Development and Temporal External Validation of an XGBoost-Based Clinical Prediction Model for PACU Hypoxemia in Elderly Thoracic Surgery Patients

Development and Temporal External Validation of an XGBoost-Based Clinical Prediction Model for PACU Hypoxemia in Elderly Thoracic Surgery Patients | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Development and Temporal External Validation of an XGBoost-Based Clinical Prediction Model for PACU Hypoxemia in Elderly Thoracic Surgery Patients Yang Xu, Anni Wu, Sheng Sun, Minzhao Sun, Xinyu Zhan, Bolun Zhao This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8723621/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Objective To develop and validate a machine learning-based risk prediction model for postoperative hypoxemia in elderly patients undergoing thoracic surgery in the post-anesthesia care unit (PACU), aiming to provide a decision-support tool for perioperative precision prevention. Methods This study involved a retrospective development cohort of 5134 elderly thoracic surgery patients from a tertiary academic medical center between 2019 and 2024(split 8:2 into training and internal validation sets), and a prospective external validation cohort of 272 patients in 2025. Feature selection was performed using LASSO regression. Six machine learning algorithms, including Logistic Regression, XGBoost, and LightGBM, were trained and compared. Model performance was assessed using area under the receiver operating characteristic curve (AUC), calibration plots, and decision curve analysis (DCA). SHAP values were used for model interpretability. Results The incidence of hypoxemia was 18.8%(967/5134). LASSO regression identified 10 independent predictors categorized into thoracic surgery site, age, BMI, preoperative albumin, hemoglobin, D-dimer, PaO₂,duration of anesthesia and surgery, intraoperative fluid volume.The AUC values across the six machine learning models ranged from 0.852 to 0.953. XGBoost demonstrated the best performance, with AUCs of 0.988 in the training set, 0.953 in the internal validation set, and 0.706 in the external validation set, indicating good generalizability. Conclusions The XGBoost-based prediction model accurately identifies elderly thoracic surgery patients at high risk for PACU hypoxemia. Incorporating multiple modifiable clinical indicators, this tool facilitates early risk stratification and proactive, continuous, multidisciplinary interventions by thoracic surgery, anesthesia, and PACU teams. Hypoxemia Thoracic Surgical Procedures Aged Postanesthesia Care Unit Machine Learning Risk Assessment Figures Figure 1 Figure 2 Figure 3 Figure 4 1.Introduction Hypoxemia is a common complication in the post-anesthesia care unit (PACU), primarily due to ventilation-perfusion mismatch, increased shunt, and impaired diffusion [ 1 ][ 2 ][ 3 ] . Its incidence after general anesthesia ranges from 2.79% to 21.79% [ 4 ][ 5 ][ 6 ] , with approximately 69.8% of episodes occurring within the first 30 minutes of PACU arrival [ 7 ] . Timely intervention is critical, as persistent hypoxemia can lead to arrhythmias, cerebral ischemia, multi-organ dysfunction, and increased mortality [ 8 ][ 9 ][ 10 ] . Despite strategies like lung-protective ventilation and high-flow nasal oxygen [ 11 ][ 12 ][ 13 ][ 14 ] , management remains reactive, often initiated only after overt desaturation. This delay fails to address early, silent respiratory deterioration, potentially leading to avoidable organ injury and inefficient resource use. Elderly patients undergoing thoracic surgery represent a particularly high-risk subgroup. The combination of one-lung ventilation, lung resection, and chest wall disruption, superimposed on age-related decline in pulmonary compliance and physiological reserve, results in hypoxemia incidences reported from 22.1% to over 40% [ 15 ][ 16 ] . A major clinical challenge is the lack of an objective tool to rapidly identify high-risk patients at the operating room-to-PACU transition. This impedes seamless handover and proactive, risk-stratified care. While risk factors like advanced age, obesity, smoking, and fluid balance are known [ 16 ][ 17 ][ 18 ][ 19 ][ 20 ] , and machine learning (ML) shows promise for clinical prediction, existing models lack specificity for this populatio [ 21 ] n. General models [ 22 ] are unvalidated in thoracic surgery, whereas models for other specific cohorts [ 23 ][ 24 ][ 25 ] demonstrate the utility of population-tailored prediction. A significant gap exists: no validated model predicts PACU hypoxemia specifically for elderly thoracic surgery patients. To address this, we leveraged a large-scale dataset and employed a comparative modeling approach, using logistic regression and five ML algorithms to develop and validate a risk prediction model for this high-risk cohort. We integrated the SHAP (Shapley Additive exPlanations) framework to ensure interpretability. The goal is to provide a practical decision-support tool for the perioperative team, facilitating a shift from reactive management to proactive, coordinated intervention to improve outcomes. 2.Methods 2.1 Study Design and Ethical Approval This was a retrospective cohort study for model development, with subsequent prospective temporal validation. The protocol was approved by the Institutional Review Board of Dalian Medical University (Approval No. : 2025-041) and adhered to the Declaration of Helsinki. 2.2 Study Population and Selection Criteria Patients aged ≥ 60 years who underwent non-cardiac thoracic surgery under general anesthesia with endotracheal intubation and were admitted directly to the PACU between January 2019 and July 2025 were included. Exclusion criteria were: 1) direct postoperative transfer to ICU due to pre-existing critical conditions (e.g., need for ECMO, end-stage cardiopulmonary disease); 2) > 20% missing key data or irreconcilable data inconsistencies.Cardiac/vascular surgery patients (routinely admitted to specialized ICUs) were not part of the target population. The screening process is shown in Fig. 1 A. 2.3 Outcome Definition The primary outcome was PACU hypoxemia, defined as meeting any of the following while breathing room air: SpO₂ < 90%, PaO₂ < 60 mmHg, or PaO₂/FiO₂ ≤ 300 mmHg [ 1 ][ 2 ][ 3 ] . 2.4 Sample size calculation This study followed the TRIPOD guidelines for transparent reporting of prediction models. The sample size was calculated using Riley et al.’s method to ensure statistical adequacy.The minimum sample size was estimated as: n = p / [π(1 - π)(1 - R²)] Where: p = 30 predictors; π = hypoxemia incidence (41.2%, from prior research); R² = 0.20 (common clinical model value).The base calculation yielded 155 cases.Applying the Events Per Variable (EPV ≥ 10) principle for robustness, at least 300 hypoxemia events were required (30 predictors × 10). Given the 41.2% incidence, the total sample size needed was:n = 300 / 0.412 ≈ 729 cases. 2.5 Data and Cohorts After double-entry verification and de-identification, 30 predictor variables were prespecified based on literature and expert consultation (Eq.A.1). Two cohorts were established:(1) Development Cohort : 5,134 patients (January 2019 – December 2024) for model training/internal validation. (2) External Temporal Validation Cohort 272 consecutive prospective patients (January – July 2025), used to assess generalizability per the PROBAST framework [ 26 ] . 2.6 Data Preprocessing and Modeling Missing data (< 5%) followed an MCAR pattern; complete-case analysis was used. Continuous variables were normalized (min-max scaling). The development cohort was split into training (80%) and internal validation (20%) sets. A two-step strategy was employed: (1) Feature Selection Least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation selected a parsimonious predictor subset (Fig. 1 B). (2) Model Construction Six algorithms—Logistic Regression, XGBoost, SVM, RF, KNN, and LightGBM—were trained using the selected features. Hyperparameters were optimized via grid search. Analyses used R (v4.5.0); p < 0.05 was significant. 2.7 Model Evaluation and Interpretability Performance was evaluated using the AUC, calibration plots (Hosmer-Lemeshow test, Brier score), sensitivity, specificity, accuracy, NPV, and F1-score. Generalizability was tested via external temporal validation. The SHAP (Shapley Additive exPlanations) framework was applied to provide global feature importance rankings and individualized visual explanations for predictions (Fig. 1 C), enhancing model interpretability. 3.Results 3.1 Study Population and Baseline Characteristics The development cohort included 5,134 elderly thoracic surgery patients (January 2019 – December 2024). The incidence of PACU hypoxemia was 18.8% (967/5134). Patients were randomly split into a training set (n = 4108, 80%) and an internal validation set (n = 1026, 20%). Baseline characteristics were balanced between sets (all P > 0.05; Eq.A2Table1 ). 3.2 Feature Selection via LASSO Regression LASSO regression selected 10 predictors with non-zero coefficients from 30 candidate variables (Fig. 2 ): age, BMI, preoperative serum albumin, hemoglobin, D-dimer, preoperative PaO₂, surgical site (right lung), anesthesia duration, operative time, and total intraoperative fluid volume. 3.3 Model Performance and Validation Among six ML algorithms, XGBoost demonstrated superior discrimination. AUCs in internal validation ranged from 0.852 to 0.953 (Fig. 3A). The XGBoost model achieved an AUC of 0.988 (95% CI, 0.985–0.991) in training and 0.953 (95% CI, 0.937–0.969) in internal validation. Performance in the external temporal validation cohort was an AUC of 0.706 (95% CI, 0.623–0.789). In internal validation, accuracy was 90.3%, sensitivity 88.3%, and F1-score 0.760 ( Eq.A3Table 2 ). Decision curve analysis confirmed the highest net clinical benefit for XGBoost across relevant thresholds (Fig. 3B). Figure 3. Discrimination and clinical utility assessment of machine learning models in the internal validation cohort. A, Receiver operating characteristic (ROC) curves illustrating the discriminative ability of the six evaluated models. The corresponding AUC values (95% CIs) are listed in the legend. B, Decision curve analysis (DCA) comparing the standardized net benefit of the models against default strategies of treating all or no patients across a range of threshold probabilities. Models above the "All" and "None" lines demonstrate potential clinical value. Abbreviations: AUC, area under the curve; CI, confidence interval; KNN, K-nearest neighbors; ROC, receiver operating characteristic; SVM, support vector machine; XGBoost, extreme gradient boosting. Table 1 Baseline Characteristics of the Study Population Variable Category Variable Level / Description Total (N = 5134) Hypoxemia Group (n = 967) Non-Hypoxemia Group (n = 4167) P Value (Hypoxemia vs Non-Hypoxemia) Training Set (n = 4108) Internal Validation Set (n = 1026) P Value (Training vs Internal) External Validation Set (n = 272) Demographic Age (years) 67.00 [64.00, 71.00] 69.00 [65.00, 73.00] 67.00 [64.00, 71.00] < 0.001 67.00 [64.00, 71.00] 67.00 [64.00, 71.00] 0.151 67.00 [63.00, 71.00] Gender Male 2338 (45.5%) 453 (46.8%) 1885 (45.2%) 0.365 1881 (45.8%) 457 (44.5%) 0.473 137 (50.4%) Female 2796 (54.5%) 514 (53.2%) 2282 (54.8%) 2227 (54.2%) 569 (55.5%) 135 (49.6%) Comorbidities Smoking history Yes 1521 (29.6%) 317 (32.8%) 1204 (28.9%) 0.017 1218 (29.6%) 303 (29.5%) 0.941 53 (19.5%) No 3613 (70.4%) 650 (67.2%) 2963 (71.1%) 2890 (70.4%) 723 (70.5%) 219 (80.5%) Alcohol drinking history Yes 651 (12.7%) 133 (13.8%) 518 (12.4%) 0.265 531 (12.9%) 120 (11.7%) 0.290 24 (8.8%) No 4483 (87.3%) 834 (86.2%) 3649 (87.6%) 3577 (87.1%) 906 (88.3%) 248 (91.2%) Heart disease Yes 257 (5.0%) 78 (8.1%) 179 (4.3%) < 0.001 205 (5.0%) 52 (5.1%) 0.918 43 (15.8%) No 4877 (95.0%) 889 (91.9%) 3988 (95.7%) 3903 (95.0%) 974 (94.9%) 229 (84.2%) COPD Yes 27 (0.5%) 11 (1.1%) 16 (0.4%) 0.004 23 (0.6%) 4 (0.4%) 0.634 6 (2.2%) No 5107 (99.5%) 956 (98.9%) 4151 (99.6%) 4085 (99.4%) 1022 (99.6%) 266 (97.8%) Hypertension Yes 1273 (24.8%) 354 (36.6%) 919 (22.1%) < 0.001 987 (24.0%) 286 (27.9%) 0.011 98 (36.0%) No 3861 (75.2%) 613 (63.4%) 3248 (77.9%) 3121 (76.0%) 740 (72.1%) 174 (64.0%) Asthma Yes 20 (0.4%) 6 (0.6%) 14 (0.3%) 0.201 14 (0.3%) 6 (0.6%) 0.262 3 (1.1%) No 5114 (99.6%) 961 (99.4%) 4153 (99.7%) 4094 (99.7%) 1020 (99.4%) 269 (98.9%) Diabetes Yes 579 (11.3%) 140 (14.5%) 439 (10.5%) < 0.001 458 (11.1%) 121 (11.8%) 0.559 50 (18.4%) No 4555 (88.7%) 827 (85.5%) 3728 (89.5%) 3650 (88.9%) 905 (88.2%) 222 (81.6%) Renal insufficiency Yes 81 (1.6%) 26 (2.7%) 55 (1.3%) 0.002 60 (1.5%) 21 (2.0%) 0.178 8 (2.9%) No 5053 (98.4%) 941 (97.3%) 4112 (98.7%) 4048 (98.5%) 1005 (98.0%) 264 (97.1%) Preoperative Status Pleural effusion Yes 29 (0.6%) 5 (0.5%) 24 (0.6%) 0.826 26 (0.6%) 3 (0.3%) 0.247 1 (0.4%) No 5105 (99.4%) 962 (99.5%) 4143 (99.4%) 4082 (99.4%) 1023 (99.7%) 271 (99.6%) Pulmonary infection Yes 65 (1.3%) 13 (1.3%) 52 (1.2%) 0.809 53 (1.3%) 12 (1.2%) 0.757 1 (0.4%) No 5069 (98.7%) 954 (98.7%) 4115 (98.8%) 4055 (98.7%) 1014 (98.8%) 271 (99.6%) Pulmonary bulla Yes 58 (1.1%) 13 (1.3%) 45 (1.1%) 0.483 51 (1.2%) 7 (0.7%) 0.129 7 (2.6%) No 5076 (98.9%) 954 (98.7%) 4122 (98.9%) 4057 (98.8%) 1019 (99.3%) 265 (97.4%) Laboratory Values ALB (g/L) 34.98 [34.02, 40.91] 34.75 [33.94, 40.72] 39.01 [34.04, 40.94] 0.001 34.97 [34.02, 40.93] 39.02 [34.01, 40.87] 0.636 138.00 [129.00, 149.00] HGB (g/L) 134.00 [122.00, 145.00] 118.00 [112.00, 138.50] 135.00 [126.00, 145.50] < 0.001 134.00 [122.00, 145.00] 133.00 [122.00, 144.00] 0.219 41.86 [39.74, 44.23] D-Dimer (mg/L) 0.50 [0.40, 0.80] 0.70 [0.41, 0.80] 0.50 [0.40, 0.79] 0.149 0.50 [0.40, 0.80] 0.50 [0.40, 0.79] 0.751 0.35 [0.29, 0.46] PaO₂ (mmHg) 91.05 [83.73, 98.90] 86.50 [77.70, 97.00] 91.80 [85.20, 99.30] < 0.001 91.20 [83.90, 99.30] 90.30 [83.45, 98.00] 0.052 87.60 [80.07, 95.32] PaCO₂ (mmHg) 39.90 [38.30, 41.50] 40.00 [38.40, 41.60] 39.90 [38.30, 41.50] 0.184 39.90 [38.30, 41.50] 39.90 [38.20, 41.60] 0.877 39.55 [37.40, 41.30] BMI (kg/m²) 24.22 [22.15, 26.22] 25.56 [23.44, 27.72] 23.88 [21.97, 25.81] < 0.001 24.22 [22.15, 26.22] 24.22 [22.16, 26.12] 0.921 24.09 [22.30, 26.29] Surgical Classification ASA Grade I 16 (0.3%) 1 (0.1%) 15 (0.4%) < 0.001 15 (0.4%) 1 (0.1%) 0.413 1 (0.4%) II 4112 (80.1%) 699 (72.3%) 3413 (81.9%) 3284 (79.9%) 828 (80.7%) 222 (81.6%) III 1006 (19.6%) 267 (27.6%) 739 (17.7%) 809 (19.7%) 197 (19.2%) 49 (18.0%) Surgical type Emergency operation 41 (0.8%) 10 (1.0%) 31 (0.7%) 0.361 36 (0.9%) 5 (0.5%) 0.210 1 (0.4%) Elective surgery 5093 (99.2%) 957 (99.0%) 4136 (99.3%) 4072 (99.1%) 1021 (99.5%) 271 (99.6%) Intraoperative Details Blood loss (mL) 20.00 [20.00, 50.00] 20.00 [20.00, 50.00] 20.00 [20.00, 50.00] 0.402 20.00 [20.00, 50.00] 20.00 [20.00, 50.00] 0.984 80.00 [60.00, 125.00] Infusion volume (mL) 1200.00 [1100.00, 1600.00] 1205.00 [1103.00, 1650.00] 1200.00 [1100.00, 1600.00] < 0.001 1200.00 [1100.00, 1600.00] 1200.00 [1100.00, 1600.00] 0.622 1200.00 [1128.75, 1601.25] Surgical time (hours) 1.33 [0.92, 1.83] 1.33 [0.92, 1.83] 1.33 [0.92, 1.83] 0.383 1.33 [0.92, 1.83] 1.33 [0.92, 1.78] 0.415 1.50 [0.94, 2.00] Anesthesia time (hours) 1.78 [1.33, 2.33] 1.75 [1.33, 2.33] 1.78 [1.33, 2.33] 0.344 1.78 [1.33, 2.33] 1.75 [1.33, 2.25] 0.442 1.92 [1.38, 2.50] Blood transfusion Yes 39 (0.8%) 11 (1.1%) 28 (0.7%) 0.133 31 (0.8%) 8 (0.8%) 0.934 4 (1.5%) No 5095 (99.2%) 956 (98.9%) 4139 (99.3%) 4077 (99.2%) 1018 (99.2%) 268 (98.5%) Surgical level‡ II(0) 9 (0.2%) 3 (0.3%) 6 (0.1%) 0.192 8 (0.2%) 1 (0.1%) 0.688 1 (0.4%) III(1) 272 (5.3%) 43 (4.4%) 229 (5.5%) 213 (5.2%) 59 (5.8%) 11 (4.0%) V(2) 4853 (94.5%) 921 (95.2%) 3932 (94.4%) 3887 (94.6%) 966 (94.2%) 260 (95.6%) Operative method‡ OT 44 (0.9%) 10 (1.0%) 34 (0.8%) 0.507 36 (0.9%) 8 (0.8%) 0.764 1 (0.4%) VATS 5090 (99.1%) 957 (99.0%) 4133 (99.2%) 4072 (99.1%) 1018 (99.2%) 271 (99.6%) Surgical site Right lung 2705 (52.7%) 185 (19.1%) 2520 (60.5%) < 0.001 2150 (52.3%) 555 (54.1%) 0.757 135 (49.6%) Left lung 2308 (45.0%) 732 (75.7%) 1576 (37.8%) 1863 (45.4%) 445 (43.4%) 128 (47.1%) sophagus 12 (0.2%) 6 (0.6%) 6 (0.1%) 10 (0.2%) 2 (0.2%) 1 (0.4%) IV (Multiple) 89 (1.7%) 36 (3.7%) 53 (1.3%) 70 (1.7%) 19 (1.9%) 1 (0.4%) Mediastinum 20 (0.4%) 8 (0.8%) 12 (0.3%) 15 (0.4%) 5 (0.5%) 7 (2.6%) Cut level I 4732 (92.2%) 890 (92.0%) 3842 (92.2%) 0.448 3794 (92.4%) 938 (91.4%) 0.247 180 (66.2%) II 48 (0.9%) 6 (0.6%) 42 (1.0%) 34 (0.8%) 14 (1.4%) 10 (3.7%) III 354 (6.9%) 71 (7.3%) 283 (6.8%) 280 (6.8%) 74 (7.2%) 82 (30.1%) Table Notes :Data are presented as median [interquartile range] for continuous variables and n (%) for categorical variables.M ± SD, mean ± standard ;deviationCategories with very low frequency (e.g., ASA III, Surgical Level 0, etc.) are included for completeness. Some variable names (Surgical_type, Operative_method) appear to measure similar concepts (OT vs. VATS); consider clarifying or merging in the manuscript text based on actual definitions. Abbreviations : ALB, albumin; HGB, hemoglobin; BMI, body mass index; COPD, chronic obstructive pulmonary disease; ASA, American Society of Anesthesiologists; VATS, video-assisted thoracic surgery. Table 2 Performance of machine learning models in the internal test dataset. Model AUC (95% CI) Accuracy (95% CI) Sensitivity (95% CI) Specificity (95% CI) F1 Score XGBoost 0.962 (0.950–0.974) 0.903 (0.885–0.921) 0.883 (0.860–0.906) 0.907 (0.889–0.925) 0.76 Decision Tree 0.945 (0.930–0.960) 0.919 (0.900–0.938) 0.771 (0.745–0.797) 0.950 (0.935–0.965) 0.769 SVM 0.930 (0.915–0.945) 0.895 (0.875–0.915) 0.855 (0.830–0.880) 0.903 (0.885–0.921) 0.739 Random Forest 0.925 (0.910–0.940) 0.896 (0.876–0.916) 0.849 (0.825–0.873) 0.906 (0.888–0.924) 0.74 KNN 0.850 (0.830–0.870) 0.806 (0.780–0.832) 0.827 (0.800–0.854) 0.802 (0.775–0.829) 0.598 Logistic Reg. 0.780 (0.760–0.800) 0.716 (0.690–0.742) 0.838 (0.815–0.861) 0.691 (0.665–0.717) 0.508 3.4 Model Interpretability Shapley additive explanations (SHAP) analysis elucidated the global importance and directional influence of each predictor in the optimal XGBoost model. Ranked by mean absolute SHAP value (Fig. 4A) , the most influential feature was surgical site (right lung), followed by preoperative serum albumin level, D-dimer level, hemoglobin concentration, BMI, age, preoperative PaO₂, total intraoperative fluid volume, anesthesia duration, and operative time. The SHAP summary plot (Fig. 4B) illustrated that right lung surgery, elevated D-dimer, higher BMI, advanced age, and greater intraoperative burden (as indicated by fluid volume and duration) were associated with an increased risk of hypoxemia (positive SHAP values). Conversely, higher preoperative serum albumin, hemoglobin, and PaO₂ levels were associated with a decreased risk (negative SHAP values). Figure 4. SHAP visualization of feature contributions to PACU hypoxemia prediction. (A) SHAP summary plot. Features are listed vertically in descending order of global importance. The horizontal location shows the individual SHAP value (impact on model output); color codes the original feature value (red=high, blue = low). Positive SHAP values indicate a positive association with PACU hypoxemia risk. (B) Global feature importance ranking. Bars represent the mean absolute SHAP value for each feature across all samples, quantifying average predictive contribution. 4. Discussion This study developed and validated a risk prediction model for post-anesthesia care unit (PACU) hypoxemia in elderly patients undergoing thoracic surgery by systematically evaluating six machine learning algorithms. Our results demonstrate that the XGBoost model significantly outperformed other algorithms in terms of discriminative ability, achieving an area under the receiver operating characteristic curve (AUC) of 0.953 in the internal validation cohort, while also maintaining high accuracy and sensitivity. This performance advantage can be attributed to XGBoost’s gradient boosting framework, which is particularly adept at modeling the high-dimensional, non-linear interactions inherent in complex perioperative physiological data [ 27 ][ 28 ][ 29 ] . During external temporal validation, the model achieved an AUC of 0.706 (95% CI: 0.623–0.789). While a predictable performance attenuation was observed compared to the development cohort—a phenomenon commonly attributed to increased population heterogeneity and the limited sample size of external validation sets [ 26 ] —this result confirms the model’s retention of clinically meaningful discriminatory capacity. It further provides preliminary evidence supporting its potential generalizability across institutions. The final model integrates ten predictors spanning three clinically relevant domains: preoperative physiological reserve (HGB, ALB, PaO₂, D-dimer level, BMI, age), surgical profile (thoracic surgical site), and intraoperative stress burden (operative duration, anesthesia duration, total intraoperative fluid volume). These features collectively reflect the unique clinical profile of thoracic surgery while maintaining high practical utility; notably, 60% represent modifiable or intervenable parameters readily extractable from electronic medical record (EMR) systems. This composition supports the model’s potential for seamless integration into clinical workflows as a practical decision-support tool. By providing a standardized, evidence-based risk metric, it could establish a common reference framework for communication and coordinated action among thoracic surgeons, anesthesiologists, and PACU teams, thereby facilitating precision management throughout the perioperative continuum. Feature importance analysis of the XGBoost model indicates that the pathogenesis of PACU hypoxemia in elderly thoracic surgery patients is multifactorial and complex, involving three principal mechanisms: anatomical predisposition, physiological reserve depletion, and iatrogenic stress imposed during the perioperative period. First, anatomical predisposition represents a dominant determinant in the model. Specifically, the feature “surgical site (right lung)” emerged as the primary predictor, accounting for the highest mean SHAP value weight (18.9%). This finding underscores the significant and distinct impact of one-lung ventilation (OLV) on pulmonary hemodynamics and ventilation-perfusion matching in the elderly population. Notably, the right lung, which accounts for approximately 55% of total lung volume and receives a proportionally greater share of pulmonary blood flow, is particularly vulnerable to severe ventilation/perfusion (V/Q) mismatch when subjected to OLV [ 30 ] . The present analysis confirms that this well-established pathophysiological alteration is significantly exacerbated in the elderly surgical population [ 31 ] . Furthermore, age-related decline in lung parenchymal elastic recoil further compromises the ability to maintain adequate oxygenation during right-sided thoracic procedures. Consequently, this physiological compromise implies that the risk of hypoxemia is substantially higher for right-lung surgery compared to left-lung surgery in this population. Therefore, more rigorous lung-protective ventilation strategies should be employed, and the duration of OLV should be minimized whenever feasible. Second, beyond anatomical factors, preoperative nutritional and metabolic reserves constitute a patient's intrinsic physiological resilience. Accordingly, the significant model weight assigned to preoperative serum albumin (ALB) and hemoglobin (HGB) levels corroborates that “physiological frailty” is a key determinant of postoperative hypoxemia in the elderly. This association is linked not only to impaired oxygen-carrying capacity and reduced colloid osmotic pressure [ 32 ] but also reflects a chronic state of nutritional depletion commonly observed in this population. Moreover, the inclusion of D-dimer—a sensitive marker of thromboinflammation—further delineates a fragile physiological baseline in the elderly, characterized by a prothrombotic state, chronic malnutrition, and diminished oxygen-carrying capacity. Critically, the confluence of age-related endothelial dysfunction and surgical stress can readily promote microthrombus formation and increase pulmonary shunting [ 33 ][ 34 ] . These findings collectively highlight that D-dimer assessment in elderly surgical patients should extend beyond its traditional role in macrovascular embolism prevention to also serve as a sensitive indicator for evaluating impaired pulmonary microcirculatory efficiency. Third, and of particular clinical relevance, a principal innovative contribution of this study is the proposal of refined "safety thresholds" specific to elderly thoracic surgery patients. The model-identified thresholds for operative duration, anesthesia duration, and total fluid infusion volume are substantially more conservative than those previously established for younger cohorts, thereby underscoring the markedly reduced tolerance of the elderly to iatrogenic stress. Regarding temporal thresholds, our analysis found that an operative duration exceeding 100 minutes significantly elevates hypoxemia risk—a "warning threshold" considerably lower than the 4-hour benchmark proposed for broader surgical populations [ 35 ] . This discrepancy suggests that the compensatory reserve of pulmonary function in the elderly is severely constrained under the combined stress of one-lung ventilation and surgical trauma. Similarly, an anesthesia duration exceeding 118 minutes emerged as a high-risk threshold, further supporting the notion that prolonged mechanical ventilation may induce diaphragmatic dysfunction in this vulnerable population [ 36 ] . Regarding fluid management, the intraoperative fluid infusion threshold (> 1200 mL) identified in our model is also more restrictive than the 1500 mL benchmark suggested by Chau et al. for general thoracic surgery [ 37 ] . This stricter limit logically reflects the critically reduced cardiopulmonary reserve in elderly patients subjected to one-lung ventilation, which diminishes their tolerance to volume overload. Specifically, excessive crystalloid administration may not only precipitate pulmonary edema but also exacerbate glycocalyx damage, thereby impairing alveolar-capillary diffusion capacity. In summary, the more conservative safety thresholds proposed in this study help delineate a refined operational “red line” for geriatric thoracic anesthesia management, suggesting a paradigm shift toward restrictive and precision-guided strategies for fluid administration and temporal control. Furthermore, our findings caution against the potential misinterpretation of “normal” laboratory values. Although the identified preoperative PaO₂ risk threshold (92 mmHg) falls within the conventionally defined normal range, its selection by the model signifies its predictive relevance. This highlights a clinically underappreciated phenomenon: in elderly patients with diminished physiological reserve [ 30 ] , values at the lower end of the normal spectrum may indicate a state of marginal compensation. Consequently, preoperative evaluation should be particularly vigilant in identifying such subtle signs of compromised reserve that may be obscured by ostensibly normal readings. In conclusion, the predictive model developed in this study not only corroborates established risk factors [ 18 ][ 19 ] but, more significantly, quantifies their relative contributions to precisely delineate the specific pathophysiological vulnerabilities of the elderly thoracic surgery population. By elucidating factors ranging from procedure-specific anatomical risks to intraoperative management thresholds that are more stringent than conventional guidelines, this work provides perioperative teams with an evidence-based, quantifiable early-warning framework tailored to the distinct physiology of geriatric patients. Currently, the clinical management of hypoxemia in elderly thoracic surgery patients within the PACU is predominantly constrained by a “passive response” paradigm. This reactive model not only risks missing the critical early intervention window [ 8 ][ 9 ] , but also fails to provide proactive protection for vulnerable, high-risk patients, potentially entrapping them in a cascade of adverse postoperative outcomes [ 11 ][ 12 ][ 13 ][ 14 ] . Therefore, the fundamental value of the prediction model developed in this study lies in its potential to catalyze a paradigm shift in clinical management from reactive rescue to proactive, continuous risk mitigation. Specifically, the model incorporates factors spanning the entire perioperative continuum, thereby providing a data-driven rationale for dismantling traditional silos and facilitating multidisciplinary team (MDT)-based, continuous care: • Preoperative Optimization Phase: Building upon the identified risk factors, thoracic surgeons can utilize this model as an early screening tool to identify patients with significant physiological vulnerabilities. Consequently, for high-impact modifiable risk factors highlighted by the model—such as hypoalbuminemia, anemia, and hypercoagulable states—targeted prehabilitation strategies can be implemented to mitigate baseline risk prior to surgery. • Intraoperative Precision Management Phase: This phase represents the model's core application for dynamic risk mitigation. Here, informed by the preoperative risk assessment, the anesthesia and surgical teams can implement real-time, model-guided management. This involves direct intervention on key intraoperative predictors, such as adopting a restrictive fluid strategy and minimizing operative and anesthetic duration—strategies directly supported by the stricter thresholds identified in our analysis. Consequently, this approach transforms perioperative risk management from a static prediction into an actionable, dynamic control process. • Postoperative Vigilance Phase: Given the computational efficiency of the model, anesthesiologists can provide PACU teams with a quantifiable risk score during the critical postoperative handover, thereby replacing subjective clinical impressions with objective data. This empowers the receiving team to promptly identify patients at “hidden high risk” and implement preventive interventions—such as early initiation of high-flow nasal oxygen therapy and intensified monitoring—within the crucial first 30 minutes of PACU admission [ 13 ] . Ultimately, such proactive measures can effectively interrupt the trajectory toward hypoxemia. Building upon its clinical utility, this study is equally innovative in its forward-looking design for integration into intelligent healthcare systems. Notably, the model demonstrates strong scalability within modern medical informatics infrastructure. Therefore, its future implementation within Anesthesia Information Management Systems (AIMS) or hospital Electronic Health Record (EHR) platforms would enable automated data extraction and real-time risk computation. For instance,during the surgical closure phase, the system could automatically synthesize preoperative and intraoperative data to generate and forward a risk notification to the PACU in advance of patient arrival. This “digital handover” would substantially reduce clinical cognitive burden and crystallize the envisioned intelligent management paradigm: “information precedes the patient, and preparedness precedes the event.” Ultimately, and most significantly, this research transcends the provision of a mere predictive algorithm; it offers a comprehensive and pragmatic framework for perioperative quality improvement. By quantifying and visualizing risk across the continuum of care, it mandates and facilitates seamless, evidence-based coordination across the entire surgical pathway. Consequently, this integrated approach—from risk identification to proactive management within a smart system—holds the potential to translate predictive insights into tangible clinical actions, thereby delivering substantial improvements in patient outcomes and considerable health economic value. Limitations and Future Directions Notwithstanding the promising findings, several limitations of this study warrant consideration to contextualize the model's application. First, the retrospective design, while enabling analysis of a large single-center cohort, inherently carries the risk of unmeasured selection bias. Furthermore, although the data were derived from a national-level regional medical center—which ensured high-quality documentation—the patient population encompassed a relatively high proportion of elderly and complex cases. Consequently, this demographic focus may constrain the model's direct generalizability to other healthcare settings, such as primary or community hospitals, where patient risk profiles and surgical complexities may differ. Second, the external temporal validation cohort had a limited sample size. Specifically, stringent inclusion/exclusion criteria combined with a defined accrual period resulted in a modest cohort, which yielded relatively wide confidence intervals for the model's performance metrics. While the obtained AUC of 0.706 supports the model's fundamental discriminative capacity, its precise calibration and stability across broader, more heterogeneous populations require further verification. Therefore, future research must prioritize prospective, multicenter external validation. Iterative refinement of the model using larger, geographically diverse datasets will be essential to enhance its clinical robustness and practical utility across varied healthcare environments. Ultimately, such efforts are crucial for translating this predictive framework into a universally reliable tool within the envisioned intelligent perioperative ecosystem. 5. Conclusion This study establishes an XGBoost model as an effective tool for predicting PACU hypoxemia risk in elderly thoracic surgery patients. It integrates ten predictors spanning three vulnerability domains: anatomical-physiological predisposition, diminished metabolic-nutritional reserve, and cumulative iatrogenic stress. Crucially, 60% of the predictors are modifiable, enabling the model to function as a practical, interdisciplinary decision-support tool. This facilitates a shift from reactive management to a proactive, preventive strategy across the perioperative continuum. Declarations Ethics approval and consent to participate This study was approved by the Institutional Review Board of Dalian Medical University (Approval No.: 2025-041). The need for informed consent was waived due to the retrospective design. CRediT authorship contribution statement Yang Xu : Writing – review & editing, Writing – original draft, Investigation,Conceptualization. Anni Wu : Writing – original draft, Software,Methodology,Investigation. Sheng Sun : Investigation, Data curation, Conceptualization. Minzhao Sun : Investigation. Xinyu Zhan : Investigation. Bolun Zhao : Writing – review & editing, Writing – original draft, Methodology, Conceptualization. Funding This work was supported by Education Department of Liaoning Province, China (grant number LJ212410161006);Key Laboratory of Geriatric Long-term Care(Naval Medical University)，Ministry of Education(grant number LNYB-2023-11) Declaration of competing interest The authors declare that they have no competing interests. Acknowledgements The authors thank the clinical teams at the Second Affiliated Hospital of Dalian Medical University for their support. Clinical trial number N ot applicable. Data availability The datasets generated and analysed during the current study are not publicly available due to patient confidentiality and ethical restrictions, but are available from the corresponding author on reasonable request. References Toffaletti JG, Rackley CR. Monitoring oxygen status[J]. Adv Clin Chem. 2016;77:103–24. https://doi.org/10.1016/bs. acc.2016.06.003. Zhuo S, Daniel I, Sessler JE, Dalton PJ, Devereaux A, Shahinyan, Amanda J, Naylor MT, Hutcherson PS, Finnegan V, Tandon. Saeed Darvish-Kazem, Shaan Chugh, Hussain Alzayer, Andrea Kurz; Postoperative Hypoxemia Is Common and Persistent: A Prospective Blinded Observational StudyAnesthesia and analgesia 2015;121(3):709–715 10.1213/ANE.0000000000000836 Leone M, Einav S, Chiumello D, Constantin J-M, De Robertis E. Marcelo Gama De Abreu, Cesare Gregoretti, Samir Jaber, Salvatore Maurizio Maggiore, Paolo Pelosi, Massimiliano Sorbello, Arash Afshari; Noninvasive respiratory support in the hypoxaemic peri-operative/periprocedural patient: A joint ESA/ESICM guidelineEuropean journal of anaesthesiology 2020 04;37(4):265–279 10.1097/EJA.0000000000001166 Jaber S, Molinari N, De Jong A. New method of preoxygenation for orotracheal intubation in patients with hypoxaemic acute respiratory failure in the intensive care unit,non-invasive ventilation combined with apnoeic oxygenation by high flow nasal oxygen:the randomised OPTINIV study protocol[J].BMJ Open,2016,6(8):292–8. Naveed Siddiqui C, Arzola J, Teresi G, Fox L, Guerina. Zeev Friedman; Predictors of desaturation in the postoperative anesthesia care unit: an observational studyJournal of clinical anesthesia 2013;25(8):612–7 10.1016/j.jclinane.2013.04.018 Zhang L, Tang S, Zhang X, Han W, Xue Y, Tian Y, Huang Y. Incidence of hypoxemia in a post-anesthesia care unit and relevant risk factors. Basic Med Clin. 2020;40:1242–46. Chen Yun Z, Jing Z. Risk factors for hypoxemia in patients in the post anesthesia care unit and nursing countermeasures[J]. J Nurs Sci. 2022;37(13):26–9. Alparslan Turan X, Pu. Marcelo Gama de Abreu, Daniel I Sessler; Tidal Volume, Positive End-expiratory Pressure, and Postoperative Hypoxemia: ReplyAnesthesiology 2023 08 01;139(2):234 10.1097/ALN.0000000000004555 Han S, Zeng T, Xie Y, Liu C, Yang L, Yang M. The effect of postoperative hypotension on surgical patients' adverse clinical outcomes: A systematic review and meta-analysis. J Clin Anesth. 2025;107:111987. 10.1016/j.jclinane.2025.111987 . Epub 2025 Aug 30. PMID: 40886448. Remie Saab BP, Wu E, Rivas A, Chiu S, Lozovoskiy C, Ma D, Yang. Alparslan Turan, Daniel I Sessler; Failure to detect ward hypoxaemia and hypotension: contributions of insufficient assessment frequency and patient arousal during nursing assessmentsBritish. J Anaesth. 2021;127(5):760–8. 10.1016/j.bja.2021.06.014 . Andualem AA, Yesuf KA. Incidence and associated factors of postoperative hypoxemia among adult elective surgical patients at Dessie Comprehensive Specialized Hospital: an observational study[J]. Ann Med Surg (Lond). 2022;78:103747. https://doi.org/10.1016/j.amsu.2022.103747 . Jing Tan Y, Feng H, Liu Y, Zhou J, Cai Y, Jiang P, Li L, Gu Z. Song; Effect of press needle stimulation on postoperative pulmonary complications in video-assisted thoracoscopic surgery patients: a randomized controlled trialAnnals of medicine 2025;57(1):2560684 10.1080/07853890.2025.2560684 Fogagnolo A, Grasso S, Dres M, Azzolina D, Dalla Corte F, Dolci G, Tamburini N, De Paoli G, Murgolo F, Pedarzani E, Andalò A, Volta CA, Savino S. Impact of early high flow nasal oxygen on diaphragmatic function and pulmonary complications after thoracic surgery: A randomized clinical trial. J Clin Anesth. 2025;106:111945. 10.1016/j.jclinane.2025.111945 . Epub 2025 Jul 23. PMID: 40706477. Rawley M, Harris E, Pospishil L, Thompson JA. Christian Falyar; Assessing Provider Adherence To A Lung Protective Ventilation Protocol In Patients Undergoing Thoracic Surgery Using One-Lung VentilationAANA journal 2022;90(6):439–445. Bruins SD. Pauline Meng Choo Leong, Shin Yi Ng; Retrospective review of critical incidents in the post-anaesthesia care unit at a major tertiary hospitalSingapore. Med J 2017 08;58(8):497–501 10.11622/smedj.2016126 Huang Qiumei W, Mingming L. Risk factors for hypoxemia in elderly patients after general anesthesia in postanesthesia care unit[J]. J Clin Anesthesiology. 2023;39(06):582–5. Zheng L, Wu X, Gu W, Wang R, Wang J, He H, Wang Z, Yi B. Yi Zhang; Development and validation of a hypoxemia prediction model in middle-aged and elderly outpatients undergoing painless gastroscopyScientific reports 2025;15(1):17965 10.1038/s41598-025-02540-8 DuanXZ. ZhangX,TongDK,JiF,XuKH,HeRZ.Risk factors for and predictive nomogram of post operative hypoxaemia in elderly patients with femoral neck fractures.JIntMedRes.2020;48(10):300060520945132. 10.1177/0300060520945132 .PMID:33028126;PMCID:PMC7550957. WangD,DingX. SuY,YangP,DuX,SunM,HuangX,YueZ,SunF,XieF,LiuC.Incidence,Risk Factors,and Outcomes of Severe Hypoxemia After Cardiac Surgery.Front CardiovascMed.2022Jun28;9:934533. 10.3389/fcvm.2022.934533 .PMID:35837609;PMCID:PMC9273816. Li H, Zhang Y, Cai J, Wang H, Wei R. Risk Factors of Hypoxemia in the Postanesthesia Care Unit After General Anesthesia in Children. J perianesthesia nursing: official J Am Soc PeriAnesthesia Nurses. 2023;38:799–803. Tana C, Siniscalchi C, Cerundolo N, Meschi T, Martelletti P, Tana M, Moffa L. William Wells-Gatnik, Francesco Cipollone, Maria Adele Giamberardino; Smart aging: integrating AI into elderly healthcareBMC geriatrics 2025;25(1):1024 10.1186/s12877-025-06723-w Cristian A-B, Oliver-Forniés P, Pascual-Bellosta A, Ortega-Lucea S. José Manuel Ramírez-Rodriguez, Javier Martínez-Ubieto, Research Group in Anaesthesia, Resuscitation, and Perioperative Medicine of Institute for Health Research Aragón (ISS Aragón); Model for predicting early and late-onset postoperative pulmonary complications in perioperative patients receiving neuromuscular blockade: a secondary analysisScientific reports 2023 03 31;13(1):5234 10.1038/s41598-023-32017-5 Wang D, Ding X, Su Y, Yang P, Du X, Sun M, Huang X, Yue Z. Fuqiang Sun, Fei Xie, and Chao Liu. 2022. 'Incidence, Risk Factors, and Outcomes of Severe Hypoxemia After Cardiac Surgery'. Front Cardiovasc Med, 9: 934533. He Y, Chen Y, Zhang H, Zhou Y. Qiang Fu; Factors associated with postoperative occurrence of hypoxemia in patients undergoing total knee replacement surgery and its predictive model creation and validationMedicine 2025;104(42):e45100 10.1097/MD.0000000000045100 Duan X-Z, Zhang X, Tong D-K, Ji F, Xu K-H, Rong-Zhi H. Risk factors for and predictive nomogram of postoperative hypoxaemia in elderly patients with femoral neck fractures. J Int Med Res. 2020;48:300060520945132. Richard D, Riley JA, Hayden EW, Steyerberg, Karel GM, Moons K, Abrams PA, Kyzas. Núria Malats, Andrew Briggs, Sara Schroter, Douglas G Altman, Harry Hemingway, PROGRESS Group; Prognosis Research Strategy (PROGRESS) 2: prognostic factor researchPLoS medicine 2013;10(2):e1001380 10.1371/journal.pmed.1001380 He J, Wang X, Zhu P, Wang X, Zhang Y, Zhao J, Sun W, Hu K, He W, Xie J. Identification and validation of an explainable early-stage chronic kidney disease prediction model: a multicenter retrospective study. EClinicalMedicine. 2025;84:103286. PMID: 40567347; PMCID: PMC12192353. Liang D, Wang L, Zhong P, Lin J, Chen L, Chen Q, Liu S, Luo Z, Ke C, Lai Y. Perspective: Global Burden of Iodine Deficiency: Insights and Projections to 2050 Using XGBoost and SHAP. Adv Nutr. 2025;16(3):100384. 10.1016/j.advnut.2025.100384 . Epub 2025 Feb 4. PMID: 39914495; PMCID: PMC11909719. Guan C, Gong A, Zhao Y, Yin C, Geng L, Liu L, Yang X, Lu J, Xiao B. Interpretable machine learning model for new-onset atrial fibrillation prediction in critically ill patients: a multi-center study. Crit Care. 2024;28(1):349. 10.1186/s13054-024-05138-0 . PMID: 39473013; PMCID: PMC11523862. Okamoto K, Kochi T, Yoshizaki S, Iiyori N. A Kochi; [Arterial oxygenation during one lung ventilation in obese patients]Masui. Japanese J anesthesiology. 1999;48(2):168–71. Weiss YG, Merin G, Koganov E, Ribo A, Oppenheim-Eden A, Medalion B, Peruanski M, Reider E, Bar-Ziv J, Hanson WC. R Pizov; Postcardiopulmonary bypass hypoxemia: a prospective study on incidence, risk factors, and clinical significanceJournal of cardiothoracic and vascular anesthesia 2000;14(5):506–13 10.1053/jcan.2000.9488 Massey AC. Microcytic anemia. Differential diagnosis and management of iron deficiency anemiaThe. Med Clin North Am. 1992;76(3):549–66. 10.1016/s0025-7125(16)30339-x . Geraldes R, Esiri MM, DeLuca GC, et al. Age-related small vessel disease: a potential contributor to neurodegeneration in multiple sclerosis[J]. Brain Pathol. 2017;27(6):707–22. https://doi.org/10.1111/bpa.12460 . Yu H-Y, Bai Y-P, Song X-C, Miao F, Guo X-L, Wang X, Wen-Ling S. Factors Associated with Acute Pulmonary Embolism in Patients with hypoxemia After off-Pump Coronary Artery Bypass Grafting: A Case-Control Study. J multidisciplinary Healthc. 2024;17:573–83. Peter M, Odor S, Bampoe D, Gilhooly B, Creagh-Brown S, Ramani. ed.) Moonesinghe; Perioperative interventions for prevention of postoperative pulmonary complications: systematic review and meta-analysisBMJ (Clinical research 2020;368:m540 10.1136/bmj.m540 Ewan C, Goligher AH, Jonkman J, Dianti K, Vaporidi, Jeremy R, Beitler, Bhakti K, Patel T, Yoshida S, Jaber M, Dres. Tommaso Mauri, Giacomo Bellani, Alexandre Demoule, Laurent Brochard, Leo Heunks; Clinical strategies for implementing lung and diaphragm-protective ventilation: avoiding insufficient and excessive effortIntensive care medicine 2020 12;46(12):2314–26 10.1007/s00134-020-06288-9 Chau EH, Slinger P. Perioperative fluid management for pulmonary resection surgery and esophagectomy[J]. Semin Cardiothorac Vasc Anesth. 2014;18(1):36–44. https://doi.org/10.1177/1089253213491014 . Additional Declarations No competing interests reported. Supplementary Files Appendix1.docx Appendix2Table1BaselineCharacteristics.docx Appendix3Table2ModelPerformance.docx GlossaryofAbbreviations.docx CentralPicture.png Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8723621","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":591241275,"identity":"c16c54ce-a719-4d87-8bca-61f31b497965","order_by":0,"name":"Yang Xu","email":"","orcid":"","institution":"Dalian Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Xu","suffix":""},{"id":591241276,"identity":"14fd6575-40f1-4471-914d-696eb699160f","order_by":1,"name":"Anni Wu","email":"","orcid":"","institution":"Dalian Medical University","correspondingAuthor":false,"prefix":"","firstName":"Anni","middleName":"","lastName":"Wu","suffix":""},{"id":591241277,"identity":"74ee53e1-b457-4029-b3cb-3216d743c813","order_by":2,"name":"Sheng Sun","email":"","orcid":"","institution":"Dalian Medical University","correspondingAuthor":false,"prefix":"","firstName":"Sheng","middleName":"","lastName":"Sun","suffix":""},{"id":591241278,"identity":"a28ac179-859e-4dc6-9f7c-e2d941f2afc6","order_by":3,"name":"Minzhao Sun","email":"","orcid":"","institution":"Dalian Medical University","correspondingAuthor":false,"prefix":"","firstName":"Minzhao","middleName":"","lastName":"Sun","suffix":""},{"id":591241279,"identity":"6bdb8b74-e131-45d8-a850-0b062f199881","order_by":4,"name":"Xinyu Zhan","email":"","orcid":"","institution":"Dalian Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xinyu","middleName":"","lastName":"Zhan","suffix":""},{"id":591241280,"identity":"63b7dc78-3f7b-4931-a305-707b29890181","order_by":5,"name":"Bolun Zhao","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6ElEQVRIie3PsWrDMBCA4QMVaxF4vaG4r3DFECfghzlR0JSGQpZs7eQpD6DH6CPYHM3k0tVjH6AUQZcOhdbOlsVKt0L1D0KC+5AEkEr9wQgMwOu0U6ptA9XFeYSnnW5s5+9c+Qti+lJMEBsllX4+IO+kqDyT1KQYtDw9zpHVfuOQeykvB2ZZU7YB49ww+7B2vUDbOOuR25GYLaBZzJOXtyO592gfZEnjGiXD8Zaa0QgIEMXJyr/fLLmvr71uoNsTl1nsL1V+2w1hh1eo8o/w+fVd5FoOs2TsAk+OWWR8SoUzhlKpVOo/9wMnQ0xozV/VJAAAAABJRU5ErkJggg==","orcid":"","institution":"Dalian Medical University","correspondingAuthor":true,"prefix":"","firstName":"Bolun","middleName":"","lastName":"Zhao","suffix":""}],"badges":[],"createdAt":"2026-01-28 16:54:07","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8723621/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8723621/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":102854297,"identity":"cd1e4215-8dd7-4bd0-a4c3-524acb0074ba","added_by":"auto","created_at":"2026-02-17 14:47:49","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":5890882,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eStudy design and analytical flowchart.\u003c/strong\u003e A, Derivation of the study cohorts showing inclusion and exclusion criteria for elderly thoracic surgery patients. B, Feature selection process utilizing LASSO regression with 10-fold cross-validation to identify the optimal predictive subset. C, Model development and validation pipeline. The development cohort was randomly split (8:2) for training and internal validation across six machine learning algorithms. Model performance was assessed via discrimination, calibration, and decision curve analysis, followed by temporal external validation in a subsequent 2025 cohort. Abbreviations: AUC, area under the receiver operating characteristic curve; LASSO, least absolute shrinkage and selection operator; NPV, negative predictive value; PACU, post-anesthesia care unit; SHAP, Shapley additive explanations.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-8723621/v1/08d677fe9e4255a29c5c4dc4.png"},{"id":102854300,"identity":"909e3f59-b0c5-48be-a599-131dee7cea16","added_by":"auto","created_at":"2026-02-17 14:47:50","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":4482738,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFeature selection via LASSO regression with 10-fold cross-validation. \u003c/strong\u003eA: Coefffcient path of the LASSO regression model as a function of the regularisation parameter, λ. Co-variation of the 30 variables when the regularisation parameter λ is varied. \u0026nbsp;B: Binomial deviance as a function of the logarithm of the λ parameter for LASSO regression. LASSO, Least Absolute Shrinkage and Selection Operator.The regularisation parameter λ screening process\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-8723621/v1/c2268da2341a9231e6a171d0.png"},{"id":102963138,"identity":"7359ce73-1b68-4199-bdb2-a80c20eaf860","added_by":"auto","created_at":"2026-02-19 04:13:49","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":4816448,"visible":true,"origin":"","legend":"\u003cp\u003eDiscrimination and clinical utility assessment of machine learning models in the internal validation cohort. A, Receiver operating characteristic (ROC) curves illustrating the discriminative ability of the six evaluated models. The corresponding AUC values (95% CIs) are listed in the legend. B, Decision curve analysis (DCA) comparing the standardized net benefit of the models against default strategies of treating all or no patients across a range of threshold probabilities. Models above the \"All\" and \"None\" lines demonstrate potential clinical value. Abbreviations: AUC, area under the curve; CI, confidence interval; KNN, K-nearest neighbors; ROC, receiver operating characteristic; SVM, support vector machine; XGBoost, extreme gradient boosting.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8723621/v1/4e52bf7449841f3e4042e442.png"},{"id":102963154,"identity":"82e693be-cada-4cce-800f-bd3835b4bb11","added_by":"auto","created_at":"2026-02-19 04:13:57","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":4962858,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSHAP visualization of feature contributions to PACU hypoxemia prediction.\u003c/strong\u003e(A) SHAP summary plot. Features are listed vertically in descending order of global importance. The horizontal location shows the individual SHAP value (impact on model output); color codes the original feature value (red=high, blue=low). Positive SHAP values indicate a positive association with \u003cstrong\u003ePACU hypoxemia\u003c/strong\u003e risk. (B) Global feature importance ranking. Bars represent the mean absolute SHAP value for each feature across all samples, quantifying average predictive contribution.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-8723621/v1/512025022f1770647f12588d.png"},{"id":109102122,"identity":"c6d377e3-32d8-4663-ab13-3d048d7444c5","added_by":"auto","created_at":"2026-05-12 14:31:14","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":20483322,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8723621/v1/d7b96ea3-11ea-446b-ae30-fef033d25e68.pdf"},{"id":102854292,"identity":"add8c0bf-5bf4-4f27-ab15-b07d0a9c2624","added_by":"auto","created_at":"2026-02-17 14:47:49","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":15685,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix1.docx","url":"https://assets-eu.researchsquare.com/files/rs-8723621/v1/9f01e7123b016f696999b50a.docx"},{"id":102854293,"identity":"32c6932c-4e6a-419b-9bb7-a921688d4def","added_by":"auto","created_at":"2026-02-17 14:47:49","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":29310,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix2Table1BaselineCharacteristics.docx","url":"https://assets-eu.researchsquare.com/files/rs-8723621/v1/cc17e93d180384c12fdde5c3.docx"},{"id":102854294,"identity":"ed1ce16a-9b28-482e-8770-56a94602852e","added_by":"auto","created_at":"2026-02-17 14:47:49","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":13475,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix3Table2ModelPerformance.docx","url":"https://assets-eu.researchsquare.com/files/rs-8723621/v1/3d5ee9f8d519977653ffdd1a.docx"},{"id":102854299,"identity":"6e3884b1-a70c-43ac-abca-d937d51ad2f7","added_by":"auto","created_at":"2026-02-17 14:47:49","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":12392,"visible":true,"origin":"","legend":"","description":"","filename":"GlossaryofAbbreviations.docx","url":"https://assets-eu.researchsquare.com/files/rs-8723621/v1/b6362a6e41a47547e7c65e47.docx"},{"id":102854298,"identity":"06138840-f71b-464b-8b4e-eb4507db8393","added_by":"auto","created_at":"2026-02-17 14:47:49","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":5890882,"visible":true,"origin":"","legend":"","description":"","filename":"CentralPicture.png","url":"https://assets-eu.researchsquare.com/files/rs-8723621/v1/17799df8f7dcb4e898c471a6.png"}],"financialInterests":"No competing interests reported.","formattedTitle":"Development and Temporal External Validation of an XGBoost-Based Clinical Prediction Model for PACU Hypoxemia in Elderly Thoracic Surgery Patients","fulltext":[{"header":"1.Introduction","content":"\u003cp\u003eHypoxemia is a common complication in the post-anesthesia care unit (PACU), primarily due to ventilation-perfusion mismatch, increased shunt, and impaired diffusion\u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e][\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e][\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/sup\u003e. Its incidence after general anesthesia ranges from 2.79% to 21.79% \u003csup\u003e[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e][\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e][\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/sup\u003e, with approximately 69.8% of episodes occurring within the first 30 minutes of PACU arrival\u003csup\u003e[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e. Timely intervention is critical, as persistent hypoxemia can lead to arrhythmias, cerebral ischemia, multi-organ dysfunction, and increased mortality\u003csup\u003e[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e][\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e][\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eDespite strategies like lung-protective ventilation and high-flow nasal oxygen \u003csup\u003e[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e][\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e][\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e][\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/sup\u003e, management remains reactive, often initiated only after overt desaturation. This delay fails to address early, silent respiratory deterioration, potentially leading to avoidable organ injury and inefficient resource use.\u003c/p\u003e \u003cp\u003eElderly patients undergoing thoracic surgery represent a particularly high-risk subgroup. The combination of one-lung ventilation, lung resection, and chest wall disruption, superimposed on age-related decline in pulmonary compliance and physiological reserve, results in hypoxemia incidences reported from 22.1% to over 40%\u003csup\u003e[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e][\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/sup\u003e. A major clinical challenge is the lack of an objective tool to rapidly identify high-risk patients at the operating room-to-PACU transition. This impedes seamless handover and proactive, risk-stratified care.\u003c/p\u003e \u003cp\u003eWhile risk factors like advanced age, obesity, smoking, and fluid balance are known\u003csup\u003e[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e][\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e][\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e][\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e][\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/sup\u003e, and machine learning (ML) shows promise for clinical prediction, existing models lack specificity for this populatio\u003csup\u003e[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003en. General models\u003csup\u003e[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/sup\u003eare unvalidated in thoracic surgery, whereas models for other specific cohorts\u003csup\u003e[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e][\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e][\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]\u003c/sup\u003e demonstrate the utility of population-tailored prediction. A significant gap exists: no validated model predicts PACU hypoxemia specifically for elderly thoracic surgery patients.\u003c/p\u003e \u003cp\u003eTo address this, we leveraged a large-scale dataset and employed a comparative modeling approach, using logistic regression and five ML algorithms to develop and validate a risk prediction model for this high-risk cohort. We integrated the SHAP (Shapley Additive exPlanations) framework to ensure interpretability. The goal is to provide a practical decision-support tool for the perioperative team, facilitating a shift from reactive management to proactive, coordinated intervention to improve outcomes.\u003c/p\u003e"},{"header":"2.Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Study Design and Ethical Approval\u003c/h2\u003e \u003cp\u003eThis was a retrospective cohort study for model development, with subsequent prospective temporal validation. The protocol was approved by the Institutional Review Board of Dalian Medical University (Approval No. : 2025-041) and adhered to the Declaration of Helsinki.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Study Population and Selection Criteria\u003c/h2\u003e \u003cp\u003ePatients aged\u0026thinsp;\u0026ge;\u0026thinsp;60 years who underwent non-cardiac thoracic surgery under general anesthesia with endotracheal intubation and were admitted directly to the PACU between January 2019 and July 2025 were included. Exclusion criteria were: 1) direct postoperative transfer to ICU due to pre-existing critical conditions (e.g., need for ECMO, end-stage cardiopulmonary disease); 2)\u0026thinsp;\u0026gt;\u0026thinsp;20% missing key data or irreconcilable data inconsistencies.Cardiac/vascular surgery patients (routinely admitted to specialized ICUs) were not part of the target population. The screening process is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Outcome Definition\u003c/h2\u003e \u003cp\u003eThe primary outcome was PACU hypoxemia, defined as meeting any of the following while breathing room air: SpO₂ \u0026lt; 90%, PaO₂ \u0026lt; 60 mmHg, or PaO₂/FiO₂ \u0026le; 300 mmHg \u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e][\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e][\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4 Sample size calculation\u003c/h2\u003e \u003cp\u003e This study followed the TRIPOD guidelines for transparent reporting of prediction models. The sample size was calculated using Riley et al.\u0026rsquo;s method to ensure statistical adequacy.The minimum sample size was estimated as:\u003c/p\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;p / [π(1 - π)(1 - R\u0026sup2;)]\u003c/p\u003e \u003cp\u003eWhere: p\u0026thinsp;=\u0026thinsp;30 predictors; π\u0026thinsp;=\u0026thinsp;hypoxemia incidence (41.2%, from prior research); R\u0026sup2; = 0.20 (common clinical model value).The base calculation yielded 155 cases.Applying the Events Per Variable (EPV\u0026thinsp;\u0026ge;\u0026thinsp;10) principle for robustness, at least 300 hypoxemia events were required (30 predictors \u0026times; 10). Given the 41.2% incidence, the total sample size needed was:n\u0026thinsp;=\u0026thinsp;300 / 0.412\u0026thinsp;\u0026asymp;\u0026thinsp;729 cases.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2.5 Data and Cohorts\u003c/h2\u003e \u003cp\u003eAfter double-entry verification and de-identification, 30 predictor variables were prespecified based on literature and expert consultation \u003cb\u003e(Eq.A.1).\u003c/b\u003e\u003c/p\u003e \u003cp\u003eTwo cohorts were established:(1) \u003cb\u003eDevelopment Cohort\u003c/b\u003e: 5,134 patients (January 2019 \u0026ndash; December 2024) for model training/internal validation.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003e(2)\u003cb\u003eExternal Temporal Validation Cohort\u003c/b\u003e\u003c/strong\u003e \u003cp\u003e272 consecutive prospective patients (January \u0026ndash; July 2025), used to assess generalizability per the PROBAST framework\u003csup\u003e[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e2.6 Data Preprocessing and Modeling\u003c/h2\u003e \u003cp\u003eMissing data (\u0026lt;\u0026thinsp;5%) followed an MCAR pattern; complete-case analysis was used. Continuous variables were normalized (min-max scaling).\u003c/p\u003e \u003cp\u003eThe development cohort was split into training (80%) and internal validation (20%) sets. A two-step strategy was employed:\u003c/p\u003e \u003cp\u003e \u003cstrong\u003e(1)\u003cb\u003eFeature Selection\u003c/b\u003e\u003c/strong\u003e \u003cp\u003eLeast absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation selected a parsimonious predictor subset (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB).\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003e(2)\u003cb\u003eModel Construction\u003c/b\u003e\u003c/strong\u003e \u003cp\u003eSix algorithms\u0026mdash;Logistic Regression, XGBoost, SVM, RF, KNN, and LightGBM\u0026mdash;were trained using the selected features. Hyperparameters were optimized via grid search. Analyses used R (v4.5.0); p\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was significant.\u003c/p\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e2.7 Model Evaluation and Interpretability\u003c/h2\u003e \u003cp\u003ePerformance was evaluated using the AUC, calibration plots (Hosmer-Lemeshow test, Brier score), sensitivity, specificity, accuracy, NPV, and F1-score. Generalizability was tested via external temporal validation.\u003c/p\u003e \u003cp\u003eThe SHAP (Shapley Additive exPlanations) framework was applied to provide global feature importance rankings and individualized visual explanations for predictions (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC), enhancing model interpretability.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"3.Results","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Study Population and Baseline Characteristics\u003c/h2\u003e \u003cp\u003eThe development cohort included 5,134 elderly thoracic surgery patients (January 2019 \u0026ndash; December 2024). The incidence of PACU hypoxemia was 18.8% (967/5134). Patients were randomly split into a training set (n\u0026thinsp;=\u0026thinsp;4108, 80%) and an internal validation set (n\u0026thinsp;=\u0026thinsp;1026, 20%). Baseline characteristics were balanced between sets (all P\u0026thinsp;\u0026gt;\u0026thinsp;0.05; \u003cb\u003eEq.A2Table1\u003c/b\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Feature Selection via LASSO Regression\u003c/h2\u003e \u003cp\u003eLASSO regression selected 10 predictors with non-zero coefficients from 30 candidate variables (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e): age, BMI, preoperative serum albumin, hemoglobin, D-dimer, preoperative PaO₂, surgical site (right lung), anesthesia duration, operative time, and total intraoperative fluid volume.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003e3.3 Model Performance and Validation\u003c/b\u003e \u003c/p\u003e \u003cp\u003eAmong six ML algorithms, XGBoost demonstrated superior discrimination. AUCs in internal validation ranged from 0.852 to 0.953 (Fig.\u0026nbsp;3A). The XGBoost model achieved an AUC of 0.988 (95% CI, 0.985\u0026ndash;0.991) in training and 0.953 (95% CI, 0.937\u0026ndash;0.969) in internal validation. Performance in the external temporal validation cohort was an AUC of 0.706 (95% CI, 0.623\u0026ndash;0.789). In internal validation, accuracy was 90.3%, sensitivity 88.3%, and F1-score 0.760 ( \u003cb\u003eEq.A3Table 2\u003c/b\u003e). Decision curve analysis confirmed the highest net clinical benefit for XGBoost across relevant thresholds (Fig.\u0026nbsp;3B).\u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;3. Discrimination and clinical utility assessment of machine learning models in the internal validation cohort. A, Receiver operating characteristic (ROC) curves illustrating the discriminative ability of the six evaluated models. The corresponding AUC values (95% CIs) are listed in the legend. B, Decision curve analysis (DCA) comparing the standardized net benefit of the models against default strategies of treating all or no patients across a range of threshold probabilities. Models above the \"All\" and \"None\" lines demonstrate potential clinical value. Abbreviations: AUC, area under the curve; CI, confidence interval; KNN, K-nearest neighbors; ROC, receiver operating characteristic; SVM, support vector machine; XGBoost, extreme gradient boosting.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eBaseline Characteristics of the Study Population\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"11\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c11\" colnum=\"11\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable Category\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLevel / Description\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTotal (N\u0026thinsp;=\u0026thinsp;5134)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eHypoxemia Group (n\u0026thinsp;=\u0026thinsp;967)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eNon-Hypoxemia Group (n\u0026thinsp;=\u0026thinsp;4167)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eP Value (Hypoxemia vs Non-Hypoxemia)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eTraining Set (n\u0026thinsp;=\u0026thinsp;4108)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eInternal Validation Set (n\u0026thinsp;=\u0026thinsp;1026)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c10\"\u003e \u003cp\u003eP Value (Training vs Internal)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c11\"\u003e \u003cp\u003eExternal Validation Set (n\u0026thinsp;=\u0026thinsp;272)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDemographic\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAge (years)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e67.00 [64.00, 71.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e69.00 [65.00, 73.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e67.00 [64.00, 71.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e67.00 [64.00, 71.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e67.00 [64.00, 71.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.151\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e67.00 [63.00, 71.00]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGender\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2338 (45.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e453 (46.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1885 (45.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.365\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e1881 (45.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e457 (44.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.473\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e137 (50.4%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2796 (54.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e514 (53.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2282 (54.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e2227 (54.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e569 (55.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e135 (49.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eComorbidities\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSmoking history\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1521 (29.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e317 (32.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1204 (28.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e0.017\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e1218 (29.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e303 (29.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.941\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e53 (19.5%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3613 (70.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e650 (67.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2963 (71.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e2890 (70.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e723 (70.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e219 (80.5%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAlcohol drinking history\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e651 (12.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e133 (13.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e518 (12.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.265\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e531 (12.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e120 (11.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.290\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e24 (8.8%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4483 (87.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e834 (86.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e3649 (87.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e3577 (87.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e906 (88.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e248 (91.2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHeart disease\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e257 (5.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e78 (8.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e179 (4.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e205 (5.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e52 (5.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.918\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e43 (15.8%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4877 (95.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e889 (91.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e3988 (95.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e3903 (95.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e974 (94.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e229 (84.2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCOPD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e27 (0.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e11 (1.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e16 (0.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e0.004\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e23 (0.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e4 (0.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.634\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e6 (2.2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5107 (99.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e956 (98.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e4151 (99.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e4085 (99.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1022 (99.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e266 (97.8%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHypertension\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1273 (24.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e354 (36.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e919 (22.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e987 (24.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e286 (27.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e\u003cb\u003e0.011\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e98 (36.0%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3861 (75.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e613 (63.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e3248 (77.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e3121 (76.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e740 (72.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e174 (64.0%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAsthma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e20 (0.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e6 (0.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e14 (0.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.201\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e14 (0.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e6 (0.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.262\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e3 (1.1%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5114 (99.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e961 (99.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e4153 (99.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e4094 (99.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1020 (99.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e269 (98.9%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDiabetes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e579 (11.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e140 (14.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e439 (10.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e458 (11.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e121 (11.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.559\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e50 (18.4%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4555 (88.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e827 (85.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e3728 (89.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e3650 (88.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e905 (88.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e222 (81.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRenal insufficiency\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e81 (1.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e26 (2.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e55 (1.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e0.002\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e60 (1.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e21 (2.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.178\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e8 (2.9%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5053 (98.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e941 (97.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e4112 (98.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e4048 (98.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1005 (98.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e264 (97.1%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePreoperative Status\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePleural effusion\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e29 (0.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e5 (0.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e24 (0.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.826\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e26 (0.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e3 (0.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.247\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e1 (0.4%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5105 (99.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e962 (99.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e4143 (99.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e4082 (99.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1023 (99.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e271 (99.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePulmonary infection\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e65 (1.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e13 (1.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e52 (1.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.809\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e53 (1.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e12 (1.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.757\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e1 (0.4%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5069 (98.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e954 (98.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e4115 (98.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e4055 (98.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1014 (98.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e271 (99.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePulmonary bulla\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e58 (1.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e13 (1.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e45 (1.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.483\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e51 (1.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e7 (0.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.129\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e7 (2.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5076 (98.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e954 (98.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e4122 (98.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e4057 (98.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1019 (99.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e265 (97.4%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eLaboratory Values\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eALB (g/L)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e34.98 [34.02, 40.91]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e34.75 [33.94, 40.72]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e39.01 [34.04, 40.94]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e34.97 [34.02, 40.93]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e39.02 [34.01, 40.87]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.636\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e138.00 [129.00, 149.00]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHGB (g/L)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e134.00 [122.00, 145.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e118.00 [112.00, 138.50]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e135.00 [126.00, 145.50]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e134.00 [122.00, 145.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e133.00 [122.00, 144.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.219\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e41.86 [39.74, 44.23]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eD-Dimer (mg/L)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.50 [0.40, 0.80]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.70 [0.41, 0.80]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.50 [0.40, 0.79]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.149\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.50 [0.40, 0.80]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.50 [0.40, 0.79]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.751\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.35 [0.29, 0.46]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePaO₂ (mmHg)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e91.05 [83.73, 98.90]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e86.50 [77.70, 97.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e91.80 [85.20, 99.30]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e91.20 [83.90, 99.30]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e90.30 [83.45, 98.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.052\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e87.60 [80.07, 95.32]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePaCO₂ (mmHg)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e39.90 [38.30, 41.50]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e40.00 [38.40, 41.60]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e39.90 [38.30, 41.50]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.184\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e39.90 [38.30, 41.50]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e39.90 [38.20, 41.60]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.877\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e39.55 [37.40, 41.30]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBMI (kg/m\u0026sup2;)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e24.22 [22.15, 26.22]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e25.56 [23.44, 27.72]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e23.88 [21.97, 25.81]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e24.22 [22.15, 26.22]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e24.22 [22.16, 26.12]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.921\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e24.09 [22.30, 26.29]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSurgical Classification\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eASA Grade\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e16 (0.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1 (0.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e15 (0.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e15 (0.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1 (0.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.413\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e1 (0.4%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eII\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4112 (80.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e699 (72.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e3413 (81.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e3284 (79.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e828 (80.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e222 (81.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eIII\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1006 (19.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e267 (27.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e739 (17.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e809 (19.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e197 (19.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e49 (18.0%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSurgical type\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEmergency operation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e41 (0.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e10 (1.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e31 (0.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.361\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e36 (0.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e5 (0.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.210\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e1 (0.4%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eElective surgery\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5093 (99.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e957 (99.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e4136 (99.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e4072 (99.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1021 (99.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e271 (99.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eIntraoperative Details\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBlood loss (mL)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e20.00 [20.00, 50.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e20.00 [20.00, 50.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e20.00 [20.00, 50.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.402\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e20.00 [20.00, 50.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e20.00 [20.00, 50.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.984\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e80.00 [60.00, 125.00]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eInfusion volume (mL)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1200.00 [1100.00, 1600.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1205.00 [1103.00, 1650.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1200.00 [1100.00, 1600.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e1200.00 [1100.00, 1600.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1200.00 [1100.00, 1600.00]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.622\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e1200.00 [1128.75, 1601.25]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSurgical time (hours)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.33 [0.92, 1.83]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.33 [0.92, 1.83]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.33 [0.92, 1.83]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.383\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e1.33 [0.92, 1.83]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1.33 [0.92, 1.78]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.415\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e1.50 [0.94, 2.00]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAnesthesia time (hours)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.78 [1.33, 2.33]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.75 [1.33, 2.33]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.78 [1.33, 2.33]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.344\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e1.78 [1.33, 2.33]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1.75 [1.33, 2.25]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.442\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e1.92 [1.38, 2.50]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBlood transfusion\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e39 (0.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e11 (1.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e28 (0.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.133\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e31 (0.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e8 (0.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.934\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e4 (1.5%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5095 (99.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e956 (98.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e4139 (99.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e4077 (99.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1018 (99.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e268 (98.5%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSurgical level\u0026Dagger;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eII(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e9 (0.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e3 (0.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e6 (0.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.192\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e8 (0.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1 (0.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.688\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e1 (0.4%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eIII(1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e272 (5.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e43 (4.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e229 (5.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e213 (5.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e59 (5.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e11 (4.0%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eV(2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4853 (94.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e921 (95.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e3932 (94.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e3887 (94.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e966 (94.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e260 (95.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOperative method\u0026Dagger;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eOT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e44 (0.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e10 (1.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e34 (0.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.507\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e36 (0.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e8 (0.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.764\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e1 (0.4%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eVATS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5090 (99.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e957 (99.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e4133 (99.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e4072 (99.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e1018 (99.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e271 (99.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSurgical site\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRight lung\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2705 (52.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e185 (19.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2520 (60.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e2150 (52.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e555 (54.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.757\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e135 (49.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLeft lung\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2308 (45.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e732 (75.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1576 (37.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e1863 (45.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e445 (43.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e128 (47.1%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003esophagus\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e12 (0.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e6 (0.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e6 (0.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e10 (0.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e2 (0.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e1 (0.4%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eIV (Multiple)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e89 (1.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e36 (3.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e53 (1.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e70 (1.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e19 (1.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e1 (0.4%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMediastinum\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e20 (0.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e8 (0.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e12 (0.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e15 (0.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e5 (0.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e7 (2.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCut level\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4732 (92.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e890 (92.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e3842 (92.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.448\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e3794 (92.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e938 (91.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.247\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e180 (66.2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eII\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e48 (0.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e6 (0.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e42 (1.0%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e34 (0.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e14 (1.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e10 (3.7%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eIII\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e354 (6.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e71 (7.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e283 (6.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e280 (6.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e74 (7.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e82 (30.1%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"11\"\u003e\u003cb\u003eTable Notes\u003c/b\u003e:Data are presented as \u003cb\u003emedian [interquartile range]\u003c/b\u003e for continuous variables and \u003cb\u003en (%)\u003c/b\u003e for categorical variables.M\u0026thinsp;\u0026plusmn;\u0026thinsp;SD, mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard ;deviationCategories with very low frequency (e.g., ASA III, Surgical Level 0, etc.) are included for completeness. Some variable names (Surgical_type, Operative_method) appear to measure similar concepts (OT vs. VATS); consider clarifying or merging in the manuscript text based on actual definitions.\u003cb\u003eAbbreviations\u003c/b\u003e: ALB, albumin; HGB, hemoglobin; BMI, body mass index; COPD, chronic obstructive pulmonary disease; ASA, American Society of Anesthesiologists; VATS, video-assisted thoracic surgery.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePerformance of machine learning models in the internal test dataset.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAUC (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAccuracy (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSensitivity (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSpecificity (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eF1 Score\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eXGBoost\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.962 (0.950\u0026ndash;0.974)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.903 (0.885\u0026ndash;0.921)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.883 (0.860\u0026ndash;0.906)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.907 (0.889\u0026ndash;0.925)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.76\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDecision Tree\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.945 (0.930\u0026ndash;0.960)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.919 (0.900\u0026ndash;0.938)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.771 (0.745\u0026ndash;0.797)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.950 (0.935\u0026ndash;0.965)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.769\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSVM\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.930 (0.915\u0026ndash;0.945)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.895 (0.875\u0026ndash;0.915)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.855 (0.830\u0026ndash;0.880)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.903 (0.885\u0026ndash;0.921)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e0.739\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eRandom Forest\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.925 (0.910\u0026ndash;0.940)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.896 (0.876\u0026ndash;0.916)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.849 (0.825\u0026ndash;0.873)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.906 (0.888\u0026ndash;0.924)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e0.74\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eKNN\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.850 (0.830\u0026ndash;0.870)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.806 (0.780\u0026ndash;0.832)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.827 (0.800\u0026ndash;0.854)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.802 (0.775\u0026ndash;0.829)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e0.598\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eLogistic Reg.\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.780 (0.760\u0026ndash;0.800)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.716 (0.690\u0026ndash;0.742)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.838 (0.815\u0026ndash;0.861)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.691 (0.665\u0026ndash;0.717)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e0.508\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Model Interpretability\u003c/h2\u003e \u003cp\u003eShapley additive explanations (SHAP) analysis elucidated the global importance and directional influence of each predictor in the optimal XGBoost model. Ranked by mean absolute SHAP value \u003cb\u003e(Fig.\u0026nbsp;4A)\u003c/b\u003e, the most influential feature was surgical site (right lung), followed by preoperative serum albumin level, D-dimer level, hemoglobin concentration, BMI, age, preoperative PaO₂, total intraoperative fluid volume, anesthesia duration, and operative time. The SHAP summary plot \u003cb\u003e(Fig.\u0026nbsp;4B)\u003c/b\u003e illustrated that right lung surgery, elevated D-dimer, higher BMI, advanced age, and greater intraoperative burden (as indicated by fluid volume and duration) were associated with an increased risk of hypoxemia (positive SHAP values). Conversely, higher preoperative serum albumin, hemoglobin, and PaO₂ levels were associated\u003c/p\u003e \u003cp\u003ewith a decreased risk (negative SHAP values).\u003c/p\u003e \u003cp\u003e \u003cb\u003eFigure\u0026nbsp;4. SHAP visualization of feature contributions to PACU hypoxemia prediction.\u003c/b\u003e (A) SHAP summary plot. Features are listed vertically in descending order of global importance. The horizontal location shows the individual SHAP value (impact on model output); color codes the original feature value (red=high, blue\u0026thinsp;=\u0026thinsp;low). Positive SHAP values indicate a positive association with \u003cb\u003ePACU hypoxemia\u003c/b\u003e risk. (B) Global feature importance ranking. Bars represent the mean absolute SHAP value for each feature across all samples, quantifying average predictive contribution.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eThis study developed and validated a risk prediction model for post-anesthesia care unit (PACU) hypoxemia in elderly patients undergoing thoracic surgery by systematically evaluating six machine learning algorithms. Our results demonstrate that the XGBoost model significantly outperformed other algorithms in terms of discriminative ability, achieving an area under the receiver operating characteristic curve (AUC) of 0.953 in the internal validation cohort, while also maintaining high accuracy and sensitivity. This performance advantage can be attributed to XGBoost\u0026rsquo;s gradient boosting framework, which is particularly adept at modeling the high-dimensional, non-linear interactions inherent in complex perioperative physiological data\u003csup\u003e[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e][\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e][\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eDuring external temporal validation, the model achieved an AUC of 0.706 (95% CI: 0.623\u0026ndash;0.789). While a predictable performance attenuation was observed compared to the development cohort\u0026mdash;a phenomenon commonly attributed to increased population heterogeneity and the limited sample size of external validation sets\u003csup\u003e[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]\u003c/sup\u003e\u0026mdash;this result confirms the model\u0026rsquo;s retention of clinically meaningful discriminatory capacity. It further provides preliminary evidence supporting its potential generalizability across institutions.\u003c/p\u003e \u003cp\u003eThe final model integrates ten predictors spanning three clinically relevant domains: preoperative physiological reserve (HGB, ALB, PaO₂, D-dimer level, BMI, age), surgical profile (thoracic surgical site), and intraoperative stress burden (operative duration, anesthesia duration, total intraoperative fluid volume). These features collectively reflect the unique clinical profile of thoracic surgery while maintaining high practical utility; notably, 60% represent modifiable or intervenable parameters readily extractable from electronic medical record (EMR) systems. This composition supports the model\u0026rsquo;s potential for seamless integration into clinical workflows as a practical decision-support tool. By providing a standardized, evidence-based risk metric, it could establish a common reference framework for communication and coordinated action among thoracic surgeons, anesthesiologists, and PACU teams, thereby facilitating precision management throughout the perioperative continuum.\u003c/p\u003e \u003cp\u003eFeature importance analysis of the XGBoost model indicates that the pathogenesis of PACU hypoxemia in elderly thoracic surgery patients is multifactorial and complex, involving three principal mechanisms: anatomical predisposition, physiological reserve depletion, and iatrogenic stress imposed during the perioperative period.\u003c/p\u003e \u003cp\u003eFirst, anatomical predisposition represents a dominant determinant in the model. Specifically, the feature \u0026ldquo;surgical site (right lung)\u0026rdquo; emerged as the primary predictor, accounting for the highest mean SHAP value weight (18.9%). This finding underscores the significant and distinct impact of one-lung ventilation (OLV) on pulmonary hemodynamics and ventilation-perfusion matching in the elderly population. Notably, the right lung, which accounts for approximately 55% of total lung volume and receives a proportionally greater share of pulmonary blood flow, is particularly vulnerable to severe ventilation/perfusion (V/Q) mismatch when subjected to OLV\u003csup\u003e[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]\u003c/sup\u003e. The present analysis confirms that this well-established pathophysiological alteration is significantly exacerbated in the elderly surgical population\u003csup\u003e[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]\u003c/sup\u003e. Furthermore, age-related decline in lung parenchymal elastic recoil further compromises the ability to maintain adequate oxygenation during right-sided thoracic procedures. Consequently, this physiological compromise implies that the risk of hypoxemia is substantially higher for right-lung surgery compared to left-lung surgery in this population. Therefore, more rigorous lung-protective ventilation strategies should be employed, and the duration of OLV should be minimized whenever feasible.\u003c/p\u003e \u003cp\u003eSecond, beyond anatomical factors, preoperative nutritional and metabolic reserves constitute a patient's intrinsic physiological resilience. Accordingly, the significant model weight assigned to preoperative serum albumin (ALB) and hemoglobin (HGB) levels corroborates that \u0026ldquo;physiological frailty\u0026rdquo; is a key determinant of postoperative hypoxemia in the elderly. This association is linked not only to impaired oxygen-carrying capacity and reduced colloid osmotic pressure\u003csup\u003e[\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]\u003c/sup\u003e but also reflects a chronic state of nutritional depletion commonly observed in this population. Moreover, the inclusion of D-dimer\u0026mdash;a sensitive marker of thromboinflammation\u0026mdash;further delineates a fragile physiological baseline in the elderly, characterized by a prothrombotic state, chronic malnutrition, and diminished oxygen-carrying capacity. Critically, the confluence of age-related endothelial dysfunction and surgical stress can readily promote microthrombus formation and increase pulmonary shunting\u003csup\u003e[\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e][\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]\u003c/sup\u003e. These findings collectively highlight that D-dimer assessment in elderly surgical patients should extend beyond its traditional role in macrovascular embolism prevention to also serve as a sensitive indicator for evaluating impaired pulmonary microcirculatory efficiency.\u003c/p\u003e \u003cp\u003eThird, and of particular clinical relevance, a principal innovative contribution of this study is the proposal of refined \"safety thresholds\" specific to elderly thoracic surgery patients. The model-identified thresholds for operative duration, anesthesia duration, and total fluid infusion volume are substantially more conservative than those previously established for younger cohorts, thereby underscoring the markedly reduced tolerance of the elderly to iatrogenic stress. Regarding temporal thresholds, our analysis found that an operative duration exceeding 100 minutes significantly elevates hypoxemia risk\u0026mdash;a \"warning threshold\" considerably lower than the 4-hour benchmark proposed for broader surgical populations\u003csup\u003e[\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]\u003c/sup\u003e. This discrepancy suggests that the compensatory reserve of pulmonary function in the elderly is severely constrained under the combined stress of one-lung ventilation and surgical trauma. Similarly, an anesthesia duration exceeding 118 minutes emerged as a high-risk threshold, further supporting the notion that prolonged mechanical ventilation may induce diaphragmatic dysfunction in this vulnerable population \u003csup\u003e[\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]\u003c/sup\u003e. Regarding fluid management, the intraoperative fluid infusion threshold (\u0026gt;\u0026thinsp;1200 mL) identified in our model is also more restrictive than the 1500 mL benchmark suggested by Chau et al. for general thoracic surgery \u003csup\u003e[\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]\u003c/sup\u003e. This stricter limit logically reflects the critically reduced cardiopulmonary reserve in elderly patients subjected to one-lung ventilation, which diminishes their tolerance to volume overload. Specifically, excessive crystalloid administration may not only precipitate pulmonary edema but also exacerbate glycocalyx damage, thereby impairing alveolar-capillary diffusion capacity.\u003c/p\u003e \u003cp\u003eIn summary, the more conservative safety thresholds proposed in this study help delineate a refined operational \u0026ldquo;red line\u0026rdquo; for geriatric thoracic anesthesia management, suggesting a paradigm shift toward restrictive and precision-guided strategies for fluid administration and temporal control. Furthermore, our findings caution against the potential misinterpretation of \u0026ldquo;normal\u0026rdquo; laboratory values. Although the identified preoperative PaO₂ risk threshold (92 mmHg) falls within the conventionally defined normal range, its selection by the model signifies its predictive relevance. This highlights a clinically underappreciated phenomenon: in elderly patients with diminished physiological reserve\u003csup\u003e[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]\u003c/sup\u003e, values at the lower end of the normal spectrum may indicate a state of marginal compensation. Consequently, preoperative evaluation should be particularly vigilant in identifying such subtle signs of compromised reserve that may be obscured by ostensibly normal readings.\u003c/p\u003e \u003cp\u003eIn conclusion, the predictive model developed in this study not only corroborates established risk factors\u003csup\u003e[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e][\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/sup\u003e but, more significantly, quantifies their relative contributions to precisely delineate the specific pathophysiological vulnerabilities of the elderly thoracic surgery population. By elucidating factors ranging from procedure-specific anatomical risks to intraoperative management thresholds that are more stringent than conventional guidelines, this work provides perioperative teams with an evidence-based, quantifiable early-warning framework tailored to the distinct physiology of geriatric patients.\u003c/p\u003e \u003cp\u003eCurrently, the clinical management of hypoxemia in elderly thoracic surgery patients within the PACU is predominantly constrained by a \u0026ldquo;passive response\u0026rdquo; paradigm. This reactive model not only risks missing the critical early intervention window\u003csup\u003e[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e][\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/sup\u003e, but also fails to provide proactive protection for vulnerable, high-risk patients, potentially entrapping them in a cascade of adverse postoperative outcomes \u003csup\u003e[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e][\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e][\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e][\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eTherefore, the fundamental value of the prediction model developed in this study lies in its potential to catalyze a paradigm shift in clinical management from reactive rescue to proactive, continuous risk mitigation. Specifically, the model incorporates factors spanning the entire perioperative continuum, thereby providing a data-driven rationale for dismantling traditional silos and facilitating multidisciplinary team (MDT)-based, continuous care:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003e\u0026bull; Preoperative Optimization Phase: Building upon the identified risk factors, thoracic surgeons can utilize this model as an early screening tool to identify patients with significant physiological vulnerabilities. Consequently, for high-impact modifiable risk factors highlighted by the model\u0026mdash;such as hypoalbuminemia, anemia, and hypercoagulable states\u0026mdash;targeted prehabilitation strategies can be implemented to mitigate baseline risk prior to surgery.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e\u0026bull; Intraoperative Precision Management Phase: This phase represents the model's core application for dynamic risk mitigation. Here, informed by the preoperative risk assessment, the anesthesia and surgical teams can implement real-time, model-guided management. This involves direct intervention on key intraoperative predictors, such as adopting a restrictive fluid strategy and minimizing operative and anesthetic duration\u0026mdash;strategies directly supported by the stricter thresholds identified in our analysis. Consequently, this approach transforms perioperative risk management from a static prediction into an actionable, dynamic control process.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e\u0026bull; Postoperative Vigilance Phase: Given the computational efficiency of the model, anesthesiologists can provide PACU teams with a quantifiable risk score during the critical postoperative handover, thereby replacing subjective clinical impressions with objective data. This empowers the receiving team to promptly identify patients at \u0026ldquo;hidden high risk\u0026rdquo; and implement preventive interventions\u0026mdash;such as early initiation of high-flow nasal oxygen therapy and intensified monitoring\u0026mdash;within the crucial first 30 minutes of PACU admission \u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e. Ultimately, such proactive measures can effectively interrupt the trajectory toward hypoxemia.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eBuilding upon its clinical utility, this study is equally innovative in its forward-looking design for integration into intelligent healthcare systems. Notably, the model demonstrates strong scalability within modern medical informatics infrastructure. Therefore, its future implementation within Anesthesia Information Management Systems (AIMS) or hospital Electronic Health Record (EHR) platforms would enable automated data extraction and real-time risk computation. For instance,during the surgical closure phase, the system could automatically synthesize preoperative and intraoperative data to generate and forward a risk notification to the PACU in advance of patient arrival. This \u0026ldquo;digital handover\u0026rdquo; would substantially reduce clinical cognitive burden and crystallize the envisioned intelligent management paradigm: \u0026ldquo;information precedes the patient, and preparedness precedes the event.\u0026rdquo;\u003c/p\u003e \u003cp\u003eUltimately, and most significantly, this research transcends the provision of a mere predictive algorithm; it offers a comprehensive and pragmatic framework for perioperative quality improvement. By quantifying and visualizing risk across the continuum of care, it mandates and facilitates seamless, evidence-based coordination across the entire surgical pathway. Consequently, this integrated approach\u0026mdash;from risk identification to proactive management within a smart system\u0026mdash;holds the potential to translate predictive insights into tangible clinical actions, thereby delivering substantial improvements in patient outcomes and considerable health economic value.\u003c/p\u003e \u003cp\u003e \u003cb\u003eLimitations and Future Directions\u003c/b\u003e \u003c/p\u003e \u003cp\u003eNotwithstanding the promising findings, several limitations of this study warrant consideration to contextualize the model's application. First, the retrospective design, while enabling analysis of a large single-center cohort, inherently carries the risk of unmeasured selection bias. Furthermore, although the data were derived from a national-level regional medical center\u0026mdash;which ensured high-quality documentation\u0026mdash;the patient population encompassed a relatively high proportion of elderly and complex cases. Consequently, this demographic focus may constrain the model's direct generalizability to other healthcare settings, such as primary or community hospitals, where patient risk profiles and surgical complexities may differ.\u003c/p\u003e \u003cp\u003eSecond, the external temporal validation cohort had a limited sample size. Specifically, stringent inclusion/exclusion criteria combined with a defined accrual period resulted in a modest cohort, which yielded relatively wide confidence intervals for the model's performance metrics. While the obtained AUC of 0.706 supports the model's fundamental discriminative capacity, its precise calibration and stability across broader, more heterogeneous populations require further verification.\u003c/p\u003e \u003cp\u003eTherefore, future research must prioritize prospective, multicenter external validation. Iterative refinement of the model using larger, geographically diverse datasets will be essential to enhance its clinical robustness and practical utility across varied healthcare environments. Ultimately, such efforts are crucial for translating this predictive framework into a universally reliable tool within the envisioned intelligent perioperative ecosystem.\u003c/p\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eThis study establishes an XGBoost model as an effective tool for predicting PACU hypoxemia risk in elderly thoracic surgery patients. It integrates ten predictors spanning three vulnerability domains: anatomical-physiological predisposition, diminished metabolic-nutritional reserve, and cumulative iatrogenic stress. Crucially, 60% of the predictors are modifiable, enabling the model to function as a practical, interdisciplinary decision-support tool. This facilitates a shift from reactive management to a proactive, preventive strategy across the perioperative continuum.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was approved by the Institutional Review Board of Dalian Medical University (Approval No.: 2025-041). The need for informed consent was waived due to the retrospective design.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCRediT authorship contribution statement\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eYang Xu\u003c/strong\u003e: Writing \u0026ndash; review \u0026amp; editing, Writing \u0026ndash; original draft, Investigation,Conceptualization.\u0026nbsp;\u003cstrong\u003eAnni Wu\u003c/strong\u003e:\u0026nbsp;Writing\u0026nbsp;\u0026ndash;\u0026nbsp;original draft, Software,Methodology,Investigation.\u0026nbsp;\u003cstrong\u003eSheng Sun\u003c/strong\u003e:\u0026nbsp;Investigation, Data curation, Conceptualization. \u003cstrong\u003eMinzhao\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eSun\u003c/strong\u003e:\u0026nbsp;Investigation.\u0026nbsp;\u003cstrong\u003eXinyu Zhan\u003c/strong\u003e :\u0026nbsp;Investigation.\u0026nbsp;\u003cstrong\u003eBolun Zhao\u003c/strong\u003e:\u0026nbsp;Writing\u0026nbsp;\u0026ndash;\u0026nbsp;review\u0026nbsp;\u0026amp;\u0026nbsp;editing, Writing\u0026nbsp;\u0026ndash;\u0026nbsp;original draft, Methodology, Conceptualization.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by\u0026nbsp;Education Department of Liaoning Province, China (grant number LJ212410161006);Key Laboratory of Geriatric Long-term Care(Naval Medical University)，Ministry of Education(grant number LNYB-2023-11)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of competing interest\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors thank the clinical teams at the Second Affiliated Hospital of Dalian Medical University for their support.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical trial number\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;N\u003c/strong\u003eot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated and analysed during the current study are not publicly available due to patient confidentiality and ethical restrictions, but are available from the corresponding author on reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eToffaletti JG, Rackley CR. Monitoring oxygen status[J]. Adv Clin Chem. 2016;77:103\u0026ndash;24. https://doi.org/10.1016/bs. acc.2016.06.003.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhuo S, Daniel I, Sessler JE, Dalton PJ, Devereaux A, Shahinyan, Amanda J, Naylor MT, Hutcherson PS, Finnegan V, Tandon. Saeed Darvish-Kazem, Shaan Chugh, Hussain Alzayer, Andrea Kurz; Postoperative Hypoxemia Is Common and Persistent: A Prospective Blinded Observational StudyAnesthesia and analgesia 2015;121(3):709\u0026ndash;715 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1213/ANE.0000000000000836\u003c/span\u003e\u003cspan address=\"10.1213/ANE.0000000000000836\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeone M, Einav S, Chiumello D, Constantin J-M, De Robertis E. Marcelo Gama De Abreu, Cesare Gregoretti, Samir Jaber, Salvatore Maurizio Maggiore, Paolo Pelosi, Massimiliano Sorbello, Arash Afshari; Noninvasive respiratory support in the hypoxaemic peri-operative/periprocedural patient: A joint ESA/ESICM guidelineEuropean journal of anaesthesiology 2020 04;37(4):265\u0026ndash;279 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1097/EJA.0000000000001166\u003c/span\u003e\u003cspan address=\"10.1097/EJA.0000000000001166\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJaber S, Molinari N, De Jong A. New method of preoxygenation for orotracheal intubation in patients with hypoxaemic acute respiratory failure in the intensive care unit,non-invasive ventilation combined with apnoeic oxygenation by high flow nasal oxygen:the randomised OPTINIV study protocol[J].BMJ Open,2016,6(8):292\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNaveed Siddiqui C, Arzola J, Teresi G, Fox L, Guerina. Zeev Friedman; Predictors of desaturation in the postoperative anesthesia care unit: an observational studyJournal of clinical anesthesia 2013;25(8):612\u0026ndash;7 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jclinane.2013.04.018\u003c/span\u003e\u003cspan address=\"10.1016/j.jclinane.2013.04.018\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang L, Tang S, Zhang X, Han W, Xue Y, Tian Y, Huang Y. Incidence of hypoxemia in a post-anesthesia care unit and relevant risk factors. Basic Med Clin. 2020;40:1242\u0026ndash;46.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen Yun Z, Jing Z. Risk factors for hypoxemia in patients in the post anesthesia care unit and nursing countermeasures[J]. J Nurs Sci. 2022;37(13):26\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlparslan Turan X, Pu. Marcelo Gama de Abreu, Daniel I Sessler; Tidal Volume, Positive End-expiratory Pressure, and Postoperative Hypoxemia: ReplyAnesthesiology 2023 08 01;139(2):234 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1097/ALN.0000000000004555\u003c/span\u003e\u003cspan address=\"10.1097/ALN.0000000000004555\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHan S, Zeng T, Xie Y, Liu C, Yang L, Yang M. The effect of postoperative hypotension on surgical patients' adverse clinical outcomes: A systematic review and meta-analysis. J Clin Anesth. 2025;107:111987. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jclinane.2025.111987\u003c/span\u003e\u003cspan address=\"10.1016/j.jclinane.2025.111987\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Epub 2025 Aug 30. PMID: 40886448.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRemie Saab BP, Wu E, Rivas A, Chiu S, Lozovoskiy C, Ma D, Yang. Alparslan Turan, Daniel I Sessler; Failure to detect ward hypoxaemia and hypotension: contributions of insufficient assessment frequency and patient arousal during nursing assessmentsBritish. J Anaesth. 2021;127(5):760\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.bja.2021.06.014\u003c/span\u003e\u003cspan address=\"10.1016/j.bja.2021.06.014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAndualem AA, Yesuf KA. Incidence and associated factors of postoperative hypoxemia among adult elective surgical patients at Dessie Comprehensive Specialized Hospital: an observational study[J]. Ann Med Surg (Lond). 2022;78:103747. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.amsu.2022.103747\u003c/span\u003e\u003cspan address=\"10.1016/j.amsu.2022.103747\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJing Tan Y, Feng H, Liu Y, Zhou J, Cai Y, Jiang P, Li L, Gu Z. Song; Effect of press needle stimulation on postoperative pulmonary complications in video-assisted thoracoscopic surgery patients: a randomized controlled trialAnnals of medicine 2025;57(1):2560684 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1080/07853890.2025.2560684\u003c/span\u003e\u003cspan address=\"10.1080/07853890.2025.2560684\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFogagnolo A, Grasso S, Dres M, Azzolina D, Dalla Corte F, Dolci G, Tamburini N, De Paoli G, Murgolo F, Pedarzani E, Andal\u0026ograve; A, Volta CA, Savino S. Impact of early high flow nasal oxygen on diaphragmatic function and pulmonary complications after thoracic surgery: A randomized clinical trial. J Clin Anesth. 2025;106:111945. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jclinane.2025.111945\u003c/span\u003e\u003cspan address=\"10.1016/j.jclinane.2025.111945\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Epub 2025 Jul 23. PMID: 40706477.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRawley M, Harris E, Pospishil L, Thompson JA. Christian Falyar; Assessing Provider Adherence To A Lung Protective Ventilation Protocol In Patients Undergoing Thoracic Surgery Using One-Lung VentilationAANA journal 2022;90(6):439\u0026ndash;445.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBruins SD. Pauline Meng Choo Leong, Shin Yi Ng; Retrospective review of critical incidents in the post-anaesthesia care unit at a major tertiary hospitalSingapore. Med J 2017 08;58(8):497\u0026ndash;501 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.11622/smedj.2016126\u003c/span\u003e\u003cspan address=\"10.11622/smedj.2016126\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang Qiumei W, Mingming L. Risk factors for hypoxemia in elderly patients after general anesthesia in postanesthesia care unit[J]. J Clin Anesthesiology. 2023;39(06):582\u0026ndash;5.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZheng L, Wu X, Gu W, Wang R, Wang J, He H, Wang Z, Yi B. Yi Zhang; Development and validation of a hypoxemia prediction model in middle-aged and elderly outpatients undergoing painless gastroscopyScientific reports 2025;15(1):17965 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-025-02540-8\u003c/span\u003e\u003cspan address=\"10.1038/s41598-025-02540-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDuanXZ. ZhangX,TongDK,JiF,XuKH,HeRZ.Risk factors for and predictive nomogram of post operative hypoxaemia in elderly patients with femoral neck fractures.JIntMedRes.2020;48(10):300060520945132.\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1177/0300060520945132\u003c/span\u003e\u003cspan address=\"10.1177/0300060520945132\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.PMID:33028126;PMCID:PMC7550957.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWangD,DingX. SuY,YangP,DuX,SunM,HuangX,YueZ,SunF,XieF,LiuC.Incidence,Risk Factors,and Outcomes of Severe Hypoxemia After Cardiac Surgery.Front CardiovascMed.2022Jun28;9:934533.\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fcvm.2022.934533\u003c/span\u003e\u003cspan address=\"10.3389/fcvm.2022.934533\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.PMID:35837609;PMCID:PMC9273816.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi H, Zhang Y, Cai J, Wang H, Wei R. Risk Factors of Hypoxemia in the Postanesthesia Care Unit After General Anesthesia in Children. J perianesthesia nursing: official J Am Soc PeriAnesthesia Nurses. 2023;38:799\u0026ndash;803.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTana C, Siniscalchi C, Cerundolo N, Meschi T, Martelletti P, Tana M, Moffa L. William Wells-Gatnik, Francesco Cipollone, Maria Adele Giamberardino; Smart aging: integrating AI into elderly healthcareBMC geriatrics 2025;25(1):1024 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12877-025-06723-w\u003c/span\u003e\u003cspan address=\"10.1186/s12877-025-06723-w\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCristian A-B, Oliver-Forni\u0026eacute;s P, Pascual-Bellosta A, Ortega-Lucea S. Jos\u0026eacute; Manuel Ram\u0026iacute;rez-Rodriguez, Javier Mart\u0026iacute;nez-Ubieto, Research Group in Anaesthesia, Resuscitation, and Perioperative Medicine of Institute for Health Research Arag\u0026oacute;n (ISS Arag\u0026oacute;n); Model for predicting early and late-onset postoperative pulmonary complications in perioperative patients receiving neuromuscular blockade: a secondary analysisScientific reports 2023 03 31;13(1):5234 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-023-32017-5\u003c/span\u003e\u003cspan address=\"10.1038/s41598-023-32017-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang D, Ding X, Su Y, Yang P, Du X, Sun M, Huang X, Yue Z. Fuqiang Sun, Fei Xie, and Chao Liu. 2022. 'Incidence, Risk Factors, and Outcomes of Severe Hypoxemia After Cardiac Surgery'. Front Cardiovasc Med, 9: 934533.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHe Y, Chen Y, Zhang H, Zhou Y. Qiang Fu; Factors associated with postoperative occurrence of hypoxemia in patients undergoing total knee replacement surgery and its predictive model creation and validationMedicine 2025;104(42):e45100 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1097/MD.0000000000045100\u003c/span\u003e\u003cspan address=\"10.1097/MD.0000000000045100\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDuan X-Z, Zhang X, Tong D-K, Ji F, Xu K-H, Rong-Zhi H. Risk factors for and predictive nomogram of postoperative hypoxaemia in elderly patients with femoral neck fractures. J Int Med Res. 2020;48:300060520945132.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRichard D, Riley JA, Hayden EW, Steyerberg, Karel GM, Moons K, Abrams PA, Kyzas. N\u0026uacute;ria Malats, Andrew Briggs, Sara Schroter, Douglas G Altman, Harry Hemingway, PROGRESS Group; Prognosis Research Strategy (PROGRESS) 2: prognostic factor researchPLoS medicine 2013;10(2):e1001380 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1371/journal.pmed.1001380\u003c/span\u003e\u003cspan address=\"10.1371/journal.pmed.1001380\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHe J, Wang X, Zhu P, Wang X, Zhang Y, Zhao J, Sun W, Hu K, He W, Xie J. Identification and validation of an explainable early-stage chronic kidney disease prediction model: a multicenter retrospective study. EClinicalMedicine. 2025;84:103286. PMID: 40567347; PMCID: PMC12192353.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiang D, Wang L, Zhong P, Lin J, Chen L, Chen Q, Liu S, Luo Z, Ke C, Lai Y. Perspective: Global Burden of Iodine Deficiency: Insights and Projections to 2050 Using XGBoost and SHAP. Adv Nutr. 2025;16(3):100384. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.advnut.2025.100384\u003c/span\u003e\u003cspan address=\"10.1016/j.advnut.2025.100384\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Epub 2025 Feb 4. PMID: 39914495; PMCID: PMC11909719.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuan C, Gong A, Zhao Y, Yin C, Geng L, Liu L, Yang X, Lu J, Xiao B. Interpretable machine learning model for new-onset atrial fibrillation prediction in critically ill patients: a multi-center study. Crit Care. 2024;28(1):349. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s13054-024-05138-0\u003c/span\u003e\u003cspan address=\"10.1186/s13054-024-05138-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 39473013; PMCID: PMC11523862.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOkamoto K, Kochi T, Yoshizaki S, Iiyori N. A Kochi; [Arterial oxygenation during one lung ventilation in obese patients]Masui. Japanese J anesthesiology. 1999;48(2):168\u0026ndash;71.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWeiss YG, Merin G, Koganov E, Ribo A, Oppenheim-Eden A, Medalion B, Peruanski M, Reider E, Bar-Ziv J, Hanson WC. R Pizov; Postcardiopulmonary bypass hypoxemia: a prospective study on incidence, risk factors, and clinical significanceJournal of cardiothoracic and vascular anesthesia 2000;14(5):506\u0026ndash;13 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1053/jcan.2000.9488\u003c/span\u003e\u003cspan address=\"10.1053/jcan.2000.9488\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMassey AC. Microcytic anemia. Differential diagnosis and management of iron deficiency anemiaThe. Med Clin North Am. 1992;76(3):549\u0026ndash;66. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/s0025-7125(16)30339-x\u003c/span\u003e\u003cspan address=\"10.1016/s0025-7125(16)30339-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGeraldes R, Esiri MM, DeLuca GC, et al. Age-related small vessel disease: a potential contributor to neurodegeneration in multiple sclerosis[J]. Brain Pathol. 2017;27(6):707\u0026ndash;22. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/bpa.12460\u003c/span\u003e\u003cspan address=\"10.1111/bpa.12460\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYu H-Y, Bai Y-P, Song X-C, Miao F, Guo X-L, Wang X, Wen-Ling S. Factors Associated with Acute Pulmonary Embolism in Patients with hypoxemia After off-Pump Coronary Artery Bypass Grafting: A Case-Control Study. J multidisciplinary Healthc. 2024;17:573\u0026ndash;83.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePeter M, Odor S, Bampoe D, Gilhooly B, Creagh-Brown S, Ramani. ed.) Moonesinghe; Perioperative interventions for prevention of postoperative pulmonary complications: systematic review and meta-analysisBMJ (Clinical research 2020;368:m540 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/bmj.m540\u003c/span\u003e\u003cspan address=\"10.1136/bmj.m540\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEwan C, Goligher AH, Jonkman J, Dianti K, Vaporidi, Jeremy R, Beitler, Bhakti K, Patel T, Yoshida S, Jaber M, Dres. Tommaso Mauri, Giacomo Bellani, Alexandre Demoule, Laurent Brochard, Leo Heunks; Clinical strategies for implementing lung and diaphragm-protective ventilation: avoiding insufficient and excessive effortIntensive care medicine 2020 12;46(12):2314\u0026ndash;26 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00134-020-06288-9\u003c/span\u003e\u003cspan address=\"10.1007/s00134-020-06288-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChau EH, Slinger P. Perioperative fluid management for pulmonary resection surgery and esophagectomy[J]. Semin Cardiothorac Vasc Anesth. 2014;18(1):36\u0026ndash;44. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/1089253213491014\u003c/span\u003e\u003cspan address=\"10.1177/1089253213491014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Hypoxemia, Thoracic Surgical Procedures, Aged, Postanesthesia Care Unit, Machine Learning, Risk Assessment","lastPublishedDoi":"10.21203/rs.3.rs-8723621/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8723621/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjective\u003c/h2\u003e \u003cp\u003eTo develop and validate a machine learning-based risk prediction model for postoperative hypoxemia in elderly patients undergoing thoracic surgery in the post-anesthesia care unit (PACU), aiming to provide a decision-support tool for perioperative precision prevention.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eThis study involved a retrospective development cohort of 5134 elderly thoracic surgery patients from a tertiary academic medical center between 2019 and 2024(split 8:2 into training and internal validation sets), and a prospective external validation cohort of 272 patients in 2025. Feature selection was performed using LASSO regression. Six machine learning algorithms, including Logistic Regression, XGBoost, and LightGBM, were trained and compared. Model performance was assessed using area under the receiver operating characteristic curve (AUC), calibration plots, and decision curve analysis (DCA). SHAP values were used for model interpretability.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe incidence of hypoxemia was 18.8%(967/5134). LASSO regression identified 10 independent predictors categorized into thoracic surgery site, age, BMI, preoperative albumin, hemoglobin, D-dimer, PaO₂,duration of anesthesia and surgery, intraoperative fluid volume.The AUC values across the six machine learning models ranged from 0.852 to 0.953. XGBoost demonstrated the best performance, with AUCs of 0.988 in the training set, 0.953 in the internal validation set, and 0.706 in the external validation set, indicating good generalizability.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eThe XGBoost-based prediction model accurately identifies elderly thoracic surgery patients at high risk for PACU hypoxemia. Incorporating multiple modifiable clinical indicators, this tool facilitates early risk stratification and proactive, continuous, multidisciplinary interventions by thoracic surgery, anesthesia, and PACU teams.\u003c/p\u003e","manuscriptTitle":"Development and Temporal External Validation of an XGBoost-Based Clinical Prediction Model for PACU Hypoxemia in Elderly Thoracic Surgery Patients","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-17 14:47:44","doi":"10.21203/rs.3.rs-8723621/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"7d4ef804-f3e3-4765-be20-e3a4bb86465e","owner":[],"postedDate":"February 17th, 2026","published":true,"recentEditorialEvents":[{"type":"decision","content":"Rejected","date":"2026-05-12T14:09:47+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-12T02:45:27+00:00","index":192,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-05-12T14:28:03+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-17 14:47:44","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8723621","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8723621","identity":"rs-8723621","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}