A Clinically Interpretable Machine Learning Framework for Mortality Prediction in Critically Ill Orthopedic Patients

A Clinically Interpretable Machine Learning Framework for Mortality Prediction in Critically Ill Orthopedic Patients | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article A Clinically Interpretable Machine Learning Framework for Mortality Prediction in Critically Ill Orthopedic Patients Ting Zhang, MeiLing Li, Hong Lu, Wei Liu, Zhen Tan This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8620797/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 12 You are reading this latest preprint version Abstract Background The global burden of geriatric orthopedic conditions (e.g., hip fractures, complex trauma) is rising, posing significant challenges to critical care medicine. Existing prognostic tools rely on single indicators or traditional scoring systems, lacking sufficient accuracy and interpretability in ICUs. This study aims to develop an interpretable machine learning (ML) framework with high predictive performance for individualized risk stratification. Methods Critically ill orthopedic patients ≥ 50 from MIMIC-IV/III were enrolled (excluding those with malignant tumors or incomplete data). MIMIC-IV was split 7:3 into training/testing sets. LASSO Cox regression with 20-fold cross-validation enabled dimension reduction and feature selection. Nine ML models were compared; the optimal model was selected by accuracy and AUROC. SHAP quantified feature impacts and individual decision processes. MIMIC-III served for external validation. Results 6,488 patients were included, with 11.8% in-hospital mortality in MIMIC-IV. Eight core features (age, APSIII, SOFA, blood glucose, WBC, lactate, body temperature, CRRT) were identified. Logistic regression performed best (AUROC = 0.82) and achieved AUC = 0.81 (95% CI: 0.78–0.83) in external validation. Conclusion This interpretable mortality prediction model for critically ill orthopedic patients aids preoperative risk assessment and postoperative ICU monitoring, supporting targeted early interventions to improve outcomes. Health sciences/Biomarkers Health sciences/Diseases Health sciences/Health care Health sciences/Medical research Health sciences/Risk factors Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction The rising global prevalence of orthopedic conditions and injuries among the elderly presents a substantial public health challenge [ 1 ], while also holding significant clinical importance in critical care medicine—particularly for cases such as hip fractures, osteoporotic fractures, and complex multiple traumas in this population [ 2 , 3 ]. Such patients often require post-operative or acute-phase management in the ICU. These conditions are not only surgical emergencies but also represent critical states involving multisystem dysfunction, the outcomes of their management directly impact patient length of hospitalization, long-term quality of life and in-hospital mortality [ 4 – 6 ]. With the intensification of population aging, the number of those orthopedic patients continues to rise [ 7 ]. In-hospital mortality among orthopedic patients is a key indicator for measuring the quality of medical care and perioperative safety management [ 8 ]. Although orthopedic surgeries are predominantly elective, and the overall prognosis for patients is favorable, the mortality risk in certain patient populations—such as with multiple traumas, complex joint revisions, or severe comorbidities—cannot be overlooked [ 9 ]. Mortality events often result from postoperative critical conditions such as deep vein thrombosis, pulmonary embolism, cardiopulmonary complications, and severe infections [ 5 ]. These events tend to be insidious and sudden, posing significant challenges for early identification and intervention. However, current research on the prognosis of orthopedic patients primarily focuses on single clinical indicators (such as age, race, or the number of comorbidities) and lacks a multidimensional integrated model [ 10 ]. Traditional risk scoring systems (e.g.,Charlson Comorbidity Index) exhibit limited predictive performance for orthopedic patients in the ICU, making it difficult to meet the clinical demand for individualized risk stratification and early intervention [ 11 ]. Therefore, developing a ML framework that combines high predictive accuracy with strong clinical interpretability is of urgent significance for improving the level of perioperative safety management for orthopedic patients. An ideal framework should not only accurately identify high-risk patients but also clearly elucidate the driving factors behind the risks—for example, which key features (such as combined abnormalities in specific laboratory indicators, interactions between age and comorbidities, specific thresholds for surgery duration, etc.) collectively contribute to the high-risk assessment. Accordingly, this study aims to develop and validate a clinically interpretable machine learning framework for predicting in-hospital mortality in orthopedic patients. By systematically integrating multidimensional clinical data and employing explainable artificial intelligence techniques—with a particular focus on SHAP—we seek to achieve a balance between model predictive accuracy and clinical transparency. The framework is designed not only to generate individualized risk assessments but also to provide clinicians with interpretable, evidence-based insights into the key factors driving each prediction. We envision this tool as both an early warning system for optimizing clinical resource allocation and a transparent decision-support companion, thereby advancing perioperative orthopedic care toward more precise, intelligent, and patient-centered outcomes. Results 1. The characteristics of included participants Between 2008 and 2019, a total of 5,304 patients who underwent orthopedic-related diseases and orthopedic surgeries were identified from the MIMIC database. After excluding 534 ineligible patients, 4,770 patients were ultimately included in the analysis (Figure 1). Then, participants were randomly divided into the training set and the testing set in a ratio of 7:3 (3,339 vs 1,431). Table 1 demonstrated the characteristics of patients. The median age of patients was 67(52,80) years old, 2,508 (53%) were female patients, 3,042 (64%) had hypertension, 3,801(80%) had type 2 diabetes. The median length of ICU stay was 2.81(1.73, 5.66) days. 4,209(88.2%) patients were discharged from ICU, 561(11.8%) died in the ICU. The median SOFA score was 4(2,6), the median APSIII score was 40.00 (30.00,52.00), the median GCS score was 15.00(13.00, 15.00). Compared with survivors, non-survivors had significantly elevated age, glucose, lactate, ICU length of stay, organ failure and severity scores (SOFA, APSIII, SAPSII, OASIS), inflammatory and coagulation markers (WBC, INR, PT), and renal function parameters (creatinine, BUN). Conversely, body weight, SBP, DBP, MAP, temperature, PO2, PH, hemoglobin, and red blood cell count were significantly reduced. All differences were statistically significant (P < 0.05). Table 1. Baseline characteristics of included patients. Variables Total (n = 4770) Survivor (n = 4209) Non-survivor (n = 561) P value Age (years) 67 (52, 80) 65 (50, 79) 76 (64, 87) < 0.001 Weight (kg) 74.2 (62.2, 89.6) 74.6 (62.5, 90) 72.4 (60, 85) < 0.001 ICU Day (days) 2.81 (1.73, 5.66) 2.73 (1.69, 5.3) 3.85 (2.1, 7.89) < 0.001 Temperature(℃) 36.8 (36.5, 37.1) 36.8 (36.5, 37.2) 36.6 (36.4, 37) < 0.001 Gender, n (%) Male Female 2262 (47) 2508 (53) 1988 (47) 2221 (53) 274 (49) 287 (51) 0.502 Hypertension, n (%) Yes No 3042 (64) 1728 (36) 2677 (64) 1532 (36) 365 (65) 196 (35) 0.529 T2DM, n (%) Yes No 3801 (80) 969 (20) 3387 (80) 822 (20) 414 (74) 147 (26) < 0.001 AKI, n (%) Yes No 3555 (75) 1215 (25) 3268 (78) 941 (22) 287 (51) 274 (49) < 0.001 LC, n (%) Yes No 4503 (94) 267 (6) 4001 (95) 208 (5) 502 (89) 59 (11) < 0.001 CVA, n (%) Yes No 4424 (93) 346 (7) 3914 (93) 295 (7) 510 (91) 51 (9) 0.089 CRRT, n (%) Yes No 4630 (97) 140 (3) 4131 (98) 78 (2) 499 (89) 62 (11) < 0.001 CABG, n (%) Yes No 4712 (99) 58 (1) 4154 (99) 55 (1) 558 (99) 3 (1) 0.173 HR (times/minute) 88 (75, 103) 88 (75, 102) 90 (75, 105) 0.159 SBP (mmHg) 123 (108, 140) 123 (109, 140) 118 (101, 136) < 0.001 DBP (mmHg) 69 (58, 81) 69 (58, 81) 66 (54, 79) < 0.001 MAP (mmHg) 83 (71, 95) 83 (72, 95) 79 (68, 94) < 0.001 SOFA 4 (2, 6) 3 (2, 5) 7 (4, 10) < 0.001 APSIII 40 (30, 52) 39 (29, 49) 56 (43, 74) < 0.001 SAPSII 33 (25, 42) 32 (24, 41) 46 (36, 56) < 0.001 OASIS 32 (26, 37) 31 (26, 36) 38 (32, 43) < 0.001 GCS 15 (13, 15) 15 (13, 15) 15 (13, 15) 0.409 Laboratory tests WBC (m/uL) 11.2 (8.3, 15.1) 11.1 (8.2, 14.8) 12.3 (8.9, 17.3) < 0.001 RBC (10¹²/L) 3.55 (3.03, 4.11) 3.57 (3.05, 4.11) 3.44 (2.87, 4.13) 0.002 Platelet (K/uL) 192 (144, 251) 195 (148, 253) 173 (123, 232) < 0.001 Hemoglobin (g/dL) 10.7 (9.1, 12.4) 10.8 (9.2, 12.4) 10.3 (8.7, 12.3) < 0.001 INR 1.2 (1.1, 1.4) 1.2 (1.1, 1.4) 1.3 (1.1, 1.7) < 0.001 PT(s) 13.2 (12, 15.2) 13.1 (12, 14.9) 14.3(12.7, 18.7) < 0.001 Creatinine (mg/dL) 0.9 (0.7, 1.2) 0.9 (0.7, 1.2) 1.2 (0.8, 1.7) < 0.001 BUN (mg/dL) 17 (12, 26) 16 (12, 25) 24 (17, 37) < 0.001 PH 7.37 (7.31, 7.42) 7.37 (7.31, 7.42) 7.34 (7.26, 7.41) < 0.001 PO2 (mmHg) 107 (60, 190) 109 (61, 192) 91 (50, 175) < 0.001 PCO2 (mmHg) 42 (37, 48) 42 (37, 48) 42 (35, 49) 0.642 Lactate(mmol/L) 1.7 (1.2, 2.5) 1.6 (1.1, 2.4) 2 (1.4, 3.6) < 0.001 Glucose (mmol/L) 7.2(6, 9) 7.1 (5.9, 8.8) 8.1 (6.3, 10.8) < 0.001 Note: T2DM: Type 2 Diabetes Mellitus; AKI: Acute Kidney Injury; LC: Lung Cancer; CVA: Cerebrovascular Accident; CRRT: Continuous Renal Replacement Therapy; CABG: Coronary Artery Bypass Grafting; HR: Heart Rate; SBP: Systolic Blood Pressure; DBP: Diastolic Blood Pressure; MAP: Mean Arterial Pressure; SOFA: Sequential Organ Failure Assessment; APSIII: Acute Physiology Score III; SAPSII: Simplified Acute Physiology Score II; OASIS: Oxford Acute Severity of Illness Score; GCS: Glasgow Coma Scale; WBC: White Blood Cell; RBC: Red Blood Cell; INR: International Normalized Ratio; PT: Prothrombin Time; BUN: Blood Urea Nitrogen; PH: Potential of Hydrogen; PO2: Partial Pressure of Dioxide; PCO2: Partial Pressure of Carbon Dioxide; 2. Selection of Predictor Variables Feature variables for in-hospital mortality in orthopedic critical illness were identified using the LASSO regression algorithm, as shown in Figure 2. Specifically, Figure 2A presents the predictors selected by LASSO (shrinkage parameter, λ = 0.01890518; lambda.1se). Eventually, the coefficients of all predictors shrink to zero, and the curve reaching zero indicates that the corresponding feature has been eliminated by LASSO regression. The LASSO regression adopted 20-fold cross-validation based on the minimum criterion to select the optimal parameters, and plotted the curve of the relationship between binomial deviance and lambda, with a total of 10 feature variables screened out (Figure 2B). The Boruta algorithm identified 29 feature variables, while the Recursive Feature Elimination (RFE) method selected 13 feature variables (Figure 3). A Venn diagram was used to extract the intersection of the results from the three algorithms, yielding 8 important feature variables as follows: Age, APSIII, SOFA, Glucose, WBC, Lactate, Temperature, and CRRT (Figure 4). 3. Model performance After the key variables were identified, 9 ML models were constructed to predict the mortality risk of inpatients with orthopedic critical illness. The AUROC was used to analyze the predictive efficacy of these 9 ML models for the mortality of inpatients with orthopedic critical illness. The 9 algorithms were SVM, KNN, GBM, NN, Logistics, AdaBoost, XGBoost, LightGBM, and CatBoost, respectively. As shown in Table 2, in the training set, the KNN model achieved an AUROC of 1.00, indicating severe overfitting. The remaining models all demonstrated satisfactory performance. In the test group, an identical peak AUC of 82% was attained by the SVM, NN, and logistic regression models; meanwhile, the neural network and logistic regression models jointly achieved the highest sensitivity of 89%. Moreover, the Logistic regression model outperformed the NN model in terms of accuracy, and thus was finally selected as the optimal model for this study. Table 2. The performance of 9 machine learning models for predicting in-hospital mortality in patients. Indicators SVM KNN GBM NN Logistics AdaBoost XGBoost LightGBM CatBoost Training group Sensitivity 0.71 1.00 0.72 0.80 0.68 0.51 0.86 0.87 0.69 Specificity 0.77 1.00 0.74 0.71 0.80 0.89 0.68 0.91 0.78 F1-Score 0.73 1.00 0.77 0.76 0.73 0.62 0.79 0.89 0.76 Accuracy 0.74 1.00 0.76 0.75 0.74 0.70 0.77 0.90 0.74 AUROC 0.80 1.00 0.84 0.82 0.81 0.75 0.84 0.95 0.82 Testing group Sensitivity 0.72 0.45 0.87 0.89 0.89 0.50 0.76 0.80 0.86 Specificity 0.77 0.76 0.62 0.62 0.73 0.83 0.73 0.60 0.65 F1-Score 0.42 0.28 0.37 0.38 0.38 0.36 0.27 0.33 0.24 Accuracy 0.76 0.72 0.65 0.65 0.66 0.80 0.73 0.62 0.67 AUROC 0.82 0.60 0.80 0.82 0.82 0.72 0.81 0.76 0.81 Note: Neural Networks (NN), Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Gradient Boosting Machine (GBM), Logistic regression, Adaptive Boosting (AdaBoost), Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), Categorical Boosting (CatBoost). 4. Model explainability To better understand the relationship between the model and data, the SHAP algorithm was adopted to conduct interpretability analysis on the Logistic model, yielding the importance ranking and contribution degree of each predictor variable to the model’s prediction results. The bar plot was generated by ranking features in descending order of their mean absolute SHAP values, where this ranking represents the contribution degree of each feature to the overall model (Figure 5A). A larger absolute SHAP value indicates that the feature is more important and has a greater impact on the model output. The top 8 ranked features are Age, APSIII, SOFA, Glucose, WBC, Lactate, Temperature, and CRRT. The swarm plot in Figure 5B illustrates the distribution of SHAP values for these 8 feature variables: except for Temperature, higher values of Age, APSIII, and other features are associated with higher short-term mortality rates. Two specific cases were presented to demonstrate the contribution of each feature to mortality risk prediction at the individual level. In Case A (Figure 6A), an orthopedic critically ill patient aged 80 years with a SOFA score of 1, APSIII score of 33, WBC count of 5.5 m/uL, and glucose level of 6.11 mmol/L had a low risk of in-hospital mortality (< 0.01%) and did not actually experience death during hospitalization. In contrast, an orthopedic critically ill patient with a SOFA score of 8, APSIII score of 70, and glucose level of 6.22 mmol/L had a high risk of in-hospital mortality (32.2%) and actually experienced in-hospital death (Figure 6B). 5. External Validation To rigorously evaluate the generalization capability of the model, the logistic regression model was applied to an independent MIMIC-III test set. As shown in Figure 7, the AUC of the model on the external test set was 0.81 (95% CI: 0.78–0.83), while other performance metrics are presented in Table 3. Table 3. The performance of Logistic Regression for predicting in-hospital mortality in patients in external validation. Indicators Sensitivity Specificity F1-Score Accuracy Logistics 0.79 0.70 0.47 0.71 Methods 1. Database This retrospective study utilized the data from the publicly available health record dataset MIMIC-IV (2008–2019) and III (2001–2012) [12] . In this study, ethical approval was granted with a waiver of informed consent, as it involved the analysis of anonymized retrospective data. 2. Study population In the data processing phase, patients with orthopedic-related diseases and those undergoing orthopedic surgeries in the MIMIC IV and III database were accurately identified and classified using specific codes, structured query language was applied to extract patient data, including socio-demographic characteristics, vital signs, laboratory parameters, treatments and outcomes. To determine the study sample, strict inclusion and exclusion criteria were applied. Inclusion criteria: (1) Age ≥ 50 years old; (2) Patients admitted with fractures including hip fracture, femoral shaft fracture and spinal fracture, or those who underwent orthopedic surgeries such as hip arthroplasty and spinal internal fixation; (3) ICU stay of ≥ 24 hours. Exclusion criteria: (1) First admission to the ICU; (2) Diagnosed with malignant tumors and suffers from end-stage diseases. (3) Variables with missing data rates greater than 30%. The primary outcome was in-hospital mortality, defined as death from any cause before discharge. Based on this outcome, patients were categorized into survival and non-survival groups. 3. Statistical analysis All statistical analyses were performed using R software (version 4.5.0). Missing data for variables are common in the MIMIC database. Multiple imputation was used to handle missing values; Variables with a missing rate exceeding 30% were excluded. Specifically, for variables with a missing rate of less than 30%, multiple imputation was performed using the mice package in R software. Continuous variables are presented as medians with interquartile ranges. The Mann-Whitney U test was used for comparisons between the two groups. Categorical data were expressed as frequency [n (%)]. And inter-group differences were compared using the chi-square test or Fisher’s exact test. The LASSO algorithm was applied to screen for the optimal predictive features. All statistical tests were two-tailed, and a value of P < 0.05 was considered statistically significant. 4. Machine learning model building In our study, nine common algorithms, i.e., Neural Networks (NN), Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Gradient Boosting Machine (GBM), Logistic regression, Adaptive Boosting (AdaBoost), Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), Categorical Boosting (CatBoost), were applied to build models for predicting in-hospital mortality of patients with orthopedic-related diseases and those undergoing orthopedic surgeries. During the search process, the final models were evaluated using the confusion matrix metrics such as sensitivity, specificity, accuracy, F1 score and the AUROC to identify the model with the best predictive performance. 5. Tools for interpreting machine learning SHAP is a visualization method based on game theory for interpreting the outputs of machine learning models [13] . In this study, the SHAP algorithm was employed to analyze the variables of the optimal model, so as to quantitatively visualize the relationships between risk factors and outcomes and further enhance the credibility of the model. 6. External Validation To rigorously evaluate the model's generalization performance, this study additionally extracted data from MIMIC-III (2001–2012) for orthopedic critical care patients meeting the same inclusion and exclusion criteria (a total of 1,718 patients) to serve as an independent external test set. The trained optimal model was applied to this external test set to examine its performance on unseen data. Declarations Data Availability The MIMIC series critical care databases (including MIMIC-IV and MIMIC-III) have been deposited in the PhysioNet public repository. The latest version of MIMIC-IV (version 2.2) can be accessed at the following URL: https://physionet.org/content/mimiciv/2.2/; the MIMIC-III database (version 1.4) can be accessed at the following URL: https://physionet.org/content/mimic-iii-clinical-database/1.4/. Author contributions statement Conceptualization: T.Z., M.L., W.L., H.L., Z.T.; Data Curation: T.Z., M.L. (data extraction and manuscript structure organization); W.L., H.L. (data sorting, cleaning, and validation); Formal Analysis: W.L., H.L. (statistical analysis and machine learning model validation); Visualization: Z.T. (figure preparation and formatting); Literature Search: Z.T. (literature screening and reference collation); Writing - Original Draft: All authors; Writing - Review & Editing: All authors. Additional information All authors declare no competing interests References Peterson, B. E. et al. Orthopedic Trauma and Aging: It Isn't Just About Mortality. Geriatr. Orthop. Surg. Rehabil . 6 (1), 33–36 (2015). Zhong, H. et al. The Use of Critical Care Services After Orthopedic Surgery at a High-Volume Orthopedic Medical Center: A Retrospective Study. Hss j. 18 (3), 344–350 (2022). Tang, Y. & Li HGuo, Z. Prediction of ICU admission after orthopedic surgery in elderly patients. Pak J. Med. Sci. 37 (4), 1179–1184 (2021). Anil, U. et al. Orthopedic pelvic and extremity injuries increase overall hospital length of stay but not in-hospital complications or mortality in trauma ICU patients: Orthopedic Injuries in Trauma ICU Patients. Injury 55 (11), 111843 (2024). Fernandez-Bustamante, A. et al. Postoperative Pulmonary Complications, Early Mortality, and Hospital Stay Following Noncardiothoracic Surgery: A Multicenter Study by the Perioperative Research Network Investigators. JAMA Surg. 152 (2), 157–166 (2017). Olofsson, E. et al. Association of depressive disorders and dementia with mortality among older people with hip fracture. BMC Geriatr. 23 (1), 135 (2023). Hawley, S. et al. Incidence and number of fragility fractures of the hip in South Africa: estimated projections from 2020 to 2050. Osteoporos. Int. 33 (12), 2575–2583 (2022). Shabani, F. et al. Hip fractures sustained in hospital: comorbidities and outcome. Postgrad. Med. J. 91 (1072), 61–64 (2015). Groff, H. et al. Causes of in-hospital mortality after hip fractures in the elderly. Hip Int. 30 (2), 204–209 (2020). Schoenfeld, A. J. et al. The influence of race and ethnicity on complications and mortality after orthopedic surgery: a systematic review of the literature. Med. Care . 52 (9), 842–851 (2014). Gundtoft, P. H. et al. The ability of comorbidity indices to predict mortality in an orthopedic setting: a systematic review. Syst. Rev. 10 (1), 234 (2021). Johnson, A. E. W. et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci. Data . 10 (1), 1 (2023). Petch, J. & Di SNelson, W. Opening the Black Box: The Promise and Limitations of Explainable Machine Learning in Cardiology. Can. J. Cardiol. 38 (2), 204–213 (2022). Jin, J. et al. Development and validation of a multidimensional predictive model for 28-day mortality in ICU patients with bloodstream infections: a cohort study. Front. Cell. Infect. Microbiol. 15 , 1569748 (2025). Maxwell, M. J. & Moran CGMoppett, I. K. Development and validation of a preoperative scoring system to predict 30 day mortality in patients undergoing hip fracture surgery. Br. J. Anaesth. 101 (4), 511–517 (2008). Solumsmoen, S. et al. Morbidity and mortality following degenerative spine surgery in a prospective cohort of 1687 consecutive surgical procedures. Acta Neurochir. (Wien) . 163 (1), 281–287 (2021). Han, T. et al. Development and validation of an artificial intelligence mobile application for predicting 30-day mortality in critically ill patients with orthopaedic trauma. Int. J. Med. Inf. 184 , 105383 (2024). Hibino, M. et al. Impact of Sex and Age on Trends of Mortality From Infective Endocarditis in High-Income Countries. JACC Adv. 4 (8), 101900 (2025). Moon, S. G. et al. Trends of Epilepsy-Related Mortality in South Korea. J. Clin. Neurol. 19 (2), 174–178 (2023). Knaus, W. A. et al. The APACHE III prognostic system. Risk prediction of hospital mortality for critically ill hospitalized adults. Chest 100 (6), 1619–1636 (1991). Vincent, J. L. et al. The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the Working Group on Sepsis-Related Problems of the European Society of Intensive Care Medicine. Intensive Care Med. 22 (7), 707–710 (1996). Liu, C. et al. The predictive value of SOFA and APSIII scores for 28-day mortality risk in SIMI: a cohort study based on the MIMIC-IV database. Front. Cell. Infect. Microbiol. 15 , 1574625 (2025). Chen, H. et al. Risk factors and nomogram for postoperative delirium in elderly orthopedic patients in the Intensive Care Unit: Retrospective study based on MIMIC-IV. Int. J. Orthop. Trauma. Nurs. 58 , 101196 (2025). Stollings, J. L. et al. Delirium in critical illness: clinical manifestations, outcomes, and management. Intensive Care Med. 47 (10), 1089–1103 (2021). Slooter, A. J. & Van De Leur RRZaal, I. J. Delirium in critically ill patients. Handb. Clin. Neurol. 141 , 449–466 (2017). Heuer, A. et al. Outcomes in very elderly ICU patients surgically treated for proximal femur fractures. Sci. Rep. 14 (1), 1376 (2024). Mifsud, S. & Schembri ELGruppetta, M. Stress-induced hyperglycaemia. Br. J. Hosp. Med. (Lond) . 79 (11), 634–639 (2018). Leto, R. et al. Admission hyperglycaemia is associated with higher mortality in patients with hip fracture. Eur. J. Emerg. Med. 22 (2), 99–102 (2015). Chang, C. H. et al. Diabetes, Glycemic Control, and Risk of Infection Morbidity and Mortality: A Cohort Study. Open. Forum Infect. Dis. 6 (10), ofz358 (2019). Broman, M. E. et al. The Relationship Between Heart Rate and Body Temperature in Critically Ill Patients. Crit. Care Med. 49 (3), e327–e331 (2021). Biegus, J. et al. Clinical, respiratory, haemodynamic, and metabolic determinants of lactate in heart failure. Kardiol Pol. 77 (1), 47–52 (2019). Bakker, J. et al. Serial blood lactate levels can predict the development of multiple organ failure following septic shock. Am. J. Surg. 171 (2), 221–226 (1996). Nishida, Y. et al. Association between preoperative lactate level and early complications after surgery for isolated extremity fracture. BMC Musculoskelet. Disord . 25 (1), 314 (2024). Mittwede, P. N. et al. Oxidative stress contributes to orthopedic trauma-induced acute kidney injury in obese rats. Am. J. Physiol. Ren. Physiol. 308 (2), F157–163 (2015). Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 20 Apr, 2026 Reviews received at journal 18 Apr, 2026 Reviewers agreed at journal 06 Apr, 2026 Reviewers agreed at journal 04 Apr, 2026 Reviewers agreed at journal 12 Mar, 2026 Reviews received at journal 20 Feb, 2026 Reviewers agreed at journal 11 Feb, 2026 Reviewers invited by journal 05 Feb, 2026 Editor assigned by journal 05 Feb, 2026 Editor invited by journal 02 Feb, 2026 Submission checks completed at journal 30 Jan, 2026 First submitted to journal 30 Jan, 2026 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-8620797","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":586989320,"identity":"9d363ae0-c3df-4a97-8bbb-977b0940d079","order_by":0,"name":"Ting Zhang","email":"data:image/png;base64,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","orcid":"","institution":"The General Hospital of Western Theater Command PLA","correspondingAuthor":true,"prefix":"","firstName":"Ting","middleName":"","lastName":"Zhang","suffix":""},{"id":586989321,"identity":"e9c97287-065f-406f-a401-b75288384cc4","order_by":1,"name":"MeiLing Li","email":"","orcid":"","institution":"The General Hospital of Western Theater Command PLA","correspondingAuthor":false,"prefix":"","firstName":"MeiLing","middleName":"","lastName":"Li","suffix":""},{"id":586989322,"identity":"6127bd1e-ef61-4629-8f37-ab39c9f0a338","order_by":2,"name":"Hong Lu","email":"","orcid":"","institution":"The General Hospital of Western Theater Command PLA","correspondingAuthor":false,"prefix":"","firstName":"Hong","middleName":"","lastName":"Lu","suffix":""},{"id":586989323,"identity":"052d5e38-934b-4c3c-a875-7d11f0d64bab","order_by":3,"name":"Wei Liu","email":"","orcid":"","institution":"The General Hospital of Western Theater Command PLA","correspondingAuthor":false,"prefix":"","firstName":"Wei","middleName":"","lastName":"Liu","suffix":""},{"id":586989324,"identity":"7b2ebd37-4fc8-4da2-a5bf-c727922fb0c0","order_by":4,"name":"Zhen Tan","email":"","orcid":"","institution":"The General Hospital of Western Theater Command PLA","correspondingAuthor":false,"prefix":"","firstName":"Zhen","middleName":"","lastName":"Tan","suffix":""}],"badges":[],"createdAt":"2026-01-16 16:09:04","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8620797/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8620797/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":102316273,"identity":"757b2fbe-93f9-424c-955d-574fa13bcebe","added_by":"auto","created_at":"2026-02-10 12:43:23","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":533084,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8620797/v1/f0a34af896729fb527fe87bc.png"},{"id":102316126,"identity":"dd81eb9e-29e2-419a-ab7c-cba30517b9d4","added_by":"auto","created_at":"2026-02-10 12:42:31","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":779786,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8620797/v1/93edbcb02efe458e40a91b81.png"},{"id":102316138,"identity":"75bebf77-d8e7-493a-a9ab-a09fdda0969c","added_by":"auto","created_at":"2026-02-10 12:42:40","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":407276,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-8620797/v1/1209791bd3c32208f3df40ae.png"},{"id":102316123,"identity":"a0802009-8699-443d-8894-6c7959789bf5","added_by":"auto","created_at":"2026-02-10 12:42:30","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":272958,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-8620797/v1/364e03fc651d75bf2585e763.png"},{"id":102316130,"identity":"a1f7207f-826f-45ec-98f2-5f8d726b0fc1","added_by":"auto","created_at":"2026-02-10 12:42:32","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":537216,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-8620797/v1/96ddad85355754839d756c42.png"},{"id":102316140,"identity":"81ec407c-4db7-417d-8cd2-2a2ad8e3a22b","added_by":"auto","created_at":"2026-02-10 12:42:41","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":501404,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-8620797/v1/552a181340bc0b0ab82ebdb0.png"},{"id":102316163,"identity":"43d2b23b-9543-45c8-9bc7-93efc0faadcd","added_by":"auto","created_at":"2026-02-10 12:42:51","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":616026,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend.\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-8620797/v1/f216f1db0a8824f3472abcd8.png"},{"id":102316287,"identity":"8f137b26-ac54-45f2-9273-980f5589b936","added_by":"auto","created_at":"2026-02-10 12:43:32","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4733727,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8620797/v1/b7240769-a3fd-4857-a65e-84af08d6ad2f.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Clinically Interpretable Machine Learning Framework for Mortality Prediction in Critically Ill Orthopedic Patients","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe rising global prevalence of orthopedic conditions and injuries among the elderly presents a substantial public health challenge [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e], while also holding significant clinical importance in critical care medicine\u0026mdash;particularly for cases such as hip fractures, osteoporotic fractures, and complex multiple traumas in this population [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Such patients often require post-operative or acute-phase management in the ICU. These conditions are not only surgical emergencies but also represent critical states involving multisystem dysfunction, the outcomes of their management directly impact patient length of hospitalization, long-term quality of life and in-hospital mortality [\u003cspan additionalcitationids=\"CR5\" citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. With the intensification of population aging, the number of those orthopedic patients continues to rise [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. In-hospital mortality among orthopedic patients is a key indicator for measuring the quality of medical care and perioperative safety management [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Although orthopedic surgeries are predominantly elective, and the overall prognosis for patients is favorable, the mortality risk in certain patient populations\u0026mdash;such as with multiple traumas, complex joint revisions, or severe comorbidities\u0026mdash;cannot be overlooked [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Mortality events often result from postoperative critical conditions such as deep vein thrombosis, pulmonary embolism, cardiopulmonary complications, and severe infections [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. These events tend to be insidious and sudden, posing significant challenges for early identification and intervention.\u003c/p\u003e \u003cp\u003eHowever, current research on the prognosis of orthopedic patients primarily focuses on single clinical indicators (such as age, race, or the number of comorbidities) and lacks a multidimensional integrated model [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Traditional risk scoring systems (e.g.,Charlson Comorbidity Index) exhibit limited predictive performance for orthopedic patients in the ICU, making it difficult to meet the clinical demand for individualized risk stratification and early intervention [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Therefore, developing a ML framework that combines high predictive accuracy with strong clinical interpretability is of urgent significance for improving the level of perioperative safety management for orthopedic patients. An ideal framework should not only accurately identify high-risk patients but also clearly elucidate the driving factors behind the risks\u0026mdash;for example, which key features (such as combined abnormalities in specific laboratory indicators, interactions between age and comorbidities, specific thresholds for surgery duration, etc.) collectively contribute to the high-risk assessment.\u003c/p\u003e \u003cp\u003eAccordingly, this study aims to develop and validate a clinically interpretable machine learning framework for predicting in-hospital mortality in orthopedic patients. By systematically integrating multidimensional clinical data and employing explainable artificial intelligence techniques\u0026mdash;with a particular focus on SHAP\u0026mdash;we seek to achieve a balance between model predictive accuracy and clinical transparency. The framework is designed not only to generate individualized risk assessments but also to provide clinicians with interpretable, evidence-based insights into the key factors driving each prediction. We envision this tool as both an early warning system for optimizing clinical resource allocation and a transparent decision-support companion, thereby advancing perioperative orthopedic care toward more precise, intelligent, and patient-centered outcomes.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003e1. The characteristics of included participants\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBetween 2008 and 2019, a total of 5,304 patients who underwent orthopedic-related diseases and orthopedic surgeries were identified from the MIMIC database. After excluding 534 ineligible patients, 4,770 patients were ultimately included in the analysis (Figure 1). Then, participants were randomly divided into the training set and the testing set in a ratio of 7:3 (3,339 vs 1,431). Table 1 demonstrated the characteristics of patients. The median age of patients was 67(52,80) years old, 2,508 (53%) were female patients, 3,042 (64%) had hypertension, 3,801(80%) had type 2 diabetes. The median length of ICU stay was 2.81(1.73, 5.66) days. 4,209(88.2%) patients were discharged from ICU, 561(11.8%) died in the ICU. The median SOFA score was 4(2,6), the median APSIII score was 40.00 (30.00,52.00), the median GCS score was 15.00(13.00, 15.00). Compared with survivors, non-survivors had significantly elevated age, glucose, lactate, ICU length of stay, organ failure and severity scores (SOFA, APSIII, SAPSII, OASIS), inflammatory and coagulation markers (WBC, INR, PT), and renal function parameters (creatinine, BUN). Conversely, body weight, SBP, DBP, MAP, temperature, PO2, PH, hemoglobin, and red blood cell count were significantly reduced. All differences were statistically significant (P \u0026lt; 0.05).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1.\u003c/strong\u003e Baseline characteristics of included patients.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"593\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eVariables\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTotal\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(n = 4770)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSurvivor\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(n = 4209)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNon-survivor\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(n = 561)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003eP\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;value\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eAge (years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e67 (52, 80)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e65 (50, 79)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e76 (64, 87)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eWeight (kg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e74.2 (62.2, 89.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e74.6 (62.5, 90)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e72.4 (60, 85)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eICU Day (days)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e2.81 (1.73, 5.66)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e2.73 (1.69, 5.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e3.85 (2.1, 7.89)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eTemperature(℃)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e36.8 (36.5, 37.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e36.8 (36.5, 37.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e36.6 (36.4, 37)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eGender, n (%)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; Male\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; Female\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e2262 (47)\u003c/p\u003e\n \u003cp\u003e2508 (53)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e1988 (47)\u003c/p\u003e\n \u003cp\u003e2221 (53)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e274 (49)\u003c/p\u003e\n \u003cp\u003e287 (51)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e0.502\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eHypertension, n (%)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; Yes\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; No\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e3042 (64)\u003c/p\u003e\n \u003cp\u003e1728 (36)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e2677 (64)\u003c/p\u003e\n \u003cp\u003e1532 (36)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e365 (65)\u003c/p\u003e\n \u003cp\u003e196 (35)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e0.529\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eT2DM, n (%)\u003c/p\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; No\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e3801 (80)\u003c/p\u003e\n \u003cp\u003e969 (20)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e3387 (80)\u003c/p\u003e\n \u003cp\u003e822 (20)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e414 (74)\u003c/p\u003e\n \u003cp\u003e147 (26)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eAKI, n (%)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; Yes\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; No\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e3555 (75)\u003c/p\u003e\n \u003cp\u003e1215 (25)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e3268 (78)\u003c/p\u003e\n \u003cp\u003e941 (22)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e287 (51)\u003c/p\u003e\n \u003cp\u003e274 (49)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eLC, n (%)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; Yes\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; No\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e4503 (94)\u003c/p\u003e\n \u003cp\u003e267 (6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e4001 (95)\u003c/p\u003e\n \u003cp\u003e208 (5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e502 (89)\u003c/p\u003e\n \u003cp\u003e59 (11)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eCVA, n (%)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; Yes\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; No\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e4424 (93)\u003c/p\u003e\n \u003cp\u003e346 (7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e3914 (93)\u003c/p\u003e\n \u003cp\u003e295 (7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e510 (91)\u003c/p\u003e\n \u003cp\u003e51 (9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e0.089\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eCRRT, n (%)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; Yes\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; No\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e4630 (97)\u003c/p\u003e\n \u003cp\u003e140 (3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e4131 (98)\u003c/p\u003e\n \u003cp\u003e78 (2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e499 (89)\u003c/p\u003e\n \u003cp\u003e62 (11)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eCABG, n (%)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; Yes\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; No\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e4712 (99)\u003c/p\u003e\n \u003cp\u003e58 (1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e4154 (99)\u003c/p\u003e\n \u003cp\u003e55 (1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e558 (99)\u003c/p\u003e\n \u003cp\u003e3 (1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e0.173\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eHR (times/minute)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e88 (75, 103)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e88 (75, 102)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e90 (75, 105)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e0.159\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eSBP (mmHg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e123 (108, 140)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e123 (109, 140)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e118 (101, 136)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eDBP (mmHg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e69 (58, 81)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e69 (58, 81)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e66 (54, 79)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eMAP (mmHg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e83 (71, 95)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e83 (72, 95)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e79 (68, 94)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eSOFA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e4 (2, 6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e3 (2, 5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e7 (4, 10)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eAPSIII\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e40 (30, 52)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e39 (29, 49)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e56 (43, 74)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eSAPSII\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e33 (25, 42)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e32 (24, 41)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e46 (36, 56)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eOASIS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e32 (26, 37)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e31 (26, 36)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e38 (32, 43)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eGCS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e15 (13, 15)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e15 (13, 15)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e15 (13, 15)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e0.409\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\" valign=\"top\" style=\"width: 593px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLaboratory tests\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eWBC (m/uL)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e11.2 (8.3, 15.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e11.1 (8.2, 14.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e12.3 (8.9, 17.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eRBC (10\u0026sup1;\u0026sup2;/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e3.55 (3.03, 4.11)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e3.57 (3.05, 4.11)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e3.44 (2.87, 4.13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e0.002\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003ePlatelet (K/uL)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e192 (144, 251)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e195 (148, 253)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e173 (123, 232)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eHemoglobin (g/dL)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e10.7 (9.1, 12.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e10.8 (9.2, 12.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e10.3 (8.7, 12.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eINR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e1.2 (1.1, 1.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e1.2 (1.1, 1.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e1.3 (1.1, 1.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003ePT(s)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e13.2 (12, 15.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e13.1 (12, 14.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e14.3(12.7, 18.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eCreatinine (mg/dL)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e0.9 (0.7, 1.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e0.9 (0.7, 1.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e1.2 (0.8, 1.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eBUN (mg/dL)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e17 (12, 26)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e16 (12, 25)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e24 (17, 37)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003ePH\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e7.37 (7.31, 7.42)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e7.37 (7.31, 7.42)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e7.34 (7.26, 7.41)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003ePO2 (mmHg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e107 (60, 190)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e109 (61, 192)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e91 (50, 175)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003ePCO2 (mmHg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e42 (37, 48)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e42 (37, 48)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e42 (35, 49)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e0.642\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eLactate(mmol/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e1.7 (1.2, 2.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e1.6 (1.1, 2.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e2 (1.4, 3.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 149px;\"\u003e\n \u003cp\u003eGlucose (mmol/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e7.2(6, 9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e7.1 (5.9, 8.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e\n \u003cp\u003e8.1 (6.3, 10.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026lt; 0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\" valign=\"top\" style=\"width: 593px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003eNote:\u003c/em\u003e\u003c/strong\u003e T2DM: Type 2 Diabetes Mellitus; AKI: Acute Kidney Injury; LC: Lung Cancer; \u0026nbsp;CVA: Cerebrovascular Accident; CRRT: Continuous Renal Replacement Therapy; \u0026nbsp; \u0026nbsp; CABG: Coronary Artery Bypass Grafting; HR: Heart Rate; SBP: Systolic Blood Pressure; DBP: Diastolic Blood Pressure; MAP: Mean Arterial Pressure; SOFA: Sequential Organ Failure Assessment; APSIII: Acute Physiology Score III; SAPSII: Simplified Acute Physiology Score II; OASIS: Oxford Acute Severity of Illness Score; GCS: Glasgow Coma Scale; WBC: White Blood Cell; RBC: Red Blood Cell; INR: International Normalized Ratio; PT: Prothrombin Time; BUN: Blood Urea Nitrogen; PH: Potential of Hydrogen; PO2: Partial Pressure of Dioxide; PCO2: Partial Pressure of Carbon Dioxide;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e2.\u0026nbsp;\u003c/strong\u003e \u003cstrong\u003eSelection of Predictor Variables\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFeature variables for in-hospital mortality in orthopedic critical illness were identified using the LASSO regression algorithm, as shown in Figure 2. Specifically, Figure 2A presents the predictors selected by LASSO (shrinkage parameter, \u0026lambda; = 0.01890518; lambda.1se). Eventually, the coefficients of all predictors shrink to zero, and the curve reaching zero indicates that the corresponding feature has been eliminated by LASSO regression. The LASSO regression adopted 20-fold cross-validation based on the minimum criterion to select the optimal parameters, and plotted the curve of the relationship between binomial deviance and lambda, with a total of 10 feature variables screened out (Figure 2B). The Boruta algorithm identified 29 feature variables, while the Recursive Feature Elimination (RFE) method selected 13 feature variables (Figure 3). A Venn diagram was used to extract the intersection of the results from the three algorithms, yielding 8 important feature variables as follows: Age, APSIII, SOFA, Glucose, WBC, Lactate, Temperature, and CRRT (Figure 4).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3. \u0026nbsp;Model performance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAfter the key variables were identified, 9 ML models were constructed to predict the mortality risk of inpatients with orthopedic critical illness. The AUROC was used to analyze the predictive efficacy of these 9 ML models for the mortality of inpatients with orthopedic critical illness. The 9 algorithms were SVM, KNN, GBM, NN, Logistics, AdaBoost, XGBoost, LightGBM, and CatBoost, respectively. As shown in Table 2, in the training set, the KNN model achieved an AUROC of 1.00, indicating severe overfitting. The remaining models all demonstrated satisfactory performance. In the test group, an identical peak AUC of 82% was attained by the SVM, NN, and logistic regression models; meanwhile, the neural network and logistic regression models jointly achieved the highest sensitivity of 89%. Moreover, the Logistic regression model outperformed the NN model in terms of accuracy, and thus was finally selected as the optimal model for this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2. The performance of 9 machine learning models for predicting in-hospital mortality in patients.\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"906\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIndicators\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSVM\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eKNN\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eGBM\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNN\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLogistics\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAdaBoost\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eXGBoost\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLightGBM\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCatBoost\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"10\" valign=\"top\" style=\"width: 906px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTraining group\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSensitivity\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003e0.71\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.72\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e0.80\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.68\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.51\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.86\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSpecificity\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003e0.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.74\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e0.71\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.80\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.68\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eF1-Score\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003e0.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e0.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.79\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.76\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAccuracy\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003e0.74\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e0.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.74\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.74\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAUROC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003e0.80\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e0.82\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.81\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.82\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"10\" valign=\"top\" style=\"width: 906px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTesting group\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSensitivity\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003e0.72\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e0.45\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e0.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.80\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.86\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSpecificity\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003e0.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e0.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.83\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.60\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.65\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eF1-Score\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003e0.42\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e0.28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e0.38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.27\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.24\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAccuracy\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003e0.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e0.72\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.65\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e0.65\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.80\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAUROC\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003e0.82\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e0.60\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.80\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e0.82\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.82\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.72\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e0.81\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 94px;\"\u003e\n \u003cp\u003e0.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 67px;\"\u003e\n \u003cp\u003e0.81\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"10\" valign=\"top\" style=\"width: 906px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003eNote:\u0026nbsp;\u003c/em\u003e\u003c/strong\u003eNeural Networks (NN), Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Gradient Boosting Machine (GBM), Logistic regression, Adaptive Boosting (AdaBoost), Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), Categorical Boosting (CatBoost).\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4. \u0026nbsp;Model explainability\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo better understand the relationship between the model and data, the SHAP algorithm was adopted to conduct interpretability analysis on the Logistic model, yielding the importance ranking and contribution degree of each predictor variable to the model\u0026rsquo;s prediction results. The bar plot was generated by ranking features in descending order of their mean absolute SHAP values, where this ranking represents the contribution degree of each feature to the overall model (Figure 5A). A larger absolute SHAP value indicates that the feature is more important and has a greater impact on the model output. The top 8 ranked features are Age, APSIII, SOFA, Glucose, WBC, Lactate, Temperature, and CRRT. The swarm plot in Figure 5B illustrates the distribution of SHAP values for these 8 feature variables: except for Temperature, higher values of Age, APSIII, and other features are associated with higher short-term mortality rates. Two specific cases were presented to demonstrate the contribution of each feature to mortality risk prediction at the individual level. In Case A (Figure 6A), an orthopedic critically ill patient aged 80 years with a SOFA score of 1, APSIII score of 33, WBC count of 5.5 m/uL, and glucose level of 6.11 mmol/L had a low risk of in-hospital mortality (\u0026lt; 0.01%) and did not actually experience death during hospitalization. In contrast, an orthopedic critically ill patient with a SOFA score of 8, APSIII score of 70, and glucose level of 6.22 mmol/L had a high risk of in-hospital mortality (32.2%) and actually experienced in-hospital death (Figure 6B).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e5. External Validation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo rigorously evaluate the generalization capability of the model, the logistic regression model was applied to an independent MIMIC-III test set. As shown in Figure 7, the AUC of the model on the external test set was 0.81 (95% CI: 0.78\u0026ndash;0.83), while other performance metrics are presented in Table 3.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 3.\u0026nbsp;\u003c/strong\u003eThe performance of Logistic Regression for predicting in-hospital mortality in patients\u0026nbsp;in\u0026nbsp;external validation.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"520\" class=\"fr-table-selection-hover\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 99px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIndicators\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 110px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSensitivity\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSpecificity\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eF1-Score\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAccuracy\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 99px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLogistics\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 110px;\"\u003e\n \u003cp\u003e0.79\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e0.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e0.47\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e0.71\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"},{"header":"Methods","content":"\u003cp\u003e \u003cb\u003e1. Database\u003c/b\u003e \u003c/p\u003e\u003cp\u003eThis retrospective study utilized the data from the publicly available health record dataset MIMIC-IV (2008\u0026ndash;2019) and III (2001\u0026ndash;2012) \u003csup\u003e[12]\u003c/sup\u003e. In this study, ethical approval was granted with a waiver of informed consent, as it involved the analysis of anonymized retrospective data.\u003c/p\u003e\n\u003ch3\u003e2. Study population\u003c/h3\u003e\n\u003cp\u003eIn the data processing phase, patients with orthopedic-related diseases and those undergoing orthopedic surgeries in the MIMIC IV and III database were accurately identified and classified using specific codes, structured query language was applied to extract patient data, including socio-demographic characteristics, vital signs, laboratory parameters, treatments and outcomes. To determine the study sample, strict inclusion and exclusion criteria were applied. Inclusion criteria: (1) Age\u0026thinsp;\u0026ge;\u0026thinsp;50 years old; (2) Patients admitted with fractures including hip fracture, femoral shaft fracture and spinal fracture, or those who underwent orthopedic surgeries such as hip arthroplasty and spinal internal fixation; (3) ICU stay of \u0026ge;\u0026thinsp;24 hours. Exclusion criteria: (1) First admission to the ICU; (2) Diagnosed with malignant tumors and suffers from end-stage diseases. (3) Variables with missing data rates greater than 30%. The primary outcome was in-hospital mortality, defined as death from any cause before discharge. Based on this outcome, patients were categorized into survival and non-survival groups.\u003c/p\u003e\n\u003ch3\u003e3. Statistical analysis\u003c/h3\u003e\n\u003cp\u003eAll statistical analyses were performed using R software (version 4.5.0). Missing data for variables are common in the MIMIC database. Multiple imputation was used to handle missing values; Variables with a missing rate exceeding 30% were excluded. Specifically, for variables with a missing rate of less than 30%, multiple imputation was performed using the mice package in R software. Continuous variables are presented as medians with interquartile ranges. The Mann-Whitney U test was used for comparisons between the two groups. Categorical data were expressed as frequency [n (%)]. And inter-group differences were compared using the chi-square test or Fisher\u0026rsquo;s exact test. The LASSO algorithm was applied to screen for the optimal predictive features. All statistical tests were two-tailed, and a value of P\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e\n\u003ch3\u003e4. Machine learning model building\u003c/h3\u003e\n\u003cp\u003eIn our study, nine common algorithms, i.e., Neural Networks (NN), Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Gradient Boosting Machine (GBM), Logistic regression, Adaptive Boosting (AdaBoost), Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), Categorical Boosting (CatBoost), were applied to build models for predicting in-hospital mortality of patients with orthopedic-related diseases and those undergoing orthopedic surgeries. During the search process, the final models were evaluated using the confusion matrix metrics such as sensitivity, specificity, accuracy, F1 score and the AUROC to identify the model with the best predictive performance.\u003c/p\u003e\n\u003ch3\u003e5. Tools for interpreting machine learning\u003c/h3\u003e\n\u003cp\u003eSHAP is a visualization method based on game theory for interpreting the outputs of machine learning models \u003csup\u003e[13]\u003c/sup\u003e. In this study, the SHAP algorithm was employed to analyze the variables of the optimal model, so as to quantitatively visualize the relationships between risk factors and outcomes and further enhance the credibility of the model.\u003c/p\u003e\n\u003ch3\u003e6. External Validation\u003c/h3\u003e\n\u003cp\u003e To rigorously evaluate the model's generalization performance, this study additionally extracted data from MIMIC-III (2001\u0026ndash;2012) for orthopedic critical care patients meeting the same inclusion and exclusion criteria (a total of 1,718 patients) to serve as an independent external test set. The trained optimal model was applied to this external test set to examine its performance on unseen data.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe MIMIC series critical care databases (including MIMIC-IV and MIMIC-III) have been deposited in the PhysioNet public repository. The latest version of MIMIC-IV (version 2.2) can be accessed at the following URL: https://physionet.org/content/mimiciv/2.2/; the MIMIC-III database (version 1.4) can be accessed at the following URL: https://physionet.org/content/mimic-iii-clinical-database/1.4/.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor\u003c/strong\u003e \u003cstrong\u003econtributions\u003c/strong\u003e \u003cstrong\u003estatement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConceptualization: T.Z., M.L., W.L., H.L., Z.T.;\u003c/p\u003e\n\u003cp\u003eData Curation: T.Z., M.L. (data extraction and manuscript structure organization); W.L.,\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eH.L. (data sorting, cleaning, and validation);\u003c/p\u003e\n\u003cp\u003eFormal Analysis: W.L., H.L. (statistical analysis and machine learning model validation);\u003c/p\u003e\n\u003cp\u003eVisualization: Z.T. (figure preparation and formatting);\u003c/p\u003e\n\u003cp\u003eLiterature Search: Z.T. (literature screening and reference collation);\u003c/p\u003e\n\u003cp\u003eWriting - Original Draft: All authors;\u003c/p\u003e\n\u003cp\u003eWriting - Review \u0026amp; Editing: All authors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAdditional\u003c/strong\u003e \u003cstrong\u003einformation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors declare no competing interests\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003ePeterson, B. E. et al. Orthopedic Trauma and Aging: It Isn't Just About Mortality. \u003cem\u003eGeriatr. Orthop. Surg. Rehabil\u003c/em\u003e. \u003cb\u003e6\u003c/b\u003e (1), 33\u0026ndash;36 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhong, H. et al. The Use of Critical Care Services After Orthopedic Surgery at a High-Volume Orthopedic Medical Center: A Retrospective Study. \u003cem\u003eHss j.\u003c/em\u003e \u003cb\u003e18\u003c/b\u003e (3), 344\u0026ndash;350 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTang, Y. \u0026amp; Li HGuo, Z. Prediction of ICU admission after orthopedic surgery in elderly patients. \u003cem\u003ePak J. Med. Sci.\u003c/em\u003e \u003cb\u003e37\u003c/b\u003e (4), 1179\u0026ndash;1184 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnil, U. et al. Orthopedic pelvic and extremity injuries increase overall hospital length of stay but not in-hospital complications or mortality in trauma ICU patients: Orthopedic Injuries in Trauma ICU Patients. \u003cem\u003eInjury\u003c/em\u003e \u003cb\u003e55\u003c/b\u003e (11), 111843 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFernandez-Bustamante, A. et al. Postoperative Pulmonary Complications, Early Mortality, and Hospital Stay Following Noncardiothoracic Surgery: A Multicenter Study by the Perioperative Research Network Investigators. \u003cem\u003eJAMA Surg.\u003c/em\u003e \u003cb\u003e152\u003c/b\u003e (2), 157\u0026ndash;166 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOlofsson, E. et al. Association of depressive disorders and dementia with mortality among older people with hip fracture. \u003cem\u003eBMC Geriatr.\u003c/em\u003e \u003cb\u003e23\u003c/b\u003e (1), 135 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHawley, S. et al. Incidence and number of fragility fractures of the hip in South Africa: estimated projections from 2020 to 2050. \u003cem\u003eOsteoporos. Int.\u003c/em\u003e \u003cb\u003e33\u003c/b\u003e (12), 2575\u0026ndash;2583 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShabani, F. et al. Hip fractures sustained in hospital: comorbidities and outcome. \u003cem\u003ePostgrad. Med. J.\u003c/em\u003e \u003cb\u003e91\u003c/b\u003e (1072), 61\u0026ndash;64 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGroff, H. et al. Causes of in-hospital mortality after hip fractures in the elderly. \u003cem\u003eHip Int.\u003c/em\u003e \u003cb\u003e30\u003c/b\u003e (2), 204\u0026ndash;209 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchoenfeld, A. J. et al. The influence of race and ethnicity on complications and mortality after orthopedic surgery: a systematic review of the literature. \u003cem\u003eMed. Care\u003c/em\u003e. \u003cb\u003e52\u003c/b\u003e (9), 842\u0026ndash;851 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGundtoft, P. H. et al. The ability of comorbidity indices to predict mortality in an orthopedic setting: a systematic review. \u003cem\u003eSyst. Rev.\u003c/em\u003e \u003cb\u003e10\u003c/b\u003e (1), 234 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJohnson, A. E. W. et al. MIMIC-IV, a freely accessible electronic health record dataset. \u003cem\u003eSci. Data\u003c/em\u003e. \u003cb\u003e10\u003c/b\u003e (1), 1 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePetch, J. \u0026amp; Di SNelson, W. Opening the Black Box: The Promise and Limitations of Explainable Machine Learning in Cardiology. \u003cem\u003eCan. J. Cardiol.\u003c/em\u003e \u003cb\u003e38\u003c/b\u003e (2), 204\u0026ndash;213 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJin, J. et al. Development and validation of a multidimensional predictive model for 28-day mortality in ICU patients with bloodstream infections: a cohort study. \u003cem\u003eFront. Cell. Infect. Microbiol.\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e, 1569748 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMaxwell, M. J. \u0026amp; Moran CGMoppett, I. K. Development and validation of a preoperative scoring system to predict 30 day mortality in patients undergoing hip fracture surgery. \u003cem\u003eBr. J. Anaesth.\u003c/em\u003e \u003cb\u003e101\u003c/b\u003e (4), 511\u0026ndash;517 (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSolumsmoen, S. et al. Morbidity and mortality following degenerative spine surgery in a prospective cohort of 1687 consecutive surgical procedures. \u003cem\u003eActa Neurochir. (Wien)\u003c/em\u003e. \u003cb\u003e163\u003c/b\u003e (1), 281\u0026ndash;287 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHan, T. et al. Development and validation of an artificial intelligence mobile application for predicting 30-day mortality in critically ill patients with orthopaedic trauma. \u003cem\u003eInt. J. Med. Inf.\u003c/em\u003e \u003cb\u003e184\u003c/b\u003e, 105383 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHibino, M. et al. Impact of Sex and Age on Trends of Mortality From Infective Endocarditis in High-Income Countries. \u003cem\u003eJACC Adv.\u003c/em\u003e \u003cb\u003e4\u003c/b\u003e (8), 101900 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoon, S. G. et al. Trends of Epilepsy-Related Mortality in South Korea. \u003cem\u003eJ. Clin. Neurol.\u003c/em\u003e \u003cb\u003e19\u003c/b\u003e (2), 174\u0026ndash;178 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKnaus, W. A. et al. The APACHE III prognostic system. Risk prediction of hospital mortality for critically ill hospitalized adults. \u003cem\u003eChest\u003c/em\u003e \u003cb\u003e100\u003c/b\u003e (6), 1619\u0026ndash;1636 (1991).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVincent, J. L. et al. The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the Working Group on Sepsis-Related Problems of the European Society of Intensive Care Medicine. \u003cem\u003eIntensive Care Med.\u003c/em\u003e \u003cb\u003e22\u003c/b\u003e (7), 707\u0026ndash;710 (1996).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu, C. et al. The predictive value of SOFA and APSIII scores for 28-day mortality risk in SIMI: a cohort study based on the MIMIC-IV database. \u003cem\u003eFront. Cell. Infect. Microbiol.\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e, 1574625 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen, H. et al. Risk factors and nomogram for postoperative delirium in elderly orthopedic patients in the Intensive Care Unit: Retrospective study based on MIMIC-IV. \u003cem\u003eInt. J. Orthop. Trauma. Nurs.\u003c/em\u003e \u003cb\u003e58\u003c/b\u003e, 101196 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStollings, J. L. et al. Delirium in critical illness: clinical manifestations, outcomes, and management. \u003cem\u003eIntensive Care Med.\u003c/em\u003e \u003cb\u003e47\u003c/b\u003e (10), 1089\u0026ndash;1103 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSlooter, A. J. \u0026amp; Van De Leur RRZaal, I. J. Delirium in critically ill patients. \u003cem\u003eHandb. Clin. Neurol.\u003c/em\u003e \u003cb\u003e141\u003c/b\u003e, 449\u0026ndash;466 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHeuer, A. et al. Outcomes in very elderly ICU patients surgically treated for proximal femur fractures. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cb\u003e14\u003c/b\u003e (1), 1376 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMifsud, S. \u0026amp; Schembri ELGruppetta, M. Stress-induced hyperglycaemia. \u003cem\u003eBr. J. Hosp. Med. (Lond)\u003c/em\u003e. \u003cb\u003e79\u003c/b\u003e (11), 634\u0026ndash;639 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeto, R. et al. Admission hyperglycaemia is associated with higher mortality in patients with hip fracture. \u003cem\u003eEur. J. Emerg. Med.\u003c/em\u003e \u003cb\u003e22\u003c/b\u003e (2), 99\u0026ndash;102 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChang, C. H. et al. Diabetes, Glycemic Control, and Risk of Infection Morbidity and Mortality: A Cohort Study. \u003cem\u003eOpen. Forum Infect. Dis.\u003c/em\u003e \u003cb\u003e6\u003c/b\u003e (10), ofz358 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBroman, M. E. et al. The Relationship Between Heart Rate and Body Temperature in Critically Ill Patients. \u003cem\u003eCrit. Care Med.\u003c/em\u003e \u003cb\u003e49\u003c/b\u003e (3), e327\u0026ndash;e331 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBiegus, J. et al. Clinical, respiratory, haemodynamic, and metabolic determinants of lactate in heart failure. \u003cem\u003eKardiol Pol.\u003c/em\u003e \u003cb\u003e77\u003c/b\u003e (1), 47\u0026ndash;52 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBakker, J. et al. Serial blood lactate levels can predict the development of multiple organ failure following septic shock. \u003cem\u003eAm. J. Surg.\u003c/em\u003e \u003cb\u003e171\u003c/b\u003e (2), 221\u0026ndash;226 (1996).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNishida, Y. et al. Association between preoperative lactate level and early complications after surgery for isolated extremity fracture. \u003cem\u003eBMC Musculoskelet. Disord\u003c/em\u003e. \u003cb\u003e25\u003c/b\u003e (1), 314 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMittwede, P. N. et al. Oxidative stress contributes to orthopedic trauma-induced acute kidney injury in obese rats. \u003cem\u003eAm. J. Physiol. Ren. Physiol.\u003c/em\u003e \u003cb\u003e308\u003c/b\u003e (2), F157\u0026ndash;163 (2015).\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":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8620797/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8620797/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eThe global burden of geriatric orthopedic conditions (e.g., hip fractures, complex trauma) is rising, posing significant challenges to critical care medicine. Existing prognostic tools rely on single indicators or traditional scoring systems, lacking sufficient accuracy and interpretability in ICUs. This study aims to develop an interpretable machine learning (ML) framework with high predictive performance for individualized risk stratification.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eCritically ill orthopedic patients\u0026thinsp;\u0026ge;\u0026thinsp;50 from MIMIC-IV/III were enrolled (excluding those with malignant tumors or incomplete data). MIMIC-IV was split 7:3 into training/testing sets. LASSO Cox regression with 20-fold cross-validation enabled dimension reduction and feature selection. Nine ML models were compared; the optimal model was selected by accuracy and AUROC. SHAP quantified feature impacts and individual decision processes. MIMIC-III served for external validation.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003e6,488 patients were included, with 11.8% in-hospital mortality in MIMIC-IV. Eight core features (age, APSIII, SOFA, blood glucose, WBC, lactate, body temperature, CRRT) were identified. Logistic regression performed best (AUROC\u0026thinsp;=\u0026thinsp;0.82) and achieved AUC\u0026thinsp;=\u0026thinsp;0.81 (95% CI: 0.78\u0026ndash;0.83) in external validation.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eThis interpretable mortality prediction model for critically ill orthopedic patients aids preoperative risk assessment and postoperative ICU monitoring, supporting targeted early interventions to improve outcomes.\u003c/p\u003e","manuscriptTitle":"A Clinically Interpretable Machine Learning Framework for Mortality Prediction in Critically Ill Orthopedic Patients","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-10 12:39:45","doi":"10.21203/rs.3.rs-8620797/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-04-20T07:55:04+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-18T14:37:49+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"261334206749029470476467025327295268402","date":"2026-04-06T13:42:13+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"21589799535534569878038736801916428079","date":"2026-04-04T17:26:47+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"83150503468962446670153100321073090864","date":"2026-03-12T20:18:15+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-20T18:08:51+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"188972695157569176020070118468991396561","date":"2026-02-11T09:37:23+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-02-05T21:24:19+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-02-05T21:21:22+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-02-02T17:50:56+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-01-30T13:50:39+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2026-01-30T13:12:05+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"44537565-30bf-41e9-aa0c-392df74b5f21","owner":[],"postedDate":"February 10th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":62456062,"name":"Health sciences/Biomarkers"},{"id":62456063,"name":"Health sciences/Diseases"},{"id":62456064,"name":"Health sciences/Health care"},{"id":62456065,"name":"Health sciences/Medical research"},{"id":62456066,"name":"Health sciences/Risk factors"}],"tags":[],"updatedAt":"2026-05-13T20:09:19+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-10 12:39:45","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8620797","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8620797","identity":"rs-8620797","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}