Machine Learning Prediction of MACE in Older Chinese Adults Integrating Traditional and Geriatric-Specific Risk Factors: A CHARLS Cohort Analysis

Machine Learning Prediction of MACE in Older Chinese Adults Integrating Traditional and Geriatric-Specific Risk Factors: A CHARLS Cohort Analysis | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Machine Learning Prediction of MACE in Older Chinese Adults Integrating Traditional and Geriatric-Specific Risk Factors: A CHARLS Cohort Analysis Jingwei Li, Zhongyang Song, Fan Zou, Yiming Hu, Qian Xu, Guoxiong Hao, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6906133/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 5 You are reading this latest preprint version Abstract Background Cardiovascular disease (CVD) poses a substantial health burden on China's aging population. Existing cardiovascular risk models often perform poorly in older Chinese adults and rarely integrate geriatric-specific non-traditional factors. This study aimed to develop and validate machine learning-based models incorporating both traditional and non-traditional risk factors for predicting major adverse cardiovascular events (MACE) among older Chinese adults. Methods Data from 4,580 participants aged ≥ 60 years without baseline MACE were obtained from the China Health and Retirement Longitudinal Study (CHARLS, 2011–2018). Incident MACE (myocardial infarction or stroke) was self-reported during a median follow-up of approximately 7 years. Candidate predictors included demographics, health behaviors, clinical measures, anthropometric indices, biomarkers, depressive symptoms (CES-D score), and functional limitations (Activities of Daily Living, ADL). Missing data were handled via Multiple Imputation by Chained Equations (MICE, 5 imputations). Logistic Regression (LR), Random Forest (RF), and XGBoost models were trained using stratified 70/30 splits for training and testing sets. Hyperparameter tuning employed a grid search with limited complexity. Model performance was evaluated by discrimination (AUC), calibration (Brier score and calibration plots), and clinical utility (Decision Curve Analysis, DCA). Exploratory non-linear relationships were assessed using generalized additive models (GAMs). Results were benchmarked against the Framingham Risk Score (FRS-CVD), and an LR-based nomogram was developed. Results Incident MACE occurred in 28.7% of participants. The LR model demonstrated the highest discrimination (mean AUC = 0.649), closely followed by XGBoost (mean AUC = 0.645); both significantly outperformed RF (mean AUC = 0.632) and the FRS-CVD benchmark (mean AUC = 0.504). LR and XGBoost models showed good calibration and superior net benefit in DCA. Significant independent predictors in the LR model included hypertension history (OR = 1.85), diabetes (OR = 1.34), age (OR = 1.02/year), systolic blood pressure (OR = 1.006/mmHg), high education level (OR = 1.60), waist circumference (OR = 1.01/cm), depressive symptoms (CES-D score, OR = 1.03/point), and ADL limitations (OR = 1.11/limitation). GAM analysis revealed significant non-linear relationships for age and waist circumference. Conclusion Machine learning models integrating traditional and non-traditional factors effectively predict MACE risk in older Chinese adults, outperforming the standard FRS. Central obesity, depressive symptoms, and functional impairments were significant predictors, underscoring the importance of holistic cardiovascular risk assessment in geriatric populations. The developed nomogram offers a practical clinical tool pending external validation. Major Adverse Cardiovascular Events Risk Prediction Machine Learning Chinese Population Central Obesity Depression CHARLS Nomogram Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 1. Introduction Cardiovascular disease (CVD), encompassing major adverse cardiovascular events (MACE) such as myocardial infarction and stroke, remains a leading cause of morbidity and mortality worldwide, imposing a substantial burden on healthcare systems and societal resources( 1 ). This burden is particularly pronounced among older adults, as the prevalence of key CVD risk factors increases markedly with age. In China, which has one of the fastest-aging populations globally, the rising incidence of CVD and MACE among older adults presents an urgent public health challenge( 2 , 3 ). Consequently, effective prevention strategies critically depend on accurate risk stratification to identify high-risk individuals who would benefit most from targeted interventions. Traditional CVD risk prediction models, such as the widely used Framingham Risk Score (FRS) and SCORE, have been instrumental in clinical practice( 4 ). However, these models were primarily developed and validated using data from Western, predominantly middle-aged populations. As a result, their predictive accuracy and calibration are often suboptimal when applied to diverse ethnic groups, particularly Chinese populations, potentially leading to risk misestimation( 4 , 5 ). Moreover, conventional models frequently overlook risk factors that become increasingly relevant in older adults. Geriatric syndromes (e.g., functional limitations assessed by Activities of Daily Living [ADL]) and psychosocial factors (e.g., depression, social isolation), now recognized as independent contributors to cardiovascular health, are typically absent from standard risk calculators. The advent of machine learning (ML) offers new opportunities to enhance CVD risk prediction. ML algorithms excel at handling high-dimensional datasets, capturing complex non-linear relationships, and identifying novel predictor interactions often overlooked by traditional regression methods( 6 ). Recent studies have demonstrated that ML approaches can significantly outperform conventional statistical techniques in predicting cardiovascular risk( 7 ). Simultaneously, an expanding body of evidence has linked non-traditional risk factors—such as central obesity (better captured by waist circumference than by BMI), depressive symptoms, functional limitations (e.g., ADL impairments), and frailty markers (e.g., low grip strength)—to an increased risk of MACEs, particularly among older adults. Integrating these readily available yet underutilized factors through advanced modeling techniques holds considerable promise for developing more accurate and clinically relevant risk prediction tools tailored to the aging population( 8 ). The China Health and Retirement Longitudinal Study (CHARLS) provides a unique and valuable resource to address this research gap. As a nationally representative, longitudinal cohort of middle-aged and older adults in China, CHARLS offers comprehensive data encompassing demographics, health behaviors, clinical assessments, biomarkers, functional measures (e.g., ADL, grip strength), and psychosocial indicators (e.g., the Center for Epidemiologic Studies Depression Scale [CES-D] score). Despite the richness of this dataset, few studies have developed and validated MACE risk prediction models specifically for older Chinese adults that integrate both traditional and non-traditional risk factors. Moreover, contemporary machine learning approaches combined with rigorous evaluation metrics—extending beyond traditional discrimination analysis—remain underutilized. Therefore, the primary objective of this study was to develop and validate robust ML models for predicting the approximate 7-year risk of MACE among Chinese adults aged 60 years and older, using data from the CHARLS cohort. The secondary objectives were to: ( 1 ) identify key traditional and non-traditional predictors most strongly associated with MACE risk using interpretable ML methods; ( 2 ) comprehensively evaluate model performance based on discrimination (area under the receiver operating characteristic curve [AUC]), calibration (Brier score, calibration plots), and clinical utility (decision curve analysis); ( 3 ) compare the developed models with a standard benchmark, the FRS; ( 4 ) explore potential non-linear relationships between continuous predictors and MACE risk; and ( 5 ) construct a practical nomogram based on the best-performing interpretable model to facilitate individualized risk estimation. 2. Methods 2.1 Study Design, Data Source, and Cohort Selection This prospective cohort study utilized data from the CHARLS dataset, a nationally representative longitudinal survey targeting individuals aged 45 years and older in China. The present analysis was based on the 2011 baseline wave (Wave 1), which included 17,708 participants. Participants aged 60 years or older at baseline were eligible for inclusion, resulting in 7,560 individuals. Exclusion criteria were subsequently applied. Participants with a self-reported physician-diagnosed history of myocardial infarction or stroke at or prior to the baseline interview were excluded (n = 1,447). This yielded 6,113 participants aged ≥ 60 years and free of prevalent MACE for follow-up. Incident MACE events were identified using data from subsequent CHARLS waves in 2013 (Wave 2), 2015 (Wave 3), and 2018 (Wave 4). Participants were further excluded if their incident MACE status could not be determined during follow-up (e.g., due to loss to follow-up or missing outcome data; n = 1,533). The final analytical cohort thus comprised 4,580 participants who met the inclusion criteria, were free of MACE at baseline, and had complete outcome data between 2011 and 2018.CHARLS collects comprehensive information, including demographics, socioeconomic status, health behaviors and conditions, biomarkers, functional measures (e.g., ADL, grip strength), and psychosocial indicators (e.g., CES-D scores) through face-to-face interviews and physical examinations. The CHARLS study was approved by the Institutional Review Board of Peking University (IRB00001052-11015), and all participants provided written informed consent. 2.2 Outcome Ascertainment The primary outcome of this study was the first occurrence of an incident MACE. MACE was defined as a composite of self-reported, physician-diagnosed myocardial infarction or stroke during the follow-up period. Incident MACE events were identified based on participant self-reports, in accordance with standard CHARLS protocols. Although self-reports may introduce recall bias, previous validation studies( 9 ) have demonstrated reasonable reliability of self-reported cardiovascular events in CHARLS. The follow-up period spanned approximately seven years from the baseline interview. As established during cohort selection, participants were excluded from the final analytical cohort if their incident MACE status could not be definitively determined by the Wave 4 assessment (e.g., due to loss to follow-up without a prior reported event or missing outcome data). 2.3 Predictor Variables Candidate predictor variables were assessed at baseline during the 2011 wave of CHARLS. Variables were categorized into six key domains: ( 1 ) Demographics : age (continuous, years); sex (female/male); education level (categorized as less than primary, primary school, middle school, or high school or higher); marital status (married/other); and residence (urban/rural); ( 2 ) Health Behaviors : current smoking status (yes/no); current alcohol consumption (yes/no); and physical activity level (estimated metabolic equivalent of task [MET] minutes per week, continuous); ( 3 ) Anthropometric Measures : body mass index (BMI; kg/m², continuous) and waist circumference (cm, continuous); ( 4 ) Clinical Measures and Medical History : measured systolic and diastolic blood pressure (SBP/DBP; mmHg, continuous); self-reported physician-diagnosed hypertension (yes/no); and diabetes (yes/no); ( 5 ) Biomarkers : total cholesterol, high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol, triglycerides (all mg/dL); C-reactive protein (CRP; mg/L); and creatinine (mg/dL); and ( 6 ) Psychosocial and Geriatric Factors : depressive symptoms assessed by the CES-D-10; continuous); functional status assessed by the number of limitations in six Activities of Daily Living (ADL limitations; continuous); and maximum grip strength (kg, continuous). 2.4 Missing Data Handling Missing data were observed across several predictor variables, particularly for biomarkers, physical activity, anthropometric measures, and blood pressure. To address missingness, multiple imputation by chained equations (MICE) was performed using the mice package (version 3.17.0) in R (version 4.4.1). This approach assumes that the data were missing at random (MAR). Five imputed datasets (m = 5) were generated. The imputation models included all candidate predictor variables and the MACE outcome to preserve potential associations during the imputation process. Subsequent statistical analyses were conducted independently on each of the five imputed datasets, and the results were pooled using Rubin’s rules where applicable. 2.5 Statistical Analysis and Predictive Modeling All statistical analyses were conducted using R (version 4.4.1). A two-sided P-value < 0.05 was considered statistically significant. Prior to model development, each of the five imputed datasets was randomly partitioned into a training set (70%) and a test set (30%), stratified by MACE outcomes to maintain balanced class distributions. A fixed random seed was set to ensure reproducibility. Partitioning was applied consistently across all imputations, ensuring that the same individuals were assigned to the training and test sets within each imputed dataset. Three predictive modeling approaches were trained on each imputed training set: logistic regression (LR), random forest (RF), and extreme gradient boosting (XGBoost)( 10 , 11 ). Model training was performed using the caret package (version 7.0.1). For the RF and XGBoost models, hyperparameters were tuned using a basic grid search with a tuning length of 3 due to computational constraints. Future studies may explore more extensive hyperparameter optimization techniques, such as Bayesian optimization or randomized grid search, to further improve model performance. The models were optimized to maximize the area under the receiver operating characteristic curve (AUC). Model performance was primarily evaluated on the independent test sets, with results subsequently pooled across the five imputations. Discrimination was assessed using the AUC, with the average AUC and standard deviation reported across imputations. Calibration, reflecting the agreement between predicted probabilities and observed event rates, was evaluated both visually and quantitatively: pooled calibration plots were generated by grouping predictions into deciles, and the pooled Brier score was calculated (lower values indicating better overall accuracy)( 12 ). Clinical utility was assessed using decision curve analysis (DCA), by plotting the pooled average net benefit against varying risk thresholds( 13 ). The performance of the developed LR, RF, and XGBoost models was compared internally and benchmarked against the Framingham Risk Score for cardiovascular disease (FRS-CVD), calculated based on baseline variables from the test sets, while acknowledging limitations arising from differences in outcome definitions and follow-up durations( 2 , 4 , 14 ). Model interpretation focused primarily on the LR model, which demonstrated the best performance based on the average AUC. Pooled coefficients, odds ratios (ORs), 95% confidence intervals (CIs), and P-values were obtained using the mice::pool() function, applying Rubin’s rules. Predictor importance in the LR model was ranked based on the absolute magnitude of the pooled Z-statistic. Significant ORs were visualized using a forest plot. To gain additional insights potentially missed by the linear model, the XGBoost model—the best-performing non-linear model—was interpreted using SHapley Additive exPlanations (SHAP) values. SHAP values were calculated via the xgboost package (version 1.7.9.1) and averaged across imputations. Pooled SHAP summary plots and dependence plots were generated using the shapviz package (version 0.9.7) to identify key features and visualize their effects, a methodology previously applied successfully in cardiovascular risk prediction studies( 15 ). To explore potential non-linear associations between key continuous predictors (age, waist circumference, SBP, CES-D score, and maximum grip strength) and MACE risk, generalized additive models (GAMs) with penalized smoothing splines (specifically s (..., k = 5)) were fitted to the first imputed dataset using the mgcv package (version 1.9.1), adjusting for other significant covariates identified in the LR model. Smooth function plots were generated using the gratia package (version 0.10.0). The effective degrees of freedom (EDF) and P-values from the GAM summaries were examined to assess the presence and significance of non-linearity. Two sensitivity analyses were conducted to assess the robustness of the primary findings: (A) a complete-case analysis, excluding participants with any missing predictor data prior to modeling; and (B) an exclusion analysis, removing the predictor with the highest proportion of missing values (physical activity, measured as MET-minutes/week) prior to imputation and modeling. Model performance, evaluated by average AUC, from these scenarios was compared to that of the main analysis. Finally, a nomogram was developed based on the final pooled LR model coefficients (using key predictors identified by Z-statistic magnitude and refitted on the first imputed dataset) to provide a practical tool for individual risk estimation. The nomogram was constructed using the rms package (version 6.9.0). 2.6 Software Used All statistical analyses were conducted using R (version 4.4.1). Key packages utilized included mice (version 3.17.0) for multiple imputation, caret (version 7.0.1) for model training and hyperparameter tuning, and pROC (version 1.18.5) for receiver operating characteristic (ROC) analysis. The rms package (version 6.9.0) was used for nomogram development, while mgcv (version 1.9.1) and gratia (version 0.10.0) were employed for generalized additive model (GAM) analysis. For XGBoost modeling and interpretation using SHapley Additive exPlanations (SHAP), the xgboost (version 1.7.9.1) and shapviz (version 0.9.7) packages were applied. Decision curve analysis was performed using the dcurves package (version 0.5.0). Data manipulation was conducted using dplyr (version 1.1.4) and tidyr (version 1.3.1). Visualization was achieved with ggplot2 (version 3.5.1) and patchwork (version 1.3.0). Descriptive tables were generated with gtsummary (version 2.2.0), and tabular results were exported using writexl (version 1.5.4). 3. Results 3.1 Cohort Characteristics The final analytical cohort comprised 4,580 participants aged 60 years or older from the 2011 baseline wave of CHARLS, who were free of MACE at baseline and followed until the 2018 wave or the occurrence of an incident MACE. Baseline characteristics, stratified by incident MACE status over the approximately 7-year follow-up period, are summarized in Table 1 . The mean age of the cohort was 67.1 years (SD = 6.2), and 41.1% of participants were male. During follow-up, 1,315 individuals (28.7%) experienced an incident MACE. Compared to those who remained MACE-free, individuals who developed MACE were significantly older, had a higher prevalence of hypertension and diabetes, exhibited larger waist circumference, and reported higher CES-D scores and more limitations in ADLs at baseline (all P < 0.05; see Table 1 ). Table 1 Baseline Characteristics of Study Participants (n = 4580), Stratified by Incident MACE Status. Characteristic Overall (N = 4580) No MACE (N = 3265) MACE (N = 1315) P-value Age (years) 67.1 (6.2) 66.4 (5.8) 68.9 (6.7) < 0.001 Sex 0.13 Female 2042 (45%) 1479 (45%) 563 (43%) Male 1884 (41%) 1316 (40%) 568 (43%) NA 654 (14%) 470 (14%) 184 (14%) Education < 0.001 Less than primary 2676 (58%) 1833 (56%) 843 (64%) Primary school 1108 (24%) 826 (25%) 282 (21%) Middle school 506 (11%) 379 (12%) 127 (10%) High school+ 279 (6.1%) 219 (6.7%) 60 (4.6%) NA 11 (< 1%) 8 (< 1%) 3 (< 1%) Marital Status 0.11 Other 878 (19%) 604 (18%) 274 (21%) Married 3697 (81%) 2657 (81%) 1040 (79%) NA 5 (< 1%) 4 (< 1%) 1 (< 1%) Residence 0.002 Urban 1582 (35%) 1169 (36%) 413 (31%) Rural 2998 (65%) 2096 (64%) 902 (69%) Current Smoking 0.06 No 3082 (67%) 2231 (68%) 851 (65%) Yes 1381 (30%) 965 (30%) 416 (32%) NA 117 (2.6%) 69 (2.1%) 48 (3.7%) Current Drinking 0.001 No 3090 (67%) 2156 (66%) 934 (71%) Yes 1458 (32%) 1087 (33%) 371 (28%) NA 32 (< 1%) 22 (< 1%) 10 (< 1%) Physical Activity (MET-min/week) 7269.1 (6432.5) 7386.5 (6452.8) 6976.1 (6369.4) NA BMI (kg/m^2) 23.6 (18.8) 23.5 (16.4) 23.7 (23.8) NA Waist Circumference (cm) 83.9 (10.6) 83.3 (10.4) 85.2 (10.9) < 0.001 Systolic BP (mmHg) 133.1 (21.5) 131.9 (20.8) 136.1 (22.8) < 0.001 Diastolic BP (mmHg) 74.3 (11.5) 74.0 (11.3) 75.0 (11.9) 0.009 History of Hypertension < 0.001 No 3266 (71%) 2452 (75%) 814 (62%) Yes 1262 (28%) 777 (24%) 485 (37%) NA 52 (1.1%) 36 (1.1%) 16 (1.2%) History of Diabetes < 0.001 No 4262 (93%) 3054 (94%) 1208 (92%) Yes 248 (5.4%) 176 (5.4%) 72 (5.5%) NA 70 (1.5%) 35 (1.1%) 35 (2.7%) Total Cholesterol (mg/dL) 194.9 (35.2) 194.5 (34.8) 195.9 (36.2) 0.14 HDL Cholesterol (mg/dL) 52.3 (14.0) 52.6 (14.0) 51.4 (13.9) 0.007 LDL Cholesterol (mg/dL) 118.1 (29.9) 117.6 (29.5) 119.3 (30.8) 0.1 Triglycerides (mg/dL) 128.1 (89.4) 125.5 (77.8) 134.4 (111.6) 0.015 CRP (mg/L) 2.9 (5.8) 2.8 (5.8) 3.0 (5.8) 0.47 Creatinine (mg/dL) 0.8 (0.2) 0.8 (0.2) 0.8 (0.2) 0.65 CES-D Score 8.7 (6.3) 8.0 (6.0) 10.3 (6.7) < 0.001 ADL Limitations (Count) 0.4 (1.0) 0.3 (0.9) 0.8 (1.3) < 0.001 Max Grip Strength (kg) 29.8 (11.4) 30.3 (11.6) 28.6 (10.9) < 0.001 Data presented as Mean (SD) for continuous variables or n (%) for categorical variables, based on the first imputed dataset. P-values compare MACE vs No MACE groups using appropriate statistical tests (e.g., t-test/Wilcoxon, Chi-squared/Fisher's). Abbreviations: SD, standard deviation; MACE, Major Adverse Cardiovascular Event; BMI, Body Mass Index; BP, Blood Pressure; HDL, High-Density Lipoprotein; LDL, Low-Density Lipoprotein; CRP, C-Reactive Protein; CES-D, Center for Epidemiologic Studies Depression Scale; ADL, Activities of Daily Living; MET, Metabolic Equivalent of Task. 3.2 Model Performance Comparison The predictive performance of the LR, RF, and XGBoost models was compared against each other and the benchmark Framingham Risk Score for cardiovascular disease (FRS-CVD), using results pooled across the five imputed test sets (Table 2 ). In terms of discrimination, the LR model achieved the highest average area under the receiver operating characteristic curve (AUC) (mean = 0.649, SD = 0.003), closely followed by XGBoost (mean = 0.645, SD = 0.009). Both models significantly outperformed the RF model (mean = 0.632, SD = 0.006) and especially the FRS-CVD benchmark (mean = 0.504, SD = 0.007). Representative ROC curves from one imputed test set visually illustrate the relative performance ranking (Fig. 2). Table 2 Predictive Performance Comparison of Models for ~ 7-Year Incident MACE. Model AUC (SD) Brier Score Logistic Regression 0.649 (nan) 0.193 Random Forest 0.632 (nan) 0.197 XGBoost 0.645 (nan) 0.194 Framingham (CVD) 0.504 (nan) NA Performance metrics averaged over 5 imputed test sets. AUC indicates Area Under the Receiver Operating Characteristic Curve; SD, Standard Deviation; Brier Score measures overall accuracy (lower is better); NA, Not Available. Models: LR, Logistic Regression; RF, Random Forest; XGBoost, Extreme Gradient Boosting; FRS-CVD, Framingham Risk Score for 10-year general Cardiovascular Disease (used as benchmark, limitations apply). Calibration assessment (Fig. 3) showed that both the LR and XGBoost models were reasonably well-calibrated, with predicted probabilities closely aligning with observed event rates. The RF model slightly underestimated risk at higher predicted probability levels, while the FRS-CVD score exhibited poor calibration, deviating substantially from the line of identity. Consistent with the AUC findings, the pooled Brier scores were lowest for LR (mean = 0.184) and XGBoost (mean = 0.184), indicating superior overall accuracy compared to RF (mean = 0.185) and FRS-CVD (mean = 0.215) (Table 2 ). Furthermore, decision curve analysis (DCA) demonstrated that the LR and XGBoost models provided greater net clinical benefit across a wide range of probability thresholds, outperforming both the RF model and the FRS-CVD score, as well as the default strategies of treating all or none (Fig. 4). 3.3 Key Predictors in the Logistic Regression Model The pooled multivariable logistic regression analysis identified several factors independently associated with the risk of incident MACE over the approximately 7-year follow-up period (Table 3 ; Fig. 5). Predictor importance, ranked by the absolute magnitude of the pooled Z-statistics (Appendix Table A1), indicated that the most influential variables included history of hypertension, CES-D score, education level (high school or above vs. less than primary), number of ADL limitations, waist circumference, age, SBP, and history of diabetes. After mutual adjustment, several traditional risk factors remained significantly associated with increased MACE risk: history of hypertension (OR = 1.85, 95% CI: 1.58–2.16), history of diabetes (OR = 1.34, 95% CI: 1.01–1.78), per year increase in age (OR = 1.02, 95% CI: 1.00–1.03), and per mmHg increase in SBP (OR = 1.006, 95% CI: 1.001–1.011). In addition, non-traditional and socioeconomic predictors also demonstrated significant associations: high school or higher education level (vs. less than primary school; OR = 1.60, 95% CI: 1.20–2.14), waist circumference (per cm; OR = 1.01, 95% CI: 1.00–1.02), CES-D score (per point; OR = 1.03, 95% CI: 1.02–1.04), and number of ADL limitations (per unit; OR = 1.11, 95% CI: 1.04–1.19). Notably, waist circumference, depressive symptoms (CES-D score), and functional limitations (ADL) remained significant predictors even after adjusting for conventional cardiovascular risk factors (Table 3 ). Table 3 Pooled Multivariable Logistic Regression Analysis of Predictors for ~ 7-Year Incident MACE. Predictor Odds Ratio (OR) 95% CI P-value Age (years) 1.02 1.00-1.03 0.015 Sex: Male (vs Female) 0.86 0.67–1.11 0.24 Education = Primary (vs < Primary) 1.05 0.88–1.24 0.61 Education = Middle (vs < Primary) 1.26 1.00-1.59 0.05 Education = High+ (vs < Primary) 1.6 1.20–2.14 0.002 Marital Status = Married (vs Other) 1.08 0.91–1.29 0.38 Residence = Rural (vs Urban) 0.89 0.76–1.04 0.13 Current Smoking = Yes (vs No) 0.93 0.78–1.12 0.45 Current Drinking = Yes (vs No) 0.86 0.73–1.02 0.09 Physical Activity (MET-min/week) 1 1.00–1.00 0.71 BMI (kg/m^2) 1 0.99-1.00 0.68 Waist Circumference (cm) 1.01 1.00-1.02 0.011 Systolic BP (mmHg) 1.01 1.00-1.01 0.03 Diastolic BP (mmHg) 1 0.99–1.01 0.51 History of Hypertension = Yes (vs No) 1.85 1.58–2.16 < 0.001 History of Diabetes = Yes (vs No) 1.34 1.01–1.78 0.04 Total Cholesterol (mg/dL) 1 0.99–1.01 0.65 HDL Cholesterol (mg/dL) 1 1.00-1.01 0.31 LDL Cholesterol (mg/dL) 1 0.99–1.01 0.74 Triglycerides (mg/dL) 1 1.00–1.00 0.89 CRP (mg/L) 1.01 1.00-1.02 0.13 Creatinine (mg/dL) 0.76 0.50–1.15 0.19 CES-D Score 1.03 1.02–1.04 < 0.001 ADL Limitations (Count) 1.11 1.04–1.19 0.002 Max Grip Strength (kg) 1.01 1.00-1.02 0.3 Results pooled across 5 imputed datasets using Rubin's Rules. OR indicates Odds Ratio; CI, Confidence Interval; MACE, Major Adverse Cardiovascular Event. Reference groups for categorical predictors: Sex = Female; Education = < Primary; Marital Status = Other; Residence = Urban; Smoking = No; Drinking = No; Hypertension = No; Diabetes = No. 3.4 Exploration of Non-Linear Relationships Exploratory analysis using generalized additive models (GAMs) revealed significant non-linear associations between the log-odds of MACE and both age (effective degrees of freedom [EDF] = 2.27, P = 0.003) and waist circumference (EDF = 2.28, P < 0.001) (Table 4 ; Fig. 7). In contrast, the relationships of systolic blood pressure (SBP) (EDF = 1.0, P = 0.04) and CES-D score (EDF = 1.0, P < 0.001) with MACE risk were effectively linear but remained statistically significant. Maximum grip strength was not significantly associated with MACE risk in the adjusted GAM model (EDF = 1.0, P = 0.53). Table 4 Approximate Significance of Smooth Terms from Exploratory Generalized Additive Model (GAM). Smooth Term EDF Ref.df Chi.sq P-value Non-Linear Smooth(Age) 2.27 2.7 14.63 0.003 TRUE Smooth(Waist Circum.) 2.28 2.8 25.06 < 0.001 TRUE Smooth(Systolic BP) 1 1 4.3 0.04 FALSE Smooth(CES-D Score) 1 1 33.37 < 0.001 FALSE Smooth(Max Grip Strength) 1 1 0.4 0.53 FALSE Results from GAM fitted on the first imputed dataset. EDF indicates Effective Degrees of Freedom; Ref.df, Reference Degrees of Freedom; Chi.sq, Chi-squared statistic. 'Non-Linear' column indicates if EDF > 1.1, suggesting a non-linear relationship. P-value assesses the significance of the smooth term's contribution. 3.5 Interpretation Insights from the XGBoost Model Interpretation of the XGBoost model using pooled SHapley Additive exPlanations (SHAP) values revealed feature importance rankings that largely aligned with findings from the primary logistic regression model (Figure S2 ). The top-ranked predictors based on mean absolute SHAP values included history of hypertension, CES-D score, waist circumference, age, systolic blood pressure (SBP), number of ADL limitations, and higher education level (high school or above vs. less than primary). This consistency in predictor importance across distinct modeling approaches reinforces the robustness of these key risk factors. Notably, the high ranking of non-traditional predictors—waist circumference, depressive symptoms (CES-D score), and functional limitations (ADL limitations)—in the XGBoost model further supports their relevance in predicting MACE risk in this older Chinese cohort. 3.6 Sensitivity Analysis Results The robustness of model performance was evaluated through two sensitivity analyses (Figure S1 ). First, exclusion of the variable totmet (which had > 60% missingness) prior to multiple imputation resulted in minimal changes in the average AUCs for LR (mean = 0.650), RF (mean = 0.624), and XGBoost (mean = 0.643) compared to the main analysis, suggesting that the primary findings were robust to the exclusion of this highly incomplete predictor. In contrast, a complete-case analysis—which reduced the analytical sample to 977 participants (21.3% of the original cohort)—resulted in substantially lower AUC values for all models (LR: 0.543; RF: 0.558; XGBoost: 0.553), potentially reflecting the value of multiple imputation in preserving statistical power and minimizing selection bias. Importantly, the relative performance ranking among models (LR ≈ XGBoost > RF) remained generally consistent across both sensitivity scenarios. 3.7 Nomogram for Risk Prediction Based on the final multivariable logistic regression model incorporating key predictors—age, sex, education level, waist circumference, SBP, history of hypertension, history of diabetes, CES-D score, and number of ADL limitations—a nomogram was developed to facilitate individualized clinical risk estimation (Fig. 6). This graphical tool enables clinicians to estimate an individual's approximate 7-year risk of MACE by summing points corresponding to their specific risk factor profile. Each predictor is assigned a point value proportional to its relative contribution to MACE risk, and the total score maps to a predicted probability of incident MACE. The nomogram serves as a practical, interpretable aid for risk stratification and shared decision-making in older Chinese adults. 4. Discussion This study leveraged data from the China Health and Retirement Longitudinal Study (CHARLS)—a rich, nationally representative longitudinal cohort—to develop and evaluate machine learning models for predicting incident MACE over a 7-year follow-up among older Chinese adults. The LR model incorporating both traditional and non-traditional risk factors demonstrated the best overall performance, achieving moderate discrimination (mean AUC = 0.649) and good calibration. Notably, it significantly outperformed the standard Framingham Risk Score for cardiovascular disease (FRS-CVD), which yielded poor discrimination (mean AUC = 0.504) in this population Importantly, several non-traditional factors—waist circumference, depressive symptoms (CES-D score), and functional limitations (ADL limitations)—were identified as independent predictors of MACE, alongside established risk factors including age, hypertension, diabetes, and systolic blood pressure. Exploratory analyses using GAMs further revealed statistically significant non-linear relationships between MACE risk and both age and waist circumference. A major contribution of this study lies in elucidating the predictive importance of non-traditional yet readily accessible factors for MACE risk among older Chinese adults. The significant association of waist circumference, as opposed to BMI, highlights the role of central obesity—distinct from general adiposity—as a key risk factor (Table 3 ; Table 4 ; Fig. 5; Figure S2 ; Appendix Table A1). Central obesity is known to be closely linked to metabolic syndrome, systemic inflammation, and insulin resistance, which provide plausible biological pathways for its association with cardiovascular events( 16 , 17 )、. The independent predictive value of the CES-D score (Table 3 ; Table 4 ; Fig. 5; Figure S2 ; Appendix Table A1) reinforces the critical link between mental and cardiovascular health in later life. Depression may contribute to increased MACE risk through behavioral mechanisms (e.g., reduced treatment adherence, physical inactivity, and smoking) and physiological pathways (e.g., hypothalamic–pituitary–adrenal axis dysregulation, systemic inflammation, and endothelial dysfunction)( 18 – 20 ). Furthermore, the observed association between limitations in ADLs and MACE risk (Table 3 ; Fig. 5; Figure S2 ) underscores the prognostic relevance of functional status in aging populations. ADL impairments may reflect underlying frailty, unrecognized disease burden, or diminished physiological reserve, all of which increase vulnerability to adverse cardiovascular outcomes( 21 ). These findings support the adoption of a more holistic approach to cardiovascular risk assessment in older adults—one that incorporates measures of central obesity, mental health, and functional status alongside conventional risk factors. Notably, a higher education level (high school or above versus less than primary) was associated with increased MACE risk in the pooled logistic regression model (Table 3 ), a finding that may seem counterintuitive in light of previous literature, where higher education is often viewed as protective( 21 ). This discrepancy may reflect complex interactions with unmeasured socioeconomic determinants, differences in health behaviors, or survival bias within this aging cohort. It also raises important questions about how educational attainment categories function as proxies for health advantage in different demographic and cultural contexts, warranting further investigation. In addition, grip strength—while included in the generalized additive model—did not emerge as a significant linear or non-linear predictor in the adjusted analysis (Table 4 ), suggesting limited utility of this measure in predicting MACE risk in this population. Our exploratory GAM analysis (Table 4 ; Fig. 7) revealed significant non-linear relationships between the log-odds of MACE and both age (EDF = 2.3) and waist circumference (EDF = 2.3). These results suggest that the risk of MACE does not increase linearly with age or waist circumference—a nuance that may be overlooked by conventional linear models such as the FRS or basic logistic regression models without non-linear terms (e.g., spline functions)( 22 – 24 ). Although the logistic regression model slightly outperformed XGBoost in terms of average AUC, the confirmation of non-linearity supports the further exploration of machine learning models capable of capturing complex, non-linear relationships( 25 ), or the refinement of traditional regression models through the incorporation of flexible approaches such as restricted cubic splines. In addition, the strong agreement in predictor importance rankings between the pooled logistic regression model (Appendix Table A1) and the SHAP analysis of the XGBoost model (Figure S2 ) reinforces the robustness of key predictors—hypertension, CES-D score, waist circumference, age, systolic blood pressure, and ADL limitations—consistent with prior studies investigating cardiovascular risk stratification using feature importance metrics( 26 ). This study has several notable strengths. First, it is based on a large, nationally representative longitudinal cohort (CHARLS), ensuring the generalizability of findings to the older adult population in China. Second, it uniquely integrates both traditional and non-traditional risk factors relevant to aging, such as central obesity, depressive symptoms, and functional status. Third, a comparative evaluation of multiple modeling approaches—LR, RF, and XGBoost—was conducted( 27 , 28 ), with performance assessed comprehensively using not only discrimination (AUC) but also calibration and decision curve analysis (DCA)( 29 ). Fourth, interpretable methods such as SHAP and GAMs were applied to enhance understanding of model outputs( 30 ). The use of SHAP values, in particular, represents an important methodological advancement over traditional feature importance metrics by offering consistent, individualized estimates of predictor contributions( 31 ). Finally, a nomogram was developed based on the best-performing interpretable model (LR) (Fig. 6), offering a practical and user-friendly tool for individualized clinical risk estimation and decision-making in older Chinese adults. Despite its strengths, this study has several limitations that should be considered when interpreting the findings. First, the reliance on self-reported MACE outcomes may introduce recall bias and outcome misclassification. Second, the handling of missing data—particularly prevalent for biomarker and physical activity variables—required multiple imputation under the unverifiable assumption of missing at random (MAR)( 32 ). Third, the approximate 7-year follow-up period differs from the conventional 10-year horizon used in established risk scores such as the FRS, which limits direct comparability. Additionally, the FRS itself was originally developed for a different population (predominantly white, middle-aged American adults) and employed a distinct composite outcome definition, further complicating comparisons. Fourth, the nomogram construction and GAM analyses were conducted using only the first imputed dataset due to computational constraints, and thus do not fully reflect the uncertainty inherent in multiple imputation procedures. Fifth, although a broad range of predictors was included, the possibility of residual confounding from unmeasured variables remains, as is common in observational studies. Sixth, hyperparameter tuning for the machine learning models was limited (tuneLength = 3), which may have constrained the full potential performance of models like XGBoost, particularly when evaluated across multiple imputed datasets. Finally, while the internal performance of the models was promising, external validation in independent cohorts of older Chinese adults is essential before considering clinical application—an important prerequisite for implementing machine learning-based risk prediction models in practice( 33 ). 5. Conclusion In conclusion, this study successfully developed and comprehensively evaluated risk prediction models for incident MACE using data from a large, nationally representative cohort of older Chinese adults (CHARLS). The best-performing model—a logistic regression incorporating traditional risk factors alongside waist circumference, depressive symptoms (CES-D score), and limitations in ADLs—demonstrated moderate discrimination (mean AUC = 0.649) and good calibration, significantly outperforming the standard FRS benchmark (mean AUC = 0.504)( 34 ). These findings underscore the independent predictive value of central obesity (as measured by waist circumference), mental health (CES-D score), and functional status (A DL limitations) in cardiovascular risk estimation, supporting a more holistic and aging-sensitive approach to risk stratification in clinical practice( 17 , 35 ). Public health strategies targeting this rapidly aging population should consider interventions aimed at reducing central adiposity, promoting mental well-being, and preserving functional capacity alongside conventional CVD prevention measures( 36 – 38 ). The developed nomogram (Fig. 6) provides a practical tool for individualized risk estimation and may facilitate shared decision-making in clinical settings; however, its clinical utility requires further confirmation through external validation in independent populations( 39 ). Future research priorities include validating the models in external and ethnically diverse cohorts, developing dynamic prediction frameworks using repeated longitudinal measurements, applying advanced machine learning techniques to investigate interactions and nonlinearities among predictors( 40 ), and evaluating the incremental predictive value of additional biomarkers and social determinants of health. Declarations Data Availability Statement The data that support the findings of this study are derived from the China Health and Retirement Longitudinal Study (CHARLS). Access to the data is restricted and was granted under a data use agreement with Peking University for the purposes of this study. Therefore, the datasets are not publicly available without prior approval. Researchers interested in accessing the CHARLS data may do so upon reasonable request and completion of a data use application through the official CHARLS website: http://charls.pku.edu.cn. The R code used to conduct the statistical analyses in this study is available from the corresponding author upon reasonable request. Conflict of Interest The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Author Contributions JL designed the study, performed the data analysis, developed the models, interpreted the results, and drafted the initial manuscript. ZS contributed to the study methodology, investigation process, and provided critical revisions to the manuscript. FZ evaluated the psychological indicators derived from the CHARLS dataset. QX, GH, GW and XL assisted with software implementation, data curation, formal analysis, visualization, and manuscript review. YH, ZZ and XH contributed to the study conceptualization, supervised various aspects of the work, and critically reviewed the manuscript. ZZ* conceived and designed the overall study, provided supervision and resources, administered the project, potentially secured funding, and critically reviewed and edited the final manuscript. All authors read and approved the final version of the manuscript. Funding This research was supported by the National Natural Science Foundation (Grant No. 81660730). Acknowledgments The authors thank all the members of the CHALRS for their contributions and the participants who contributed their data. Ethics Statement The CHARLS study was approved by the Institutional Review Board of Peking University (IRB00001052-11015), and all participants provided written informed consent prior to participation. The present study was conducted in accordance with the ethical standards of the Declaration of Helsinki and its later amendments. Ethics approval committee: Institutional Review Board of Peking University. Consent to participate: All participants provided written informed consent. Clinical trial number: not applicable. References Mensah GA, Fuster V, Murray CJL, Roth GA. Global Burden of Cardiovascular Diseases and Risks, 1990-2022. Journal of the American College of Cardiology. 2023;82(25):2350-473. D'Agostino RB, Sr., Grundy S, Sullivan LM, Wilson P. Validation of the Framingham coronary heart disease prediction scores: results of a multiple ethnic groups investigation. Jama. 2001;286(2):180-7. Wang L, Song L, Li D, Zhou Z, Chen S, Yang Y, et al. Ideal Cardiovascular Health Metric and Its Change With Lifetime Risk of Cardiovascular Diseases: A Prospective Cohort Study. Journal of the American Heart Association. 2021;10(22):e022502. Liu J, Hong Y, D'Agostino RB, Sr., Wu Z, Wang W, Sun J, et al. Predictive value for the Chinese population of the Framingham CHD risk assessment tool compared with the Chinese Multi-Provincial Cohort Study. Jama. 2004;291(21):2591-9. Barzi F, Patel A, Gu D, Sritara P, Lam TH, Rodgers A, et al. Cardiovascular risk prediction tools for populations in Asia. Journal of epidemiology and community health. 2007;61(2):115-21. Chen JH, Asch SM. Machine Learning and Prediction in Medicine - Beyond the Peak of Inflated Expectations. The New England journal of medicine. 2017;376(26):2507-9. Ambale-Venkatesh B, Yang X, Wu CO, Liu K, Hundley WG, McClelland R, et al. Cardiovascular Event Prediction by Machine Learning: The Multi-Ethnic Study of Atherosclerosis. Circulation research. 2017;121(9):1092-101. Wu Q, Kling JM. Depression and the Risk of Myocardial Infarction and Coronary Death: A Meta-Analysis of Prospective Cohort Studies. Medicine. 2016;95(6):e2815. Li W, Shen C, Kong W, Zhou X, Fan H, Zhang Y, et al. Association between the triglyceride glucose-body mass index and future cardiovascular disease risk in a population with Cardiovascular-Kidney-Metabolic syndrome stage 0-3: a nationwide prospective cohort study. Cardiovascular diabetology. 2024;23(1):292. Kakadiaris IA, Vrigkas M, Yen AA, Kuznetsova T, Budoff M, Naghavi M. Machine Learning Outperforms ACC / AHA CVD Risk Calculator in MESA. Journal of the American Heart Association. 2018;7(22):e009476. Narain R, Saxena S, Goyal AK. Cardiovascular risk prediction: a comparative study of Framingham and quantum neural network based approach. Patient preference and adherence. 2016;10:1259-70. Xi Y, Wang H, Sun N. Machine learning outperforms traditional logistic regression and offers new possibilities for cardiovascular risk prediction: A study involving 143,043 Chinese patients with hypertension. Frontiers in cardiovascular medicine. 2022;9:1025705. Sadatsafavi M, Adibi A, Puhan M, Gershon A, Aaron SD, Sin DD. Moving beyond AUC: decision curve analysis for quantifying net benefit of risk prediction models. The European respiratory journal. 2021;58(5). Wang K, Tian J, Zheng C, Yang H, Ren J, Liu Y, et al. Interpretable prediction of 3-year all-cause mortality in patients with heart failure caused by coronary heart disease based on machine learning and SHAP. Computers in biology and medicine. 2021;137:104813. Han Y, Wang S. Disability risk prediction model based on machine learning among Chinese healthy older adults: results from the China Health and Retirement Longitudinal Study. Frontiers in public health. 2023;11:1271595. Cong X, Liu S, Wang W, Ma J, Li J. Combined consideration of body mass index and waist circumference identifies obesity patterns associated with risk of stroke in a Chinese prospective cohort study. BMC public health. 2022;22(1):347. Powell-Wiley TM, Poirier P, Burke LE, Després JP, Gordon-Larsen P, Lavie CJ, et al. Obesity and Cardiovascular Disease: A Scientific Statement From the American Heart Association. Circulation. 2021;143(21):e984-e1010. Gan Y, Gong Y, Tong X, Sun H, Cong Y, Dong X, et al. Depression and the risk of coronary heart disease: a meta-analysis of prospective cohort studies. BMC psychiatry. 2014;14:371. Barth J, Schumacher M, Herrmann-Lingen C. Depression as a risk factor for mortality in patients with coronary heart disease: a meta-analysis. Psychosomatic medicine. 2004;66(6):802-13. Lichtman JH, Froelicher ES, Blumenthal JA, Carney RM, Doering LV, Frasure-Smith N, et al. Depression as a risk factor for poor prognosis among patients with acute coronary syndrome: systematic review and recommendations: a scientific statement from the American Heart Association. Circulation. 2014;129(12):1350-69. Nicholson A, Kuper H, Hemingway H. Depression as an aetiologic and prognostic factor in coronary heart disease: a meta-analysis of 6362 events among 146 538 participants in 54 observational studies. European heart journal. 2006;27(23):2763-74. Wang L, Lee Y, Wu Y, Zhang X, Jin C, Huang Z, et al. A prospective study of waist circumference trajectories and incident cardiovascular disease in China: the Kailuan Cohort Study. The American journal of clinical nutrition. 2021;113(2):338-47. Liu R, Dang S, Zhao Y, Yan H, Han Y, Mi B. Long-term waist circumference trajectories and body mass index with all-cause mortality in older Chinese adults: a prospective nationwide cohort study. Archives of public health = Archives belges de sante publique. 2022;80(1):94. Fauchier G, Bisson A, Bodin A, Herbert J, Semaan C, Angoulvant D, et al. Metabolically healthy obesity and cardiovascular events: A nationwide cohort study. Diabetes, obesity & metabolism. 2021;23(11):2492-501. Krittanawong C, Johnson KW, Rosenson RS, Wang Z, Aydar M, Baber U, et al. Deep learning for cardiovascular medicine: a practical primer. European heart journal. 2019;40(25):2058-73. Itchhaporia D. Artificial intelligence in cardiology. Trends in cardiovascular medicine. 2022;32(1):34-41. Motwani M, Dey D, Berman DS, Germano G, Achenbach S, Al-Mallah MH, et al. Machine learning for prediction of all-cause mortality in patients with suspected coronary artery disease: a 5-year multicentre prospective registry analysis. European heart journal. 2017;38(7):500-7. Lo-Ciganic WH, Huang JL, Zhang HH, Weiss JC, Wu Y, Kwoh CK, et al. Evaluation of Machine-Learning Algorithms for Predicting Opioid Overdose Risk Among Medicare Beneficiaries With Opioid Prescriptions. JAMA network open. 2019;2(3):e190968. Kerr KF, Brown MD, Zhu K, Janes H. Assessing the Clinical Impact of Risk Prediction Models With Decision Curves: Guidance for Correct Interpretation and Appropriate Use. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2016;34(21):2534-40. Rudin C. Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead. Nature machine intelligence. 2019;1(5):206-15. Elshawi R, Al-Mallah MH, Sakr S. On the interpretability of machine learning-based model for predicting hypertension. BMC medical informatics and decision making. 2019;19(1):146. Sterne JA, White IR, Carlin JB, Spratt M, Royston P, Kenward MG, et al. Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ (Clinical research ed). 2009;338:b2393. Siontis GC, Tzoulaki I, Siontis KC, Ioannidis JP. Comparisons of established risk prediction models for cardiovascular disease: systematic review. BMJ (Clinical research ed). 2012;344:e3318. Cameron AJ, Magliano DJ, Söderberg S. A systematic review of the impact of including both waist and hip circumference in risk models for cardiovascular diseases, diabetes and mortality. Obesity reviews : an official journal of the International Association for the Study of Obesity. 2013;14(1):86-94. Chen Z, Iona A, Parish S, Chen Y, Guo Y, Bragg F, et al. Adiposity and risk of ischaemic and haemorrhagic stroke in 0·5 million Chinese men and women: a prospective cohort study. The Lancet Global health. 2018;6(6):e630-e40. Hu L, Huang X, You C, Li J, Hong K, Li P, et al. Prevalence of overweight, obesity, abdominal obesity and obesity-related risk factors in southern China. PloS one. 2017;12(9):e0183934. Després JP. Waist circumference as a vital sign in cardiology 20 years after its initial publication in the American Journal of Cardiology. The American journal of cardiology. 2014;114(2):320-3. Yusuf S, Joseph P, Rangarajan S, Islam S, Mente A, Hystad P, et al. Modifiable risk factors, cardiovascular disease, and mortality in 155 722 individuals from 21 high-income, middle-income, and low-income countries (PURE): a prospective cohort study. Lancet (London, England). 2020;395(10226):795-808. External validation of clinical prediction models using big datasets from e-health records or IPD meta-analysis: opportunities and challenges. BMJ (Clinical research ed). 2019;365:l4379. Goldstein BA, Navar AM, Carter RE. Moving beyond regression techniques in cardiovascular risk prediction: applying machine learning to address analytic challenges. European heart journal. 2017;38(23):1805-14. Additional Declarations No competing interests reported. Supplementary Files AppendixTables.docx figures1.tif figures2.tif figures3.tif Cite Share Download PDF Status: Under Review Version 1 posted Reviewers invited by journal 15 Jul, 2025 Editor invited by journal 20 Jun, 2025 Editor assigned by journal 20 Jun, 2025 Submission checks completed at journal 20 Jun, 2025 First submitted to journal 16 Jun, 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-6906133","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":485997929,"identity":"2e62d3c5-c252-4314-83fa-073f25ae06e5","order_by":0,"name":"Jingwei Li","email":"","orcid":"","institution":"Gansu University of Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Jingwei","middleName":"","lastName":"Li","suffix":""},{"id":485997930,"identity":"c69f2261-38dc-4ef7-be5e-f0e2e92cf8bc","order_by":1,"name":"Zhongyang Song","email":"","orcid":"","institution":"Affiliated Hospital of Gansu University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Zhongyang","middleName":"","lastName":"Song","suffix":""},{"id":485997931,"identity":"2ac00801-4bf0-4ea0-8d8e-2e6387091492","order_by":2,"name":"Fan Zou","email":"","orcid":"","institution":"Northwest Normal University","correspondingAuthor":false,"prefix":"","firstName":"Fan","middleName":"","lastName":"Zou","suffix":""},{"id":485997932,"identity":"f28f3f13-c0d8-4deb-baed-925b68344c4e","order_by":3,"name":"Yiming Hu","email":"","orcid":"","institution":"Gansu University of Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Yiming","middleName":"","lastName":"Hu","suffix":""},{"id":485997933,"identity":"1a713da5-f22a-4e00-bc42-43187648b023","order_by":4,"name":"Qian Xu","email":"","orcid":"","institution":"Gansu University of Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Qian","middleName":"","lastName":"Xu","suffix":""},{"id":485997934,"identity":"0e234784-5bd1-48e4-b334-4f1377ae8b94","order_by":5,"name":"Guoxiong Hao","email":"","orcid":"","institution":"Gansu University of Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Guoxiong","middleName":"","lastName":"Hao","suffix":""},{"id":485997935,"identity":"fe03268a-c1e0-48c2-bca4-b40b1a1f97a8","order_by":6,"name":"Xiali Liang","email":"","orcid":"","institution":"Gansu University of Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Xiali","middleName":"","lastName":"Liang","suffix":""},{"id":485997936,"identity":"814d0095-7c9f-4022-8897-1693a0481bbc","order_by":7,"name":"Zexiang Zhang","email":"","orcid":"","institution":"Gansu University of Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Zexiang","middleName":"","lastName":"Zhang","suffix":""},{"id":485997937,"identity":"75a7f23b-029f-40fb-a75a-b43f25cc3f41","order_by":8,"name":"Xixi Huang","email":"","orcid":"","institution":"Gansu University of Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Xixi","middleName":"","lastName":"Huang","suffix":""},{"id":485997938,"identity":"1b4642e2-0f7a-4786-9afc-f991dcbfafeb","order_by":9,"name":"Guanwei Wang","email":"","orcid":"","institution":"Gansu University of Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Guanwei","middleName":"","lastName":"Wang","suffix":""},{"id":485997939,"identity":"61743b77-2daa-436e-b6a2-05650debdbc4","order_by":10,"name":"Zhiming Zhang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA9klEQVRIiWNgGAWjYDACCTB5gIeBGcyw4eHnbyBNS5qM5IwDxGmBcQ/bGDQk4NchP7v52cMvf+7I6LbzHn7N8+c8jwHDAcYPH3Nwa2Gcc8zcWIbnGY/ZYb40a9622zzmzA3MkjO34dbCLJFgJi0hcRiohcfMmLfhNo9lwwE2Zl48Wtgk0r9JSxhAtfD8OcdjcCABvxYeiRwzyQ8JYC3Gj3nYDhDWIiGRUybNcABiC+PctmQeyRkHm/H6RX5G+jbJH38O25udP2P84c0fO3t+/uaDHz7i0QIOAh6ov6QgDMYG/OpBSn5AtX78QVDtKBgFo2AUjEQAAOO5TxSzORlHAAAAAElFTkSuQmCC","orcid":"","institution":"Gansu Provincial Hospital of Traditional Chinese Medicine","correspondingAuthor":true,"prefix":"","firstName":"Zhiming","middleName":"","lastName":"Zhang","suffix":""}],"badges":[],"createdAt":"2025-06-16 13:38:31","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6906133/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6906133/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":87319863,"identity":"c6762c1d-fd46-4658-a126-eba3dbd5dc35","added_by":"auto","created_at":"2025-07-22 16:23:21","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2110071,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-6906133/v1/d03758da66bc3301da2dcff1.png"},{"id":87318199,"identity":"2d4c845b-92aa-4004-8b59-f6510ad9464f","added_by":"auto","created_at":"2025-07-22 16:15:21","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":2103136,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-6906133/v1/20c0d792a9df3fad1f631e48.png"},{"id":87319865,"identity":"3a410838-a3a6-4143-a609-e6b1b1006433","added_by":"auto","created_at":"2025-07-22 16:23:21","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":2242372,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-6906133/v1/f1c248a950d786bc4cadad85.png"},{"id":87319869,"identity":"d816c35d-3e86-46fd-b58f-35ba464a8973","added_by":"auto","created_at":"2025-07-22 16:23:21","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1389215,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-6906133/v1/409948f2efd113e15b879fd7.png"},{"id":87318214,"identity":"1d38a34b-97c5-4f1d-a63d-6c76a00ffd38","added_by":"auto","created_at":"2025-07-22 16:15:21","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1428615,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-6906133/v1/b3728d7846e1a9acddc767ee.png"},{"id":87318209,"identity":"65ea9388-bc76-48de-94b1-40654e44e5fb","added_by":"auto","created_at":"2025-07-22 16:15:21","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":2243330,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-6906133/v1/53f47f80e26c502a0818f60f.png"},{"id":87318202,"identity":"e0a94257-9578-46f1-b297-fa3f67cc98dd","added_by":"auto","created_at":"2025-07-22 16:15:21","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":4493,"visible":true,"origin":"","legend":"\u003cp\u003eExploratory Generalized Additive Model (GAM).\u003c/p\u003e","description":"","filename":"fig.png","url":"https://assets-eu.researchsquare.com/files/rs-6906133/v1/dd9b90fa2516b4e2a0c494ed.png"},{"id":87322365,"identity":"da224a89-0b5e-4c5d-b517-c78875747073","added_by":"auto","created_at":"2025-07-22 16:47:31","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":14138478,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6906133/v1/c41b17d1-750e-420d-be6c-4bb37aea40cf.pdf"},{"id":87318201,"identity":"e78519b3-2a7c-46e4-b06e-d568cab09a33","added_by":"auto","created_at":"2025-07-22 16:15:21","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":21089,"visible":true,"origin":"","legend":"","description":"","filename":"AppendixTables.docx","url":"https://assets-eu.researchsquare.com/files/rs-6906133/v1/b235e4d56ccd7895b7637543.docx"},{"id":87319874,"identity":"982bb8cd-9c4a-4d12-93c0-b24464138857","added_by":"auto","created_at":"2025-07-22 16:23:21","extension":"tif","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":3544506,"visible":true,"origin":"","legend":"","description":"","filename":"figures1.tif","url":"https://assets-eu.researchsquare.com/files/rs-6906133/v1/ee980a0a88f1cb72497c22f6.tif"},{"id":87318205,"identity":"f9758c5d-fa70-4e72-94e6-f01f6ee2c5f8","added_by":"auto","created_at":"2025-07-22 16:15:21","extension":"tif","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":1180416,"visible":true,"origin":"","legend":"","description":"","filename":"figures2.tif","url":"https://assets-eu.researchsquare.com/files/rs-6906133/v1/65b1de2f65bc78126c708a2d.tif"},{"id":87320645,"identity":"3d3a5a69-7767-4ee8-8ca1-c57ce5b68ec0","added_by":"auto","created_at":"2025-07-22 16:31:21","extension":"tif","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":1120414,"visible":true,"origin":"","legend":"","description":"","filename":"figures3.tif","url":"https://assets-eu.researchsquare.com/files/rs-6906133/v1/696a1d1285311b36a4728543.tif"}],"financialInterests":"No competing interests reported.","formattedTitle":"Machine Learning Prediction of MACE in Older Chinese Adults Integrating Traditional and Geriatric-Specific Risk Factors: A CHARLS Cohort Analysis","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eCardiovascular disease (CVD), encompassing major adverse cardiovascular events (MACE) such as myocardial infarction and stroke, remains a leading cause of morbidity and mortality worldwide, imposing a substantial burden on healthcare systems and societal resources(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). This burden is particularly pronounced among older adults, as the prevalence of key CVD risk factors increases markedly with age. In China, which has one of the fastest-aging populations globally, the rising incidence of CVD and MACE among older adults presents an urgent public health challenge(\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e). Consequently, effective prevention strategies critically depend on accurate risk stratification to identify high-risk individuals who would benefit most from targeted interventions.\u003c/p\u003e\u003cp\u003eTraditional CVD risk prediction models, such as the widely used Framingham Risk Score (FRS) and SCORE, have been instrumental in clinical practice(\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e). However, these models were primarily developed and validated using data from Western, predominantly middle-aged populations. As a result, their predictive accuracy and calibration are often suboptimal when applied to diverse ethnic groups, particularly Chinese populations, potentially leading to risk misestimation(\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). Moreover, conventional models frequently overlook risk factors that become increasingly relevant in older adults. Geriatric syndromes (e.g., functional limitations assessed by Activities of Daily Living [ADL]) and psychosocial factors (e.g., depression, social isolation), now recognized as independent contributors to cardiovascular health, are typically absent from standard risk calculators.\u003c/p\u003e\u003cp\u003eThe advent of machine learning (ML) offers new opportunities to enhance CVD risk prediction. ML algorithms excel at handling high-dimensional datasets, capturing complex non-linear relationships, and identifying novel predictor interactions often overlooked by traditional regression methods(\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). Recent studies have demonstrated that ML approaches can significantly outperform conventional statistical techniques in predicting cardiovascular risk(\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e). Simultaneously, an expanding body of evidence has linked non-traditional risk factors\u0026mdash;such as central obesity (better captured by waist circumference than by BMI), depressive symptoms, functional limitations (e.g., ADL impairments), and frailty markers (e.g., low grip strength)\u0026mdash;to an increased risk of MACEs, particularly among older adults. Integrating these readily available yet underutilized factors through advanced modeling techniques holds considerable promise for developing more accurate and clinically relevant risk prediction tools tailored to the aging population(\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe China Health and Retirement Longitudinal Study (CHARLS) provides a unique and valuable resource to address this research gap. As a nationally representative, longitudinal cohort of middle-aged and older adults in China, CHARLS offers comprehensive data encompassing demographics, health behaviors, clinical assessments, biomarkers, functional measures (e.g., ADL, grip strength), and psychosocial indicators (e.g., the Center for Epidemiologic Studies Depression Scale [CES-D] score). Despite the richness of this dataset, few studies have developed and validated MACE risk prediction models specifically for older Chinese adults that integrate both traditional and non-traditional risk factors. Moreover, contemporary machine learning approaches combined with rigorous evaluation metrics\u0026mdash;extending beyond traditional discrimination analysis\u0026mdash;remain underutilized.\u003c/p\u003e\u003cp\u003eTherefore, the primary objective of this study was to develop and validate robust ML models for predicting the approximate 7-year risk of MACE among Chinese adults aged 60 years and older, using data from the CHARLS cohort. The secondary objectives were to: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) identify key traditional and non-traditional predictors most strongly associated with MACE risk using interpretable ML methods; (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) comprehensively evaluate model performance based on discrimination (area under the receiver operating characteristic curve [AUC]), calibration (Brier score, calibration plots), and clinical utility (decision curve analysis); (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) compare the developed models with a standard benchmark, the FRS; (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) explore potential non-linear relationships between continuous predictors and MACE risk; and (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e) construct a practical nomogram based on the best-performing interpretable model to facilitate individualized risk estimation.\u003c/p\u003e"},{"header":"2. Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Study Design, Data Source, and Cohort Selection\u003c/h2\u003e\u003cp\u003eThis prospective cohort study utilized data from the CHARLS dataset, a nationally representative longitudinal survey targeting individuals aged 45 years and older in China. The present analysis was based on the 2011 baseline wave (Wave 1), which included 17,708 participants. Participants aged 60 years or older at baseline were eligible for inclusion, resulting in 7,560 individuals. Exclusion criteria were subsequently applied. Participants with a self-reported physician-diagnosed history of myocardial infarction or stroke at or prior to the baseline interview were excluded (n\u0026thinsp;=\u0026thinsp;1,447). This yielded 6,113 participants aged\u0026thinsp;\u0026ge;\u0026thinsp;60 years and free of prevalent MACE for follow-up.\u003c/p\u003e\u003cp\u003eIncident MACE events were identified using data from subsequent CHARLS waves in 2013 (Wave 2), 2015 (Wave 3), and 2018 (Wave 4). Participants were further excluded if their incident MACE status could not be determined during follow-up (e.g., due to loss to follow-up or missing outcome data; n\u0026thinsp;=\u0026thinsp;1,533). The final analytical cohort thus comprised 4,580 participants who met the inclusion criteria, were free of MACE at baseline, and had complete outcome data between 2011 and 2018.CHARLS collects comprehensive information, including demographics, socioeconomic status, health behaviors and conditions, biomarkers, functional measures (e.g., ADL, grip strength), and psychosocial indicators (e.g., CES-D scores) through face-to-face interviews and physical examinations. The CHARLS study was approved by the Institutional Review Board of Peking University (IRB00001052-11015), and all participants provided written informed consent.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 Outcome Ascertainment\u003c/h2\u003e\u003cp\u003eThe primary outcome of this study was the first occurrence of an incident MACE. MACE was defined as a composite of self-reported, physician-diagnosed myocardial infarction or stroke during the follow-up period. Incident MACE events were identified based on participant self-reports, in accordance with standard CHARLS protocols. Although self-reports may introduce recall bias, previous validation studies(\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e) have demonstrated reasonable reliability of self-reported cardiovascular events in CHARLS. The follow-up period spanned approximately seven years from the baseline interview. As established during cohort selection, participants were excluded from the final analytical cohort if their incident MACE status could not be definitively determined by the Wave 4 assessment (e.g., due to loss to follow-up without a prior reported event or missing outcome data).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3 Predictor Variables\u003c/h2\u003e\u003cp\u003eCandidate predictor variables were assessed at baseline during the 2011 wave of CHARLS. Variables were categorized into six key domains: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) \u003cb\u003eDemographics\u003c/b\u003e: age (continuous, years); sex (female/male); education level (categorized as less than primary, primary school, middle school, or high school or higher); marital status (married/other); and residence (urban/rural); (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) \u003cb\u003eHealth Behaviors\u003c/b\u003e: current smoking status (yes/no); current alcohol consumption (yes/no); and physical activity level (estimated metabolic equivalent of task [MET] minutes per week, continuous); (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) \u003cb\u003eAnthropometric Measures\u003c/b\u003e: body mass index (BMI; kg/m\u0026sup2;, continuous) and waist circumference (cm, continuous); (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) \u003cb\u003eClinical Measures and Medical History\u003c/b\u003e: measured systolic and diastolic blood pressure (SBP/DBP; mmHg, continuous); self-reported physician-diagnosed hypertension (yes/no); and diabetes (yes/no); (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e) \u003cb\u003eBiomarkers\u003c/b\u003e: total cholesterol, high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol, triglycerides (all mg/dL); C-reactive protein (CRP; mg/L); and creatinine (mg/dL); and (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e) \u003cb\u003ePsychosocial and Geriatric Factors\u003c/b\u003e: depressive symptoms assessed by the CES-D-10; continuous); functional status assessed by the number of limitations in six Activities of Daily Living (ADL limitations; continuous); and maximum grip strength (kg, continuous).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e2.4 Missing Data Handling\u003c/h2\u003e\u003cp\u003eMissing data were observed across several predictor variables, particularly for biomarkers, physical activity, anthropometric measures, and blood pressure. To address missingness, multiple imputation by chained equations (MICE) was performed using the \u003cb\u003emice\u003c/b\u003e package (version 3.17.0) in R (version 4.4.1). This approach assumes that the data were missing at random (MAR). Five imputed datasets (m\u0026thinsp;=\u0026thinsp;5) were generated. The imputation models included all candidate predictor variables and the MACE outcome to preserve potential associations during the imputation process. Subsequent statistical analyses were conducted independently on each of the five imputed datasets, and the results were pooled using Rubin\u0026rsquo;s rules where applicable.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e2.5 Statistical Analysis and Predictive Modeling\u003c/h2\u003e\u003cp\u003eAll statistical analyses were conducted using R (version 4.4.1). A two-sided P-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant. Prior to model development, each of the five imputed datasets was randomly partitioned into a training set (70%) and a test set (30%), stratified by MACE outcomes to maintain balanced class distributions. A fixed random seed was set to ensure reproducibility. Partitioning was applied consistently across all imputations, ensuring that the same individuals were assigned to the training and test sets within each imputed dataset. Three predictive modeling approaches were trained on each imputed training set: logistic regression (LR), random forest (RF), and extreme gradient boosting (XGBoost)(\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e). Model training was performed using the caret package (version 7.0.1). For the RF and XGBoost models, hyperparameters were tuned using a basic grid search with a tuning length of 3 due to computational constraints. Future studies may explore more extensive hyperparameter optimization techniques, such as Bayesian optimization or randomized grid search, to further improve model performance.\u003c/p\u003e\u003cp\u003eThe models were optimized to maximize the area under the receiver operating characteristic curve (AUC). Model performance was primarily evaluated on the independent test sets, with results subsequently pooled across the five imputations. Discrimination was assessed using the AUC, with the average AUC and standard deviation reported across imputations. Calibration, reflecting the agreement between predicted probabilities and observed event rates, was evaluated both visually and quantitatively: pooled calibration plots were generated by grouping predictions into deciles, and the pooled Brier score was calculated (lower values indicating better overall accuracy)(\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). Clinical utility was assessed using decision curve analysis (DCA), by plotting the pooled average net benefit against varying risk thresholds(\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e). The performance of the developed LR, RF, and XGBoost models was compared internally and benchmarked against the Framingham Risk Score for cardiovascular disease (FRS-CVD), calculated based on baseline variables from the test sets, while acknowledging limitations arising from differences in outcome definitions and follow-up durations(\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eModel interpretation focused primarily on the LR model, which demonstrated the best performance based on the average AUC. Pooled coefficients, odds ratios (ORs), 95% confidence intervals (CIs), and P-values were obtained using the \u003cem\u003emice::pool()\u003c/em\u003e function, applying Rubin\u0026rsquo;s rules. Predictor importance in the LR model was ranked based on the absolute magnitude of the pooled Z-statistic. Significant ORs were visualized using a forest plot. To gain additional insights potentially missed by the linear model, the \u003cem\u003eXGBoost\u003c/em\u003e model\u0026mdash;the best-performing non-linear model\u0026mdash;was interpreted using SHapley Additive exPlanations (SHAP) values. SHAP values were calculated via the \u003cem\u003exgboost\u003c/em\u003e package (version 1.7.9.1) and averaged across imputations. Pooled SHAP summary plots and dependence plots were generated using the \u003cem\u003eshapviz\u003c/em\u003e package (version 0.9.7) to identify key features and visualize their effects, a methodology previously applied successfully in cardiovascular risk prediction studies(\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eTo explore potential non-linear associations between key continuous predictors (age, waist circumference, SBP, CES-D score, and maximum grip strength) and MACE risk, generalized additive models (GAMs) with penalized smoothing splines (specifically \u003cem\u003es\u003c/em\u003e (..., \u003cem\u003ek\u003c/em\u003e\u0026thinsp;=\u0026thinsp;5)) were fitted to the first imputed dataset using the \u003cem\u003emgcv\u003c/em\u003e package (version 1.9.1), adjusting for other significant covariates identified in the LR model. Smooth function plots were generated using the \u003cem\u003egratia\u003c/em\u003e package (version 0.10.0). The effective degrees of freedom (EDF) and P-values from the GAM summaries were examined to assess the presence and significance of non-linearity. Two sensitivity analyses were conducted to assess the robustness of the primary findings: (A) a complete-case analysis, excluding participants with any missing predictor data prior to modeling; and (B) an exclusion analysis, removing the predictor with the highest proportion of missing values (physical activity, measured as MET-minutes/week) prior to imputation and modeling. Model performance, evaluated by average AUC, from these scenarios was compared to that of the main analysis. Finally, a nomogram was developed based on the final pooled LR model coefficients (using key predictors identified by Z-statistic magnitude and refitted on the first imputed dataset) to provide a practical tool for individual risk estimation. The nomogram was constructed using the \u003cem\u003erms\u003c/em\u003e package (version 6.9.0).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e2.6 Software Used\u003c/h2\u003e\u003cp\u003eAll statistical analyses were conducted using R (version 4.4.1). Key packages utilized included \u003cb\u003emice\u003c/b\u003e (version 3.17.0) for multiple imputation, \u003cb\u003ecaret\u003c/b\u003e (version 7.0.1) for model training and hyperparameter tuning, and \u003cb\u003epROC\u003c/b\u003e (version 1.18.5) for receiver operating characteristic (ROC) analysis. The \u003cb\u003erms\u003c/b\u003e package (version 6.9.0) was used for nomogram development, while \u003cb\u003emgcv\u003c/b\u003e (version 1.9.1) and \u003cb\u003egratia\u003c/b\u003e (version 0.10.0) were employed for generalized additive model (GAM) analysis. For XGBoost modeling and interpretation using SHapley Additive exPlanations (SHAP), the \u003cb\u003exgboost\u003c/b\u003e (version 1.7.9.1) and \u003cb\u003eshapviz\u003c/b\u003e (version 0.9.7) packages were applied. Decision curve analysis was performed using the \u003cb\u003edcurves\u003c/b\u003e package (version 0.5.0). Data manipulation was conducted using \u003cb\u003edplyr\u003c/b\u003e (version 1.1.4) and \u003cb\u003etidyr\u003c/b\u003e (version 1.3.1). Visualization was achieved with \u003cb\u003eggplot2\u003c/b\u003e (version 3.5.1) and \u003cb\u003epatchwork\u003c/b\u003e (version 1.3.0). Descriptive tables were generated with \u003cb\u003egtsummary\u003c/b\u003e (version 2.2.0), and tabular results were exported using \u003cb\u003ewritexl\u003c/b\u003e (version 1.5.4).\u003c/p\u003e\u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Cohort Characteristics\u003c/h2\u003e\u003cp\u003eThe final analytical cohort comprised 4,580 participants aged 60 years or older from the 2011 baseline wave of CHARLS, who were free of MACE at baseline and followed until the 2018 wave or the occurrence of an incident MACE. Baseline characteristics, stratified by incident MACE status over the approximately 7-year follow-up period, are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. The mean age of the cohort was 67.1 years (SD\u0026thinsp;=\u0026thinsp;6.2), and 41.1% of participants were male. During follow-up, 1,315 individuals (28.7%) experienced an incident MACE. Compared to those who remained MACE-free, individuals who developed MACE were significantly older, had a higher prevalence of hypertension and diabetes, exhibited larger waist circumference, and reported higher CES-D scores and more limitations in ADLs at baseline (all P\u0026thinsp;\u0026lt;\u0026thinsp;0.05; see Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eBaseline Characteristics of Study Participants (n\u0026thinsp;=\u0026thinsp;4580), Stratified by Incident MACE Status.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCharacteristic\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eOverall (N\u0026thinsp;=\u0026thinsp;4580)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eNo MACE (N\u0026thinsp;=\u0026thinsp;3265)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eMACE (N\u0026thinsp;=\u0026thinsp;1315)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eP-value\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge (years)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e67.1 (6.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e66.4 (5.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e68.9 (6.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSex\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.13\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFemale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2042 (45%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1479 (45%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e563 (43%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1884 (41%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1316 (40%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e568 (43%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e654 (14%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e470 (14%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e184 (14%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEducation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLess than primary\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2676 (58%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1833 (56%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e843 (64%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePrimary school\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1108 (24%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e826 (25%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e282 (21%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMiddle school\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e506 (11%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e379 (12%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e127 (10%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHigh school+\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e279 (6.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e219 (6.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e60 (4.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e11 (\u0026lt;\u0026thinsp;1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8 (\u0026lt;\u0026thinsp;1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e3 (\u0026lt;\u0026thinsp;1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMarital Status\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.11\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOther\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e878 (19%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e604 (18%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e274 (21%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMarried\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3697 (81%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2657 (81%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1040 (79%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e5 (\u0026lt;\u0026thinsp;1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4 (\u0026lt;\u0026thinsp;1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1 (\u0026lt;\u0026thinsp;1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eResidence\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.002\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eUrban\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1582 (35%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1169 (36%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e413 (31%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRural\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2998 (65%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2096 (64%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e902 (69%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCurrent Smoking\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.06\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3082 (67%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2231 (68%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e851 (65%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1381 (30%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e965 (30%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e416 (32%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e117 (2.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e69 (2.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e48 (3.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCurrent Drinking\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3090 (67%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2156 (66%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e934 (71%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1458 (32%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1087 (33%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e371 (28%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e32 (\u0026lt;\u0026thinsp;1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e22 (\u0026lt;\u0026thinsp;1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e10 (\u0026lt;\u0026thinsp;1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePhysical Activity (MET-min/week)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e7269.1 (6432.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7386.5 (6452.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e6976.1 (6369.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eNA\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBMI (kg/m^2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e23.6 (18.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e23.5 (16.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e23.7 (23.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eNA\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWaist Circumference (cm)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e83.9 (10.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e83.3 (10.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e85.2 (10.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSystolic BP (mmHg)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e133.1 (21.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e131.9 (20.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e136.1 (22.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiastolic BP (mmHg)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e74.3 (11.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e74.0 (11.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e75.0 (11.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.009\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHistory of Hypertension\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3266 (71%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2452 (75%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e814 (62%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1262 (28%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e777 (24%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e485 (37%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e52 (1.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e36 (1.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e16 (1.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHistory of Diabetes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4262 (93%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3054 (94%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1208 (92%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e248 (5.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e176 (5.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e72 (5.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e70 (1.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e35 (1.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e35 (2.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTotal Cholesterol (mg/dL)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e194.9 (35.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e194.5 (34.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e195.9 (36.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.14\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHDL Cholesterol (mg/dL)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e52.3 (14.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e52.6 (14.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e51.4 (13.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.007\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLDL Cholesterol (mg/dL)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e118.1 (29.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e117.6 (29.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e119.3 (30.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTriglycerides (mg/dL)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e128.1 (89.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e125.5 (77.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e134.4 (111.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.015\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCRP (mg/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2.9 (5.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.8 (5.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e3.0 (5.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.47\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCreatinine (mg/dL)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.8 (0.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.8 (0.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.8 (0.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.65\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCES-D Score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e8.7 (6.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8.0 (6.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e10.3 (6.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eADL Limitations (Count)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.4 (1.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.3 (0.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.8 (1.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMax Grip Strength (kg)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e29.8 (11.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e30.3 (11.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e28.6 (10.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"5\"\u003eData presented as Mean (SD) for continuous variables or n (%) for categorical variables, based on the first imputed dataset. P-values compare MACE vs No MACE groups using appropriate statistical tests (e.g., t-test/Wilcoxon, Chi-squared/Fisher's). Abbreviations: SD, standard deviation; MACE, Major Adverse Cardiovascular Event; BMI, Body Mass Index; BP, Blood Pressure; HDL, High-Density Lipoprotein; LDL, Low-Density Lipoprotein; CRP, C-Reactive Protein; CES-D, Center for Epidemiologic Studies Depression Scale; ADL, Activities of Daily Living; MET, Metabolic Equivalent of Task.\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Model Performance Comparison\u003c/h2\u003e\u003cp\u003eThe predictive performance of the LR, RF, and XGBoost models was compared against each other and the benchmark Framingham Risk Score for cardiovascular disease (FRS-CVD), using results pooled across the five imputed test sets (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). In terms of discrimination, the LR model achieved the highest average area under the receiver operating characteristic curve (AUC) (mean\u0026thinsp;=\u0026thinsp;0.649, SD\u0026thinsp;=\u0026thinsp;0.003), closely followed by XGBoost (mean\u0026thinsp;=\u0026thinsp;0.645, SD\u0026thinsp;=\u0026thinsp;0.009). Both models significantly outperformed the RF model (mean\u0026thinsp;=\u0026thinsp;0.632, SD\u0026thinsp;=\u0026thinsp;0.006) and especially the FRS-CVD benchmark (mean\u0026thinsp;=\u0026thinsp;0.504, SD\u0026thinsp;=\u0026thinsp;0.007). Representative ROC curves from one imputed test set visually illustrate the relative performance ranking (Fig.\u0026nbsp;2).\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\u003ePredictive Performance Comparison of Models for ~\u0026thinsp;7-Year Incident MACE.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAUC (SD)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eBrier Score\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLogistic Regression\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.649 (nan)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.193\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRandom Forest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.632 (nan)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.197\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eXGBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.645 (nan)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.194\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFramingham (CVD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.504 (nan)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eNA\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"3\"\u003ePerformance metrics averaged over 5 imputed test sets. AUC indicates Area Under the Receiver Operating Characteristic Curve; SD, Standard Deviation; Brier Score measures overall accuracy (lower is better); NA, Not Available. Models: LR, Logistic Regression; RF, Random Forest; XGBoost, Extreme Gradient Boosting; FRS-CVD, Framingham Risk Score for 10-year general Cardiovascular Disease (used as benchmark, limitations apply).\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eCalibration assessment (Fig.\u0026nbsp;3) showed that both the LR and XGBoost models were reasonably well-calibrated, with predicted probabilities closely aligning with observed event rates. The RF model slightly underestimated risk at higher predicted probability levels, while the FRS-CVD score exhibited poor calibration, deviating substantially from the line of identity. Consistent with the AUC findings, the pooled Brier scores were lowest for LR (mean\u0026thinsp;=\u0026thinsp;0.184) and XGBoost (mean\u0026thinsp;=\u0026thinsp;0.184), indicating superior overall accuracy compared to RF (mean\u0026thinsp;=\u0026thinsp;0.185) and FRS-CVD (mean\u0026thinsp;=\u0026thinsp;0.215) (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Furthermore, decision curve analysis (DCA) demonstrated that the LR and XGBoost models provided greater net clinical benefit across a wide range of probability thresholds, outperforming both the RF model and the FRS-CVD score, as well as the default strategies of treating all or none (Fig.\u0026nbsp;4).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Key Predictors in the Logistic Regression Model\u003c/h2\u003e\u003cp\u003eThe pooled multivariable logistic regression analysis identified several factors independently associated with the risk of incident MACE over the approximately 7-year follow-up period (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e; Fig.\u0026nbsp;5). Predictor importance, ranked by the absolute magnitude of the pooled Z-statistics (Appendix Table A1), indicated that the most influential variables included history of hypertension, CES-D score, education level (high school or above vs. less than primary), number of ADL limitations, waist circumference, age, SBP, and history of diabetes. After mutual adjustment, several traditional risk factors remained significantly associated with increased MACE risk: history of hypertension (OR\u0026thinsp;=\u0026thinsp;1.85, 95% CI: 1.58\u0026ndash;2.16), history of diabetes (OR\u0026thinsp;=\u0026thinsp;1.34, 95% CI: 1.01\u0026ndash;1.78), per year increase in age (OR\u0026thinsp;=\u0026thinsp;1.02, 95% CI: 1.00\u0026ndash;1.03), and per mmHg increase in SBP (OR\u0026thinsp;=\u0026thinsp;1.006, 95% CI: 1.001\u0026ndash;1.011). In addition, non-traditional and socioeconomic predictors also demonstrated significant associations: high school or higher education level (vs. less than primary school; OR\u0026thinsp;=\u0026thinsp;1.60, 95% CI: 1.20\u0026ndash;2.14), waist circumference (per cm; OR\u0026thinsp;=\u0026thinsp;1.01, 95% CI: 1.00\u0026ndash;1.02), CES-D score (per point; OR\u0026thinsp;=\u0026thinsp;1.03, 95% CI: 1.02\u0026ndash;1.04), and number of ADL limitations (per unit; OR\u0026thinsp;=\u0026thinsp;1.11, 95% CI: 1.04\u0026ndash;1.19). Notably, waist circumference, depressive symptoms (CES-D score), and functional limitations (ADL) remained significant predictors even after adjusting for conventional cardiovascular risk factors (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003ePooled Multivariable Logistic Regression Analysis of Predictors for ~\u0026thinsp;7-Year Incident MACE.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePredictor\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eOdds Ratio (OR)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003e95% CI\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eP-value\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge (years)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.00-1.03\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.015\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSex: Male (vs Female)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.86\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.67\u0026ndash;1.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.24\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEducation\u0026thinsp;=\u0026thinsp;Primary (vs\u0026thinsp;\u0026lt;\u0026thinsp;Primary)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.05\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.88\u0026ndash;1.24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.61\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEducation\u0026thinsp;=\u0026thinsp;Middle (vs\u0026thinsp;\u0026lt;\u0026thinsp;Primary)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.26\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.00-1.59\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.05\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEducation\u0026thinsp;=\u0026thinsp;High+ (vs\u0026thinsp;\u0026lt;\u0026thinsp;Primary)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.20\u0026ndash;2.14\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.002\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMarital Status\u0026thinsp;=\u0026thinsp;Married (vs Other)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.08\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.91\u0026ndash;1.29\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.38\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eResidence\u0026thinsp;=\u0026thinsp;Rural (vs Urban)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.89\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.76\u0026ndash;1.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.13\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCurrent Smoking\u0026thinsp;=\u0026thinsp;Yes (vs No)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.93\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.78\u0026ndash;1.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.45\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCurrent Drinking\u0026thinsp;=\u0026thinsp;Yes (vs No)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.86\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.73\u0026ndash;1.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.09\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePhysical Activity (MET-min/week)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.00\u0026ndash;1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.71\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBMI (kg/m^2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.99-1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.68\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWaist Circumference (cm)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.00-1.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.011\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSystolic BP (mmHg)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.00-1.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.03\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiastolic BP (mmHg)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.99\u0026ndash;1.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.51\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHistory of Hypertension\u0026thinsp;=\u0026thinsp;Yes (vs No)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.85\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.58\u0026ndash;2.16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHistory of Diabetes\u0026thinsp;=\u0026thinsp;Yes (vs No)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.34\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.01\u0026ndash;1.78\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.04\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTotal Cholesterol (mg/dL)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.99\u0026ndash;1.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.65\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHDL Cholesterol (mg/dL)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.00-1.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.31\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLDL Cholesterol (mg/dL)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.99\u0026ndash;1.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.74\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTriglycerides (mg/dL)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.00\u0026ndash;1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.89\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCRP (mg/L)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.00-1.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.13\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCreatinine (mg/dL)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.76\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.50\u0026ndash;1.15\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.19\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCES-D Score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.03\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.02\u0026ndash;1.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eADL Limitations (Count)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.04\u0026ndash;1.19\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.002\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMax Grip Strength (kg)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.00-1.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.3\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"4\"\u003eResults pooled across 5 imputed datasets using Rubin's Rules. OR indicates Odds Ratio; CI, Confidence Interval; MACE, Major Adverse Cardiovascular Event. Reference groups for categorical predictors: Sex\u0026thinsp;=\u0026thinsp;Female; Education\u0026thinsp;=\u0026thinsp;\u0026lt;\u0026thinsp;Primary; Marital Status\u0026thinsp;=\u0026thinsp;Other; Residence\u0026thinsp;=\u0026thinsp;Urban; Smoking\u0026thinsp;=\u0026thinsp;No; Drinking\u0026thinsp;=\u0026thinsp;No; Hypertension\u0026thinsp;=\u0026thinsp;No; Diabetes\u0026thinsp;=\u0026thinsp;No.\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e3.4 Exploration of Non-Linear Relationships\u003c/h2\u003e\u003cp\u003eExploratory analysis using generalized additive models (GAMs) revealed significant non-linear associations between the log-odds of MACE and both age (effective degrees of freedom [EDF]\u0026thinsp;=\u0026thinsp;2.27, P\u0026thinsp;=\u0026thinsp;0.003) and waist circumference (EDF\u0026thinsp;=\u0026thinsp;2.28, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001) (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e; Fig.\u0026nbsp;7). In contrast, the relationships of systolic blood pressure (SBP) (EDF\u0026thinsp;=\u0026thinsp;1.0, P\u0026thinsp;=\u0026thinsp;0.04) and CES-D score (EDF\u0026thinsp;=\u0026thinsp;1.0, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001) with MACE risk were effectively linear but remained statistically significant. Maximum grip strength was not significantly associated with MACE risk in the adjusted GAM model (EDF\u0026thinsp;=\u0026thinsp;1.0, P\u0026thinsp;=\u0026thinsp;0.53).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eApproximate Significance of Smooth Terms from Exploratory Generalized Additive Model (GAM).\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=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSmooth Term\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEDF\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eRef.df\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eChi.sq\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eP-value\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eNon-Linear\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSmooth(Age)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2.27\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e14.63\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.003\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eTRUE\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSmooth(Waist Circum.)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2.28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e25.06\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eTRUE\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSmooth(Systolic BP)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e4.3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eFALSE\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSmooth(CES-D Score)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e33.37\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eFALSE\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSmooth(Max Grip Strength)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.53\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eFALSE\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"6\"\u003eResults from GAM fitted on the first imputed dataset. EDF indicates Effective Degrees of Freedom; Ref.df, Reference Degrees of Freedom; Chi.sq, Chi-squared statistic. 'Non-Linear' column indicates if EDF\u0026thinsp;\u0026gt;\u0026thinsp;1.1, suggesting a non-linear relationship. P-value assesses the significance of the smooth term's contribution.\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e3.5 Interpretation Insights from the XGBoost Model\u003c/h2\u003e\u003cp\u003eInterpretation of the XGBoost model using pooled SHapley Additive exPlanations (SHAP) values revealed feature importance rankings that largely aligned with findings from the primary logistic regression model (Figure \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e). The top-ranked predictors based on mean absolute SHAP values included history of hypertension, CES-D score, waist circumference, age, systolic blood pressure (SBP), number of ADL limitations, and higher education level (high school or above vs. less than primary). This consistency in predictor importance across distinct modeling approaches reinforces the robustness of these key risk factors. Notably, the high ranking of non-traditional predictors\u0026mdash;waist circumference, depressive symptoms (CES-D score), and functional limitations (ADL limitations)\u0026mdash;in the XGBoost model further supports their relevance in predicting MACE risk in this older Chinese cohort.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003e3.6 Sensitivity Analysis Results\u003c/h2\u003e\u003cp\u003eThe robustness of model performance was evaluated through two sensitivity analyses (Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). First, exclusion of the variable \u003cem\u003etotmet\u003c/em\u003e (which had\u0026thinsp;\u0026gt;\u0026thinsp;60% missingness) prior to multiple imputation resulted in minimal changes in the average AUCs for LR (mean\u0026thinsp;=\u0026thinsp;0.650), RF (mean\u0026thinsp;=\u0026thinsp;0.624), and XGBoost (mean\u0026thinsp;=\u0026thinsp;0.643) compared to the main analysis, suggesting that the primary findings were robust to the exclusion of this highly incomplete predictor. In contrast, a complete-case analysis\u0026mdash;which reduced the analytical sample to 977 participants (21.3% of the original cohort)\u0026mdash;resulted in substantially lower AUC values for all models (LR: 0.543; RF: 0.558; XGBoost: 0.553), potentially reflecting the value of multiple imputation in preserving statistical power and minimizing selection bias. Importantly, the relative performance ranking among models (LR\u0026thinsp;\u0026asymp;\u0026thinsp;XGBoost\u0026thinsp;\u0026gt;\u0026thinsp;RF) remained generally consistent across both sensitivity scenarios.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003e3.7 Nomogram for Risk Prediction\u003c/h2\u003e\u003cp\u003eBased on the final multivariable logistic regression model incorporating key predictors\u0026mdash;age, sex, education level, waist circumference, SBP, history of hypertension, history of diabetes, CES-D score, and number of ADL limitations\u0026mdash;a nomogram was developed to facilitate individualized clinical risk estimation (Fig.\u0026nbsp;6). This graphical tool enables clinicians to estimate an individual's approximate 7-year risk of MACE by summing points corresponding to their specific risk factor profile. Each predictor is assigned a point value proportional to its relative contribution to MACE risk, and the total score maps to a predicted probability of incident MACE. The nomogram serves as a practical, interpretable aid for risk stratification and shared decision-making in older Chinese adults.\u003c/p\u003e\u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eThis study leveraged data from the China Health and Retirement Longitudinal Study (CHARLS)\u0026mdash;a rich, nationally representative longitudinal cohort\u0026mdash;to develop and evaluate machine learning models for predicting incident MACE over a 7-year follow-up among older Chinese adults. The LR model incorporating both traditional and non-traditional risk factors demonstrated the best overall performance, achieving moderate discrimination (mean AUC\u0026thinsp;=\u0026thinsp;0.649) and good calibration. Notably, it significantly outperformed the standard Framingham Risk Score for cardiovascular disease (FRS-CVD), which yielded poor discrimination (mean AUC\u0026thinsp;=\u0026thinsp;0.504) in this population Importantly, several non-traditional factors\u0026mdash;waist circumference, depressive symptoms (CES-D score), and functional limitations (ADL limitations)\u0026mdash;were identified as independent predictors of MACE, alongside established risk factors including age, hypertension, diabetes, and systolic blood pressure. Exploratory analyses using GAMs further revealed statistically significant non-linear relationships between MACE risk and both age and waist circumference.\u003c/p\u003e\u003cp\u003eA major contribution of this study lies in elucidating the predictive importance of non-traditional yet readily accessible factors for MACE risk among older Chinese adults. The significant association of waist circumference, as opposed to BMI, highlights the role of central obesity\u0026mdash;distinct from general adiposity\u0026mdash;as a key risk factor (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e; Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e; Fig.\u0026nbsp;5; Figure \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e; Appendix Table A1). Central obesity is known to be closely linked to metabolic syndrome, systemic inflammation, and insulin resistance, which provide plausible biological pathways for its association with cardiovascular events(\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e)、. The independent predictive value of the CES-D score (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e; Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e; Fig.\u0026nbsp;5; Figure \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e; Appendix Table A1) reinforces the critical link between mental and cardiovascular health in later life. Depression may contribute to increased MACE risk through behavioral mechanisms (e.g., reduced treatment adherence, physical inactivity, and smoking) and physiological pathways (e.g., hypothalamic\u0026ndash;pituitary\u0026ndash;adrenal axis dysregulation, systemic inflammation, and endothelial dysfunction)(\u003cspan additionalcitationids=\"CR19\" citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e). Furthermore, the observed association between limitations in ADLs and MACE risk (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e; Fig.\u0026nbsp;5; Figure \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e) underscores the prognostic relevance of functional status in aging populations. ADL impairments may reflect underlying frailty, unrecognized disease burden, or diminished physiological reserve, all of which increase vulnerability to adverse cardiovascular outcomes(\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThese findings support the adoption of a more holistic approach to cardiovascular risk assessment in older adults\u0026mdash;one that incorporates measures of central obesity, mental health, and functional status alongside conventional risk factors. Notably, a higher education level (high school or above versus less than primary) was associated with increased MACE risk in the pooled logistic regression model (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e), a finding that may seem counterintuitive in light of previous literature, where higher education is often viewed as protective(\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e). This discrepancy may reflect complex interactions with unmeasured socioeconomic determinants, differences in health behaviors, or survival bias within this aging cohort. It also raises important questions about how educational attainment categories function as proxies for health advantage in different demographic and cultural contexts, warranting further investigation. In addition, grip strength\u0026mdash;while included in the generalized additive model\u0026mdash;did not emerge as a significant linear or non-linear predictor in the adjusted analysis (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e), suggesting limited utility of this measure in predicting MACE risk in this population.\u003c/p\u003e\u003cp\u003eOur exploratory GAM analysis (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e; Fig.\u0026nbsp;7) revealed significant non-linear relationships between the log-odds of MACE and both age (EDF\u0026thinsp;=\u0026thinsp;2.3) and waist circumference (EDF\u0026thinsp;=\u0026thinsp;2.3). These results suggest that the risk of MACE does not increase linearly with age or waist circumference\u0026mdash;a nuance that may be overlooked by conventional linear models such as the FRS or basic logistic regression models without non-linear terms (e.g., spline functions)(\u003cspan additionalcitationids=\"CR23\" citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e). Although the logistic regression model slightly outperformed XGBoost in terms of average AUC, the confirmation of non-linearity supports the further exploration of machine learning models capable of capturing complex, non-linear relationships(\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e), or the refinement of traditional regression models through the incorporation of flexible approaches such as restricted cubic splines. In addition, the strong agreement in predictor importance rankings between the pooled logistic regression model (Appendix Table A1) and the SHAP analysis of the XGBoost model (Figure \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e) reinforces the robustness of key predictors\u0026mdash;hypertension, CES-D score, waist circumference, age, systolic blood pressure, and ADL limitations\u0026mdash;consistent with prior studies investigating cardiovascular risk stratification using feature importance metrics(\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThis study has several notable strengths. First, it is based on a large, nationally representative longitudinal cohort (CHARLS), ensuring the generalizability of findings to the older adult population in China. Second, it uniquely integrates both traditional and non-traditional risk factors relevant to aging, such as central obesity, depressive symptoms, and functional status. Third, a comparative evaluation of multiple modeling approaches\u0026mdash;LR, RF, and XGBoost\u0026mdash;was conducted(\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e), with performance assessed comprehensively using not only discrimination (AUC) but also calibration and decision curve analysis (DCA)(\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e). Fourth, interpretable methods such as SHAP and GAMs were applied to enhance understanding of model outputs(\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e). The use of SHAP values, in particular, represents an important methodological advancement over traditional feature importance metrics by offering consistent, individualized estimates of predictor contributions(\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e). Finally, a nomogram was developed based on the best-performing interpretable model (LR) (Fig.\u0026nbsp;6), offering a practical and user-friendly tool for individualized clinical risk estimation and decision-making in older Chinese adults.\u003c/p\u003e\u003cp\u003eDespite its strengths, this study has several limitations that should be considered when interpreting the findings. First, the reliance on self-reported MACE outcomes may introduce recall bias and outcome misclassification. Second, the handling of missing data\u0026mdash;particularly prevalent for biomarker and physical activity variables\u0026mdash;required multiple imputation under the unverifiable assumption of missing at random (MAR)(\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e). Third, the approximate 7-year follow-up period differs from the conventional 10-year horizon used in established risk scores such as the FRS, which limits direct comparability. Additionally, the FRS itself was originally developed for a different population (predominantly white, middle-aged American adults) and employed a distinct composite outcome definition, further complicating comparisons. Fourth, the nomogram construction and GAM analyses were conducted using only the first imputed dataset due to computational constraints, and thus do not fully reflect the uncertainty inherent in multiple imputation procedures. Fifth, although a broad range of predictors was included, the possibility of residual confounding from unmeasured variables remains, as is common in observational studies. Sixth, hyperparameter tuning for the machine learning models was limited (tuneLength\u0026thinsp;=\u0026thinsp;3), which may have constrained the full potential performance of models like XGBoost, particularly when evaluated across multiple imputed datasets. Finally, while the internal performance of the models was promising, external validation in independent cohorts of older Chinese adults is essential before considering clinical application\u0026mdash;an important prerequisite for implementing machine learning-based risk prediction models in practice(\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e).\u003c/p\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eIn conclusion, this study successfully developed and comprehensively evaluated risk prediction models for incident MACE using data from a large, nationally representative cohort of older Chinese adults (CHARLS). The best-performing model\u0026mdash;a logistic regression incorporating traditional risk factors alongside waist circumference, depressive symptoms (CES-D score), and limitations in ADLs\u0026mdash;demonstrated moderate discrimination (mean AUC\u0026thinsp;=\u0026thinsp;0.649) and good calibration, significantly outperforming the standard FRS benchmark (mean AUC\u0026thinsp;=\u0026thinsp;0.504)(\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThese findings underscore the independent predictive value of central obesity (as measured by waist circumference), mental health (CES-D score), and functional status (A DL limitations) in cardiovascular risk estimation, supporting a more holistic and aging-sensitive approach to risk stratification in clinical practice(\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e). Public health strategies targeting this rapidly aging population should consider interventions aimed at reducing central adiposity, promoting mental well-being, and preserving functional capacity alongside conventional CVD prevention measures(\u003cspan additionalcitationids=\"CR37\" citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe developed nomogram (Fig.\u0026nbsp;6) provides a practical tool for individualized risk estimation and may facilitate shared decision-making in clinical settings; however, its clinical utility requires further confirmation through external validation in independent populations(\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e). Future research priorities include validating the models in external and ethnically diverse cohorts, developing dynamic prediction frameworks using repeated longitudinal measurements, applying advanced machine learning techniques to investigate interactions and nonlinearities among predictors(\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e), and evaluating the incremental predictive value of additional biomarkers and social determinants of health.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eData Availability Statement\u003c/p\u003e\n\u003cp\u003eThe data that support the findings of this study are derived from the China Health and Retirement Longitudinal Study (CHARLS). Access to the data is restricted and was granted under a data use agreement with Peking University for the purposes of this study. Therefore, the datasets are not publicly available without prior approval. Researchers interested in accessing the CHARLS data may do so upon reasonable request and completion of a data use application through the official CHARLS website: http://charls.pku.edu.cn. The R code used to conduct the statistical analyses in this study is available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003eConflict of Interest\u003c/p\u003e\n\u003cp\u003eThe authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.\u003c/p\u003e\n\u003cp\u003eAuthor Contributions\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eJL\u003c/em\u003e designed the study, performed the data analysis, developed the models, interpreted the results, and drafted the initial manuscript. \u003cem\u003eZS\u003c/em\u003e contributed to the study methodology, investigation process, and provided critical revisions to the manuscript. \u003cem\u003eFZ\u003c/em\u003e evaluated the psychological indicators derived from the CHARLS dataset.\u003cem\u003e\u0026nbsp;QX, GH, GW\u0026nbsp;\u003c/em\u003eand \u003cem\u003eXL\u003c/em\u003e assisted with software implementation, data curation, formal analysis, visualization, and manuscript review. \u003cem\u003eYH, ZZ and XH\u0026nbsp;\u003c/em\u003econtributed to the study conceptualization, supervised various aspects of the work, and critically reviewed the manuscript. \u003cem\u003eZZ*\u003c/em\u003e conceived and designed the overall study, provided supervision and resources, administered the project, potentially secured funding, and critically reviewed and edited the final manuscript. All authors read and approved the final version of the manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFunding\u003c/p\u003e\n\u003cp\u003eThis research was supported by the National Natural Science Foundation (Grant No. 81660730).\u003c/p\u003e\n\u003cp\u003eAcknowledgments\u003c/p\u003e\n\u003cp\u003eThe authors thank all the members of the CHALRS for their contributions and the participants who contributed their data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe CHARLS study was approved by the Institutional Review Board of Peking University (IRB00001052-11015), and all participants provided written informed consent prior to participation. The present study was conducted in accordance with the ethical standards of the Declaration of Helsinki and its later amendments.\u003c/p\u003e\n\u003cp\u003eEthics approval committee: Institutional Review Board of Peking University.\u003c/p\u003e\n\u003cp\u003eConsent to participate: All participants provided written informed consent.\u003c/p\u003e\n\u003cp\u003eClinical trial number: not applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eMensah GA, Fuster V, Murray CJL, Roth GA. Global Burden of Cardiovascular Diseases and Risks, 1990-2022. Journal of the American College of Cardiology. 2023;82(25):2350-473.\u003c/li\u003e\n\u003cli\u003eD\u0026apos;Agostino RB, Sr., Grundy S, Sullivan LM, Wilson P. Validation of the Framingham coronary heart disease prediction scores: results of a multiple ethnic groups investigation. Jama. 2001;286(2):180-7.\u003c/li\u003e\n\u003cli\u003eWang L, Song L, Li D, Zhou Z, Chen S, Yang Y, et al. Ideal Cardiovascular Health Metric and Its Change With Lifetime Risk of Cardiovascular Diseases: A Prospective Cohort Study. Journal of the American Heart Association. 2021;10(22):e022502.\u003c/li\u003e\n\u003cli\u003eLiu J, Hong Y, D\u0026apos;Agostino RB, Sr., Wu Z, Wang W, Sun J, et al. Predictive value for the Chinese population of the Framingham CHD risk assessment tool compared with the Chinese Multi-Provincial Cohort Study. Jama. 2004;291(21):2591-9.\u003c/li\u003e\n\u003cli\u003eBarzi F, Patel A, Gu D, Sritara P, Lam TH, Rodgers A, et al. Cardiovascular risk prediction tools for populations in Asia. Journal of epidemiology and community health. 2007;61(2):115-21.\u003c/li\u003e\n\u003cli\u003eChen JH, Asch SM. Machine Learning and Prediction in Medicine - Beyond the Peak of Inflated Expectations. The New England journal of medicine. 2017;376(26):2507-9.\u003c/li\u003e\n\u003cli\u003eAmbale-Venkatesh B, Yang X, Wu CO, Liu K, Hundley WG, McClelland R, et al. Cardiovascular Event Prediction by Machine Learning: The Multi-Ethnic Study of Atherosclerosis. Circulation research. 2017;121(9):1092-101.\u003c/li\u003e\n\u003cli\u003eWu Q, Kling JM. Depression and the Risk of Myocardial Infarction and Coronary Death: A Meta-Analysis of Prospective Cohort Studies. Medicine. 2016;95(6):e2815.\u003c/li\u003e\n\u003cli\u003eLi W, Shen C, Kong W, Zhou X, Fan H, Zhang Y, et al. Association between the triglyceride glucose-body mass index and future cardiovascular disease risk in a population with Cardiovascular-Kidney-Metabolic syndrome stage 0-3: a nationwide prospective cohort study. Cardiovascular diabetology. 2024;23(1):292.\u003c/li\u003e\n\u003cli\u003eKakadiaris IA, Vrigkas M, Yen AA, Kuznetsova T, Budoff M, Naghavi M. Machine Learning Outperforms ACC / AHA CVD Risk Calculator in MESA. Journal of the American Heart Association. 2018;7(22):e009476.\u003c/li\u003e\n\u003cli\u003eNarain R, Saxena S, Goyal AK. Cardiovascular risk prediction: a comparative study of Framingham and quantum neural network based approach. Patient preference and adherence. 2016;10:1259-70.\u003c/li\u003e\n\u003cli\u003eXi Y, Wang H, Sun N. Machine learning outperforms traditional logistic regression and offers new possibilities for cardiovascular risk prediction: A study involving 143,043 Chinese patients with hypertension. Frontiers in cardiovascular medicine. 2022;9:1025705.\u003c/li\u003e\n\u003cli\u003eSadatsafavi M, Adibi A, Puhan M, Gershon A, Aaron SD, Sin DD. Moving beyond AUC: decision curve analysis for quantifying net benefit of risk prediction models. The European respiratory journal. 2021;58(5).\u003c/li\u003e\n\u003cli\u003eWang K, Tian J, Zheng C, Yang H, Ren J, Liu Y, et al. Interpretable prediction of 3-year all-cause mortality in patients with heart failure caused by coronary heart disease based on machine learning and SHAP. Computers in biology and medicine. 2021;137:104813.\u003c/li\u003e\n\u003cli\u003eHan Y, Wang S. Disability risk prediction model based on machine learning among Chinese healthy older adults: results from the China Health and Retirement Longitudinal Study. Frontiers in public health. 2023;11:1271595.\u003c/li\u003e\n\u003cli\u003eCong X, Liu S, Wang W, Ma J, Li J. Combined consideration of body mass index and waist circumference identifies obesity patterns associated with risk of stroke in a Chinese prospective cohort study. BMC public health. 2022;22(1):347.\u003c/li\u003e\n\u003cli\u003ePowell-Wiley TM, Poirier P, Burke LE, Despr\u0026eacute;s JP, Gordon-Larsen P, Lavie CJ, et al. Obesity and Cardiovascular Disease: A Scientific Statement From the American Heart Association. Circulation. 2021;143(21):e984-e1010.\u003c/li\u003e\n\u003cli\u003eGan Y, Gong Y, Tong X, Sun H, Cong Y, Dong X, et al. Depression and the risk of coronary heart disease: a meta-analysis of prospective cohort studies. BMC psychiatry. 2014;14:371.\u003c/li\u003e\n\u003cli\u003eBarth J, Schumacher M, Herrmann-Lingen C. Depression as a risk factor for mortality in patients with coronary heart disease: a meta-analysis. Psychosomatic medicine. 2004;66(6):802-13.\u003c/li\u003e\n\u003cli\u003eLichtman JH, Froelicher ES, Blumenthal JA, Carney RM, Doering LV, Frasure-Smith N, et al. Depression as a risk factor for poor prognosis among patients with acute coronary syndrome: systematic review and recommendations: a scientific statement from the American Heart Association. Circulation. 2014;129(12):1350-69.\u003c/li\u003e\n\u003cli\u003eNicholson A, Kuper H, Hemingway H. Depression as an aetiologic and prognostic factor in coronary heart disease: a meta-analysis of 6362 events among 146 538 participants in 54 observational studies. European heart journal. 2006;27(23):2763-74.\u003c/li\u003e\n\u003cli\u003eWang L, Lee Y, Wu Y, Zhang X, Jin C, Huang Z, et al. A prospective study of waist circumference trajectories and incident cardiovascular disease in China: the Kailuan Cohort Study. The American journal of clinical nutrition. 2021;113(2):338-47.\u003c/li\u003e\n\u003cli\u003eLiu R, Dang S, Zhao Y, Yan H, Han Y, Mi B. Long-term waist circumference trajectories and body mass index with all-cause mortality in older Chinese adults: a prospective nationwide cohort study. Archives of public health = Archives belges de sante publique. 2022;80(1):94.\u003c/li\u003e\n\u003cli\u003eFauchier G, Bisson A, Bodin A, Herbert J, Semaan C, Angoulvant D, et al. Metabolically healthy obesity and cardiovascular events: A nationwide cohort study. Diabetes, obesity \u0026amp; metabolism. 2021;23(11):2492-501.\u003c/li\u003e\n\u003cli\u003eKrittanawong C, Johnson KW, Rosenson RS, Wang Z, Aydar M, Baber U, et al. Deep learning for cardiovascular medicine: a practical primer. European heart journal. 2019;40(25):2058-73.\u003c/li\u003e\n\u003cli\u003eItchhaporia D. Artificial intelligence in cardiology. Trends in cardiovascular medicine. 2022;32(1):34-41.\u003c/li\u003e\n\u003cli\u003eMotwani M, Dey D, Berman DS, Germano G, Achenbach S, Al-Mallah MH, et al. Machine learning for prediction of all-cause mortality in patients with suspected coronary artery disease: a 5-year multicentre prospective registry analysis. European heart journal. 2017;38(7):500-7.\u003c/li\u003e\n\u003cli\u003eLo-Ciganic WH, Huang JL, Zhang HH, Weiss JC, Wu Y, Kwoh CK, et al. Evaluation of Machine-Learning Algorithms for Predicting Opioid Overdose Risk Among Medicare Beneficiaries With Opioid Prescriptions. JAMA network open. 2019;2(3):e190968.\u003c/li\u003e\n\u003cli\u003eKerr KF, Brown MD, Zhu K, Janes H. Assessing the Clinical Impact of Risk Prediction Models With Decision Curves: Guidance for Correct Interpretation and Appropriate Use. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2016;34(21):2534-40.\u003c/li\u003e\n\u003cli\u003eRudin C. Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead. Nature machine intelligence. 2019;1(5):206-15.\u003c/li\u003e\n\u003cli\u003eElshawi R, Al-Mallah MH, Sakr S. On the interpretability of machine learning-based model for predicting hypertension. BMC medical informatics and decision making. 2019;19(1):146.\u003c/li\u003e\n\u003cli\u003eSterne JA, White IR, Carlin JB, Spratt M, Royston P, Kenward MG, et al. Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ (Clinical research ed). 2009;338:b2393.\u003c/li\u003e\n\u003cli\u003eSiontis GC, Tzoulaki I, Siontis KC, Ioannidis JP. Comparisons of established risk prediction models for cardiovascular disease: systematic review. BMJ (Clinical research ed). 2012;344:e3318.\u003c/li\u003e\n\u003cli\u003eCameron AJ, Magliano DJ, S\u0026ouml;derberg S. A systematic review of the impact of including both waist and hip circumference in risk models for cardiovascular diseases, diabetes and mortality. Obesity reviews : an official journal of the International Association for the Study of Obesity. 2013;14(1):86-94.\u003c/li\u003e\n\u003cli\u003eChen Z, Iona A, Parish S, Chen Y, Guo Y, Bragg F, et al. Adiposity and risk of ischaemic and haemorrhagic stroke in 0\u0026middot;5 million Chinese men and women: a prospective cohort study. The Lancet Global health. 2018;6(6):e630-e40.\u003c/li\u003e\n\u003cli\u003eHu L, Huang X, You C, Li J, Hong K, Li P, et al. Prevalence of overweight, obesity, abdominal obesity and obesity-related risk factors in southern China. PloS one. 2017;12(9):e0183934.\u003c/li\u003e\n\u003cli\u003eDespr\u0026eacute;s JP. Waist circumference as a vital sign in cardiology 20 years after its initial publication in the American Journal of Cardiology. The American journal of cardiology. 2014;114(2):320-3.\u003c/li\u003e\n\u003cli\u003eYusuf S, Joseph P, Rangarajan S, Islam S, Mente A, Hystad P, et al. Modifiable risk factors, cardiovascular disease, and mortality in 155 722 individuals from 21 high-income, middle-income, and low-income countries (PURE): a prospective cohort study. Lancet (London, England). 2020;395(10226):795-808.\u003c/li\u003e\n\u003cli\u003eExternal validation of clinical prediction models using big datasets from e-health records or IPD meta-analysis: opportunities and challenges. BMJ (Clinical research ed). 2019;365:l4379.\u003c/li\u003e\n\u003cli\u003eGoldstein BA, Navar AM, Carter RE. Moving beyond regression techniques in cardiovascular risk prediction: applying machine learning to address analytic challenges. European heart journal. 2017;38(23):1805-14.\u003c/li\u003e\n\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":"bmc-geriatrics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bgtc","sideBox":"Learn more about [BMC Geriatrics](http://bmcgeriatr.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bgtc/default.aspx","title":"BMC Geriatrics","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Major Adverse Cardiovascular Events, Risk Prediction, Machine Learning, Chinese Population, Central Obesity, Depression, CHARLS, Nomogram","lastPublishedDoi":"10.21203/rs.3.rs-6906133/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6906133/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e\u003cp\u003eCardiovascular disease (CVD) poses a substantial health burden on China's aging population. Existing cardiovascular risk models often perform poorly in older Chinese adults and rarely integrate geriatric-specific non-traditional factors. This study aimed to develop and validate machine learning-based models incorporating both traditional and non-traditional risk factors for predicting major adverse cardiovascular events (MACE) among older Chinese adults.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eData from 4,580 participants aged\u0026thinsp;\u0026ge;\u0026thinsp;60 years without baseline MACE were obtained from the China Health and Retirement Longitudinal Study (CHARLS, 2011\u0026ndash;2018). Incident MACE (myocardial infarction or stroke) was self-reported during a median follow-up of approximately 7 years. Candidate predictors included demographics, health behaviors, clinical measures, anthropometric indices, biomarkers, depressive symptoms (CES-D score), and functional limitations (Activities of Daily Living, ADL). Missing data were handled via Multiple Imputation by Chained Equations (MICE, 5 imputations). Logistic Regression (LR), Random Forest (RF), and XGBoost models were trained using stratified 70/30 splits for training and testing sets. Hyperparameter tuning employed a grid search with limited complexity. Model performance was evaluated by discrimination (AUC), calibration (Brier score and calibration plots), and clinical utility (Decision Curve Analysis, DCA). Exploratory non-linear relationships were assessed using generalized additive models (GAMs). Results were benchmarked against the Framingham Risk Score (FRS-CVD), and an LR-based nomogram was developed.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eIncident MACE occurred in 28.7% of participants. The LR model demonstrated the highest discrimination (mean AUC\u0026thinsp;=\u0026thinsp;0.649), closely followed by XGBoost (mean AUC\u0026thinsp;=\u0026thinsp;0.645); both significantly outperformed RF (mean AUC\u0026thinsp;=\u0026thinsp;0.632) and the FRS-CVD benchmark (mean AUC\u0026thinsp;=\u0026thinsp;0.504). LR and XGBoost models showed good calibration and superior net benefit in DCA. Significant independent predictors in the LR model included hypertension history (OR\u0026thinsp;=\u0026thinsp;1.85), diabetes (OR\u0026thinsp;=\u0026thinsp;1.34), age (OR\u0026thinsp;=\u0026thinsp;1.02/year), systolic blood pressure (OR\u0026thinsp;=\u0026thinsp;1.006/mmHg), high education level (OR\u0026thinsp;=\u0026thinsp;1.60), waist circumference (OR\u0026thinsp;=\u0026thinsp;1.01/cm), depressive symptoms (CES-D score, OR\u0026thinsp;=\u0026thinsp;1.03/point), and ADL limitations (OR\u0026thinsp;=\u0026thinsp;1.11/limitation). GAM analysis revealed significant non-linear relationships for age and waist circumference.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e\u003cp\u003eMachine learning models integrating traditional and non-traditional factors effectively predict MACE risk in older Chinese adults, outperforming the standard FRS. Central obesity, depressive symptoms, and functional impairments were significant predictors, underscoring the importance of holistic cardiovascular risk assessment in geriatric populations. The developed nomogram offers a practical clinical tool pending external validation.\u003c/p\u003e","manuscriptTitle":"Machine Learning Prediction of MACE in Older Chinese Adults Integrating Traditional and Geriatric-Specific Risk Factors: A CHARLS Cohort Analysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-07-22 16:15:16","doi":"10.21203/rs.3.rs-6906133/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewersInvited","content":"","date":"2025-07-15T09:35:00+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-06-20T11:57:38+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-06-20T07:15:10+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-06-20T07:12:36+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Geriatrics","date":"2025-06-16T13:33:32+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-geriatrics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bgtc","sideBox":"Learn more about [BMC Geriatrics](http://bmcgeriatr.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bgtc/default.aspx","title":"BMC Geriatrics","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"00d67a38-387e-444a-ab70-4054d813d78b","owner":[],"postedDate":"July 22nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-07-22T16:15:17+00:00","versionOfRecord":[],"versionCreatedAt":"2025-07-22 16:15:16","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6906133","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6906133","identity":"rs-6906133","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}