTRODUCTION Mortality stands as a crucial indicator for surgical outcomes, with the delivery of safe surgical care recognized as a global healthcare priority( 1 ). Current guidelines emphasize the need for preoperative risk assessment to guide treatment choices and foster collaborative decision-making( 2 ). Research indicates that postoperative mortality often results less from intraoperative events and more from an inability to "rescue" patients who encounter complications afterward( 3 ). Annually, approximately 313 million surgeries are conducted globally( 4 ), with around 77.2 million disability-adjusted life years (DALYs) potentially preventable through essential, life-saving surgical interventions( 5 ). Surgical requirements contribute to 28–32% of the global disease burden, a figure likely to grow as life expectancy increases and chronic comorbidities become more common( 6 ). Accurate risk prediction tools play a vital role in the perioperative diagnostic/therapeutic pathway, improving outcomes by effectively deploying limited resources. Once a patient is identified as having a high risk of mortality, targeted mitigation strategies, such as prompt admission to critical care units or enhanced postoperative surveillance, may effectively reduce the likelihood of adverse outcomes ( 7 ). In recent years, the advancement of precision medicine has spurred the development of new machine-learning algorithms capable of dynamically predicting clinical events by analyzing extensive and complex patient data. These advanced machine-learning models excel at detecting complex, non-linear relationships between variables and outcomes, making them especially useful for deciphering subtle signals in data-intensive clinical environments( 8 – 11 ). The objective of this study was to create and validate a machine-learning model with strong accuracy for forecasting outcomes in critically ill patients who have undergone major surgery. By utilizing a large-scale public database, we sought to design a model with robust predictive power, carefully selecting predictive features based on their clinical relevance and statistical significance. MATERIALS AND METHODS Source of Data This retrospective analysis utilized data from the MIMIC-IV, a comprehensive and updated critical care database. MIMIC-IV expands upon the previous MIMIC-III dataset, containing clinical records of patients admitted to the ICU at Beth Israel Deaconess Medical Center from 2008 to 2019. Data extraction was performed by one of the study’s authors (QZ), who had secured authorized access to MIMIC-IV through the requisite institutional protocols, ensuring both data integrity and compliance with ethical standards.. The study was documented in alignment with the recommendations outlined in the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guidelines( 12 ). Participant Selection The study population was drawn from the MIMIC-IV database and included patients who underwent major surgeries that required at least an overnight ICU stay during their hospital admission. Major surgeries were identified via ICD-9 and ICD-10 procedure codes, covering both major diagnostic and therapeutic categories (supplemental table 1 ). Exclusion criteria included: (i) patients under 18 or over 85 years of age, (ii) those with hospital stays of less than 24 hours. To maximize the predictive model’s statistical power, all eligible cases meeting these criteria in MIMIC-IV were included Data Collection and Outcome Definition Data extraction focused on clinical, demographic, and laboratory variables from the most recent preoperative records, allowing for accurate baseline assessments prior to surgery. When variables had multiple recorded values, averages were calculated to ensure representative data points. Patient-specific data included demographic details such as gender, age, ethnicity, marital status and type of admission. Vital signs recorded encompassed blood pressure, heart rate, respiratory rate, peripheral oxygen saturation [SpO2], and body temperature. Laboratory measures included results from blood gas analysis, blood counts, liver and kidney function tests, and coagulation profiles. Comorbidities were determined through ICD-9 and ICD-10 coding, with each patient's comorbidity burden assessed using the CCI. Furthermore, the SAPS-II and the SOFA were computed based on a combination of clinical and laboratory findings. Major postoperative complications were recorded, including sepsis, wound infection, acute kidney injury (AKI), ARDS, mechanical ventilation (MV), the requirement for continuous renal replacement therapy (CRRT), and vasopressor use due to shock. The primary endpoint of this study was 28-day postoperative mortality, defined as any death occurring within 28 days after the surgical procedure. Secondary endpoints included in-hospital mortality and total duration of hospital stay. This systematic collection of clinical, demographic, and laboratory data was aimed at identifying risk factors associated with adverse postoperative outcomes, contributing to a better understanding of mortality and complication patterns in surgical patients. Statistical Analysis Baseline characteristics between the survival and death groups within the MIMIC-IV were analyzed to describe and compare patient demographics and clinical profiles. Continuous variables were reported as means with standard deviations (SD) if they met normality assumptions; otherwise, they were expressed as medians with interquartile ranges (IQR). Comparisons of continuous variables between groups were performed using Student’s t-test for normally distributed data and the Wilcoxon rank-sum test for data not meeting normality assumptions. Categorical variables were summarized as frequencies and percentages, with group differences evaluated using the Chi-square test or Fisher’s exact test, as appropriate, based on expected counts. Model Development and Validation The dataset was randomly partitioned into two subsets: a training set comprising 70% of the data and an internal validation set containing the remaining 30%. This division facilitated both the construction and evaluation of predictive models within a controlled framework. A sophisticated machine-learning model, XGBoost, was developed using the "mlr3verse" package in R. Categorical variables were pre-processed using one-hot encoding, while missing values were imputed with the "imputehist" method, ensuring consistent data handling. Given the low prevalence of the outcome (death), Synthetic Minority Over-sampling Technique (SMOTE) was employed to balance class representation in the training set. To optimize model performance, nested resampling was used with a 3-fold inner and outer cross-validation approach. Hyperparameters were fine-tuned using a random search strategy within the training set to ensure optimal predictive accuracy. A comprehensive XGBoost model, incorporating all available variables, was built to predict prognosis. Following model training, the ten variables with the highest feature importance were selected to create a simplified, interpretable model for comparison purposes( 13 ). Partial dependence plots (PDPs) were generated to illustrate the marginal effects of explanatory variables on the outcome. Comparative Model Analysis For benchmarking, six additional models were developed using the training set: Random Forest, Elastic Net (Glmnet), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Logistic Regression, and LightGBM( 14 ). The same pre-processing steps (missing data imputation, encoding, and nested resampling) were consistently applied across these models( 15 ), ensuring comparability. Model performance was evaluated on the internal validation set using AUC with corresponding 95% confidence intervals (CI) to assess discrimination ability. The optimal decision threshold for each model was determined using the Youden index. Model Interpretation and Survival Analysis DCA was performed to assess the clinical utility of the XGBoost models across a range of threshold probabilities, offering insight into the net benefit for clinical decision-making. Based on risk predictions from the XGBoost model, patients in the validation set were stratified into high-risk and low-risk groups. Survival analysis was conducted to compare outcomes between these groups, providing further validation of the model’s prognostic utility. Software and Statistical Significance All statistical analyses were conducted using R software (version 4.1.2). Statistical significance was set at a p-value threshold of 0.05 for hypothesis tests. RESULTS Baseline Characteristics As illustrated in Fig. 1 , the study cohort consisted of 2,335 critically ill patients who had undergone major surgical procedures, sourced from the MIMIC-IV database. These patients were divided into two sets: a training set comprising 1,634 patients and a testing set containing 701 patients. Table 1 provides a detailed summary of the baseline characteristics for patients categorized into survival and death outcome groups within the cohort. Patients in the death group displayed significantly elevated baseline values in several key physiological and clinical metrics, including heart rate, respiratory rate, SAPS-II and SOFA scores, CCI, BUN, and WBC count, while showing reduced levels of hemoglobin and eGFR, alongside prolonged PT and APTT. Additionally, patients in this group had a higher prevalence of complications such as sepsis, AKI, ARDS, CRRT, MV, and vasopressor administration (p < 0.05), indicating a more severe clinical presentation compared to the survival group. Table 1 Characteristics of critically ill patients who underwent major surgery in the MIMIC-IV Characteristics Overall (n = 2335) Survival (n = 2202) Death (n = 133) P Gender = Female (%) 977 (41.8) 922 (41.9) 55 (41.4) 0.978 Age (median [IQR]) 64.31 [53.36, 73.68] 64.02 [53.06, 73.44] 68.75 [58.74, 77.92] < 0.001 Ethnicity (%) 0.589 Asian 73 ( 3.1) 69 ( 3.1) 4 ( 3.0) Black/African American 203 ( 8.7) 191 ( 8.7) 12 ( 9.0) Hispanic/Lation 77 ( 3.3) 75 ( 3.4) 2 ( 1.5) White 1551 (66.4) 1466 (66.6) 85 (63.9) Other and unknown 431 (18.5) 401 (18.2) 30 (22.6) Admission type (%) 0.149 Observation admit or elective 332 (14.2) 318 (14.4) 14 (10.5) Emergency 1218 (52.2) 1151 (52.3) 67 (50.4) Urgent 754 (32.3) 702 (31.9) 52 (39.1) Others 31 ( 1.3) 31 ( 1.4) 0 ( 0.0) Marital status = Married (%) 1151 (52.9) 1088 (52.9) 63 (53.8) 0.916 Temperature (median [IQR]) 36.78 [36.56, 37.06] 36.78 [36.61, 37.06] 36.72 [36.50, 37.14] 0.370 Heart rate (median [IQR]) 80.00 [69.00, 94.00] 80.00 [69.00, 94.00] 88.00 [74.00, 104.00] < 0.001 MBP (median [IQR]) 80.00 [71.00, 91.00] 81.00 [71.00, 91.00] 72.00 [65.00, 84.00] < 0.001 Respiratory rate (median [IQR]) 18.00 [16.00, 22.00] 18.00 [16.00, 22.00] 20.00 [16.00, 24.00] 0.001 SpO 2 (median [IQR]) 96.00 [94.00, 97.00] 96.00 [94.00, 97.00] 96.00 [94.00, 97.00] 0.216 GCS (median [IQR]) 15.00 [15.00, 15.00] 15.00 [15.00, 15.00] 15.00 [15.00, 15.00] 0.142 SOFA (median [IQR]) 1.00 [1.00, 3.00] 1.00 [1.00, 3.00] 3.00 [1.00, 5.00] < 0.001 SAPS Ⅱ (median [IQR]) 29.00 [22.00, 37.00] 29.00 [22.00, 36.00] 37.50 [31.00, 48.00] < 0.001 Charlson score (median [IQR]) 5.00 [3.00, 7.00] 5.00 [3.00, 7.00] 7.00 [5.00, 9.00] < 0.001 Vasopressor, n (%) 348 (14.9) 321 (14.6) 27 (20.3) 0.094 MV, n (%) 1282 (54.9) 1181 (53.6) 101 (75.9) < 0.001 CRRT, n (%) 36 ( 1.5) 27 ( 1.2) 9 ( 6.8) < 0.001 AKI, n (%) 209 ( 9.0) 185 ( 8.4) 24 (18.0) < 0.001 Complication sepsis, n (%) 299 (12.8) 246 (11.2) 53 (39.8) < 0.001 Complication wound infection, n (%) 55 ( 2.4) 51 ( 2.3) 4 ( 3.0) 0.829 Complication ARDS, n (%) 237 (10.1) 193 ( 8.8) 44 (33.1) < 0.001 Surgery during ICU, n (%) 1714 (73.4) 1602 (72.8) 112 (84.2) 0.005 BUN (median [IQR]) 18.00 [13.00, 27.00] 17.00 [12.00, 26.00] 29.00 [18.00, 47.00] < 0.001 eGFR (median [IQR]) 76.32 [52.23, 98.75] 78.33 [54.10, 98.94] 52.77 [31.67, 84.97] < 0.001 Sodium (median [IQR]) 139.00 [136.00, 141.00] 139.00 [136.00, 141.00] 137.00 [133.00, 140.00] 0.007 Potassium (median [IQR]) 4.10 [3.80, 4.40] 4.10 [3.80, 4.40] 4.20 [3.70, 4.47] 0.426 Chloride (median [IQR]) 103.00 [99.00, 106.00] 103.00 [100.00, 106.00] 102.00 [97.00, 105.00] 0.056 Bicarbonate (median [IQR]) 24.00 [22.00, 27.00] 24.00 [22.00, 27.00] 23.00 [20.00, 25.00] < 0.001 Anion Gap (median [IQR]) 14.00 [12.00, 16.00] 14.00 [12.00, 16.00] 15.00 [13.00, 18.00] < 0.001 Total calcium (median [IQR]) 8.70 [8.20, 9.10] 8.70 [8.20, 9.10] 8.50 [7.88, 9.00] 0.012 Magnesium (median [IQR]) 2.00 [1.90, 2.20] 2.00 [1.90, 2.20] 2.10 [1.90, 2.30] 0.058 Glucose (median [IQR]) 117.00 [100.00, 146.00] 117.00 [100.00, 145.00] 123.00 [107.00, 161.00] 0.039 Hemoglobin (median [IQR]) 11.50 [9.67, 13.20] 11.50 [9.70, 13.20] 10.20 [8.90, 11.90] < 0.001 WBC (median [IQR]) 9.40 [7.20, 12.90] 9.35 [7.20, 12.70] 11.80 [8.50, 15.90] < 0.001 Platelet (median [IQR]) 217.00 [164.00, 279.00] 218.00 [166.00, 279.00] 197.00 [139.00, 285.00] 0.112 PT (median [IQR]) 13.00 [11.80, 14.80] 12.90 [11.80, 14.70] 14.70 [13.10, 18.80] < 0.001 APTT (median [IQR]) 30.80 [26.90, 41.10] 30.60 [26.80, 40.60] 36.30 [29.90, 47.15] < 0.001 Phosphate (median [IQR]) 3.40 [2.90, 4.00] 3.40 [2.90, 4.00] 3.85 [3.00, 4.73] < 0.001 MBP, Mean Blood Pressure; SpO 2 , peripheral oxygen saturation; GCS, Glasgow Coma Scale; SOFA, Sequential Organ Failure Assessment; SAPS Ⅱ, Simplified Acute Physiology Score Ⅱ; MV, Mechanical Ventilation; CRRT, ContinuousRenal Replacement Therapy; AKI, Acute Kidney Injury; ARDS, Acute Respiratory Distress Syndrome; BUN, Blood Urea Nitrogen; eGFR, estimated Glomerular Filtration Rate; WBC, White Blood Cell; PT, Prothrombin Time; APTT, Activated Partial Thromboplastin Time; IQR, Interquartile range. Comparison of Predictive Models The performance metrics for each predictive model are displayed in Table 2 . Logistic Regression exhibited a baseline performance with an accuracy of 0.725 and an AUC of 0.794. Among the models, ensemble learning algorithms demonstrated superior predictive accuracy and AUC values compared to other approaches. Notably, the XGBoost model achieved the highest accuracy (0.806) and AUC (0.828), followed closely by lightGBM (accuracy: 0.795; AUC: 0.824) and the Random Forest Classifier (accuracy: 0.779; AUC: 0.819). The high discriminatory power of the XGBoost model for predicting mortality risk warranted further refinement of this model to optimize its clinical applicability. Table 2 Model prediction performance Model AUC cutoff Sensitivity(%) Specificity(%) Accuracy(%) PPV(%) NPV(%) XGboost select model-tain 0.854(0.822–0.886) 0.209 92.5(87.1–97.8) 63.8(61.5–66.1) 65.4(63.2–67.7) 13.4(12.4–14.4) 99.3(98.8–99.8) XGboost select model-test 0.824(0.762–0.886) 0.209 85(72.5–95) 65.8(62.2–69.3) 66.9(63.2–70.3) 13.1(11.1–14.9) 98.7(97.6–99.5) XGboost-tain 0.882(0.855–0.909) 0.347 81.7(74.2–89.2) 81.6(79.6–83.5) 81.6(79.7–83.4) 21.1(18.9–23.6) 98.7(98.1–99.2) XGboost-test 0.828(0.769–0.887) 0.347 62.5(47.5–77.5) 81.7(78.8–84.6) 80.6(77.6–83.6) 17.1(13.1–21.4) 97.3(96.2–98.4) LightGBM-tain 0.871(0.839–0.902) 0.372 79.6(71-87.1) 81.2(79.2–83.1) 81.1(79.2–83) 20.3(17.9–22.8) 98.5(97.9–99.1) LightGBM-test 0.824(0.767–0.881) 0.372 65(50–80) 80.3(77.3–83.2) 79.5(76.3–82.3) 16.7(12.9–20.7) 97.4(96.4–98.5) RandomForest-tain 0.871(0.84–0.902) 0.462 77.4(68.8–86) 79.4(77.3–81.4) 79.3(77.3–81.2) 18.5(16.3–20.8) 98.3(97.7–98.9) RandomForest-test 0.819(0.754–0.883) 0.462 65(50–80) 78.7(75.3–81.5) 77.9(74.7–80.9) 15.5(12-19.1) 97.4(96.3–98.4) SVM-tain 0.997(0.996–0.999) 0.005 100(100–100) 96.8(95.9–97.7) 97(96.1–97.8) 65.5(59.6–72.1) 100(100–100) SVM-test 0.719(0.66–0.778) 0.005 62.5(47.5–77.5) 71.1(67.6–74.6) 70.6(67.3–73.9) 11.6(8.9–14.3) 96.9(95.7–98.1) KNN-tain 0.999(0.997-1) 0.671 98.9(96.8–100) 99.8(99.5–100) 99.8(99.5–99.9) 96.8(92.9–100) 99.9(99.8–100) KNN-test 0.645(0.56–0.73) 0.671 20(10-32.5) 92.7(90.6–94.7) 88.6(86.6–90.6) 14.3(7-23.7) 95(94.4–95.8) Logistic regression-tain 0.847(0.815–0.88) 0.42 84.9(77.4–91.4) 73.5(71.3–75.6) 74.1(72-76.1) 16.2(14.6–17.8) 98.8(98.2–99.3) Logistic regression-test 0.794(0.728–0.86) 0.42 65(50–80) 73.1(69.7–76.4) 72.5(69.3–75.7) 12.8(9.9–15.8) 97.2(96-98.4) Glmnet-tain 0.793(0.755–0.831) 0.484 84.9(77.4–91.4) 62.9(60.5–65.3) 64.2(61.9–66.5) 12.2(11-13.3) 98.6(97.9–99.2) Glmnet-test 0.792(0.722–0.862) 0.484 85(75–95) 61.6(57.8–65.2) 62.9(59.3–66.5) 11.8(10.1–13.5) 98.5(97.5–99.5) XGboost, eXtremely Gradient Boosting; SVM, Support Vector Machine; KNN, K-Nearest Neighbor; AUC, the Receiver Operating Characteristic Curve; PPV, Positive Predictive Value; NPV, Negative Predictive Value. Development and Optimization of the XGBoost Model Feature importance and partial dependence profiles for the XGBoost model are displayed in Fig. 2 . The Charlson score emerged as the most influential predictor, with higher scores correlating with an increased risk of mortality. Using SHapley Additive exPlanations (SHAP) values, the top ten predictive features—comprising CCI, SOFA score, SAPS-II score, need for MV, presence of ARDS and sepsis, BUN levels, eGFR, respiratory rate, and marital status—were identified and used to develop a streamlined version of the XGBoost model. Although this optimized model demonstrated a slightly lower AUC of 0.824 (95% CI: 0.762–0.886) compared to the full model, it was considered to offer enhanced clinical feasibility due to its reduced complexity. Comparative Analysis of Model Performance In Fig. 3 , the AUC-ROC curve comparisons between the XGBoost model and alternative models underscore its superior discriminatory ability. To further evaluate the clinical utility of the XGBoost model, a Decision Curve Analysis (Fig. 4 ) was conducted. The analysis indicated that within a threshold probability range of 1–48%, utilizing the XGBoost model to predict patient prognosis would yield a net benefit compared to the strategies of either screening all patients or screening none. For clarity, only the results for the full XGBoost model and the streamlined XGBoost selection model are displayed. Sensitivity and specificity evaluations for these predictive models in the internal validation set are summarized in Table 2 . Risk Stratification and Survival Analysis Patients were stratified into high-risk and low-risk categories based on their calculated mortality risk scores, using the median risk score as the cutoff value. Kaplan-Meier survival curves demonstrated a clear separation between high- and low-risk groups in both the XGBoost selection model (Fig. 5 A, log-rank test, p = 0.0001) and the full XGBoost model (Fig. 5 B, log-rank test, p = 0.0005), highlighting the models’ effectiveness in distinguishing patient outcomes. DISCUSSION This study is the first to utilize machine-learning models to predict the prognosis of critically ill patients after major surgery. By developing and validating two variations of a dynamic machine-learning model, this research facilitates the identification of high-risk patients, thus providing healthcare decision-makers with actionable clinical insights. Among the models evaluated, the XGBoost algorithm achieved the highest AUC, reflecting its superior predictive accuracy. XGBoost is recognized for its efficiency in handling missing data and its ability to combine weak learners into a robust predictive ensemble ( 16 ). This method has gained widespread credibility and is frequently employed in data science competitions. For instance, in 2015, XGBoost was used in 17 of the 29 winning solutions published on Kaggle’s blog, and the top 10 winning teams in the 2015 KDD Cup utilized XGBoost ( 17 , 18 ). Typically, models incorporating a broader range of variables exhibit enhanced discriminatory ability but may suffer from reduced clinical feasibility. In light of this trade-off, our study developed two model variants, each tailored to distinct clinical settings. The full model, utilizing 58 clinical variables, achieved the highest AUC in this study and demonstrated strong prognostic capabilities. However, the extensive data requirements limit its applicability to institutions with comprehensive and structured clinical data systems. Conversely, the streamlined model, trained on only 10 selected variables, offers a pragmatic balance between accuracy and usability, making it viable for settings with more limited resources. Interpretation of the full model highlighted that multiple clinical variables are integral to assessing mortality risk. Notably, CCI emerged as the most critical predictor, followed closely by complications associated with ARDS. The predictive value of CCI has been extensively validated across diverse medical conditions, with higher CCI scores correlating strongly with increased mortality rates ( 19 ). Furthermore, as shown in Fig. 1 , patients with elevated SOFA scores demonstrated a higher risk of death, consistent with previous studies. Early monitoring of organ dysfunction within the initial days in the ICU provides a critical indicator of patient prognosis. Specifically, an increase in SOFA score within the first 48 hours of ICU admission has been associated with a predicted mortality rate of at least 50% ( 20 ). As anticipated, patients with ARDS and sepsis demonstrated poorer prognoses. Additional factors such as mechanical ventilation and SAPS-II scores were useful in assessing mortality risk. Patients with reduced renal function, indicated by elevated serum creatinine and urea nitrogen, had a higher likelihood of death. Interestingly, we found that marital status is a predictor of death. Studies across Chinese, U.S., and international cohorts have shown that married patients with cardiovascular disease and various cancers tend to have better outcomesr( 21 , 22 , 23 ). Psychosocial and socioeconomic factors, along with acute stressors, may partly explain the association between marital status and patient prognosis, though the specific underlying mechanisms remain largely unexplored. Our model demonstrates substantial potential as a tool to improve the prognosis for critically ill patients following major surgery, offering both clinical and economic advantages. Implementing this model could facilitate a more tailored approach to patient care. For patients identified as high-risk, delayed transfer from intensive care and enhanced monitoring may reduce mortality rates. This approach is particularly valuable given the scarcity of critical care beds, as decisions regarding patient discharge or transfer directly affect the efficiency of ICU resource utilization ( 24 ). Conversely, for patients assessed as low-risk, unnecessary medical expenses associated with additional testing and treatment may be avoided. The clinical utility of this predictive model will be further assessed in future prospective studies to ensure its effectiveness and practicality. Several limitations of this study should be acknowledged. First, although patients from various surgical procedures were included, the prognosis may differ significantly depending on the specific type of surgery, indicating a need for future subgroup analyses. Second, as a retrospective study, this research may be affected by missing data or input errors, even though the MIMIC-IV dataset is known for its high quality. Finally, while the model has shown promise, external validation using other datasets and prospective studies are essential to confirm its accuracy and generalizability in diverse clinical settings. CONCLUSIONS In summary, this study successfully developed two versions of an XGBoost model that outperformed alternative predictive models in forecasting outcomes for ICU patients after major surgery. These models offer potential as decision-support tools, enhancing patient care and resource allocation in critical care environments. Declarations Clinical trial number: not applicable Acknowledgements We are grateful to all those who took part in or assisted with this study project. And we thank the Massachusetts Institute of Technology and the Beth Israel Deaconess Medical Center for the MIMIC project. Ethics declarations Human ethics and consent to participate Not applicable. Consent to participate Each author consents to participate in the manuscript. Competing interests The authors declare no competing interests. Funding This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Author information Authors and Affiliations Department of Intensive Care Medicine, Shanghai Sixth People’s Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China Xian Zhang Department of Intensive Care Medicine, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China Tao Wang Science and Technology Department, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China Zhe Chen Department of Critical Care Medicine, The 960th Hospital of the PLA Joint Logistics Support Force, Shandong, China Shuliu Zhang Department of Critical Care Medicine, Seventh People's Hospital of Shanghai University of Traditional Chinese Medicine, Shanghai, China Jihong Yuan Department of Military Health Statistics, Naval Medical University, Shanghai, China YingYi Qin Department of Critical Care Medicine, The 960th Hospital of the PLA Joint Logistics Support Force, Shandong, China Yunliang Cui Contributions Contributions: (I) Conception and design: Xian Zhang, Yunliang Cui; (II) Administrative support: Jihong Yuan, YingYi Qin, Yunliang Cui; (III) Provision of study materials : Xian Zhang, Tao Wang; (IV) Collection and assembly of data: Zhe Chen, Shuliu Zhang; (V) Data analysis and interpretation: Xian Zhang, YingYi Qin; (VI) Manuscript writing: All authors; (VII) Final approval of manuscript: All authors. Corresponding author Correspondence: Yunliang Cui, YingYi Qin, Jihong Yuan. Data availability The original dataset generated during the current study is available from the corresponding author on reasonable request. References Meara JG, Leather AJ, Hagander L, Alkire BC, Alonso N, Ameh EA, et al. Global Surgery 2030: evidence and solutions for achieving health, welfare, and economic development. Lancet. 2015;386:569–624. 10.1016/S0140-6736(15)60160-X . Levine GN, O’Gara PT, Beckman JA, Al-Khatib SM, Birtcher KK, Cigarroa JE, et al. Recent Innovations,Modifications, and Evolution of ACC/AHA Clinical Practice Guidelines: An Update for Our Constituencies: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. Circulation. 2019;139:e879–86. 10.1161/CIR.0000000000000651 . Ghaferi AA, Birkmeyer JD, Dimick JB. Complications, failure to rescue, and mortality with major inpatient surgery in medicare patients. Ann Surg., Biccard BM, Madiba TE, Kluyts HL, Munlemvo DM, Madzimbamuto FD, Basenero A et al. Perioperative patient outcomes in the African Surgical Outcomes Study: a 7-day prospective observational cohort study. Lancet. 2018; 391: 1589–1598. doi:10.1016/S0140-6736(18)30001-1. Meara JG, Leather AJ, Hagander L, Alkire BC, Alonso N, Ameh EA, et al. Global Surgery 2030: evidence and solutions for achieving health, welfare, and economic development. Lancet. 2015;386(9993):569–624. 10.1016/S0140-6736(15)60160-X . Bickler SN et al. Global Burden of Surgical Conditions. Essential Surgery: Disease Control Priorities. (ed. Debas, H.T.) Chap. 2 (The International Bank for Reconstruction and Development/The World Bank, 2015). Piccirillo JF, Vlahiotis A, Barrett LB, Flood KL, Spitznagel EL, Steyerberg EW, et al. The changing prevalence of comorbidity across the age spectrum. Crit Rev Oncol/Hematol. 2008;67(2):124–32. 10.1016/j.critrevonc.2008.01.013 . Lees N, Peden CJ, Dhesi J, Quiney N, Lockwood S, Symons NR et al. The High-Risk General Surgical Patient: Raising the Standard. Updated recommendations on the Perioperative Care of the High-Risk General Surgical Patient [Internet]. 2018 [cited 2020 Jun 22]. Zhao QY, Liu LP, Luo JC, Luo YW, Luo Z. A machine-learning approach for dynamic prediction of sepsis-induced coagulopathy in critically ill patients with sepsis. Front Med. 2020;7. 10.3389/fmed.2020.637434 . Beam AL, Kohane IS. Big data and machine learning in health care. JAMA. 2018;319:1317–8. 10.1001/jama.201718391 . Luo JC, Zhao QY, Tu GW. Clinical prediction models in the precisionmedicine era: old and new algorithms. Ann Transl Med. 2020;8:274. 10.21037/atm.2020.02.63 . Ge H, Pan Q, Zhou Y, Xu P, Zhang L, Zhang J, et al. Lungmechanics of mechanically ventilated patients with COVID-19: analytics with high-granularity ventilator waveform data. Front Med. 2020;7:541. 10.3389/fmed.202000541 . Collins GS, Reitsma JB, Altman DG, Moons KG. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD Statement. BMC Med. 2015;13:1. 10.1186/s12916-014-0241-z . Zhang ZH. Variable selection with stepwise and best subset approaches. Ann Transl Med. 2016;4:136. 10.21037/atm.2016.03.35 . Luo JC, Zhao QY, Tu GW. Clinical prediction models in the precision medicine era: old and new algorithms. Ann Transl Med. 2020;8:274. 10.21037/atm.2020.02.63 . Lundberg SM, Erion G, Chen H, DeGrave A, Prutkin JM, Nair B, et al. From local explanations to global understanding with explainable AI for trees. Nat Mach Intell. 2020;2:56–67. 10.1038/s42256-019-0138-9 . Yuan KC, Tsai LW, Lee KH, Cheng YW, Chen RJ, et al. The development an artifcial intelligence algorithm for early sepsis diagnosis in the intensive care unit. Int J Med Inf. 2020;141:104176. 10.1016/j.ijmedinf.2020.104176 . Chen T, Guestrin C. XGBoost: a scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining-KDD 2016, San Francisco, CA, USA; 2016. pp. 785–94. Hou N, Li M, He L, Xie B, Wang K, et al. Predicting 30-days mortality for MIMIC-III patients with sepsis-3: a machine learning approach using XGboost. J Translational Med. 2020;18(1). 10.1186/s12967-020-02620-5 . Mary EC, Danilo C, Jenny G, Chiara P. Charlson Comorbidity Index: A Critical Review of Clinimetric Properties. Psychother Psychosom. 2022;91(1):8–35. 10.1159/000521288 . Ferreira LF. Serial Evaluation of the SOFA Score to Predict Outcome in Critically Ill Patients. JAMA. 2001;286(14):1754–8. 10.1001/jama.286.14.1754 . Dhindsa DS, Khambhati J, Schultz WM, Tahhan AS, Quyyumi AA. Marital status and outcomes in patients with cardiovascular disease.rends. Cardiovasc Med. 2020;30(4):215–20. 10.1016/j.tcm.2019.05.012 . Liang Y, Wu X, Lu C, Xiao F. Impact of marital status on the prognosis of liver cancer patients without surgery and the critical window. Annals Palliat Med. 2021;10(3):2990–9. 10.21037/apm-20-1885 . Guo Z, Gu C, Li S, Gan S, Li Y, et al. Association between Marital Status and Prognosis in Patients with Prostate Cancer: A Meta-Analysis of Observational Studies. Urol J. 2020;18(4):371–9. 10.22037/uj.v16i7.6197 . Cypress BS. Transfer out of intensive care: An evidence-based literature review. Dimensions Crit care nursing: DCCN. 2013;32(5):244–61. 10.1097/DCC.0b013e3182a07646 . Additional Declarations No competing interests reported. Supplementary Files MajorprocedureICD910.xls Cite Share Download PDF Status: Published Journal Publication published 29 Oct, 2025 Read the published version in BMC Medical Informatics and Decision Making → Version 1 posted Editorial decision: Revision requested 18 Nov, 2024 Editor assigned by journal 13 Nov, 2024 Submission checks completed at journal 13 Nov, 2024 First submitted to journal 07 Nov, 2024 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-5411439","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":379328569,"identity":"be459cfb-fef9-4184-9677-644a3457151a","order_by":0,"name":"Xian Zhang","email":"","orcid":"","institution":"Shanghai Sixth People’s Hospital, Shanghai Jiao Tong University School of Medicine","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Xian","middleName":"","lastName":"Zhang","suffix":""},{"id":379328570,"identity":"1156f6a6-d1e4-42ca-99de-6f8753030dfb","order_by":1,"name":"Tao Wang","email":"","orcid":"","institution":"Ruijin Hospital, Shanghai Jiao Tong University School of Medicine","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Tao","middleName":"","lastName":"Wang","suffix":""},{"id":379328571,"identity":"a2e165a7-cf7f-4b48-8ab3-9469713b7e91","order_by":2,"name":"Zhe Chen","email":"","orcid":"","institution":"Ruijin Hospital, Shanghai Jiao Tong University School of Medicine","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Zhe","middleName":"","lastName":"Chen","suffix":""},{"id":379328572,"identity":"0adbf219-f480-4caa-9add-6cbfa53e7ae5","order_by":3,"name":"Shuliu Zhang","email":"","orcid":"","institution":"The 960th Hospital of the PLA Joint Logistics Support Force","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Shuliu","middleName":"","lastName":"Zhang","suffix":""},{"id":379328573,"identity":"9b9540e4-2dd0-406c-a770-fc63c7b72f25","order_by":4,"name":"Jihong Yuan","email":"","orcid":"","institution":"Seventh People's Hospital of Shanghai University of Traditional Chinese Medicine","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Jihong","middleName":"","lastName":"Yuan","suffix":""},{"id":379328574,"identity":"c8565e78-74d3-4883-be83-647fd1bb94ab","order_by":5,"name":"YingYi Qin","email":"","orcid":"","institution":"Naval Medical University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"YingYi","middleName":"","lastName":"Qin","suffix":""},{"id":379328575,"identity":"33ad45f9-2df6-436b-93aa-d8d5d30b30ad","order_by":6,"name":"Yunliang Cui","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6ElEQVRIie3PsYrCQBCA4VkCq8Vi2hEh9wqBQAicYOGLjM3arIdwjZUShLO4PW31LXyEwII22qeM2FrcdVte+hM311nsX8/HzAD4fE8YDzeVsRYXYSvPK7J9N+ngiV2ZzqirjYkvWrpJBCpIgM8oLqXsVtw0OAzOh95U4BuUKp2RMBCuPukxCdayt83wnelbWlI2Bjyd944tRYpCIFu2VE3EK8Q4cRGqCUf2ASqdEg+aEJUkNRnptpRAfNiA4GF02WlMUBhTQymcv7xslkXxbefR4JjnP9b2o3D19Zj8Sfxv3Ofz+Xx3+wXhlUhELYjXTwAAAABJRU5ErkJggg==","orcid":"","institution":"The 960th Hospital of the PLA Joint Logistics Support Force","correspondingAuthor":true,"submittingAuthor":false,"prefix":"","firstName":"Yunliang","middleName":"","lastName":"Cui","suffix":""}],"badges":[],"createdAt":"2024-11-07 16:38:12","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5411439/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5411439/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s12911-025-03240-z","type":"published","date":"2025-10-29T15:57:26+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":71632922,"identity":"d838856a-8bb4-4cd6-8724-6c4f7624915b","added_by":"auto","created_at":"2024-12-17 09:48:38","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":218674,"visible":true,"origin":"","legend":"\u003cp\u003eFlow chart of patient selection. MIMIC-IV, Medical Information Mart for Intensive Care-IV; ICU, intensive care unit\u003c/p\u003e","description":"","filename":"Figure1Flowchart.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5411439/v1/582353f5087cee83377d29e3.jpg"},{"id":71632921,"identity":"70dc4aef-fad3-4cca-b757-86490e84fcad","added_by":"auto","created_at":"2024-12-17 09:48:38","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":851293,"visible":true,"origin":"","legend":"\u003cp\u003eFeature importance and SHAP\u003c/p\u003e","description":"","filename":"Figure2FeatureimportanceandSHAP.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5411439/v1/a74ce0e9ad0af979485f2703.jpg"},{"id":71632920,"identity":"bcc3e2b4-036f-4dce-8f09-8533227db0af","added_by":"auto","created_at":"2024-12-17 09:48:38","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":515361,"visible":true,"origin":"","legend":"\u003cp\u003eROC-test cohort\u003c/p\u003e","description":"","filename":"Figure3ROCtestcohort.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5411439/v1/1fba8bdd07fd7fa43eb9d733.jpg"},{"id":71634627,"identity":"1fce3d21-e7ee-444f-a1be-de1d7386bc2c","added_by":"auto","created_at":"2024-12-17 09:56:39","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":302287,"visible":true,"origin":"","legend":"\u003cp\u003eDCA-test cohort\u003c/p\u003e","description":"","filename":"Figure4DCAtestcohort.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5411439/v1/363633195ee015c438fc8f85.jpg"},{"id":71632923,"identity":"a177b920-71ef-46fd-b4da-ced31baeca42","added_by":"auto","created_at":"2024-12-17 09:48:39","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":630396,"visible":true,"origin":"","legend":"\u003cp\u003esurvival plot\u003c/p\u003e","description":"","filename":"Figure5survivalplot.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5411439/v1/6bb94d5a0ac26bd4ca18ab3a.jpg"},{"id":95039952,"identity":"ebe7493c-2722-43f0-b4bc-1921c02b6927","added_by":"auto","created_at":"2025-11-03 16:06:13","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3489639,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5411439/v1/a3a447ed-f345-481b-a569-293a0fcd77ff.pdf"},{"id":71632925,"identity":"8ec3cc16-a6cd-4b64-81ac-75273141902e","added_by":"auto","created_at":"2024-12-17 09:48:39","extension":"xls","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":16223744,"visible":true,"origin":"","legend":"","description":"","filename":"MajorprocedureICD910.xls","url":"https://assets-eu.researchsquare.com/files/rs-5411439/v1/a409d58d59e875a39d7d82e8.xls"}],"financialInterests":"No competing interests reported.","formattedTitle":"Development and Validation of a Machine-Learning Model for Predicting Prognosis in Critically Ill Patients Undergoing Major Surgery","fulltext":[{"header":"INTRODUCTION","content":"\u003cp\u003eMortality stands as a crucial indicator for surgical outcomes, with the delivery of safe surgical care recognized as a global healthcare priority(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). Current guidelines emphasize the need for preoperative risk assessment to guide treatment choices and foster collaborative decision-making(\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). Research indicates that postoperative mortality often results less from intraoperative events and more from an inability to \"rescue\" patients who encounter complications afterward(\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e). Annually, approximately 313\u0026nbsp;million surgeries are conducted globally(\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e), with around 77.2\u0026nbsp;million disability-adjusted life years (DALYs) potentially preventable through essential, life-saving surgical interventions(\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). Surgical requirements contribute to 28\u0026ndash;32% of the global disease burden, a figure likely to grow as life expectancy increases and chronic comorbidities become more common(\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). Accurate risk prediction tools play a vital role in the perioperative diagnostic/therapeutic pathway, improving outcomes by effectively deploying limited resources. Once a patient is identified as having a high risk of mortality, targeted mitigation strategies, such as prompt admission to critical care units or enhanced postoperative surveillance, may effectively reduce the likelihood of adverse outcomes (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn recent years, the advancement of precision medicine has spurred the development of new machine-learning algorithms capable of dynamically predicting clinical events by analyzing extensive and complex patient data. These advanced machine-learning models excel at detecting complex, non-linear relationships between variables and outcomes, making them especially useful for deciphering subtle signals in data-intensive clinical environments(\u003cspan additionalcitationids=\"CR9 CR10\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e). The objective of this study was to create and validate a machine-learning model with strong accuracy for forecasting outcomes in critically ill patients who have undergone major surgery. By utilizing a large-scale public database, we sought to design a model with robust predictive power, carefully selecting predictive features based on their clinical relevance and statistical significance.\u003c/p\u003e"},{"header":"MATERIALS AND METHODS","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eSource of Data\u003c/h2\u003e \u003cp\u003eThis retrospective analysis utilized data from the MIMIC-IV, a comprehensive and updated critical care database. MIMIC-IV expands upon the previous MIMIC-III dataset, containing clinical records of patients admitted to the ICU at Beth Israel Deaconess Medical Center from 2008 to 2019. Data extraction was performed by one of the study\u0026rsquo;s authors (QZ), who had secured authorized access to MIMIC-IV through the requisite institutional protocols, ensuring both data integrity and compliance with ethical standards.. The study was documented in alignment with the recommendations outlined in the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guidelines(\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eParticipant Selection\u003c/h3\u003e\n\u003cp\u003eThe study population was drawn from the MIMIC-IV database and included patients who underwent major surgeries that required at least an overnight ICU stay during their hospital admission. Major surgeries were identified via ICD-9 and ICD-10 procedure codes, covering both major diagnostic and therapeutic categories (supplemental table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Exclusion criteria included: (i) patients under 18 or over 85 years of age, (ii) those with hospital stays of less than 24 hours. To maximize the predictive model\u0026rsquo;s statistical power, all eligible cases meeting these criteria in MIMIC-IV were included\u003c/p\u003e\n\u003ch3\u003eData Collection and Outcome Definition\u003c/h3\u003e\n\u003cp\u003eData extraction focused on clinical, demographic, and laboratory variables from the most recent preoperative records, allowing for accurate baseline assessments prior to surgery. When variables had multiple recorded values, averages were calculated to ensure representative data points. Patient-specific data included demographic details such as gender, age, ethnicity, marital status and type of admission. Vital signs recorded encompassed blood pressure, heart rate, respiratory rate, peripheral oxygen saturation [SpO2], and body temperature. Laboratory measures included results from blood gas analysis, blood counts, liver and kidney function tests, and coagulation profiles. Comorbidities were determined through ICD-9 and ICD-10 coding, with each patient's comorbidity burden assessed using the CCI. Furthermore, the SAPS-II and the SOFA were computed based on a combination of clinical and laboratory findings. Major postoperative complications were recorded, including sepsis, wound infection, acute kidney injury (AKI), ARDS, mechanical ventilation (MV), the requirement for continuous renal replacement therapy (CRRT), and vasopressor use due to shock.\u003c/p\u003e \u003cp\u003eThe primary endpoint of this study was 28-day postoperative mortality, defined as any death occurring within 28 days after the surgical procedure. Secondary endpoints included in-hospital mortality and total duration of hospital stay. This systematic collection of clinical, demographic, and laboratory data was aimed at identifying risk factors associated with adverse postoperative outcomes, contributing to a better understanding of mortality and complication patterns in surgical patients.\u003c/p\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eBaseline characteristics between the survival and death groups within the MIMIC-IV were analyzed to describe and compare patient demographics and clinical profiles. Continuous variables were reported as means with standard deviations (SD) if they met normality assumptions; otherwise, they were expressed as medians with interquartile ranges (IQR). Comparisons of continuous variables between groups were performed using Student\u0026rsquo;s t-test for normally distributed data and the Wilcoxon rank-sum test for data not meeting normality assumptions. Categorical variables were summarized as frequencies and percentages, with group differences evaluated using the Chi-square test or Fisher\u0026rsquo;s exact test, as appropriate, based on expected counts.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eModel Development and Validation\u003c/h3\u003e\n\u003cp\u003eThe dataset was randomly partitioned into two subsets: a training set comprising 70% of the data and an internal validation set containing the remaining 30%. This division facilitated both the construction and evaluation of predictive models within a controlled framework. A sophisticated machine-learning model, XGBoost, was developed using the \"mlr3verse\" package in R. Categorical variables were pre-processed using one-hot encoding, while missing values were imputed with the \"imputehist\" method, ensuring consistent data handling. Given the low prevalence of the outcome (death), Synthetic Minority Over-sampling Technique (SMOTE) was employed to balance class representation in the training set.\u003c/p\u003e \u003cp\u003eTo optimize model performance, nested resampling was used with a 3-fold inner and outer cross-validation approach. Hyperparameters were fine-tuned using a random search strategy within the training set to ensure optimal predictive accuracy. A comprehensive XGBoost model, incorporating all available variables, was built to predict prognosis. Following model training, the ten variables with the highest feature importance were selected to create a simplified, interpretable model for comparison purposes(\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e). Partial dependence plots (PDPs) were generated to illustrate the marginal effects of explanatory variables on the outcome.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eComparative Model Analysis\u003c/h2\u003e \u003cp\u003eFor benchmarking, six additional models were developed using the training set: Random Forest, Elastic Net (Glmnet), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Logistic Regression, and LightGBM(\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). The same pre-processing steps (missing data imputation, encoding, and nested resampling) were consistently applied across these models(\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e), ensuring comparability. Model performance was evaluated on the internal validation set using AUC with corresponding 95% confidence intervals (CI) to assess discrimination ability. The optimal decision threshold for each model was determined using the Youden index.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eModel Interpretation and Survival Analysis\u003c/h3\u003e\n\u003cp\u003eDCA was performed to assess the clinical utility of the XGBoost models across a range of threshold probabilities, offering insight into the net benefit for clinical decision-making. Based on risk predictions from the XGBoost model, patients in the validation set were stratified into high-risk and low-risk groups. Survival analysis was conducted to compare outcomes between these groups, providing further validation of the model\u0026rsquo;s prognostic utility.\u003c/p\u003e\n\u003ch3\u003eSoftware and Statistical Significance\u003c/h3\u003e\n\u003cp\u003eAll statistical analyses were conducted using R software (version 4.1.2). Statistical significance was set at a p-value threshold of 0.05 for hypothesis tests.\u003c/p\u003e"},{"header":"RESULTS","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eBaseline Characteristics\u003c/h2\u003e \u003cp\u003eAs illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, the study cohort consisted of 2,335 critically ill patients who had undergone major surgical procedures, sourced from the MIMIC-IV database. These patients were divided into two sets: a training set comprising 1,634 patients and a testing set containing 701 patients. Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e provides a detailed summary of the baseline characteristics for patients categorized into survival and death outcome groups within the cohort. Patients in the death group displayed significantly elevated baseline values in several key physiological and clinical metrics, including heart rate, respiratory rate, SAPS-II and SOFA scores, CCI, BUN, and WBC count, while showing reduced levels of hemoglobin and eGFR, alongside prolonged PT and APTT. Additionally, patients in this group had a higher prevalence of complications such as sepsis, AKI, ARDS, CRRT, MV, and vasopressor administration (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05), indicating a more severe clinical presentation compared to the survival group.\u003c/p\u003e \u003cp\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\u003eCharacteristics of critically ill patients who underwent major surgery in the MIMIC-IV\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\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 \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCharacteristics\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOverall\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;2335)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSurvival\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;2202)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eDeath\u003c/p\u003e \u003cp\u003e(n\u0026thinsp;=\u0026thinsp;133)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eP\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGender\u0026thinsp;=\u0026thinsp;Female (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e977 (41.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e922 (41.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e55 (41.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.978\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\u003e64.31 [53.36, 73.68]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e64.02 [53.06, 73.44]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e68.75 [58.74, 77.92]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003eEthnicity (%)\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 \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.589\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAsian\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e73 ( 3.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e69 ( 3.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4 ( 3.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBlack/African American\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e203 ( 8.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e191 ( 8.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e12 ( 9.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHispanic/Lation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e77 ( 3.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e75 ( 3.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2 ( 1.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWhite\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1551 (66.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1466 (66.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e85 (63.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOther and unknown\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e431 (18.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e401 (18.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e30 (22.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAdmission type (%)\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 \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.149\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eObservation admit or elective\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e332 (14.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e318 (14.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e14 (10.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEmergency\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1218 (52.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1151 (52.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e67 (50.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUrgent\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e754 (32.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e702 (31.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e52 (39.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOthers\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e31 ( 1.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e31 ( 1.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0 ( 0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMarital status\u0026thinsp;=\u0026thinsp;Married (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1151 (52.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1088 (52.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e63 (53.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.916\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTemperature (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e36.78 [36.56, 37.06]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e36.78 [36.61, 37.06]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e36.72 [36.50, 37.14]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.370\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHeart rate (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e80.00 [69.00, 94.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e80.00 [69.00, 94.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e88.00 [74.00, 104.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003eMBP (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e80.00 [71.00, 91.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e81.00 [71.00, 91.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e72.00 [65.00, 84.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003eRespiratory rate (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e18.00 [16.00, 22.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e18.00 [16.00, 22.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e20.00 [16.00, 24.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSpO\u003csub\u003e2\u003c/sub\u003e (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e96.00 [94.00, 97.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e96.00 [94.00, 97.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e96.00 [94.00, 97.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.216\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGCS (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e15.00 [15.00, 15.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e15.00 [15.00, 15.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e15.00 [15.00, 15.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.142\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSOFA (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1.00 [1.00, 3.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1.00 [1.00, 3.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.00 [1.00, 5.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003eSAPS Ⅱ (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e29.00 [22.00, 37.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e29.00 [22.00, 36.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e37.50 [31.00, 48.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003eCharlson score (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e5.00 [3.00, 7.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e5.00 [3.00, 7.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e7.00 [5.00, 9.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003eVasopressor, n (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e348 (14.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e321 (14.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e27 (20.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.094\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMV, n (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1282 (54.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1181 (53.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e101 (75.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003eCRRT, n (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e36 ( 1.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e27 ( 1.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e9 ( 6.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003eAKI, n (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e209 ( 9.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e185 ( 8.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e24 (18.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003eComplication sepsis, n (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e299 (12.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e246 (11.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e53 (39.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003eComplication wound infection, n (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e55 ( 2.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e51 ( 2.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4 ( 3.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.829\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eComplication ARDS, n (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e237 (10.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e193 ( 8.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e44 (33.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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 during ICU, n (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1714 (73.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1602 (72.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e112 (84.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.005\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBUN (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e18.00 [13.00, 27.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e17.00 [12.00, 26.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e29.00 [18.00, 47.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003eeGFR (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e76.32 [52.23, 98.75]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e78.33 [54.10, 98.94]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e52.77 [31.67, 84.97]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003eSodium (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e139.00 [136.00, 141.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e139.00 [136.00, 141.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e137.00 [133.00, 140.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.007\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePotassium (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e4.10 [3.80, 4.40]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e4.10 [3.80, 4.40]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4.20 [3.70, 4.47]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.426\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChloride (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e103.00 [99.00, 106.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e103.00 [100.00, 106.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e102.00 [97.00, 105.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.056\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBicarbonate (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e24.00 [22.00, 27.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e24.00 [22.00, 27.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e23.00 [20.00, 25.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003eAnion Gap (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e14.00 [12.00, 16.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e14.00 [12.00, 16.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e15.00 [13.00, 18.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003eTotal calcium (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e8.70 [8.20, 9.10]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e8.70 [8.20, 9.10]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e8.50 [7.88, 9.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.012\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMagnesium (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2.00 [1.90, 2.20]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2.00 [1.90, 2.20]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2.10 [1.90, 2.30]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.058\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGlucose (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e117.00 [100.00, 146.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e117.00 [100.00, 145.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e123.00 [107.00, 161.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.039\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\u003e11.50 [9.67, 13.20]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e11.50 [9.70, 13.20]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e10.20 [8.90, 11.90]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003eWBC (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e9.40 [7.20, 12.90]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e9.35 [7.20, 12.70]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e11.80 [8.50, 15.90]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003ePlatelet (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e217.00 [164.00, 279.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e218.00 [166.00, 279.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e197.00 [139.00, 285.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.112\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePT (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e13.00 [11.80, 14.80]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e12.90 [11.80, 14.70]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e14.70 [13.10, 18.80]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003eAPTT (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e30.80 [26.90, 41.10]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e30.60 [26.80, 40.60]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e36.30 [29.90, 47.15]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\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\u003ePhosphate (median [IQR])\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.40 [2.90, 4.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.40 [2.90, 4.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.85 [3.00, 4.73]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"5\"\u003e\u003cem\u003eMBP, Mean Blood Pressure; SpO\u003c/em\u003e\u003csub\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sub\u003e, \u003cem\u003eperipheral oxygen saturation; GCS, Glasgow Coma Scale; SOFA, Sequential Organ Failure Assessment; SAPS Ⅱ, Simplified Acute Physiology Score Ⅱ; MV, Mechanical Ventilation; CRRT, ContinuousRenal Replacement Therapy; AKI, Acute Kidney Injury; ARDS, Acute Respiratory Distress Syndrome; BUN, Blood Urea Nitrogen; eGFR, estimated Glomerular Filtration Rate; WBC, White Blood Cell; PT, Prothrombin Time; APTT, Activated Partial Thromboplastin Time; IQR, Interquartile range.\u003c/em\u003e\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eComparison of Predictive Models\u003c/h2\u003e \u003cp\u003eThe performance metrics for each predictive model are displayed in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. Logistic Regression exhibited a baseline performance with an accuracy of 0.725 and an AUC of 0.794. Among the models, ensemble learning algorithms demonstrated superior predictive accuracy and AUC values compared to other approaches. Notably, the XGBoost model achieved the highest accuracy (0.806) and AUC (0.828), followed closely by lightGBM (accuracy: 0.795; AUC: 0.824) and the Random Forest Classifier (accuracy: 0.779; AUC: 0.819). The high discriminatory power of the XGBoost model for predicting mortality risk warranted further refinement of this model to optimize its clinical applicability.\u003c/p\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\u003eModel prediction performance\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\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 \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=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\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\u003eAUC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ecutoff\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSensitivity(%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSpecificity(%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAccuracy(%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003ePPV(%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eNPV(%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eXGboost select model-tain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.854(0.822\u0026ndash;0.886)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.209\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e92.5(87.1\u0026ndash;97.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e63.8(61.5\u0026ndash;66.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e65.4(63.2\u0026ndash;67.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e13.4(12.4\u0026ndash;14.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e99.3(98.8\u0026ndash;99.8)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eXGboost select model-test\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.824(0.762\u0026ndash;0.886)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.209\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e85(72.5\u0026ndash;95)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e65.8(62.2\u0026ndash;69.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e66.9(63.2\u0026ndash;70.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e13.1(11.1\u0026ndash;14.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e98.7(97.6\u0026ndash;99.5)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eXGboost-tain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.882(0.855\u0026ndash;0.909)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.347\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e81.7(74.2\u0026ndash;89.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e81.6(79.6\u0026ndash;83.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e81.6(79.7\u0026ndash;83.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e21.1(18.9\u0026ndash;23.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e98.7(98.1\u0026ndash;99.2)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eXGboost-test\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.828(0.769\u0026ndash;0.887)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.347\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e62.5(47.5\u0026ndash;77.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e81.7(78.8\u0026ndash;84.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e80.6(77.6\u0026ndash;83.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e17.1(13.1\u0026ndash;21.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e97.3(96.2\u0026ndash;98.4)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLightGBM-tain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.871(0.839\u0026ndash;0.902)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.372\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e79.6(71-87.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e81.2(79.2\u0026ndash;83.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e81.1(79.2\u0026ndash;83)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e20.3(17.9\u0026ndash;22.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e98.5(97.9\u0026ndash;99.1)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLightGBM-test\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.824(0.767\u0026ndash;0.881)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.372\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e65(50\u0026ndash;80)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e80.3(77.3\u0026ndash;83.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e79.5(76.3\u0026ndash;82.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e16.7(12.9\u0026ndash;20.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e97.4(96.4\u0026ndash;98.5)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRandomForest-tain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.871(0.84\u0026ndash;0.902)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.462\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e77.4(68.8\u0026ndash;86)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e79.4(77.3\u0026ndash;81.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e79.3(77.3\u0026ndash;81.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e18.5(16.3\u0026ndash;20.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e98.3(97.7\u0026ndash;98.9)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRandomForest-test\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.819(0.754\u0026ndash;0.883)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.462\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e65(50\u0026ndash;80)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e78.7(75.3\u0026ndash;81.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e77.9(74.7\u0026ndash;80.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e15.5(12-19.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e97.4(96.3\u0026ndash;98.4)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSVM-tain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.997(0.996\u0026ndash;0.999)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e100(100\u0026ndash;100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e96.8(95.9\u0026ndash;97.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e97(96.1\u0026ndash;97.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e65.5(59.6\u0026ndash;72.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e100(100\u0026ndash;100)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSVM-test\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.719(0.66\u0026ndash;0.778)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.005\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e62.5(47.5\u0026ndash;77.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e71.1(67.6\u0026ndash;74.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e70.6(67.3\u0026ndash;73.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e11.6(8.9\u0026ndash;14.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e96.9(95.7\u0026ndash;98.1)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKNN-tain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.999(0.997-1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.671\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e98.9(96.8\u0026ndash;100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e99.8(99.5\u0026ndash;100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e99.8(99.5\u0026ndash;99.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e96.8(92.9\u0026ndash;100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e99.9(99.8\u0026ndash;100)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKNN-test\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.645(0.56\u0026ndash;0.73)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.671\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e20(10-32.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e92.7(90.6\u0026ndash;94.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e88.6(86.6\u0026ndash;90.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e14.3(7-23.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e95(94.4\u0026ndash;95.8)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLogistic regression-tain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.847(0.815\u0026ndash;0.88)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.42\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e84.9(77.4\u0026ndash;91.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e73.5(71.3\u0026ndash;75.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e74.1(72-76.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e16.2(14.6\u0026ndash;17.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e98.8(98.2\u0026ndash;99.3)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLogistic regression-test\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.794(0.728\u0026ndash;0.86)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.42\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e65(50\u0026ndash;80)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e73.1(69.7\u0026ndash;76.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e72.5(69.3\u0026ndash;75.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e12.8(9.9\u0026ndash;15.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e97.2(96-98.4)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGlmnet-tain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.793(0.755\u0026ndash;0.831)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.484\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e84.9(77.4\u0026ndash;91.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e62.9(60.5\u0026ndash;65.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e64.2(61.9\u0026ndash;66.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e12.2(11-13.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e98.6(97.9\u0026ndash;99.2)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGlmnet-test\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.792(0.722\u0026ndash;0.862)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.484\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e85(75\u0026ndash;95)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e61.6(57.8\u0026ndash;65.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e62.9(59.3\u0026ndash;66.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e11.8(10.1\u0026ndash;13.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e98.5(97.5\u0026ndash;99.5)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"8\"\u003e\u003cem\u003eXGboost, eXtremely Gradient Boosting; SVM, Support Vector Machine; KNN, K-Nearest Neighbor; AUC, the Receiver Operating Characteristic Curve; PPV, Positive Predictive Value; NPV, Negative Predictive Value.\u003c/em\u003e\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eDevelopment and Optimization of the XGBoost Model\u003c/h2\u003e \u003cp\u003eFeature importance and partial dependence profiles for the XGBoost model are displayed in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. The Charlson score emerged as the most influential predictor, with higher scores correlating with an increased risk of mortality. Using SHapley Additive exPlanations (SHAP) values, the top ten predictive features\u0026mdash;comprising CCI, SOFA score, SAPS-II score, need for MV, presence of ARDS and sepsis, BUN levels, eGFR, respiratory rate, and marital status\u0026mdash;were identified and used to develop a streamlined version of the XGBoost model. Although this optimized model demonstrated a slightly lower AUC of 0.824 (95% CI: 0.762\u0026ndash;0.886) compared to the full model, it was considered to offer enhanced clinical feasibility due to its reduced complexity.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eComparative Analysis of Model Performance\u003c/h2\u003e \u003cp\u003eIn Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, the AUC-ROC curve comparisons between the XGBoost model and alternative models underscore its superior discriminatory ability. To further evaluate the clinical utility of the XGBoost model, a Decision Curve Analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e) was conducted. The analysis indicated that within a threshold probability range of 1\u0026ndash;48%, utilizing the XGBoost model to predict patient prognosis would yield a net benefit compared to the strategies of either screening all patients or screening none. For clarity, only the results for the full XGBoost model and the streamlined XGBoost selection model are displayed. Sensitivity and specificity evaluations for these predictive models in the internal validation set are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eRisk Stratification and Survival Analysis\u003c/h2\u003e \u003cp\u003ePatients were stratified into high-risk and low-risk categories based on their calculated mortality risk scores, using the median risk score as the cutoff value. Kaplan-Meier survival curves demonstrated a clear separation between high- and low-risk groups in both the XGBoost selection model (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA, log-rank test, p\u0026thinsp;=\u0026thinsp;0.0001) and the full XGBoost model (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB, log-rank test, p\u0026thinsp;=\u0026thinsp;0.0005), highlighting the models\u0026rsquo; effectiveness in distinguishing patient outcomes.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003eThis study is the first to utilize machine-learning models to predict the prognosis of critically ill patients after major surgery. By developing and validating two variations of a dynamic machine-learning model, this research facilitates the identification of high-risk patients, thus providing healthcare decision-makers with actionable clinical insights.\u003c/p\u003e \u003cp\u003eAmong the models evaluated, the XGBoost algorithm achieved the highest AUC, reflecting its superior predictive accuracy. XGBoost is recognized for its efficiency in handling missing data and its ability to combine weak learners into a robust predictive ensemble (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e). This method has gained widespread credibility and is frequently employed in data science competitions. For instance, in 2015, XGBoost was used in 17 of the 29 winning solutions published on Kaggle\u0026rsquo;s blog, and the top 10 winning teams in the 2015 KDD Cup utilized XGBoost (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTypically, models incorporating a broader range of variables exhibit enhanced discriminatory ability but may suffer from reduced clinical feasibility. In light of this trade-off, our study developed two model variants, each tailored to distinct clinical settings. The full model, utilizing 58 clinical variables, achieved the highest AUC in this study and demonstrated strong prognostic capabilities. However, the extensive data requirements limit its applicability to institutions with comprehensive and structured clinical data systems. Conversely, the streamlined model, trained on only 10 selected variables, offers a pragmatic balance between accuracy and usability, making it viable for settings with more limited resources.\u003c/p\u003e \u003cp\u003eInterpretation of the full model highlighted that multiple clinical variables are integral to assessing mortality risk. Notably, CCI emerged as the most critical predictor, followed closely by complications associated with ARDS. The predictive value of CCI has been extensively validated across diverse medical conditions, with higher CCI scores correlating strongly with increased mortality rates (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e). Furthermore, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, patients with elevated SOFA scores demonstrated a higher risk of death, consistent with previous studies. Early monitoring of organ dysfunction within the initial days in the ICU provides a critical indicator of patient prognosis. Specifically, an increase in SOFA score within the first 48 hours of ICU admission has been associated with a predicted mortality rate of at least 50% (\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAs anticipated, patients with ARDS and sepsis demonstrated poorer prognoses. Additional factors such as mechanical ventilation and SAPS-II scores were useful in assessing mortality risk. Patients with reduced renal function, indicated by elevated serum creatinine and urea nitrogen, had a higher likelihood of death. Interestingly, we found that marital status is a predictor of death. Studies across Chinese, U.S., and international cohorts have shown that married patients with cardiovascular disease and various cancers tend to have better outcomesr(\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e). Psychosocial and socioeconomic factors, along with acute stressors, may partly explain the association between marital status and patient prognosis, though the specific underlying mechanisms remain largely unexplored.\u003c/p\u003e \u003cp\u003eOur model demonstrates substantial potential as a tool to improve the prognosis for critically ill patients following major surgery, offering both clinical and economic advantages. Implementing this model could facilitate a more tailored approach to patient care. For patients identified as high-risk, delayed transfer from intensive care and enhanced monitoring may reduce mortality rates. This approach is particularly valuable given the scarcity of critical care beds, as decisions regarding patient discharge or transfer directly affect the efficiency of ICU resource utilization (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e). Conversely, for patients assessed as low-risk, unnecessary medical expenses associated with additional testing and treatment may be avoided. The clinical utility of this predictive model will be further assessed in future prospective studies to ensure its effectiveness and practicality.\u003c/p\u003e \u003cp\u003eSeveral limitations of this study should be acknowledged. First, although patients from various surgical procedures were included, the prognosis may differ significantly depending on the specific type of surgery, indicating a need for future subgroup analyses. Second, as a retrospective study, this research may be affected by missing data or input errors, even though the MIMIC-IV dataset is known for its high quality. Finally, while the model has shown promise, external validation using other datasets and prospective studies are essential to confirm its accuracy and generalizability in diverse clinical settings.\u003c/p\u003e"},{"header":"CONCLUSIONS","content":"\u003cp\u003eIn summary, this study successfully developed two versions of an XGBoost model that outperformed alternative predictive models in forecasting outcomes for ICU patients after major surgery. These models offer potential as decision-support tools, enhancing patient care and resource allocation in critical care environments.\u003c/p\u003e "},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eClinical trial number: not applicable\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe are grateful to all those who took part in or assisted with this study project. And we thank the Massachusetts Institute of Technology and the Beth Israel Deaconess Medical Center for the MIMIC project.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics declarations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHuman ethics and consent to participate\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003eConsent to participate\u003c/p\u003e\n\u003cp\u003eEach author consents to participate in the manuscript.\u003c/p\u003e\n\u003cp\u003eCompeting interests\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor information\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAuthors and Affiliations\u003c/p\u003e\n\u003cp\u003eDepartment of Intensive Care Medicine, Shanghai Sixth People’s Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China\u003c/p\u003e\n\u003cp\u003eXian Zhang\u003c/p\u003e\n\u003cp\u003eDepartment of Intensive Care Medicine, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China\u003c/p\u003e\n\u003cp\u003eTao Wang\u003c/p\u003e\n\u003cp\u003eScience and Technology Department, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China\u003c/p\u003e\n\u003cp\u003eZhe Chen\u003c/p\u003e\n\u003cp\u003eDepartment of Critical Care Medicine, The 960th Hospital of the PLA Joint Logistics Support Force, Shandong, China\u003c/p\u003e\n\u003cp\u003eShuliu Zhang\u003c/p\u003e\n\u003cp\u003eDepartment of Critical Care Medicine, Seventh People's Hospital of Shanghai University of Traditional Chinese Medicine, Shanghai, China\u003c/p\u003e\n\u003cp\u003eJihong Yuan\u003c/p\u003e\n\u003cp\u003eDepartment of Military Health Statistics, Naval Medical University, Shanghai, China\u003c/p\u003e\n\u003cp\u003eYingYi Qin\u003c/p\u003e\n\u003cp\u003eDepartment of Critical Care Medicine, The 960th Hospital of the PLA Joint Logistics Support Force, Shandong, China\u003c/p\u003e\n\u003cp\u003eYunliang Cui\u003c/p\u003e\n\u003cp\u003eContributions\u003c/p\u003e\n\u003cp\u003eContributions: (I) Conception and design: Xian Zhang, Yunliang Cui; (II) Administrative support: Jihong Yuan,\u0026nbsp;YingYi Qin, Yunliang Cui; (III) Provision of study materials : Xian Zhang, Tao Wang; (IV) Collection and assembly of data: Zhe Chen, Shuliu Zhang; (V) Data analysis and interpretation: Xian Zhang, YingYi Qin; (VI) Manuscript writing: All authors; (VII) Final approval of manuscript: All authors.\u003c/p\u003e\n\u003cp\u003eCorresponding author\u003c/p\u003e\n\u003cp\u003eCorrespondence: Yunliang Cui, \u0026nbsp;YingYi Qin, \u0026nbsp;Jihong Yuan.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe original dataset generated during the current study is available from the corresponding author on reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eMeara JG, Leather AJ, Hagander L, Alkire BC, Alonso N, Ameh EA, et al. Global Surgery 2030: evidence and solutions for achieving health, welfare, and economic development. Lancet. 2015;386:569\u0026ndash;624. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S0140-6736(15)60160-X\u003c/span\u003e\u003cspan address=\"10.1016/S0140-6736(15)60160-X\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLevine GN, O\u0026rsquo;Gara PT, Beckman JA, Al-Khatib SM, Birtcher KK, Cigarroa JE, et al. Recent Innovations,Modifications, and Evolution of ACC/AHA Clinical Practice Guidelines: An Update for Our Constituencies: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. Circulation. 2019;139:e879\u0026ndash;86. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1161/CIR.0000000000000651\u003c/span\u003e\u003cspan address=\"10.1161/CIR.0000000000000651\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGhaferi AA, Birkmeyer JD, Dimick JB. Complications, failure to rescue, and mortality with major inpatient surgery in medicare patients. Ann Surg., Biccard BM, Madiba TE, Kluyts HL, Munlemvo DM, Madzimbamuto FD, Basenero A et al. Perioperative patient outcomes in the African Surgical Outcomes Study: a 7-day prospective observational cohort study. Lancet. 2018; 391: 1589\u0026ndash;1598. doi:10.1016/S0140-6736(18)30001-1.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeara JG, Leather AJ, Hagander L, Alkire BC, Alonso N, Ameh EA, et al. Global Surgery 2030: evidence and solutions for achieving health, welfare, and economic development. Lancet. 2015;386(9993):569\u0026ndash;624. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S0140-6736(15)60160-X\u003c/span\u003e\u003cspan address=\"10.1016/S0140-6736(15)60160-X\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBickler SN et al. Global Burden of Surgical Conditions. Essential Surgery: Disease Control Priorities. (ed. Debas, H.T.) Chap. 2 (The International Bank for Reconstruction and Development/The World Bank, 2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePiccirillo JF, Vlahiotis A, Barrett LB, Flood KL, Spitznagel EL, Steyerberg EW, et al. The changing prevalence of comorbidity across the age spectrum. Crit Rev Oncol/Hematol. 2008;67(2):124\u0026ndash;32. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.critrevonc.2008.01.013\u003c/span\u003e\u003cspan address=\"10.1016/j.critrevonc.2008.01.013\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLees N, Peden CJ, Dhesi J, Quiney N, Lockwood S, Symons NR et al. The High-Risk General Surgical Patient: Raising the Standard. Updated recommendations on the Perioperative Care of the High-Risk General Surgical Patient [Internet]. 2018 [cited 2020 Jun 22].\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao QY, Liu LP, Luo JC, Luo YW, Luo Z. A machine-learning approach for dynamic prediction of sepsis-induced coagulopathy in critically ill patients with sepsis. Front Med. 2020;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fmed.2020.637434\u003c/span\u003e\u003cspan address=\"10.3389/fmed.2020.637434\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBeam AL, Kohane IS. Big data and machine learning in health care. JAMA. 2018;319:1317\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1001/jama.201718391\u003c/span\u003e\u003cspan address=\"10.1001/jama.201718391\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLuo JC, Zhao QY, Tu GW. Clinical prediction models in the precisionmedicine era: old and new algorithms. Ann Transl Med. 2020;8:274. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.21037/atm.2020.02.63\u003c/span\u003e\u003cspan address=\"10.21037/atm.2020.02.63\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGe H, Pan Q, Zhou Y, Xu P, Zhang L, Zhang J, et al. Lungmechanics of mechanically ventilated patients with COVID-19: analytics with high-granularity ventilator waveform data. Front Med. 2020;7:541. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fmed.202000541\u003c/span\u003e\u003cspan address=\"10.3389/fmed.202000541\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCollins GS, Reitsma JB, Altman DG, Moons KG. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD Statement. BMC Med. 2015;13:1. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12916-014-0241-z\u003c/span\u003e\u003cspan address=\"10.1186/s12916-014-0241-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang ZH. Variable selection with stepwise and best subset approaches. Ann Transl Med. 2016;4:136. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.21037/atm.2016.03.35\u003c/span\u003e\u003cspan address=\"10.21037/atm.2016.03.35\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLuo JC, Zhao QY, Tu GW. Clinical prediction models in the precision medicine era: old and new algorithms. Ann Transl Med. 2020;8:274. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.21037/atm.2020.02.63\u003c/span\u003e\u003cspan address=\"10.21037/atm.2020.02.63\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLundberg SM, Erion G, Chen H, DeGrave A, Prutkin JM, Nair B, et al. From local explanations to global understanding with explainable AI for trees. Nat Mach Intell. 2020;2:56\u0026ndash;67. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s42256-019-0138-9\u003c/span\u003e\u003cspan address=\"10.1038/s42256-019-0138-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYuan KC, Tsai LW, Lee KH, Cheng YW, Chen RJ, et al. The development an artifcial intelligence algorithm for early sepsis diagnosis in the intensive care unit. Int J Med Inf. 2020;141:104176. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ijmedinf.2020.104176\u003c/span\u003e\u003cspan address=\"10.1016/j.ijmedinf.2020.104176\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen T, Guestrin C. XGBoost: a scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining-KDD 2016, San Francisco, CA, USA; 2016. pp. 785\u0026ndash;94.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHou N, Li M, He L, Xie B, Wang K, et al. Predicting 30-days mortality for MIMIC-III patients with sepsis-3: a machine learning approach using XGboost. J Translational Med. 2020;18(1). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12967-020-02620-5\u003c/span\u003e\u003cspan address=\"10.1186/s12967-020-02620-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMary EC, Danilo C, Jenny G, Chiara P. Charlson Comorbidity Index: A Critical Review of Clinimetric Properties. Psychother Psychosom. 2022;91(1):8\u0026ndash;35. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1159/000521288\u003c/span\u003e\u003cspan address=\"10.1159/000521288\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFerreira LF. Serial Evaluation of the SOFA Score to Predict Outcome in Critically Ill Patients. JAMA. 2001;286(14):1754\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1001/jama.286.14.1754\u003c/span\u003e\u003cspan address=\"10.1001/jama.286.14.1754\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDhindsa DS, Khambhati J, Schultz WM, Tahhan AS, Quyyumi AA. Marital status and outcomes in patients with cardiovascular disease.rends. Cardiovasc Med. 2020;30(4):215\u0026ndash;20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.tcm.2019.05.012\u003c/span\u003e\u003cspan address=\"10.1016/j.tcm.2019.05.012\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiang Y, Wu X, Lu C, Xiao F. Impact of marital status on the prognosis of liver cancer patients without surgery and the critical window. Annals Palliat Med. 2021;10(3):2990\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.21037/apm-20-1885\u003c/span\u003e\u003cspan address=\"10.21037/apm-20-1885\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuo Z, Gu C, Li S, Gan S, Li Y, et al. Association between Marital Status and Prognosis in Patients with Prostate Cancer: A Meta-Analysis of Observational Studies. Urol J. 2020;18(4):371\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.22037/uj.v16i7.6197\u003c/span\u003e\u003cspan address=\"10.22037/uj.v16i7.6197\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCypress BS. Transfer out of intensive care: An evidence-based literature review. Dimensions Crit care nursing: DCCN. 2013;32(5):244\u0026ndash;61. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1097/DCC.0b013e3182a07646\u003c/span\u003e\u003cspan address=\"10.1097/DCC.0b013e3182a07646\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-informatics-and-decision-making","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"midm","sideBox":"Learn more about [BMC Medical Informatics and Decision Making](http://bmcmedinformdecismak.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/midm/default.aspx","title":"BMC Medical Informatics and Decision Making","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Major surgery, Dynamic prediction, Machine learning, eXtremely Gradient Boosting","lastPublishedDoi":"10.21203/rs.3.rs-5411439/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5411439/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eMajor surgery can result in elevated mortality rates, poorer prognoses, and extended hospital stays. This study sought to develop and validate an effective machine-learning model capable of accurately forecasting outcomes in critically ill patients who have undergone major surgery.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eUsing the publicly accessible Medical Information Mart for Intensive Care (MIMIC)-IV database, we developed and validated multiple machine-learning models to predict postoperative outcomes in critically ill patients who had at least an overnight ICU stay after major surgery. Seven predictive models were tested to forecast prognosis, with the highest-performing model selected based on its accuracy and the area under the receiver operating characteristic curve (AUC). An advanced model, eXtremely Gradient Boosting (XGBoost), was created using all variables, followed by a streamlined model built from 10 features chosen for their importance and clinical applicability. The performance of both models was assessed using Decision Curve Analysis (DCA), while survival analyses distinguished high- and low-risk groups within the internal validation sets.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eA cohort of 2,335 critically ill patients who had undergone major surgery were included in the MIMIC-IV cohorts. The full XGBoost model achieved an accuracy of 80.6% and an AUC of 0.828, indicating high predictive power. A more practical selection model with 10 features demonstrated a slightly lower AUC of 0.824 (95% CI: 0.762\u0026ndash;0.886) but offered advantages in clinical usability. The ten key features were identified based on their SHapley Additive exPlanations (SHAP) values, which included the Charlson Comorbidity Index (CCI), Simplified Acute Physiology Score (SAPS)-II, Sequential Organ Failure Assessment (SOFA) score, mechanical ventilation, ARDS and sepsis complications, blood urea nitrogen (BUN) levels, estimated glomerular filtration rate (eGFR), respiratory rate, and marital status. Additionally, survival curves showed a clear distinction between high- and low-risk groups based on predictions from both the full and selection XGBoost models.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eThis study developed two XGBoost model variants that outperform other ICU predictive methods for forecasting the prognosis of patients undergoing major surgery. These models have the potential to assist healthcare providers in making more informed decisions, thereby improving clinical outcomes in the ICU setting.\u003c/p\u003e","manuscriptTitle":"Development and Validation of a Machine-Learning Model for Predicting Prognosis in Critically Ill Patients Undergoing Major Surgery","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-12-17 09:48:34","doi":"10.21203/rs.3.rs-5411439/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2024-11-18T08:30:31+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-11-13T05:54:34+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-11-13T05:53:22+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Informatics and Decision Making","date":"2024-11-07T16:27:04+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-informatics-and-decision-making","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"midm","sideBox":"Learn more about [BMC Medical Informatics and Decision Making](http://bmcmedinformdecismak.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/midm/default.aspx","title":"BMC Medical Informatics and Decision Making","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"ae0b04b3-94ac-4bb8-a978-0c41b60920a8","owner":[],"postedDate":"December 17th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-11-03T16:00:36+00:00","versionOfRecord":{"articleIdentity":"rs-5411439","link":"https://doi.org/10.1186/s12911-025-03240-z","journal":{"identity":"bmc-medical-informatics-and-decision-making","isVorOnly":false,"title":"BMC Medical Informatics and Decision Making"},"publishedOn":"2025-10-29 15:57:26","publishedOnDateReadable":"October 29th, 2025"},"versionCreatedAt":"2024-12-17 09:48:34","video":"","vorDoi":"10.1186/s12911-025-03240-z","vorDoiUrl":"https://doi.org/10.1186/s12911-025-03240-z","workflowStages":[]},"version":"v1","identity":"rs-5411439","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5411439","identity":"rs-5411439","version":["v1"]},"buildId":"WrCJVZZCHTDjtuVLN7oU0","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}