Construction and Validation of an Interpretable Machine Learning Model for Predicting Diabetes Risk in COPD Patients

Construction and Validation of an Interpretable Machine Learning Model for Predicting Diabetes Risk in COPD Patients | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Construction and Validation of an Interpretable Machine Learning Model for Predicting Diabetes Risk in COPD Patients Lingpin Pang, Siyan Xu, Yingxin Wang, Tao Huang, Qian Xian, Wenjia Lin, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7033945/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 10 You are reading this latest preprint version Abstract Objective To develop a machine learning (ML)-based prediction model for identifying high-risk diabetic individuals among COPD patients, thereby facilitating early and personalized management of this complication. Methods Data from COPD patients in the MIMIC-IV database were split into training (70%) and validation (30%) sets. LASSO regression and logistic regression were used to screen 49 variables, and six ML algorithms were employed to construct and internally validate the prediction model. Model performance was evaluated using multiple metrics, followed by external validation. Finally, SHAP (SHapley Additive exPlanations) analysis was performed for interpretability. Results All six ML algorithms demonstrated excellent performance in the training, testing, and validation sets, as evidenced by ROC curve analysis, with LightGBM showing the best overall performance. Feature importance analysis revealed that marital status, blood glucose level, and insurance type were the top three factors influencing diabetes development in COPD patients. Conclusion This study developed an interpretable ML-based risk prediction model for diabetes in COPD patients. The model provides clinicians with a novel tool for early personalized intervention, ultimately improving patient prognosis. Biological sciences/Computational biology and bioinformatics Health sciences/Diseases Health sciences/Endocrinology Health sciences/Health care Health sciences/Medical research Health sciences/Risk factors Chronic obstructive pulmonary disease Diabetes Prediction model Machine learning Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Chronic obstructive pulmonary disease (COPD) remains a major global contributor to morbidity, mortality, and healthcare expenditures worldwide. As of 2015, approximately 73 billion people were estimated to be affected by COPD globally. With the accelerating trend of population aging, the disease burden of COPD is projected to increase substantially in the coming decades 1 . A recent systematic review suggests that COPD may serve as a precursor to diabetes progression, particularly highlighting that patients with severe to very severe COPD face substantially elevated risks of diabetes onset 2 . COPD and diabetes reciprocally serve as risk factors for each other, with epidemiological studies demonstrating increasing prevalence rates of either condition following diagnosis of the other 3 . Emerging evidence reveals intertwined pathophysiological mechanisms between COPD and type 2 diabetes mellitus (T2DM) 4 . Notably, while neutrophils are not the sole effector immune cells in COPD, atherosclerosis, or T2DM, they significantly outnumber other immune cell populations in pulmonary secretions of COPD patients. Animal and cellular models demonstrate that neutrophil-derived reactive substances can induce the full spectrum of pulmonary pathologies (ranging from chronic bronchitis to emphysema), with their levels showing significant correlation with disease severity and progressive lung function decline 5 . These findings underscore the importance of advocating for regular health screenings and lifestyle modifications in COPD patients, with particular emphasis on early intervention for T2DM prevention and management. This study aims to develop a machine learning-based risk prediction model to identify high-risk individuals for diabetes among COPD patients, thereby facilitating early personalized management of this comorbidity. Materials and Methods Study Design and Data Source This study aims to develop and validate a machine learning prediction model based on MIMIC-IV data. The research data were obtained from MIMIC-IV (Medical Information Mart for Intensive Care IV), an internationally recognized public medical database. As an upgraded version of MIMIC-III released in 2015, this database features broader temporal coverage and more refined data structure 6 . This study extracted clinical data records from 5,767 general ward patients for analysis, along with 4,711 ICU patient records from the same database for external validation. The open-source dataset was established based on COPD diagnostic codes (ICD-10: 'J44', 'J440', 'J441', 'J449') and comprehensively incorporated demographic characteristics including age, sex, race, insurance type, language, and marital status; lifestyle factors such as smoking status and drink wine; laboratory parameters comprising white blood cell count, red blood cell count, platelet count, hemoglobin level, red cell distribution width, hematocrit, sodium, potassium, total calcium, chloride, glucose, anion gap, prothrombin time (PT), partial thromboplastin time (PTT), international normalized ratio (INR), total bilirubin, alanine aminotransferase (ALT), aspartate aminotransferase (AST), blood urea nitrogen, and creatinine; documented comorbidities including hypertension, chronic kidney disease (CKD), hyperlipidemia, acute renal failure (ARF), heart failure (HF), myocardial infarction, tuberculosis, pneumonia, stroke, pulmonary embolism, COVID-19, cirrhosis, hepatitis, and malignant tumors; medication records covering antihypertensives, glucocorticoids, nephrotoxic drugs, and immunosuppressants; as well as hospitalization duration and mortality outcomes including in-hospital death status. Dr. Xishi Sun possesses certified access to this database (Record ID 12112229) and was responsible for data extraction. The study inclusion criteria were: (1) diagnosis of chronic obstructive pulmonary disease according to the International Classification of Diseases (ICD); (2) first admission to general ward; and (3) age between 18 and 100 years. Exclusion criteria comprised: (1) patients initially admitted to ICU or with multiple general ward admissions; (2) age below 18 or above 100 years; and (3) patients with clinical data containing > 10% missing values. Data Preparation After extracting the patient data, variables with more than 70% missing values were removed, and the random forest imputation method was applied to fill in the missing data. Based on the exclusion and inclusion criteria, a total of 4,710 patients were ultimately enrolled. A fixed random seed was set, and stratified random sampling was used to divide the dataset into training and testing sets at a 7:3 ratio. Additionally, data from 4,711 COPD patients in the ICU were extracted from the MIMIC-IV database for external validation. The analyses were implemented using R (version 4.4.1) and Python (version 3.12.4). Categorical variables (such as hypertension and heart failure) were processed using label encoding. Statistical Analysis Prior to statistical analysis, descriptive statistics were performed for all baseline variables. Normally distributed continuous variables were analyzed using the t-test and expressed as mean ± standard deviation (SD), while non-normally distributed continuous variables were assessed using the Mann-Whitney U test (rank-sum test) and reported as median (interquartile range, IQR). Categorical variables were compared using the chi-square (χ²) test and presented as counts (percentages). A p-value < 0.05 was considered statistically significant. Feature Selection Feature selection was performed using R software, with a p-value < 0.05 considered statistically significant. The model development and evaluation process proceeded as follows: (1) The Lasso algorithm was first applied for preliminary variable screening, employing 10-fold cross-validation to enhance reliability. At the 1 standard deviation lambda value, 17 variables were selected. (2) Multivariate and stepwise logistic regression analyses were then used to further refine these 17 variables, ultimately identifying 14 significant predictors for inclusion in subsequent analyses. These predictive variables included: age, insurance type, language, marital status, red cell distribution width (RDW), total calcium, glucose, chronic kidney injury, hyperlipidemia, heart failure, pneumonia, glucocorticoid use, nephrotoxic medication use, and drink wine, which were used to develop subsequent machine learning models. Model Development and Validation We developed and validated predictive models using six machine learning algorithms implemented through Python's scikit-learn, XGBoost, LightGBM, NGBoost, and CatBoost packages, including logistic regression, random forest, XGBoost, LightGBM, NGBoost, and CatBoost. To optimize model performance, we conducted hyperparameter tuning for all algorithms using 10-fold cross-validation combined with grid search. Model evaluation was based on comprehensive performance metrics including the area under the receiver operating characteristic curve (AUC-ROC), accuracy, precision, recall, and F1-score. To ensure model generalizability and prevent overfitting, we systematically compared ROC curves across the training set, internal test set, and external validation set. Additionally, we performed decision curve analysis on all three datasets to thoroughly assess the clinical applicability and relative advantages of each model in real-world scenarios. This rigorous validation framework allowed us to objectively evaluate model performance while maintaining clinical relevance. Model Interpretation To elucidate the contribution of individual features to the model's final output, we employed SHapley Additive exPlanations (SHAP) for interpreting high-performing models. The SHAP approach treats each feature as an independent participant and applies game theory to quantify its specific contribution to predictive outcomes 7 . Through SHAP decision plots, we transformed the model's "black-box" decision-making process into interpretable visual pathways, enhancing our understanding of the model's behavior. This method enabled equitable attribution of predictive performance across features, clearly delineating each feature's impact on individual observations. Result Baseline Clinical Data According to the inclusion and exclusion criteria of this study, we enrolled 4,710 COPD patients from general wards, among whom 1,557 (33.1%) had comorbid diabetes while 2,930 (66.9%) were non-diabetic. The baseline characteristics of all included COPD patients are detailed in Table 1 . The cohort was randomly divided into training (n = 3,298) and testing sets (n = 1,412) through stratified sampling at a 7:3 ratio. The complete study workflow is illustrated in Fig. 1 . As shown in Table 1 , when comparing COPD patients without diabetes to those with diabetes, the two groups showed no significant differences in the following parameters: age, sex, race, language, marital status, hemoglobin, Rdw, hematocrit, sodium, potassium, total calcium, chloride, glucose, anion gap, prothrombin time, partial thromboplastin time, international normalized ratio (Inr), total bilirubin, blood urea nitrogen, creatinine, length of hospitalization, and mortality status. Table 1 Baseline table for COPD patients in general wards Total (N = 4710) Non-Diabetes (N = 3153) Diabetes (N = 1557) P Age (Median [IQR]) 72.00 [64.00, 80.00] 73.00[64.00,81.00] 72.00[65.00,79.00] 0.044 Sex (%) Female 2317(49.2) 1613(51.2) 704(45.2) < 0.001 Male 2393(50.8) 1540(48.8) 853(54.8) Race (%) White 3416(72.5) 1859(59.0) 1557(100.0) < 0.001 Other 1294(27.5) 1294(41.0) 0(0.0) Insurance (%) Medicare 3347(71.1) 2213(70.2) 1134(72.8) 0.064 Other 1363(28.9) 940(29.8) 423(27.2) Language (%) English 4459(94.7) 3024(95.9) 1435(92.2) < 0.001 Other 251(5.3) 129(4.1) 122(7.8) Marital Status (%) Married 1683(35.7) 619(19.6) 1064(68.3) < 0.001 Other 3027(64.3) 2534(80.4) 493(31.7) Wbc (Median [IQR]) 9.60 [7.20, 13.00] 9.63[7.10,13.00] 9.60[7.30,13.00] 0.248 Rbc (Median [IQR]) 3.79 [3.26, 4.29] 3.78[3.26,4.28] 3.81[3.25,4.30] 0.745 Platelet count (Median [IQR]) 202.50 [154.50, 260.00] 201.50[153.00,262.00] 203.73[157.50,256.00] 0.842 Hemoglobin (Median [IQR]) 11.20 [9.60, 12.70] 11.30[9.75,12.80] 11.00[9.35,12.50] < 0.001 Rdw (Median [IQR]) 14.40 [13.40, 15.90] 14.30[13.30,15.80] 14.60[13.50,16.20] < 0.001 Hematocrit (Median [IQR]) 35.00 [30.20, 39.00] 35.20[30.50,39.15] 34.47[29.70,38.70] 0.001 Sodium (Median [IQR]) 139.00 [136.00, 141.05] 139.00[136.50,141.50] 139.00[136.00,141.00] 0.026 Potassium (Median [IQR]) 4.20 [3.80, 4.53] 4.10[3.80,4.50] 4.20[3.90,4.60] < 0.001 Total calcium (Median [IQR]) 8.74 [8.30, 9.10] 8.70[8.30,9.10] 8.80[8.40,9.15] < 0.001 Chloride (Median [IQR]) 102.00 [98.50, 105.00] 102.00[99.00,105.00] 101.50[98.00,104.60] < 0.001 Glucose (Median [IQR]) 118.50 [98.00, 150.95] 110.00[95.00,132.49] 150.00[114.50,195.67] < 0.001 Anion gap (Median [IQR]) 14.00 [11.50, 16.00] 13.50[11.00,16.00] 14.00[12.00,16.00] < 0.001 Pt (Median [IQR]) 13.63 [12.20, 16.08] 13.60[12.20,15.85] 13.83[12.33,16.50] 0.002 Ptt (Median [IQR]) 34.00 [29.10, 42.69] 33.73[29.10,41.77] 34.63[29.17,44.85] 0.006 Inr (Median [IQR]) 1.25 [1.10, 1.49] 1.25[1.10,1.46] 1.29[1.10,1.50] 0.001 Total bilirubin (Median [IQR]) 0.67 [0.46, 1.67] 0.68[0.48,1.70] 0.65[0.45,1.60] 0.044 Alt (Median [IQR]) 30.11 [19.28, 92.99] 30.11[19.33,88.73] 30.08[19.00,105.51] 0.791 Ast (Median [IQR]) 37.50 [23.69, 107.58] 37.65[24.00,103.99] 37.26[23.00,116.21] 0.772 Urea nitrogen (Median [IQR]) 19.00 [13.00, 28.67] 18.00[12.00,26.00] 22.00[15.00,34.00] < 0.001 Creatinine (Median [IQR]) 0.90 [0.70, 1.30] 0.90[0.70,1.20] 1.10[0.80,1.53] < 0.001 Antihypertensive (%) No 1719(36.5) 1265(40.1) 454(29.2) < 0.001 Yes 2991(63.5) 1888(59.9) 1103(70.8) Glucocorticoids (%) No 4118(87.4) 2741(86.9) 1377(88.4) 0.156 Yes 592(12.6) 412(13.1) 180(11.6) Nephrotoxic (%) No 1379(29.3) 1024(32.5) 355(22.8) < 0.001 Yes 3331(70.7) 2129(67.5) 1202(77.2) Immunosuppressant (%) No 4642(98.6) 3113(98.7) 1529(98.2) 0.192 Yes 68(1.4) 40(1.3) 28(1.8) Hospitalization duration (Median [IQR]) 5.98 [3.16, 10.81] 5.84[3.04,10.63] 6.40[3.55,11.36] 0.003 Mortality outcomes (%) No 3020(64.1) 2072(65.7) 948(60.9) 0.001 Yes 1690(35.9) 1081(34.3) 609(39.1) In-hospital death status (%) No 4346(92.3) 2912(92.4) 1434(92.1) 0.801 Yes 364(7.7) 241(7.6) 123(7.9) Hypertension (%) No 2848(60.5) 1847(58.6) 1001(64.3) < 0.001 Yes 1862(39.5) 1306(41.4) 556(35.7) Chronic kidney disease (%) No 3652(77.5) 2636(83.6) 1016(65.3) < 0.001 Yes 1058(22.5) 517(16.4) 541(34.7) Hyperlipidemia (%) No 2416(51.3) 1793(56.9) 623(40.0) < 0.001 Yes 2294(48.7) 1360(43.1) 934(60.0) Acute renal failure (%) No 3323(70.6) 2345(74.4) 978(62.8) < 0.001 Yes 1387(29.4) 808(25.6) 579(37.2) Heart failure (%) No 3043(64.6) 2210(70.1) 833(53.5) < 0.001 Yes 1667(35.4) 943(29.9) 724(46.5) Myocardial infarction (%) No 4166(88.5) 2833(89.9) 1333(85.6) < 0.001 Yes 544(11.5) 320(10.1) 224(14.4) Tuberculosis (%) No 4708(100.0) 3152(100.0) 1556(99.9) 1 Yes 2(0.0) 1(0.0) 1(0.1) Pneumonia (%) No 3534(75.0) 2347(74.4) 1187(76.2) 0.191 Yes 1176(25.0) 806(25.6) 370(23.8) Stroke (%) No 4328(91.9) 2930(92.9) 1398(89.8) < 0.001 Yes 382(8.1) 223(7.1) 159(10.2) Pulmonary embolism (%) No 4409(93.6) 2952(93.6) 1457(93.6) 1 Yes 301(6.4) 201(6.4) 100(6.4) Covid19 (%) No 4581(97.3) 3073(97.5) 1508(96.9) 0.266 Yes 129(2.7) 80(2.5) 49(3.1) Smoke (%) No 3436(73.0) 2272(72.1) 1164(74.8) 0.054 Yes 1274(27.0) 881(27.9) 393(25.2) Drink wine (%) No 4245(90.1) 2788(88.4) 1457(93.6) < 0.001 Yes 465(9.9) 365(11.6) 100(6.4) Cirrhosis (%) No 4426(94.0) 2967(94.1) 1459(93.7) 0.638 Yes 284(6.0) 186(5.9) 98(6.3) Hepatitis (%) No 4463(94.8) 2977(94.4) 1486(95.4) 0.158 Yes 247(5.2) 176(5.6) 71(4.6) Malignant tumors (%) No 3844(81.6) 2547(80.8) 1297(83.3) 0.039 Yes 866(18.4) 606(19.2) 260(16.7) However, statistically significant differences (P < 0.05) were observed between the two groups for several comorbidities: hypertension, chronic kidney disease, hyperlipidemia, acute renal failure, heart failure, myocardial infarction, stroke, alcohol consumption, and malignant tumors. Regarding medication use, significant differences were found for antihypertensive drugs (P = 0.032) and nephrotoxic medications (P = 0.032). No statistically significant differences were observed for the remaining variables examined. Feature Selection We conducted Lasso regression analysis incorporating all 49 variables in the dataset with 10-fold cross-validation. The analytical results are visually presented in Fig. 2 . Specifically, Fig. 2 A illustrates the variable selection trajectory of the Lasso regression model, while Fig. 2 B demonstrates the feature selection process - the left dashed line denotes the minimum lambda (lambda.min), and the right dashed line indicates the lambda within one standard error of the optimal value (lambda.1se). Based on the Lasso regression outcomes, we subsequently performed multivariate and stepwise logistic regression analyses on the 17 initially selected variables. The comprehensive results of the stepwise logistic regression analysis are detailed in Table 2 . Table 2 reveals several significant associations between clinical variables and diabetes risk in COPD patients. Advancing age demonstrated a protective effect, with each additional year associated with a 2.6% reduction in diabetes risk (OR = 0.974, 95% CI = 0.963–0.986, P < 0.001). Insurance status showed marked variations, where patients with "other" insurance types had substantially elevated diabetes risk (OR = 11.609, 95% CI = 7.203–19.385, P < 0.001), while Medicare coverage was protective. Similarly, "other" language speakers faced higher risk (OR = 10.019, 95% CI = 6.001–17.108, P < 0.001). Married patients exhibited a remarkable 97.4% risk reduction compared to unmarried individuals (OR = 0.026, 95% CI = 0.016–0.042, P < 0.001). Several laboratory parameters significantly influenced diabetes risk: each unit increase in RDW elevated risk by 7.5% (OR = 1.075, 95% CI = 1.026–1.127, P = 0.003), while higher total calcium levels increased risk by 20% per unit (OR = 1.200, 95% CI = 1.022–1.410, P = 0.026). Notably, each unit increase in glucose concentration was associated with a 2.3% risk elevation (OR = 1.023, 95% CI = 1.020–1.025, P < 0.001). Comorbid conditions substantially impacted diabetes risk: chronic kidney disease (OR = 2.370, 95% CI = 1.843–3.057, P < 0.001), hyperlipidemia (OR = 1.854, 95% CI = 1.495–2.301, P < 0.001), and heart failure (OR = 1.299, 95% CI = 1.028–1.642, P = 0.029) all increased risk, whereas pneumonia reduced risk by 36.8% (OR = 0.632, 95% CI = 0.490–0.814, P < 0.001). Lifestyle and medication factors also showed significant associations: alcohol consumption decreased risk by 43% (OR = 0.570, 95% CI = 0.388–0.829, P = 0.004), glucocorticoid use was protective (OR = 0.634, 95% CI = 0.451–0.889, P = 0.008), while nephrotoxic drug use increased risk (OR = 1.401, 95% CI = 1.104–1.780, P = 0.006). No significant association was observed between race and diabetes risk (P = 0.952). Following the aforementioned statistical analyses, a total of 14 variables were retained for developing subsequent machine learning models. These variables included: age, sex, insurance type, language, marital status, red cell distribution width (RDW), total calcium level, glucose, chronic kidney disease (CKD), hyperlipidemia, heart failure, pneumonia, alcohol consumption, glucocorticoid use, and nephrotoxic medication use. Table 2 Stepwise logistic regression analysis results B Value SE Z Value OR 95%CI P Age -0.026 0.006 -4.361 0.974 0.963–0.986 0.000 Race = Other -18.318 301.796 -0.061 0.000 0.000–0.000 0.952 Insurance = Other 2.452 0.252 9.731 11.609 7.203–19.385 0.000 Language = Other 2.304 0.267 8.644 10.019 6.001–17.108 0.000 Marital Status = Other -3.634 0.243 -14.963 0.026 0.016–0.042 0.000 Rdw 0.072 0.024 3.004 1.075 1.026–1.127 0.003 Total calcium 0.182 0.082 2.224 1.200 1.022–1.410 0.026 Glucose 0.022 0.001 15.555 1.023 1.020–1.025 0.000 Glucocorticoids = Yes -0.455 0.173 -2.635 0.634 0.451–0.889 0.008 Nephrotoxic = Yes 0.337 0.122 2.770 1.401 1.104–1.780 0.006 Chronic kidney disease = Yes 0.863 0.129 6.686 2.370 1.843–3.057 0.000 Hyperlipidemia = Yes 0.617 0.110 5.617 1.854 1.495–2.301 0.000 Heart failure = Yes 0.262 0.120 2.190 1.299 1.028–1.642 0.029 Pneumonia = Yes -0.459 0.130 -3.540 0.632 0.490–0.814 0.000 Drink Wine = Yes -0.563 0.193 -2.908 0.570 0.388–0.829 0.004 Evaluation and Comparison of Model Performance Based on the feature selection results, we employed six widely-used machine learning algorithms to develop predictive models, including Logistic Regression, Random Forest (RF), eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), Natural Gradient Boosting (NGBoost), and Categorical Boosting (CatBoost). During model development, we optimized hyperparameters through 10-fold cross-validation combined with grid search. Model performance was evaluated using multiple metrics, with the optimal hyperparameter combination ultimately selected based on the ROC-AUC (Receiver Operating Characteristic - Area Under the Curve) metric. The final model was trained using this optimal parameter set. We visualized the receiver operating characteristic (ROC) curves of the six machine learning models on the training set, test set, and external validation set in Fig. 3 A. The Random Forest model demonstrated the best performance on the training set (AUC = 0.96), but showed a significant performance drop on the test set (AUC = 0.90), indicating potential overfitting. In contrast, the LightGBM model maintained consistently excellent performance across both training (AUC = 0.93) and test sets (AUC = 0.90). Figure 3 B presents the calibration curves and prediction outcomes, revealing that all six models achieved high goodness-of-fit across the training, test, and validation sets. Figure 3 C displays the precision-recall curves, where the x-axis represents recall (the proportion of actual positives correctly identified) and the y-axis represents precision (the proportion of predicted positives that are truly positive). Higher recall indicates fewer missed cases, while higher precision indicates fewer false positives. The LightGBM model demonstrated optimal overall performance, with stable average precision across the training, test, and external validation sets. Figure 4 A presents the clinical decision curves for different models. The LightGBM model demonstrated superior net benefit across most threshold probabilities in all datasets (training, test, and validation sets). Figure 4 B provides a comprehensive summary of performance metrics across all machine learning models. In the training set, the models exhibited the following performance characteristics: Logistic Regression (Accuracy = 0.82, Precision = 0.83, Recall = 0.55, F1-score = 0.66, AUC = 0.88); Random Forest (Accuracy = 0.89, Precision = 0.88, Recall = 0.77, F1-score = 0.82, AUC = 0.96) ; XGBoost (Accuracy = 0.82, Precision = 0.83, Recall = 0.54, F1-score = 0.65, AUC = 0.91) ; LightGBM (Accuracy = 0.85, Precision = 0.80, Recall = 0.75, F1-score = 0.76, AUC = 0.93) ; NGBoost (Accuracy = 0.84, Precision = 0.79, Recall = 0.68, F1-score = 0.73, AUC = 0.92); CatBoost (Accuracy = 0.80, Precision = 0.85, Recall = 0.45, F1-score = 0.59, AUC = 0.90). When evaluated on the test set, the models demonstrated: Logistic Regression (Accuracy = 0.80, Precision = 0.81, Recall = 0.54, F1-score = 0.65, AUC = 0.87); Random Forest (Accuracy = 0.83, Precision = 0.79, Recall = 0.69, F1-score = 0.74, AUC = 0.90); XGBoost (Accuracy = 0.80, Precision = 0.82, Recall = 0.55, F1-score = 0.66, AUC = 0.90); LightGBM (Accuracy = 0.83, Precision = 0.77, Recall = 0.71, F1-score = 0.74, AUC = 0.90); NGBoost (Accuracy = 0.82, Precision = 0.78, Recall = 0.69, F1-score = 0.73, AUC = 0.90); CatBoost (Accuracy = 0.79, Precision = 0.84, Recall = 0.47, F1-score = 0.60, AUC = 0.89). Model interpretability analysis To better understand the relationship between the model and the data, we employed SHAP (SHapley Additive exPlanations) to provide an intuitive interpretation of the best-performing LightGBM model, demonstrating how these variables influence the predictions. Figure 5 A presents a beeswarm plot illustrating the importance of each feature in the model. The x-axis represents the SHAP value, which quantifies the contribution of each feature to the model's prediction. Positive SHAP values indicate an increased likelihood of the predicted outcome, while negative values suggest a decreased likelihood. Each point corresponds to a sample's SHAP value, with color intensity reflecting the original feature value (red for high values, blue for low values), allowing us to visualize the relationship between feature magnitude and its impact on predictions. Figure 5 B ranks features by their mean absolute SHAP values, where longer bars denote greater influence on the model's output. Marital status emerged as the most critical predictor (mean |SHAP| = +0.2), followed by glucose (|SHAP| = +0.11) and insurance type (|SHAP| = +0.06). Other notable contributors included chronic kidney disease (CKD) (|SHAP| = +0.03), hyperlipidemia (|SHAP| = +0.03), language (|SHAP| = +0.02), total calcium (|SHAP| = +0.02), age (|SHAP| = +0.02), and nephrotoxic drug use (|SHAP| = +0.02). The remaining five features collectively contributed a mean |SHAP| of + 0.03. Figure 5 C displays a SHAP decision plot for the top 50 samples, transforming the model's "black-box" logic into an interpretable visual pathway. The x-axis represents the model's output value (higher positive values indicate greater diabetes risk, while lower negative values suggest reduced risk), and the y-axis lists clinically relevant features ordered by importance. Figure 5 D presents a waterfall plot of an individual sample, with the color gradient (red to blue) indicating the magnitude of the feature values. Figure 5 E illustrates the SHAP force plot for the same patient with COPD, demonstrating the contribution of each clinical feature to the model’s prediction. In this plot, each bar represents the impact of a specific feature, where red denotes higher values and blue indicates lower values. Together, Figs. 5 D and 5 E provide an intuitive visualization of how distinct features influence the model’s prediction of diabetes. This SHAP-based interpretability analysis enhances our understanding of the role of individual features in shaping the output of the LightGBM predictive model. Discussion COPD is a prevalent respiratory disorder associated with high morbidity and mortality 8 . Patients with COPD exhibit a progressive decline in daily physical activity over time, which parallels worsening lung function and overall health status 9 . Diabetes mellitus is a major comorbidity of COPD 10 , and lifestyle-related risk factors are considered central to the development of both type 1 and type 2 diabetes 11 . A cohort study conducted in Taiwan revealed that 16% of COPD patients had pre-existing T2DM, while 19% were newly diagnosed with T2DM during a 10-year follow-up period, indicating an elevated prevalence and incidence of this comorbidity. Notably, the study found that diabetes—whether pre-existing or newly diagnosed—was associated with poorer clinical outcomes in COPD patients 12 . Therefore, early detection of diabetes and personalized management are critical for improving prognosis in COPD patients. This study leverages the MIMIC-IV database to propose an efficient, non-invasive diabetes risk prediction model based on demographic characteristics, laboratory indicators, and disease and medication usage. By incorporating a large patient cohort, the model ensures broad generalizability. Our objective was to develop a machine learning-based risk prediction model and validate it both internally and externally. The results demonstrate that our LightGBM model achieved an AUC of 0.85 in the training set and 0.80 in the external validation set, indicating robust predictive performance and strong applicability to external datasets. Table 2 presents the results of stepwise logistic regression analysis. Combined with SHAP analysis, we identified marital status, blood glucose levels, and insurance type as the three most significant features influencing diabetes development in COPD patients. Existing literature has extensively examined the relationship between type 2 diabetes mellitus and marital status through cross-sectional studies. Marital status independently correlates with T2DM incidence, where married individuals demonstrate significantly lower diabetes risk compared to divorced persons, despite comparable weight gain 13 . A prospective study utilizing ARIC data further revealed that persistent single status was associated with a 34% increased risk of T2DM in women 14 . The social causation theory suggests that marriage confers health-protective effects through stress reduction, anxiety alleviation, and promotion of positive health behaviors, while unmarried status may adversely impact health outcomes 15 . Insurance type also significantly contributes to the model's predictive performance. A study by Mahoney S et al. found that Medicare coverage was associated with reduced odds of undiagnosed prediabetes 16 . Limited healthcare access and screening opportunities among uninsured populations may lead to underdiagnosis of both prediabetes and diabetes. As shown in Table 2 , we observed a positive correlation between red cell distribution width (RDW) and diabetes risk. This finding aligns with existing evidence linking elevated RDW values with pro-inflammatory conditions. A retrospective study by Nada et al. demonstrated significantly higher RDW levels in diabetic patients compared to non-diabetic controls 17 . Engström G et al. further reported a significant positive association between RDW and HbA1c, with each 1-SD increase in RDW corresponding to a 0.10% rise in HbA1c 18 . Our analysis also revealed that increased serum calcium levels elevate diabetes risk, consistent with a cross-sectional study of 39,645 participants that established a causal relationship between higher serum calcium levels and increased type 2 diabetes prevalence 19 . Supporting this, Sing CW et al. found that elevated serum calcium concentrations were associated with diabetes incidence 20 . Regarding disease comorbidities, our analysis revealed that both chronic kidney disease (CKD) and heart failure (HF) significantly increased diabetes risk (Table 2 ). HF, diabetes mellitus (DM), and CKD represent a triad of frequently interrelated conditions that should be considered collectively rather than in isolation. Epidemiological data indicate that 25%-40% of HF patients have comorbid DM, while 40%-50% exhibit concurrent CKD. Notably, 16% of HF patients present with all three conditions - a high-risk combination associated with substantially elevated hospitalization and mortality rates. The pathophysiological interplay between HF, DM, and CKD involves complex mechanisms, likely mediated through shared cardiovascular/metabolic risk factors, chronic inflammation, oxidative stress, and downstream neuroendocrine pathway activation 21 . Given the demonstrated renal and cardiovascular protective effects of novel antidiabetic agents, implementing tailored therapeutic strategies addressing these comorbidities has become increasingly crucial 22 . However, further research is needed to elucidate these relationships more comprehensively. Our study additionally identified hyperlipidemia as a significant diabetes risk factor (OR = 1.854, 95%CI = 1.495–2.301, P < 0.001). Dyslipidemia constitutes a core component of metabolic syndrome in T2DM patients, where lipid abnormalities primarily arise as secondary consequences of insulin resistance. Consequently, any intervention reducing insulin resistance would be expected to beneficially impact lipid profiles. Current guidelines already recommend aggressive lipid-lowering therapy for high-risk T2DM individuals 23 . Our logistic regression results suggested reduced diabetes risk among pneumonia patients compared to non-pneumonia cases. While Brunetti VC et al. reported potential associations between T2DM and increased community-acquired pneumonia (CAP) risk, the available evidence derives from studies with substantial bias risk, necessitating higher-quality research to verify these observations 24 . The study by Marjo Heinjoki et al. found no significant difference in renal function between elderly individuals with diabetes and their non-diabetic counterparts. Furthermore, the use of potentially nephrotoxic medications appeared to play only a minor role in renal function decline among community-dwelling elderly in Finland's Inner-Savo region 25 . Interestingly, our current study demonstrated a significant positive association between nephrotoxic drug use and diabetes incidence (P < 0.05). However, existing research on the relationship between nephrotoxic medications and diabetes development remains limited. This novel finding warrants further investigation through large-scale prospective studies to elucidate potential causal mechanisms and clinical implications. As evidenced in Table 2 , our study demonstrated a reduced diabetes risk among COPD patients with a history of alcohol consumption. Rachel Golan and colleagues' systematic review synthesized existing evidence, suggesting that initiating moderate alcohol consumption appears reasonably safe for individuals with well-controlled blood glucose, particularly regarding heart rate variability and arterial plaque formation 26 . However, this finding should not be construed as an endorsement of heavy drinking. While epidemiological studies have consistently shown an inverse association between alcohol consumption and type 2 diabetes risk 27 , the dose-response relationship requires careful consideration. The research by G Caimi et al. revealed that light-to-moderate, regular alcohol intake - irrespective of beverage type - was negatively correlated with type 2 diabetes incidence[28]. Notably, moderate wine consumption may confer additional benefits by improving oxidative status in diabetic patients, suggesting this lifestyle modification could potentially mitigate diabetes progression and complication development 28 . Our study evaluated six machine learning algorithms, among which the LightGBM model demonstrated particularly promising performance. Through SHAP interpretability analysis, we gained more precise insights into how the 14 selected features influenced the predictive outputs of our developed model. Limitation We acknowledge that both our primary dataset and external validation cohort were derived exclusively from the MIMIC-IV database, which may limit the generalizability of our findings to other populations or geographic regions. To enhance the model's broader clinical applicability, future studies should incorporate more diverse datasets spanning wider geographical distributions, ethnic populations, and temporal ranges. Online Prediction Tool Development Based on the LightGBM model developed in this study, we created a user-friendly online prediction tool for clinicians to assess the risk of diabetes in COPD patients. The tool is accessible at: https://copd-diabetes-predictor-model.streamlit.app/ . Conclusion The LightGBM model developed in this study demonstrates robust predictive performance, providing clinicians with a novel screening tool for diabetes risk assessment in COPD patients. Declarations Ethics approval and consent to participate: Not applicable. Availability of data and materials: The raw data that support the findings of this study are available from the corresponding author, Xishi Sun, upon reasonable request. Competing interests: The authors declare that they have no competing interests. Funding: This study was supported by the Guang dong Medical Research Fund Project (No.:A2024728, A2024723),Zhanjiang Science and Technology Research Project in 2024 (No.: 2024B01356) and the High-level Talents Scientific Research Start-up Funds of the Affiliated Hospital of Guangdong Medical University (No.:GCC2022028). Authors' contributions: All authors listed have made a substantial, direct, and intellectual contribution to the work, and approved it for publication. Acknowledgements: Not applicable. References Mannino, D. M. & Buist, A. S. Global burden of COPD: risk factors, prevalence, and future trends. Lancet 370 , 765–773 (2007). Wannamethee, S. G. et al. Lung function and risk of type 2 diabetes and fatal and nonfatal major coronary heart disease events: possible associations with inflammation. Diabetes Care . 33 , 1990–1996 (2010). Frostegård, J. Immunity, atherosclerosis and cardiovascular disease. BMC Med. 11 , 117 (2013). Wang, T. et al. COPD and T2DM: a Mendelian randomization study. Front. Endocrinol. (Lausanne) . 15 , 1302641 (2024). Hughes, M. J., McGettrick, H. M. & Sapey, E. Shared mechanisms of multimorbidity in COPD, atherosclerosis and type-2 diabetes: the neutrophil as a potential inflammatory target. Eur. Respir Rev. 29 , 190102 (2020). Johnson, A. E. W. et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci. Data . 10 , 1 (2023). Explaining a series of models by propagating Shapley. values - PubMed. https://pubmed.ncbi.nlm.nih.gov/35922410/ Ferrera, M. C., Labaki, W. W. & Han, M. K. Advances in Chronic Obstructive Pulmonary Disease. Annu. Rev. Med. 72 , 119–134 (2021). Waschki, B. et al. Disease Progression and Changes in Physical Activity in Patients with Chronic Obstructive Pulmonary Disease. Am. J. Respir Crit. Care Med. 192 , 295–306 (2015). Prevalence of comorbidities in. patients with chronic obstructive pulmonary disease - PubMed. https://pubmed.ncbi.nlm.nih.gov/20134148/ Owen, N., Healy, G. N., Matthews, C. E. & Dunstan, D. W. Too much sitting: the population health science of sedentary behavior. Exerc. Sport Sci. Rev. 38 , 105–113 (2010). Ho, T. W. et al. Diabetes mellitus in patients with chronic obstructive pulmonary disease-The impact on mortality. PLoS One . 12 , e0175794 (2017). de Oliveira, C. M. et al. Relationship between marital status and incidence of type 2 diabetes mellitus in a Brazilian rural population: The Baependi Heart Study. PLoS One . 15 , e0236869 (2020). Schwandt, H. M., Coresh, J. & Hindin, M. J. Marital Status, Hypertension, Coronary Heart Disease, Diabetes, and Death Among African American Women and Men: Incidence and Prevalence in the Atherosclerosis Risk in Communities (ARIC) Study Participants. J. Fam Issues . 31 , 1211–1229 (2010). Jaffe, D. H., Manor, O., Eisenbach, Z. & Neumark, Y. D. The protective effect of marriage on mortality in a dynamic society. Ann. Epidemiol. 17 , 540–547 (2007). Mahoney, S. et al. Health Insurance Is Associated with Decreased Odds for Undiagnosed Prediabetes and Type 2 Diabetes in American Adults. Int. J. Environ. Res. Public. Health . 17 , 4706 (2020). Nada, A. M. Red cell distribution width in type 2 diabetic patients. Diabetes Metab. Syndr. Obes. 8 , 525–533 (2015). Engström, G. et al. Red cell distribution width, haemoglobin A1c and incidence of diabetes mellitus. J. Intern. Med. 276 , 174–183 (2014). Zhai, Z. et al. Association between serum calcium level and type 2 diabetes: An NHANES analysis and Mendelian randomization study. Diabet. Med. 40 , e15080 (2023). Sing, C. W. et al. Serum calcium and incident diabetes: an observational study and meta-analysis. Osteoporos. Int. 27 , 1747–1754 (2016). Vijay, K., Neuen, B. L. & Lerma, E. V. Heart Failure in Patients with Diabetes and Chronic Kidney Disease: Challenges and Opportunities. Cardiorenal Med. 12 , 1–10 (2022). Wu, M. Z. et al. Chronic kidney disease begets heart failure and vice versa: temporal associations between heart failure events in relation to incident chronic kidney disease in type 2 diabetes. Diabetes Obes. Metab. 25 , 707–715 (2023). Betteridge, D. J. Diabetic dyslipidaemia. Diabetes Obes. Metab. 2 (Suppl 1), S31–36 (2000). Type 2 diabetes. mellitus and risk of community-acquired pneumonia: a systematic review and meta-analysis of observational studies - PubMed. https://pubmed.ncbi.nlm.nih.gov/33495386/ Heinjoki, M. et al. Kidney function and nephrotoxic drug use among older home-dwelling persons with or without diabetes in Finland. BMC Nephrol. 21 , 11 (2020). Golan, R., Gepner, Y. & Shai, I. Wine and Health-New Evidence. Eur. J. Clin. Nutr. 72 , 55–59 (2019). Huang, J., Wang, X. & Zhang, Y. Specific types of alcoholic beverage consumption and risk of type 2 diabetes: A systematic review and meta-analysis. J. Diabetes Investig . 8 , 56–68 (2017). Caimi, G., Carollo, C. & Lo Presti, R. Diabetes mellitus: oxidative stress and wine. Curr. Med. Res. Opin. 19 , 581–586 (2003). Additional Declarations No competing interests reported. Supplementary Files Data1Generalwarddata.xlsx Date2Externalvalidationset.xlsx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 29 Sep, 2025 Reviews received at journal 18 Sep, 2025 Reviewers agreed at journal 18 Sep, 2025 Reviews received at journal 16 Sep, 2025 Reviewers agreed at journal 16 Sep, 2025 Reviewers invited by journal 11 Aug, 2025 Editor assigned by journal 11 Aug, 2025 Editor invited by journal 04 Jul, 2025 Submission checks completed at journal 03 Jul, 2025 First submitted to journal 02 Jul, 2025 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-7033945","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":500053887,"identity":"847815df-1e50-417d-a06f-a2910ad3136c","order_by":0,"name":"Lingpin Pang","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Lingpin","middleName":"","lastName":"Pang","suffix":""},{"id":500053888,"identity":"07e4f656-5424-4c96-990e-957eccb70d97","order_by":1,"name":"Siyan Xu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAyUlEQVRIiWNgGAWjYBACNmb245//GEjwMDYcPkCcFj52njQGngIbGebGYwnEaZHjZzBj4PmQZsPefMaAWIcxpD2QMDjMw9t25uONNwx2croNBLUwHjcwAGqR7Dm72XIOQ7Kx2QHCtiRIJAC1GM44u02ah+FA4jYitBhIHABqsb//5hnRWswkGwzSgIF8ho1YLTzJxgwGNkAtx4wt5xgQ4Rf5/uMHHzP8kbAHRuXDG28q7OQIakEBEjxERg2yFlJ1jIJRMApGwYgAAPhwPdzbhG9iAAAAAElFTkSuQmCC","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":true,"prefix":"","firstName":"Siyan","middleName":"","lastName":"Xu","suffix":""},{"id":500053889,"identity":"6fc8b0df-52ba-476a-aad5-c5e87677cc9e","order_by":2,"name":"Yingxin Wang","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yingxin","middleName":"","lastName":"Wang","suffix":""},{"id":500053890,"identity":"a5f13ddf-6215-4b1a-97b3-38b3ce09104d","order_by":3,"name":"Tao Huang","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Tao","middleName":"","lastName":"Huang","suffix":""},{"id":500053891,"identity":"9fd38ddd-f3eb-4985-b036-ec7461877d8f","order_by":4,"name":"Qian Xian","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Qian","middleName":"","lastName":"Xian","suffix":""},{"id":500053892,"identity":"811c9640-30f3-48a3-a673-7732ae403986","order_by":5,"name":"Wenjia Lin","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Wenjia","middleName":"","lastName":"Lin","suffix":""},{"id":500053893,"identity":"069eb26b-7a7b-4d77-b030-bfe86d100a27","order_by":6,"name":"Haowen Pang","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Haowen","middleName":"","lastName":"Pang","suffix":""},{"id":500053894,"identity":"30fc99ed-95a2-48ab-9959-333d0e79e059","order_by":7,"name":"Zhirui Chen","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Zhirui","middleName":"","lastName":"Chen","suffix":""},{"id":500053895,"identity":"e7393e3d-e2b9-407b-945e-9f8b6d368245","order_by":8,"name":"Bozhi Zhong","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Bozhi","middleName":"","lastName":"Zhong","suffix":""},{"id":500053896,"identity":"74f650d5-4ad4-404f-b9f0-6b56aafc4b04","order_by":9,"name":"Hui Miao","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Hui","middleName":"","lastName":"Miao","suffix":""},{"id":500053897,"identity":"1c8f0066-478d-4307-80e1-e3e2cd1ad8fe","order_by":10,"name":"Hui Chen","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Hui","middleName":"","lastName":"Chen","suffix":""},{"id":500053898,"identity":"5d7d9ba1-4fda-46b2-8445-1569588dd8cf","order_by":11,"name":"Xishi Sun","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xishi","middleName":"","lastName":"Sun","suffix":""},{"id":500053899,"identity":"88cf3163-d5f1-488a-a850-9adf117c7b88","order_by":12,"name":"Jie Sun","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jie","middleName":"","lastName":"Sun","suffix":""}],"badges":[],"createdAt":"2025-07-03 04:08:05","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7033945/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7033945/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":89378868,"identity":"4c27e78d-521e-4740-bc25-61dbc8cbd348","added_by":"auto","created_at":"2025-08-19 11:34:38","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":192794,"visible":true,"origin":"","legend":"\u003cp\u003eFlow Plot\u003c/p\u003e","description":"","filename":"image1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7033945/v1/a85d702b5052ae9dd0dabec5.jpeg"},{"id":89380185,"identity":"dd14f140-f321-46cf-8c6e-4869101d358b","added_by":"auto","created_at":"2025-08-19 11:42:39","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":45893,"visible":true,"origin":"","legend":"\u003cp\u003ePresentation of the results of the LASSO regression analysis. (A) LASSO regression model screening variable trajectories; (B) LASSO Regression Model Factor Selection: Left dashed line represents the optimal lambda value (lambda.min), while the right dashed line marks the lambda value within one standard error of the optimal (lambda.1se).\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-7033945/v1/10be49016c1e50e55cdc48b2.png"},{"id":89380175,"identity":"243be47f-63ca-45da-86f9-f4d4573e98d8","added_by":"auto","created_at":"2025-08-19 11:42:38","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":350885,"visible":true,"origin":"","legend":"\u003cp\u003eThe performance and comparison of six different predictive models. (A) ROC Curves for the Training Set, Test Set and Validation set; (B) Calibration Curve for the Training Set, Test Set and Validation set;(C) Precision-Recall Curve for the Training Set, Test Set and Validation set\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-7033945/v1/17a8095310b002a360b7ea99.png"},{"id":89378871,"identity":"0cdf02fa-2cf6-4421-be20-e62583ca7519","added_by":"auto","created_at":"2025-08-19 11:34:38","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":199222,"visible":true,"origin":"","legend":"\u003cp\u003e(A)DCA Curve for the Training Set, Test Set and Validation set;(B) Visualization of machine learning evaluation metrics.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-7033945/v1/c8ce57ba750dd16764b42dbb.png"},{"id":89378879,"identity":"1ba2f911-e1e9-4526-8a57-cff81f611585","added_by":"auto","created_at":"2025-08-19 11:34:38","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":301893,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP of the LightGBM model. (A) SHAP Beeswarm Plot; (B) SHAP Feature Importance Plot;(C) SHAP Decision Plot;(D) SHAP Waterfall Plot;(E) SHAP Force Plot.\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-7033945/v1/fe20d015e50a812bfb2e7738.png"},{"id":89382038,"identity":"127a518d-e8d4-4888-8eb9-77c971665a70","added_by":"auto","created_at":"2025-08-19 12:06:39","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2065036,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7033945/v1/cfbf4771-3be4-4b4f-818c-e073e88fd1cc.pdf"},{"id":89378872,"identity":"96aca466-7ffd-4345-93fc-1567fa7f4d4c","added_by":"auto","created_at":"2025-08-19 11:34:38","extension":"xlsx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":3834314,"visible":true,"origin":"","legend":"","description":"","filename":"Data1Generalwarddata.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7033945/v1/fe6b9d36e47cc8b8cacff580.xlsx"},{"id":89378876,"identity":"c7bddf3c-11c5-4012-a026-92d175acb96c","added_by":"auto","created_at":"2025-08-19 11:34:38","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":5674146,"visible":true,"origin":"","legend":"","description":"","filename":"Date2Externalvalidationset.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7033945/v1/5561b26a0f14372c9a0b561b.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Construction and Validation of an Interpretable Machine Learning Model for Predicting Diabetes Risk in COPD Patients","fulltext":[{"header":"Introduction","content":"\u003cp\u003eChronic obstructive pulmonary disease (COPD) remains a major global contributor to morbidity, mortality, and healthcare expenditures worldwide. As of 2015, approximately 73\u0026nbsp;billion people were estimated to be affected by COPD globally. With the accelerating trend of population aging, the disease burden of COPD is projected to increase substantially in the coming decades\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. A recent systematic review suggests that COPD may serve as a precursor to diabetes progression, particularly highlighting that patients with severe to very severe COPD face substantially elevated risks of diabetes onset \u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. COPD and diabetes reciprocally serve as risk factors for each other, with epidemiological studies demonstrating increasing prevalence rates of either condition following diagnosis of the other\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eEmerging evidence reveals intertwined pathophysiological mechanisms between COPD and type 2 diabetes mellitus (T2DM)\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. Notably, while neutrophils are not the sole effector immune cells in COPD, atherosclerosis, or T2DM, they significantly outnumber other immune cell populations in pulmonary secretions of COPD patients. Animal and cellular models demonstrate that neutrophil-derived reactive substances can induce the full spectrum of pulmonary pathologies (ranging from chronic bronchitis to emphysema), with their levels showing significant correlation with disease severity and progressive lung function decline \u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. These findings underscore the importance of advocating for regular health screenings and lifestyle modifications in COPD patients, with particular emphasis on early intervention for T2DM prevention and management. This study aims to develop a machine learning-based risk prediction model to identify high-risk individuals for diabetes among COPD patients, thereby facilitating early personalized management of this comorbidity.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cp\u003e\u003cb\u003eStudy Design and Data Source\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThis study aims to develop and validate a machine learning prediction model based on MIMIC-IV data. The research data were obtained from MIMIC-IV (Medical Information Mart for Intensive Care IV), an internationally recognized public medical database. As an upgraded version of MIMIC-III released in 2015, this database features broader temporal coverage and more refined data structure\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. This study extracted clinical data records from 5,767 general ward patients for analysis, along with 4,711 ICU patient records from the same database for external validation. The open-source dataset was established based on COPD diagnostic codes (ICD-10: 'J44', 'J440', 'J441', 'J449') and comprehensively incorporated demographic characteristics including age, sex, race, insurance type, language, and marital status; lifestyle factors such as smoking status and drink wine; laboratory parameters comprising white blood cell count, red blood cell count, platelet count, hemoglobin level, red cell distribution width, hematocrit, sodium, potassium, total calcium, chloride, glucose, anion gap, prothrombin time (PT), partial thromboplastin time (PTT), international normalized ratio (INR), total bilirubin, alanine aminotransferase (ALT), aspartate aminotransferase (AST), blood urea nitrogen, and creatinine; documented comorbidities including hypertension, chronic kidney disease (CKD), hyperlipidemia, acute renal failure (ARF), heart failure (HF), myocardial infarction, tuberculosis, pneumonia, stroke, pulmonary embolism, COVID-19, cirrhosis, hepatitis, and malignant tumors; medication records covering antihypertensives, glucocorticoids, nephrotoxic drugs, and immunosuppressants; as well as hospitalization duration and mortality outcomes including in-hospital death status. Dr. Xishi Sun possesses certified access to this database (Record ID 12112229) and was responsible for data extraction.\u003c/p\u003e\u003cp\u003eThe study inclusion criteria were: (1) diagnosis of chronic obstructive pulmonary disease according to the International Classification of Diseases (ICD); (2) first admission to general ward; and (3) age between 18 and 100 years. Exclusion criteria comprised: (1) patients initially admitted to ICU or with multiple general ward admissions; (2) age below 18 or above 100 years; and (3) patients with clinical data containing\u0026thinsp;\u0026gt;\u0026thinsp;10% missing values.\u003c/p\u003e\u003cp\u003e\u003cb\u003eData Preparation\u003c/b\u003e\u003c/p\u003e\u003cp\u003eAfter extracting the patient data, variables with more than 70% missing values were removed, and the random forest imputation method was applied to fill in the missing data. Based on the exclusion and inclusion criteria, a total of 4,710 patients were ultimately enrolled. A fixed random seed was set, and stratified random sampling was used to divide the dataset into training and testing sets at a 7:3 ratio. Additionally, data from 4,711 COPD patients in the ICU were extracted from the MIMIC-IV database for external validation. The analyses were implemented using R (version 4.4.1) and Python (version 3.12.4). Categorical variables (such as hypertension and heart failure) were processed using label encoding.\u003c/p\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eStatistical Analysis\u003c/h2\u003e\u003cp\u003ePrior to statistical analysis, descriptive statistics were performed for all baseline variables. Normally distributed continuous variables were analyzed using the t-test and expressed as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation (SD), while non-normally distributed continuous variables were assessed using the Mann-Whitney U test (rank-sum test) and reported as median (interquartile range, IQR). Categorical variables were compared using the chi-square (χ\u0026sup2;) test and presented as counts (percentages). A p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e\u003cp\u003e\u003cb\u003eFeature Selection\u003c/b\u003e\u003c/p\u003e\u003cp\u003eFeature selection was performed using R software, with a p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 considered statistically significant. The model development and evaluation process proceeded as follows: (1) The Lasso algorithm was first applied for preliminary variable screening, employing 10-fold cross-validation to enhance reliability. At the 1 standard deviation lambda value, 17 variables were selected. (2) Multivariate and stepwise logistic regression analyses were then used to further refine these 17 variables, ultimately identifying 14 significant predictors for inclusion in subsequent analyses. These predictive variables included: age, insurance type, language, marital status, red cell distribution width (RDW), total calcium, glucose, chronic kidney injury, hyperlipidemia, heart failure, pneumonia, glucocorticoid use, nephrotoxic medication use, and drink wine, which were used to develop subsequent machine learning models.\u003c/p\u003e\u003cp\u003e\u003cb\u003eModel Development and Validation\u003c/b\u003e\u003c/p\u003e\u003cp\u003eWe developed and validated predictive models using six machine learning algorithms implemented through Python's scikit-learn, XGBoost, LightGBM, NGBoost, and CatBoost packages, including logistic regression, random forest, XGBoost, LightGBM, NGBoost, and CatBoost.\u003c/p\u003e\u003cp\u003eTo optimize model performance, we conducted hyperparameter tuning for all algorithms using 10-fold cross-validation combined with grid search. Model evaluation was based on comprehensive performance metrics including the area under the receiver operating characteristic curve (AUC-ROC), accuracy, precision, recall, and F1-score. To ensure model generalizability and prevent overfitting, we systematically compared ROC curves across the training set, internal test set, and external validation set. Additionally, we performed decision curve analysis on all three datasets to thoroughly assess the clinical applicability and relative advantages of each model in real-world scenarios. This rigorous validation framework allowed us to objectively evaluate model performance while maintaining clinical relevance.\u003c/p\u003e\u003cp\u003e\u003cb\u003eModel Interpretation\u003c/b\u003e\u003c/p\u003e\u003cp\u003eTo elucidate the contribution of individual features to the model's final output, we employed SHapley Additive exPlanations (SHAP) for interpreting high-performing models. The SHAP approach treats each feature as an independent participant and applies game theory to quantify its specific contribution to predictive outcomes \u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. Through SHAP decision plots, we transformed the model's \"black-box\" decision-making process into interpretable visual pathways, enhancing our understanding of the model's behavior. This method enabled equitable attribution of predictive performance across features, clearly delineating each feature's impact on individual observations.\u003c/p\u003e"},{"header":"Result","content":"\u003cp\u003e\u003cb\u003eBaseline Clinical Data\u003c/b\u003e\u003c/p\u003e\u003cp\u003eAccording to the inclusion and exclusion criteria of this study, we enrolled 4,710 COPD patients from general wards, among whom 1,557 (33.1%) had comorbid diabetes while 2,930 (66.9%) were non-diabetic. The baseline characteristics of all included COPD patients are detailed in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. The cohort was randomly divided into training (n\u0026thinsp;=\u0026thinsp;3,298) and testing sets (n\u0026thinsp;=\u0026thinsp;1,412) through stratified sampling at a 7:3 ratio. The complete study workflow is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eAs shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, when comparing COPD patients without diabetes to those with diabetes, the two groups showed no significant differences in the following parameters: age, sex, race, language, marital status, hemoglobin, Rdw, hematocrit, sodium, potassium, total calcium, chloride, glucose, anion gap, prothrombin time, partial thromboplastin time, international normalized ratio (Inr), total bilirubin, blood urea nitrogen, creatinine, length of hospitalization, and mortality status.\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 table for COPD patients in general wards\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=\"left\" 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\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTotal\u003c/p\u003e\u003cp\u003e(N\u0026thinsp;=\u0026thinsp;4710)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNon-Diabetes\u003c/p\u003e\u003cp\u003e(N\u0026thinsp;=\u0026thinsp;3153)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eDiabetes\u003c/p\u003e\u003cp\u003e(N\u0026thinsp;=\u0026thinsp;1557)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eAge (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e72.00 [64.00, 80.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e73.00[64.00,81.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e72.00[65.00,79.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.044\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSex (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFemale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2317(49.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1613(51.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e704(45.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\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\u003eMale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2393(50.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1540(48.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e853(54.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRace (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eWhite\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e3416(72.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1859(59.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1557(100.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\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\u003eOther\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1294(27.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1294(41.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0(0.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eInsurance (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMedicare\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e3347(71.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2213(70.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1134(72.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.064\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\u003eOther\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1363(28.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e940(29.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e423(27.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLanguage (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEnglish\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4459(94.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3024(95.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1435(92.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\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\u003eOther\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e251(5.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e129(4.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e122(7.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMarital Status (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMarried\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1683(35.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e619(19.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1064(68.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\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\u003eOther\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e3027(64.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2534(80.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e493(31.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eWbc (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e9.60 [7.20, 13.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e9.63[7.10,13.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e9.60[7.30,13.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.248\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eRbc (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3.79 [3.26, 4.29]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3.78[3.26,4.28]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.81[3.25,4.30]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.745\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003ePlatelet count (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e202.50 [154.50, 260.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e201.50[153.00,262.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e203.73[157.50,256.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.842\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eHemoglobin (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e11.20 [9.60, 12.70]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e11.30[9.75,12.80]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e11.00[9.35,12.50]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eRdw (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e14.40 [13.40, 15.90]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e14.30[13.30,15.80]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e14.60[13.50,16.20]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eHematocrit (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e35.00 [30.20, 39.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e35.20[30.50,39.15]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e34.47[29.70,38.70]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eSodium (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e139.00 [136.00, 141.05]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e139.00[136.50,141.50]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e139.00[136.00,141.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.026\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003ePotassium (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4.20 [3.80, 4.53]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e4.10[3.80,4.50]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e4.20[3.90,4.60]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eTotal calcium (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8.74 [8.30, 9.10]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e8.70[8.30,9.10]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e8.80[8.40,9.15]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eChloride (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e102.00 [98.50, 105.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e102.00[99.00,105.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e101.50[98.00,104.60]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eGlucose (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e118.50 [98.00, 150.95]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e110.00[95.00,132.49]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e150.00[114.50,195.67]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eAnion gap (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e14.00 [11.50, 16.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e13.50[11.00,16.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e14.00[12.00,16.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003ePt (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e13.63 [12.20, 16.08]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e13.60[12.20,15.85]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e13.83[12.33,16.50]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.002\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003ePtt (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e34.00 [29.10, 42.69]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e33.73[29.10,41.77]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e34.63[29.17,44.85]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.006\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eInr (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.25 [1.10, 1.49]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.25[1.10,1.46]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.29[1.10,1.50]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eTotal bilirubin (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.67 [0.46, 1.67]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.68[0.48,1.70]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.65[0.45,1.60]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.044\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eAlt (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e30.11 [19.28, 92.99]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e30.11[19.33,88.73]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e30.08[19.00,105.51]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.791\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eAst (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e37.50 [23.69, 107.58]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e37.65[24.00,103.99]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e37.26[23.00,116.21]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.772\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eUrea nitrogen (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e19.00 [13.00, 28.67]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e18.00[12.00,26.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e22.00[15.00,34.00]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eCreatinine (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.90 [0.70, 1.30]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.90[0.70,1.20]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.10[0.80,1.53]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAntihypertensive (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1719(36.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1265(40.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e454(29.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2991(63.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1888(59.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1103(70.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGlucocorticoids (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4118(87.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2741(86.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1377(88.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.156\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e592(12.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e412(13.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e180(11.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNephrotoxic (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1379(29.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1024(32.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e355(22.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e3331(70.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2129(67.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1202(77.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eImmunosuppressant (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4642(98.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3113(98.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1529(98.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.192\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e68(1.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e40(1.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e28(1.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eHospitalization duration (Median [IQR])\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5.98 [3.16, 10.81]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e5.84[3.04,10.63]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e6.40[3.55,11.36]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.003\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMortality outcomes (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e3020(64.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2072(65.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e948(60.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.001\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1690(35.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1081(34.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e609(39.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIn-hospital death status (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4346(92.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2912(92.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1434(92.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.801\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e364(7.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e241(7.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e123(7.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHypertension (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2848(60.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1847(58.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1001(64.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1862(39.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1306(41.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e556(35.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eChronic kidney disease (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e3652(77.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2636(83.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1016(65.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1058(22.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e517(16.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e541(34.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHyperlipidemia (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2416(51.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1793(56.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e623(40.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2294(48.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1360(43.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e934(60.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAcute renal failure (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e3323(70.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2345(74.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e978(62.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1387(29.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e808(25.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e579(37.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHeart failure (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e3043(64.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2210(70.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e833(53.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1667(35.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e943(29.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e724(46.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMyocardial infarction (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4166(88.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2833(89.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1333(85.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e544(11.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e320(10.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e224(14.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTuberculosis (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4708(100.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3152(100.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1556(99.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2(0.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1(0.0)\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=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePneumonia (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e3534(75.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2347(74.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1187(76.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.191\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1176(25.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e806(25.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e370(23.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eStroke (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4328(91.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2930(92.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1398(89.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e382(8.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e223(7.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e159(10.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePulmonary embolism (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4409(93.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2952(93.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1457(93.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e301(6.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e201(6.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e100(6.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCovid19 (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4581(97.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3073(97.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1508(96.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.266\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e129(2.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e80(2.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e49(3.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSmoke (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e3436(73.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2272(72.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1164(74.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.054\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1274(27.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e881(27.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e393(25.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDrink wine (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4245(90.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2788(88.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1457(93.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e465(9.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e365(11.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e100(6.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCirrhosis (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4426(94.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2967(94.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1459(93.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.638\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e284(6.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e186(5.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e98(6.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHepatitis (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4463(94.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2977(94.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1486(95.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.158\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e247(5.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e176(5.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e71(4.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMalignant tumors (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e3844(81.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2547(80.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1297(83.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.039\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\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e866(18.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e606(19.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e260(16.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eHowever, statistically significant differences (P\u0026thinsp;\u0026lt;\u0026thinsp;0.05) were observed between the two groups for several comorbidities: hypertension, chronic kidney disease, hyperlipidemia, acute renal failure, heart failure, myocardial infarction, stroke, alcohol consumption, and malignant tumors.\u003c/p\u003e\u003cp\u003eRegarding medication use, significant differences were found for antihypertensive drugs (P\u0026thinsp;=\u0026thinsp;0.032) and nephrotoxic medications (P\u0026thinsp;=\u0026thinsp;0.032). No statistically significant differences were observed for the remaining variables examined.\u003c/p\u003e\u003cp\u003e\u003cb\u003eFeature Selection\u003c/b\u003e\u003c/p\u003e\u003cp\u003eWe conducted Lasso regression analysis incorporating all 49 variables in the dataset with 10-fold cross-validation. The analytical results are visually presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. Specifically, Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA illustrates the variable selection trajectory of the Lasso regression model, while Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB demonstrates the feature selection process - the left dashed line denotes the minimum lambda (lambda.min), and the right dashed line indicates the lambda within one standard error of the optimal value (lambda.1se). Based on the Lasso regression outcomes, we subsequently performed multivariate and stepwise logistic regression analyses on the 17 initially selected variables. The comprehensive results of the stepwise logistic regression analysis are detailed in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e reveals several significant associations between clinical variables and diabetes risk in COPD patients. Advancing age demonstrated a protective effect, with each additional year associated with a 2.6% reduction in diabetes risk (OR\u0026thinsp;=\u0026thinsp;0.974, 95% CI\u0026thinsp;=\u0026thinsp;0.963\u0026ndash;0.986, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Insurance status showed marked variations, where patients with \"other\" insurance types had substantially elevated diabetes risk (OR\u0026thinsp;=\u0026thinsp;11.609, 95% CI\u0026thinsp;=\u0026thinsp;7.203\u0026ndash;19.385, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), while Medicare coverage was protective. Similarly, \"other\" language speakers faced higher risk (OR\u0026thinsp;=\u0026thinsp;10.019, 95% CI\u0026thinsp;=\u0026thinsp;6.001\u0026ndash;17.108, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Married patients exhibited a remarkable 97.4% risk reduction compared to unmarried individuals (OR\u0026thinsp;=\u0026thinsp;0.026, 95% CI\u0026thinsp;=\u0026thinsp;0.016\u0026ndash;0.042, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001).\u003c/p\u003e\u003cp\u003eSeveral laboratory parameters significantly influenced diabetes risk: each unit increase in RDW elevated risk by 7.5% (OR\u0026thinsp;=\u0026thinsp;1.075, 95% CI\u0026thinsp;=\u0026thinsp;1.026\u0026ndash;1.127, P\u0026thinsp;=\u0026thinsp;0.003), while higher total calcium levels increased risk by 20% per unit (OR\u0026thinsp;=\u0026thinsp;1.200, 95% CI\u0026thinsp;=\u0026thinsp;1.022\u0026ndash;1.410, P\u0026thinsp;=\u0026thinsp;0.026). Notably, each unit increase in glucose concentration was associated with a 2.3% risk elevation (OR\u0026thinsp;=\u0026thinsp;1.023, 95% CI\u0026thinsp;=\u0026thinsp;1.020\u0026ndash;1.025, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001).\u003c/p\u003e\u003cp\u003eComorbid conditions substantially impacted diabetes risk: chronic kidney disease (OR\u0026thinsp;=\u0026thinsp;2.370, 95% CI\u0026thinsp;=\u0026thinsp;1.843\u0026ndash;3.057, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), hyperlipidemia (OR\u0026thinsp;=\u0026thinsp;1.854, 95% CI\u0026thinsp;=\u0026thinsp;1.495\u0026ndash;2.301, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and heart failure (OR\u0026thinsp;=\u0026thinsp;1.299, 95% CI\u0026thinsp;=\u0026thinsp;1.028\u0026ndash;1.642, P\u0026thinsp;=\u0026thinsp;0.029) all increased risk, whereas pneumonia reduced risk by 36.8% (OR\u0026thinsp;=\u0026thinsp;0.632, 95% CI\u0026thinsp;=\u0026thinsp;0.490\u0026ndash;0.814, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001).\u003c/p\u003e\u003cp\u003eLifestyle and medication factors also showed significant associations: alcohol consumption decreased risk by 43% (OR\u0026thinsp;=\u0026thinsp;0.570, 95% CI\u0026thinsp;=\u0026thinsp;0.388\u0026ndash;0.829, P\u0026thinsp;=\u0026thinsp;0.004), glucocorticoid use was protective (OR\u0026thinsp;=\u0026thinsp;0.634, 95% CI\u0026thinsp;=\u0026thinsp;0.451\u0026ndash;0.889, P\u0026thinsp;=\u0026thinsp;0.008), while nephrotoxic drug use increased risk (OR\u0026thinsp;=\u0026thinsp;1.401, 95% CI\u0026thinsp;=\u0026thinsp;1.104\u0026ndash;1.780, P\u0026thinsp;=\u0026thinsp;0.006). No significant association was observed between race and diabetes risk (P\u0026thinsp;=\u0026thinsp;0.952).\u003c/p\u003e\u003cp\u003eFollowing the aforementioned statistical analyses, a total of 14 variables were retained for developing subsequent machine learning models. These variables included: age, sex, insurance type, language, marital status, red cell distribution width (RDW), total calcium level, glucose, chronic kidney disease (CKD), hyperlipidemia, heart failure, pneumonia, alcohol consumption, glucocorticoid use, and nephrotoxic medication use.\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\u003eStepwise logistic regression analysis results\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"7\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eB Value\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSE\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eZ Value\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eOR\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003e95%CI\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.026\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.006\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-4.361\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.974\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.963\u0026ndash;0.986\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.000\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRace\u0026thinsp;=\u0026thinsp;Other\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-18.318\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e301.796\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-0.061\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.000\u0026ndash;0.000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.952\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eInsurance\u0026thinsp;=\u0026thinsp;Other\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e2.452\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.252\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e9.731\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e11.609\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e7.203\u0026ndash;19.385\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.000\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLanguage\u0026thinsp;=\u0026thinsp;Other\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e2.304\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.267\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e8.644\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e10.019\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e6.001\u0026ndash;17.108\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.000\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMarital Status\u0026thinsp;=\u0026thinsp;Other\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-3.634\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.243\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-14.963\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.026\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.016\u0026ndash;0.042\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.000\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRdw\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.072\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.024\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3.004\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.075\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.026\u0026ndash;1.127\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.003\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTotal calcium\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.182\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.082\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2.224\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.200\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.022\u0026ndash;1.410\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.026\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGlucose\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.022\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.001\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e15.555\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.023\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.020\u0026ndash;1.025\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.000\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGlucocorticoids\u0026thinsp;=\u0026thinsp;Yes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.455\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.173\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-2.635\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.634\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.451\u0026ndash;0.889\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.008\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNephrotoxic\u0026thinsp;=\u0026thinsp;Yes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.337\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.122\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2.770\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.401\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.104\u0026ndash;1.780\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.006\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eChronic kidney disease\u0026thinsp;=\u0026thinsp;Yes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.863\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.129\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e6.686\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e2.370\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.843\u0026ndash;3.057\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.000\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHyperlipidemia\u0026thinsp;=\u0026thinsp;Yes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.617\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.110\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e5.617\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.854\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.495\u0026ndash;2.301\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.000\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHeart failure\u0026thinsp;=\u0026thinsp;Yes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.262\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.120\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2.190\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.299\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1.028\u0026ndash;1.642\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.029\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePneumonia\u0026thinsp;=\u0026thinsp;Yes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.459\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.130\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-3.540\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.632\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.490\u0026ndash;0.814\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.000\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDrink Wine\u0026thinsp;=\u0026thinsp;Yes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.563\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.193\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e-2.908\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.570\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.388\u0026ndash;0.829\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.004\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eEvaluation and Comparison of Model Performance\u003c/b\u003e\u003c/p\u003e\u003cp\u003eBased on the feature selection results, we employed six widely-used machine learning algorithms to develop predictive models, including Logistic Regression, Random Forest (RF), eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), Natural Gradient Boosting (NGBoost), and Categorical Boosting (CatBoost).\u003c/p\u003e\u003cp\u003eDuring model development, we optimized hyperparameters through 10-fold cross-validation combined with grid search. Model performance was evaluated using multiple metrics, with the optimal hyperparameter combination ultimately selected based on the ROC-AUC (Receiver Operating Characteristic - Area Under the Curve) metric. The final model was trained using this optimal parameter set.\u003c/p\u003e\u003cp\u003eWe visualized the receiver operating characteristic (ROC) curves of the six machine learning models on the training set, test set, and external validation set in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA. The Random Forest model demonstrated the best performance on the training set (AUC\u0026thinsp;=\u0026thinsp;0.96), but showed a significant performance drop on the test set (AUC\u0026thinsp;=\u0026thinsp;0.90), indicating potential overfitting. In contrast, the LightGBM model maintained consistently excellent performance across both training (AUC\u0026thinsp;=\u0026thinsp;0.93) and test sets (AUC\u0026thinsp;=\u0026thinsp;0.90).\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB presents the calibration curves and prediction outcomes, revealing that all six models achieved high goodness-of-fit across the training, test, and validation sets. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC displays the precision-recall curves, where the x-axis represents recall (the proportion of actual positives correctly identified) and the y-axis represents precision (the proportion of predicted positives that are truly positive). Higher recall indicates fewer missed cases, while higher precision indicates fewer false positives. The LightGBM model demonstrated optimal overall performance, with stable average precision across the training, test, and external validation sets.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA presents the clinical decision curves for different models. The LightGBM model demonstrated superior net benefit across most threshold probabilities in all datasets (training, test, and validation sets). Figure\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB provides a comprehensive summary of performance metrics across all machine learning models. In the training set, the models exhibited the following performance characteristics: Logistic Regression (Accuracy\u0026thinsp;=\u0026thinsp;0.82, Precision\u0026thinsp;=\u0026thinsp;0.83, Recall\u0026thinsp;=\u0026thinsp;0.55, F1-score\u0026thinsp;=\u0026thinsp;0.66, AUC\u0026thinsp;=\u0026thinsp;0.88); Random Forest (Accuracy\u0026thinsp;=\u0026thinsp;0.89, Precision\u0026thinsp;=\u0026thinsp;0.88, Recall\u0026thinsp;=\u0026thinsp;0.77, F1-score\u0026thinsp;=\u0026thinsp;0.82, AUC\u0026thinsp;=\u0026thinsp;0.96) ; XGBoost (Accuracy\u0026thinsp;=\u0026thinsp;0.82, Precision\u0026thinsp;=\u0026thinsp;0.83, Recall\u0026thinsp;=\u0026thinsp;0.54, F1-score\u0026thinsp;=\u0026thinsp;0.65, AUC\u0026thinsp;=\u0026thinsp;0.91) ; LightGBM (Accuracy\u0026thinsp;=\u0026thinsp;0.85, Precision\u0026thinsp;=\u0026thinsp;0.80, Recall\u0026thinsp;=\u0026thinsp;0.75, F1-score\u0026thinsp;=\u0026thinsp;0.76, AUC\u0026thinsp;=\u0026thinsp;0.93) ; NGBoost (Accuracy\u0026thinsp;=\u0026thinsp;0.84, Precision\u0026thinsp;=\u0026thinsp;0.79, Recall\u0026thinsp;=\u0026thinsp;0.68, F1-score\u0026thinsp;=\u0026thinsp;0.73, AUC\u0026thinsp;=\u0026thinsp;0.92); CatBoost (Accuracy\u0026thinsp;=\u0026thinsp;0.80, Precision\u0026thinsp;=\u0026thinsp;0.85, Recall\u0026thinsp;=\u0026thinsp;0.45, F1-score\u0026thinsp;=\u0026thinsp;0.59, AUC\u0026thinsp;=\u0026thinsp;0.90). When evaluated on the test set, the models demonstrated: Logistic Regression (Accuracy\u0026thinsp;=\u0026thinsp;0.80, Precision\u0026thinsp;=\u0026thinsp;0.81, Recall\u0026thinsp;=\u0026thinsp;0.54, F1-score\u0026thinsp;=\u0026thinsp;0.65, AUC\u0026thinsp;=\u0026thinsp;0.87); Random Forest (Accuracy\u0026thinsp;=\u0026thinsp;0.83, Precision\u0026thinsp;=\u0026thinsp;0.79, Recall\u0026thinsp;=\u0026thinsp;0.69, F1-score\u0026thinsp;=\u0026thinsp;0.74, AUC\u0026thinsp;=\u0026thinsp;0.90); XGBoost (Accuracy\u0026thinsp;=\u0026thinsp;0.80, Precision\u0026thinsp;=\u0026thinsp;0.82, Recall\u0026thinsp;=\u0026thinsp;0.55, F1-score\u0026thinsp;=\u0026thinsp;0.66, AUC\u0026thinsp;=\u0026thinsp;0.90); LightGBM (Accuracy\u0026thinsp;=\u0026thinsp;0.83, Precision\u0026thinsp;=\u0026thinsp;0.77, Recall\u0026thinsp;=\u0026thinsp;0.71, F1-score\u0026thinsp;=\u0026thinsp;0.74, AUC\u0026thinsp;=\u0026thinsp;0.90); NGBoost (Accuracy\u0026thinsp;=\u0026thinsp;0.82, Precision\u0026thinsp;=\u0026thinsp;0.78, Recall\u0026thinsp;=\u0026thinsp;0.69, F1-score\u0026thinsp;=\u0026thinsp;0.73, AUC\u0026thinsp;=\u0026thinsp;0.90); CatBoost (Accuracy\u0026thinsp;=\u0026thinsp;0.79, Precision\u0026thinsp;=\u0026thinsp;0.84, Recall\u0026thinsp;=\u0026thinsp;0.47, F1-score\u0026thinsp;=\u0026thinsp;0.60, AUC\u0026thinsp;=\u0026thinsp;0.89).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eModel interpretability analysis\u003c/b\u003e\u003c/p\u003e\u003cp\u003eTo better understand the relationship between the model and the data, we employed SHAP (SHapley Additive exPlanations) to provide an intuitive interpretation of the best-performing LightGBM model, demonstrating how these variables influence the predictions.\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA presents a beeswarm plot illustrating the importance of each feature in the model. The x-axis represents the SHAP value, which quantifies the contribution of each feature to the model's prediction. Positive SHAP values indicate an increased likelihood of the predicted outcome, while negative values suggest a decreased likelihood. Each point corresponds to a sample's SHAP value, with color intensity reflecting the original feature value (red for high values, blue for low values), allowing us to visualize the relationship between feature magnitude and its impact on predictions.\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB ranks features by their mean absolute SHAP values, where longer bars denote greater influence on the model's output. Marital status emerged as the most critical predictor (mean |SHAP| = +0.2), followed by glucose (|SHAP| = +0.11) and insurance type (|SHAP| = +0.06). Other notable contributors included chronic kidney disease (CKD) (|SHAP| = +0.03), hyperlipidemia (|SHAP| = +0.03), language (|SHAP| = +0.02), total calcium (|SHAP| = +0.02), age (|SHAP| = +0.02), and nephrotoxic drug use (|SHAP| = +0.02). The remaining five features collectively contributed a mean |SHAP| of +\u0026thinsp;0.03.\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC displays a SHAP decision plot for the top 50 samples, transforming the model's \"black-box\" logic into an interpretable visual pathway. The x-axis represents the model's output value (higher positive values indicate greater diabetes risk, while lower negative values suggest reduced risk), and the y-axis lists clinically relevant features ordered by importance.\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD presents a waterfall plot of an individual sample, with the color gradient (red to blue) indicating the magnitude of the feature values. Figure\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eE illustrates the SHAP force plot for the same patient with COPD, demonstrating the contribution of each clinical feature to the model\u0026rsquo;s prediction. In this plot, each bar represents the impact of a specific feature, where red denotes higher values and blue indicates lower values. Together, Figs.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD and \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eE provide an intuitive visualization of how distinct features influence the model\u0026rsquo;s prediction of diabetes. This SHAP-based interpretability analysis enhances our understanding of the role of individual features in shaping the output of the LightGBM predictive model.\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eCOPD is a prevalent respiratory disorder associated with high morbidity and mortality\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Patients with COPD exhibit a progressive decline in daily physical activity over time, which parallels worsening lung function and overall health status\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. Diabetes mellitus is a major comorbidity of COPD\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e, and lifestyle-related risk factors are considered central to the development of both type 1 and type 2 diabetes\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. A cohort study conducted in Taiwan revealed that 16% of COPD patients had pre-existing T2DM, while 19% were newly diagnosed with T2DM during a 10-year follow-up period, indicating an elevated prevalence and incidence of this comorbidity. Notably, the study found that diabetes\u0026mdash;whether pre-existing or newly diagnosed\u0026mdash;was associated with poorer clinical outcomes in COPD patients\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. Therefore, early detection of diabetes and personalized management are critical for improving prognosis in COPD patients.\u003c/p\u003e\u003cp\u003eThis study leverages the MIMIC-IV database to propose an efficient, non-invasive diabetes risk prediction model based on demographic characteristics, laboratory indicators, and disease and medication usage. By incorporating a large patient cohort, the model ensures broad generalizability. Our objective was to develop a machine learning-based risk prediction model and validate it both internally and externally. The results demonstrate that our LightGBM model achieved an AUC of 0.85 in the training set and 0.80 in the external validation set, indicating robust predictive performance and strong applicability to external datasets.\u003c/p\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e presents the results of stepwise logistic regression analysis. Combined with SHAP analysis, we identified marital status, blood glucose levels, and insurance type as the three most significant features influencing diabetes development in COPD patients. Existing literature has extensively examined the relationship between type 2 diabetes mellitus and marital status through cross-sectional studies. Marital status independently correlates with T2DM incidence, where married individuals demonstrate significantly lower diabetes risk compared to divorced persons, despite comparable weight gain\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. A prospective study utilizing ARIC data further revealed that persistent single status was associated with a 34% increased risk of T2DM in women\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. The social causation theory suggests that marriage confers health-protective effects through stress reduction, anxiety alleviation, and promotion of positive health behaviors, while unmarried status may adversely impact health outcomes\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. Insurance type also significantly contributes to the model's predictive performance. A study by Mahoney S et al. found that Medicare coverage was associated with reduced odds of undiagnosed prediabetes \u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e. Limited healthcare access and screening opportunities among uninsured populations may lead to underdiagnosis of both prediabetes and diabetes.\u003c/p\u003e\u003cp\u003eAs shown in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, we observed a positive correlation between red cell distribution width (RDW) and diabetes risk. This finding aligns with existing evidence linking elevated RDW values with pro-inflammatory conditions. A retrospective study by Nada et al. demonstrated significantly higher RDW levels in diabetic patients compared to non-diabetic controls\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. Engstr\u0026ouml;m G et al. further reported a significant positive association between RDW and HbA1c, with each 1-SD increase in RDW corresponding to a 0.10% rise in HbA1c\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e. Our analysis also revealed that increased serum calcium levels elevate diabetes risk, consistent with a cross-sectional study of 39,645 participants that established a causal relationship between higher serum calcium levels and increased type 2 diabetes prevalence\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. Supporting this, Sing CW et al. found that elevated serum calcium concentrations were associated with diabetes incidence\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eRegarding disease comorbidities, our analysis revealed that both chronic kidney disease (CKD) and heart failure (HF) significantly increased diabetes risk (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). HF, diabetes mellitus (DM), and CKD represent a triad of frequently interrelated conditions that should be considered collectively rather than in isolation. Epidemiological data indicate that 25%-40% of HF patients have comorbid DM, while 40%-50% exhibit concurrent CKD. Notably, 16% of HF patients present with all three conditions - a high-risk combination associated with substantially elevated hospitalization and mortality rates. The pathophysiological interplay between HF, DM, and CKD involves complex mechanisms, likely mediated through shared cardiovascular/metabolic risk factors, chronic inflammation, oxidative stress, and downstream neuroendocrine pathway activation\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e. Given the demonstrated renal and cardiovascular protective effects of novel antidiabetic agents, implementing tailored therapeutic strategies addressing these comorbidities has become increasingly crucial\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. However, further research is needed to elucidate these relationships more comprehensively.\u003c/p\u003e\u003cp\u003eOur study additionally identified hyperlipidemia as a significant diabetes risk factor (OR\u0026thinsp;=\u0026thinsp;1.854, 95%CI\u0026thinsp;=\u0026thinsp;1.495\u0026ndash;2.301, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Dyslipidemia constitutes a core component of metabolic syndrome in T2DM patients, where lipid abnormalities primarily arise as secondary consequences of insulin resistance. Consequently, any intervention reducing insulin resistance would be expected to beneficially impact lipid profiles. Current guidelines already recommend aggressive lipid-lowering therapy for high-risk T2DM individuals\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. Our logistic regression results suggested reduced diabetes risk among pneumonia patients compared to non-pneumonia cases. While Brunetti VC et al. reported potential associations between T2DM and increased community-acquired pneumonia (CAP) risk, the available evidence derives from studies with substantial bias risk, necessitating higher-quality research to verify these observations\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eThe study by Marjo Heinjoki et al. found no significant difference in renal function between elderly individuals with diabetes and their non-diabetic counterparts. Furthermore, the use of potentially nephrotoxic medications appeared to play only a minor role in renal function decline among community-dwelling elderly in Finland's Inner-Savo region\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e. Interestingly, our current study demonstrated a significant positive association between nephrotoxic drug use and diabetes incidence (P\u0026thinsp;\u0026lt;\u0026thinsp;0.05). However, existing research on the relationship between nephrotoxic medications and diabetes development remains limited. This novel finding warrants further investigation through large-scale prospective studies to elucidate potential causal mechanisms and clinical implications.\u003c/p\u003e\u003cp\u003eAs evidenced in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, our study demonstrated a reduced diabetes risk among COPD patients with a history of alcohol consumption. Rachel Golan and colleagues' systematic review synthesized existing evidence, suggesting that initiating moderate alcohol consumption appears reasonably safe for individuals with well-controlled blood glucose, particularly regarding heart rate variability and arterial plaque formation\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. However, this finding should not be construed as an endorsement of heavy drinking. While epidemiological studies have consistently shown an inverse association between alcohol consumption and type 2 diabetes risk\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e, the dose-response relationship requires careful consideration. The research by G Caimi et al. revealed that light-to-moderate, regular alcohol intake - irrespective of beverage type - was negatively correlated with type 2 diabetes incidence[28]. Notably, moderate wine consumption may confer additional benefits by improving oxidative status in diabetic patients, suggesting this lifestyle modification could potentially mitigate diabetes progression and complication development\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eOur study evaluated six machine learning algorithms, among which the LightGBM model demonstrated particularly promising performance. Through SHAP interpretability analysis, we gained more precise insights into how the 14 selected features influenced the predictive outputs of our developed model.\u003c/p\u003e\u003cp\u003e\u003cb\u003eLimitation\u003c/b\u003e\u003c/p\u003e\u003cp\u003eWe acknowledge that both our primary dataset and external validation cohort were derived exclusively from the MIMIC-IV database, which may limit the generalizability of our findings to other populations or geographic regions. To enhance the model's broader clinical applicability, future studies should incorporate more diverse datasets spanning wider geographical distributions, ethnic populations, and temporal ranges.\u003c/p\u003e\u003cp\u003e\u003cb\u003eOnline Prediction Tool Development\u003c/b\u003e\u003c/p\u003e\u003cp\u003eBased on the LightGBM model developed in this study, we created a user-friendly online prediction tool for clinicians to assess the risk of diabetes in COPD patients. The tool is accessible at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://copd-diabetes-predictor-model.streamlit.app/\u003c/span\u003e\u003cspan address=\"https://copd-diabetes-predictor-model.streamlit.app/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThe LightGBM model developed in this study demonstrates robust predictive performance, providing clinicians with a novel screening tool for diabetes risk assessment in COPD patients.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe raw data that support the findings of this study are available from the corresponding author, Xishi Sun, upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was supported by the Guang dong Medical Research Fund Project (No.:A2024728, A2024723),Zhanjiang Science and Technology Research Project in 2024 \u0026nbsp; (No.: 2024B01356) \u0026nbsp;and the High-level Talents Scientific Research Start-up Funds of the Affiliated Hospital of Guangdong Medical University (No.:GCC2022028).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026apos; contributions:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors listed have made a substantial, direct, and intellectual contribution to the work, and approved it for publication.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eMannino, D. M. \u0026amp; Buist, A. S. Global burden of COPD: risk factors, prevalence, and future trends. \u003cem\u003eLancet\u003c/em\u003e \u003cb\u003e370\u003c/b\u003e, 765\u0026ndash;773 (2007).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWannamethee, S. G. et al. Lung function and risk of type 2 diabetes and fatal and nonfatal major coronary heart disease events: possible associations with inflammation. \u003cem\u003eDiabetes Care\u003c/em\u003e. \u003cb\u003e33\u003c/b\u003e, 1990\u0026ndash;1996 (2010).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFrosteg\u0026aring;rd, J. Immunity, atherosclerosis and cardiovascular disease. \u003cem\u003eBMC Med.\u003c/em\u003e \u003cb\u003e11\u003c/b\u003e, 117 (2013).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang, T. et al. COPD and T2DM: a Mendelian randomization study. \u003cem\u003eFront. Endocrinol. (Lausanne)\u003c/em\u003e. \u003cb\u003e15\u003c/b\u003e, 1302641 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHughes, M. J., McGettrick, H. M. \u0026amp; Sapey, E. Shared mechanisms of multimorbidity in COPD, atherosclerosis and type-2 diabetes: the neutrophil as a potential inflammatory target. \u003cem\u003eEur. Respir Rev.\u003c/em\u003e \u003cb\u003e29\u003c/b\u003e, 190102 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJohnson, A. E. W. et al. MIMIC-IV, a freely accessible electronic health record dataset. \u003cem\u003eSci. Data\u003c/em\u003e. \u003cb\u003e10\u003c/b\u003e, 1 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eExplaining a series of models by propagating Shapley. values - PubMed. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://pubmed.ncbi.nlm.nih.gov/35922410/\u003c/span\u003e\u003cspan address=\"https://pubmed.ncbi.nlm.nih.gov/35922410/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFerrera, M. C., Labaki, W. W. \u0026amp; Han, M. K. Advances in Chronic Obstructive Pulmonary Disease. \u003cem\u003eAnnu. Rev. Med.\u003c/em\u003e \u003cb\u003e72\u003c/b\u003e, 119\u0026ndash;134 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWaschki, B. et al. Disease Progression and Changes in Physical Activity in Patients with Chronic Obstructive Pulmonary Disease. \u003cem\u003eAm. J. Respir Crit. Care Med.\u003c/em\u003e \u003cb\u003e192\u003c/b\u003e, 295\u0026ndash;306 (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePrevalence of comorbidities in. patients with chronic obstructive pulmonary disease - PubMed. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://pubmed.ncbi.nlm.nih.gov/20134148/\u003c/span\u003e\u003cspan address=\"https://pubmed.ncbi.nlm.nih.gov/20134148/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOwen, N., Healy, G. N., Matthews, C. E. \u0026amp; Dunstan, D. W. Too much sitting: the population health science of sedentary behavior. \u003cem\u003eExerc. Sport Sci. Rev.\u003c/em\u003e \u003cb\u003e38\u003c/b\u003e, 105\u0026ndash;113 (2010).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHo, T. W. et al. Diabetes mellitus in patients with chronic obstructive pulmonary disease-The impact on mortality. \u003cem\u003ePLoS One\u003c/em\u003e. \u003cb\u003e12\u003c/b\u003e, e0175794 (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ede Oliveira, C. M. et al. Relationship between marital status and incidence of type 2 diabetes mellitus in a Brazilian rural population: The Baependi Heart Study. \u003cem\u003ePLoS One\u003c/em\u003e. \u003cb\u003e15\u003c/b\u003e, e0236869 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSchwandt, H. M., Coresh, J. \u0026amp; Hindin, M. J. Marital Status, Hypertension, Coronary Heart Disease, Diabetes, and Death Among African American Women and Men: Incidence and Prevalence in the Atherosclerosis Risk in Communities (ARIC) Study Participants. \u003cem\u003eJ. Fam Issues\u003c/em\u003e. \u003cb\u003e31\u003c/b\u003e, 1211\u0026ndash;1229 (2010).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJaffe, D. H., Manor, O., Eisenbach, Z. \u0026amp; Neumark, Y. D. The protective effect of marriage on mortality in a dynamic society. \u003cem\u003eAnn. Epidemiol.\u003c/em\u003e \u003cb\u003e17\u003c/b\u003e, 540\u0026ndash;547 (2007).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMahoney, S. et al. Health Insurance Is Associated with Decreased Odds for Undiagnosed Prediabetes and Type 2 Diabetes in American Adults. \u003cem\u003eInt. J. Environ. Res. Public. Health\u003c/em\u003e. \u003cb\u003e17\u003c/b\u003e, 4706 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNada, A. M. Red cell distribution width in type 2 diabetic patients. \u003cem\u003eDiabetes Metab. Syndr. Obes.\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e, 525\u0026ndash;533 (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEngstr\u0026ouml;m, G. et al. Red cell distribution width, haemoglobin A1c and incidence of diabetes mellitus. \u003cem\u003eJ. Intern. Med.\u003c/em\u003e \u003cb\u003e276\u003c/b\u003e, 174\u0026ndash;183 (2014).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhai, Z. et al. Association between serum calcium level and type 2 diabetes: An NHANES analysis and Mendelian randomization study. \u003cem\u003eDiabet. Med.\u003c/em\u003e \u003cb\u003e40\u003c/b\u003e, e15080 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSing, C. W. et al. Serum calcium and incident diabetes: an observational study and meta-analysis. \u003cem\u003eOsteoporos. Int.\u003c/em\u003e \u003cb\u003e27\u003c/b\u003e, 1747\u0026ndash;1754 (2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVijay, K., Neuen, B. L. \u0026amp; Lerma, E. V. Heart Failure in Patients with Diabetes and Chronic Kidney Disease: Challenges and Opportunities. \u003cem\u003eCardiorenal Med.\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e, 1\u0026ndash;10 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWu, M. Z. et al. Chronic kidney disease begets heart failure and vice versa: temporal associations between heart failure events in relation to incident chronic kidney disease in type 2 diabetes. \u003cem\u003eDiabetes Obes. Metab.\u003c/em\u003e \u003cb\u003e25\u003c/b\u003e, 707\u0026ndash;715 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBetteridge, D. J. Diabetic dyslipidaemia. \u003cem\u003eDiabetes Obes. Metab.\u003c/em\u003e \u003cb\u003e2\u003c/b\u003e (Suppl 1), S31\u0026ndash;36 (2000).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eType 2 diabetes. mellitus and risk of community-acquired pneumonia: a systematic review and meta-analysis of observational studies - PubMed. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://pubmed.ncbi.nlm.nih.gov/33495386/\u003c/span\u003e\u003cspan address=\"https://pubmed.ncbi.nlm.nih.gov/33495386/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHeinjoki, M. et al. Kidney function and nephrotoxic drug use among older home-dwelling persons with or without diabetes in Finland. \u003cem\u003eBMC Nephrol.\u003c/em\u003e \u003cb\u003e21\u003c/b\u003e, 11 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGolan, R., Gepner, Y. \u0026amp; Shai, I. Wine and Health-New Evidence. \u003cem\u003eEur. J. Clin. Nutr.\u003c/em\u003e \u003cb\u003e72\u003c/b\u003e, 55\u0026ndash;59 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHuang, J., Wang, X. \u0026amp; Zhang, Y. Specific types of alcoholic beverage consumption and risk of type 2 diabetes: A systematic review and meta-analysis. \u003cem\u003eJ. Diabetes Investig\u003c/em\u003e. \u003cb\u003e8\u003c/b\u003e, 56\u0026ndash;68 (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCaimi, G., Carollo, C. \u0026amp; Lo Presti, R. Diabetes mellitus: oxidative stress and wine. \u003cem\u003eCurr. Med. Res. Opin.\u003c/em\u003e \u003cb\u003e19\u003c/b\u003e, 581\u0026ndash;586 (2003).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Chronic obstructive pulmonary disease, Diabetes, Prediction model, Machine learning","lastPublishedDoi":"10.21203/rs.3.rs-7033945/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7033945/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjective\u003c/h2\u003e\u003cp\u003eTo develop a machine learning (ML)-based prediction model for identifying high-risk diabetic individuals among COPD patients, thereby facilitating early and personalized management of this complication.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eData from COPD patients in the MIMIC-IV database were split into training (70%) and validation (30%) sets. LASSO regression and logistic regression were used to screen 49 variables, and six ML algorithms were employed to construct and internally validate the prediction model. Model performance was evaluated using multiple metrics, followed by external validation. Finally, SHAP (SHapley Additive exPlanations) analysis was performed for interpretability.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eAll six ML algorithms demonstrated excellent performance in the training, testing, and validation sets, as evidenced by ROC curve analysis, with LightGBM showing the best overall performance. Feature importance analysis revealed that marital status, blood glucose level, and insurance type were the top three factors influencing diabetes development in COPD patients.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e\u003cp\u003eThis study developed an interpretable ML-based risk prediction model for diabetes in COPD patients. The model provides clinicians with a novel tool for early personalized intervention, ultimately improving patient prognosis.\u003c/p\u003e","manuscriptTitle":"Construction and Validation of an Interpretable Machine Learning Model for Predicting Diabetes Risk in COPD Patients","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-19 11:34:33","doi":"10.21203/rs.3.rs-7033945/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-09-29T06:27:56+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-18T06:20:00+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"208199219448787441715674605314521035396","date":"2025-09-18T06:15:45+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-16T10:06:46+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"149310620322341986993504339006595800365","date":"2025-09-16T07:39:16+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-08-11T08:00:28+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-11T07:56:09+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-07-04T09:43:17+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-07-04T03:37:58+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2025-07-03T03:54:19+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"92d9d96f-ee78-4292-b347-636349d27553","owner":[],"postedDate":"August 19th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":53120880,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":53120881,"name":"Health sciences/Diseases"},{"id":53120882,"name":"Health sciences/Endocrinology"},{"id":53120883,"name":"Health sciences/Health care"},{"id":53120884,"name":"Health sciences/Medical research"},{"id":53120885,"name":"Health sciences/Risk factors"}],"tags":[],"updatedAt":"2026-04-28T05:38:44+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-19 11:34:33","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7033945","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7033945","identity":"rs-7033945","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}