A Predictive Model for Headache-Related Depression in Middle- Aged and Older Women Incorporating Individual Treatment Effects

A Predictive Model for Headache-Related Depression in Middle- Aged and Older Women Incorporating Individual Treatment Effects | 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 A Predictive Model for Headache-Related Depression in Middle- Aged and Older Women Incorporating Individual Treatment Effects Xiaohui Wen, Zhijun Lin, Yihui Qian, Lin Wang, Jingtong Zhou, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7517256/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Depression commonly co-occurs with headache, especially among middle-aged and older women, increasing symptom burden and healthcare utilization. To enable early identification and targeted intervention, we developed a machine learning–based risk prediction model and deployed an interactive Shiny web application. Data from 1,930 individuals aged ≥ 45 years with self-reported headache in the 2015 wave of the China Health and Retirement Longitudinal Study were analyzed and randomly split into training and testing subsets. Depressive symptoms were assessed using a validated instrument, and seven predictors spanning clinical and lifestyle domains were selected. Ten machine learning algorithms were compared using discrimination, calibration, and decision curve analysis, with temporal validation on the 2011 CHARLS wave. The gradient boosting machine model achieved the best performance (area under the curve 0.823 training, 0.724 test, 0.785 temporal validation) and favorable calibration. Key predictors included hope, sleep duration, life satisfaction, and self-rated health. Individualized treatment effect analysis identified approximately 5% of participants most likely to benefit from interventions enhancing life satisfaction, and the Qini curve confirmed heterogeneity in treatment effects (uplift area under the curve 55.8). Targeting interventions based on predicted risk can achieve greater benefits than random allocation, supporting precision mental health strategies. This model and its Shiny tool facilitate early identification and tailored intervention for high-risk middle-aged and older women with headache, warranting prospective validation. Headache Depression Middle-aged and Older women Risk Prediction Model Machine Learning Shiny Application Mental Health ITE Figures Figure 1 Figure 2 Figure 3 1. Introduction Headache is one of the most prevalent neurological symptoms worldwide, with a notably higher burden in women than in men. Epidemiological studies indicate that active headache disorders affect approximately 52.0% of the global population, with a prevalence of 57.8% in females compared to 44.4% in males ( 1 ). The lifetime prevalence of headache remains consistently higher among women. In young women, hormonal fluctuations and menstrual cycles often influence headache frequency and severity.( 2 – 6 ). Depression has emerged as a major global public health concern, substantially impairing psychological well-being and quality of life. It has been linked to increased risks of suicidal behavior, stroke, cognitive decline, and neurodegenerative diseases such as Alzheimer’s disease( 7 ), A bidirectional relationship between chronic pain and depression is well documented, and women with chronic headache are particularly vulnerable to depressive symptoms, which may further exacerbate neurological and psychiatric outcomes.( 8 , 9 ). Despite this, research specifically examining depression risk in middle-aged and older women with headache remains limited, However, research specifically focusing on depression risk in middle-aged and older women with headache remains limited. .( 8 – 10 ). In this study, we leveraged data from the 2015 wave of the China Health and Retirement Longitudinal Study (CHARLS) to develop an ML-based prediction model for depression risk in middle-aged and older women with headache. We systematically compared ten widely used ML algorithms—including XGBoost, CatBoost, random forest, and neural networks—using ten-fold cross-validation to ensure robustness and mitigate overfitting. The model’s generalizability was assessed through temporal validation with the 2011 CHARLS cohort. To enhance clinical applicability and promote precision prevention, we incorporated Individual Treatment Effect (ITE) estimation into the analytical pipeline, enabling the identification of subgroups most likely to benefit from targeted interventions and the exploration of heterogeneity in predicted depression risk across diverse patient profiles. 2. Methods 2.1 Data Sources and Study Population This study utilized data from a large-scale waves of the CHARLS ( 11 ) for model development and validation. The training and internal validation datasets were derived from the 2015 wave of CHARLS, a nationally representative longitudinal survey of middle-aged and older Chinese adults. Female participants aged 45 years or older who reported chronic headache symptoms were included. For temporal external validation, data from the 2011 wave of CHARLS were used. Inclusion criteria were: ( 1 ) female gender; ( 2 ) age ≥ 45 years; ( 3 ) self-reported presence of headache symptoms; and ( 4 ) complete data on all predictor variables. Exclusion criteria included: ( 1 ) a history of severe neurological or psychiatric disorders; and ( 2 ) missing values in key variables. 2.2 Data Collection Dimensions 2.2.1 Depression and Headache In this study, the presence of a significant risk of depression was used as the target outcome variable. The definition and standardization of depressive status were based on the depression screening tools applied in each respective dataset. In the CHARLS dataset, depressive symptoms were assessed using the validated 10-item Center for Epidemiologic Studies Depression Scale (CES-D 10)( 12 ). The CES-D 10 consists of 10 items, each scored on a 4-point Likert scale from 0 (rarely or none of the time) to 3 (most or all of the time), with total scores ranging from 0 to 30. Higher scores indicate more severe depressive symptoms. Based on previous studies( 13 ), a score of ≥ 10 was used to define clinically significant depressive symptoms. For the sake of clarity, the term "depression" is used throughout this study to refer to depressive symptoms of varying severity. As for headache assessment, the CHARLS dataset contains only a single self-reported item regarding headache symptoms and does not further distinguish between specific headache subtypes such as chronic headache, migraine, or tension-type headache. Therefore, in this study, headache was treated as a broad clinical concept. Given prior epidemiological evidence indicating that chronic headache, migraine, and tension-type headache are the most prevalent forms among middle-aged and older adults( 1 ), it is reasonable to assume that individuals in CHARLS who reported headache symptoms potentially represent a heterogeneous group comprising various headache subtypes. 2.2.1 Demographic Characteristics This study assessed a range of sociodemographic variables, including age; sex (male/female); marital status (married/unmarried); place of residence (urban/rural); educational attainment (no formal education, primary school, junior high school, high school or above); and retirement status (yes/no). Subjective health status and life satisfaction were evaluated using a five-point Likert scale ranging from “very good” to “very poor.” Multisite pain was defined as the presence of pain in more than five body regions, based on the sample median. Hopefulness was categorized by its frequency over the past week: “<1 day,” “1–2 days,” “3–4 days,” or “5–7 days.” Behavioral factors included smoking status, alcohol consumption, and history of falls (all dichotomous), as well as average nighttime sleep duration over the past month, recorded in hours. 2.2.2 Health Status Based on previous literature and clinical expertise, we included several health-related variables potentially associated with depression. These included self-reported history of chronic conditions such as hypertension, cancer, dyslipidemia, heart disease, stroke, arthritis, psychiatric disorders or rheumatism, liver disease, kidney disease, gastrointestinal disorders, and asthma. Additionally, self-rated health and life satisfaction were assessed as subjective perceptions of overall health and life quality, categorized as "very good," "good," "fair," "poor," or "very poor." Multisite pain was defined as reporting pain in more than five body regions and treated as a binary variable. For physical examination data, the mean value of repeated measurements was used. 2.3 Preprocessing and Model Construction As illustrated in Fig. 1 , data preprocessing and model development followed a structured workflow. To prevent model overfitting due to multicollinearity and to minimize noise interference, we performed variable screening using Pearson correlation heatmaps and variance inflation factor (VIF) analysis (see Supplementary Figure S1 ). Variables with high pairwise correlations (|r| >0.8) and severe multicollinearity (VIF > 5) were excluded to reduce computational complexity and eliminate potential confounding interactions. The exclusion criteria for variables included:( 1 ) variables with more than 20% missing data, ( 2 ) absolute Pearson correlation coefficients greater than 0.8, ( 3 ) VIF values exceeding 5, and ( 4 ) variables deemed not significantly associated with depression risk based on previous studies and clinical expertise. Ultimately, 37 variables were retained for further analysis, encompassing sociodemographic characteristics, health status, medical history, lifestyle behaviors, and physiological measurements. These variables included: Age, Marital status(marry), Residence༈rural༉, Self-rated health༈srh༉, Hope for the future༈hope༉, Llife satisfaction༈satlife༉, History of falls༈falldown༉, Retirement status༈retire༉, Multisite pain༈mulpain༉,Education level༈edu༉, Disability, Hypertension༈hibpe༉, Diabetes༈diabe༉, Cancer༈cancre༉, Arthritis༈arthre༉, Chronic lung disease༈lunge༉, Cardiovascular diseases༈hearte༉, Stroke, dyslipidemia༈dyslipe༉,Liver disease༈livere༉, Kidney disease༈kidneye༉,Gastrointestinal disorders༈digeste༉, Asthma༈asthmae༉, Memory-related disorders༈memrye༉, Alcohol use༈drinkev༉,Smoking status༈smoken༉Health insurance coverage(ins), Systolic blood pressure(systo), Diastolic blood pressure(diasto), Pulse rate(pulse), Grip strength of left(lgrip) and Right hand(rgrip), Sleep duration(sleep), and Lung function(puff) (Table 1 ). Table 1 Baseline characteristics table of the raw dataset Variables Total (n = 1930) Non-depression (n = 569) Depression (n = 1361) p Feel hopeful about the Future, n (%) < 0.001 Rarely or never (< 1 day) 776 ( 40 ) 164 ( 29 ) 612 (45) Occasional (1–2 days) 300 ( 16 ) 66 ( 12 ) 234 ( 17 ) Sometimes (3–4 days) 334 ( 17 ) 78 ( 14 ) 256 ( 19 ) Almost always (5–7 days) 520 ( 27 ) 261 (46) 259 ( 19 ) Retirement status n (%) < 0.001 No 1820 (94) 516 (91) 1304 (96) Yes 110 ( 6 ) 53 ( 9 ) 57 ( 4 ) History of falls, n (%) < 0.001 No 1326 (69) 440 (77) 886 (65) Yes 604 ( 31 ) 129 ( 23 ) 475 ( 35 ) Marital status, n (%) 0.015 Unmarried 347 ( 18 ) 83 ( 15 ) 264 ( 19 ) Married 1583 (82) 486 (85) 1097 (81) Life satisfaction, n (%) < 0.001 Not at all satisfied 92 ( 5 ) 9 ( 2 ) 83 ( 6 ) Not very satisfied 252 ( 13 ) 34 ( 6 ) 218 ( 16 ) Somewhat satisfied 940 (49) 263 (46) 677 (50) Very satisfied 539 ( 28 ) 221 ( 39 ) 318 ( 23 ) Completely satisfied 107 ( 6 ) 42 ( 7 ) 65 ( 5 ) Multisite pain, n (%) < 0.001 No 612 ( 32 ) 222 ( 39 ) 390 ( 29 ) Yes 1318 (68) 347 (61) 971 (71) Place of residence, n (%) < 0.001 rural 592 ( 31 ) 219 ( 38 ) 373 ( 27 ) Urban 1338 (69) 350 (62) 988 (73) Educational attainment, n (%) < 0.001 Elementary school or below 1212 (63) 321 (56) 891 (65) Middle school 452 ( 23 ) 146 ( 26 ) 306 ( 22 ) High school 188 ( 10 ) 65 ( 11 ) 123 ( 9 ) Above high school 78 ( 4 ) 37 ( 7 ) 41 ( 3 ) Disability, n (%) < 0.001 No 1070 (55) 363 (64) 707 (52) Yes 860 (45) 206 ( 36 ) 654 (48) Self perceived health status, n (%) < 0.001 Very good 298 ( 15 ) 50 ( 9 ) 248 ( 18 ) Poor 718 ( 37 ) 167 ( 29 ) 551 ( 40 ) Fair 802 (42) 298 (52) 504 ( 37 ) Good 70 ( 4 ) 33 ( 6 ) 37 ( 3 ) Very good 42 ( 2 ) 21 ( 4 ) 21 ( 2 ) Hypertension n (%) 0.859 No 1166 (60) 346 (61) 820 (60) Yes 764 ( 40 ) 223 ( 39 ) 541 ( 40 ) Diabetes mellitus, n (%) 0.524 No 1645 (85) 490 (86) 1155 (85) Yes 285 ( 15 ) 79 ( 14 ) 206 ( 15 ) Cancer, n (%) 0.835 No 1883 (98) 554 (97) 1329 (98) Yes 47 ( 2 ) 15 ( 3 ) 32 ( 2 ) Lunge disease, n (%) 0.224 No 1560 (81) 470 (83) 1090 (80) Yes 370 ( 19 ) 99 ( 17 ) 271 ( 20 ) Heart disease, n (%) 0.128 No 1331 (69) 407 (72) 924 (68) Yes 599 ( 31 ) 162 ( 28 ) 437 ( 32 ) History of stroke n (%) 0.875 No 1855 (96) 548 (96) 1307 (96) Yes 75 ( 4 ) 21 ( 4 ) 54 ( 4 ) Arthritis, n (%) < 0.001 No 610 ( 32 ) 214 ( 38 ) 396 ( 29 ) Yes 1320 (68) 355 (62) 965 (71) Dyslipidemia, n (%) 0.853 No 1465 (76) 434 (76) 1031 (76) Yes 465 ( 24 ) 135 ( 24 ) 330 ( 24 ) Liver disease, n (%) 0.915 No 1743 (90) 515 (91) 1228 (90) Yes 187 ( 10 ) 54 ( 9 ) 133 ( 10 ) Kidney disease, n (%) 0.005 No 1635 (85) 503 (88) 1132 (83) Yes 295 ( 15 ) 66 ( 12 ) 229 ( 17 ) Digestive disease, n (%) 0.004 No 926 (48) 302 (53) 624 (46) Yes 1004 (52) 267 (47) 737 (54) Asthma, n (%) 0.113 No 1766 (92) 530 (93) 1236 (91) Yes 164 ( 8 ) 39 ( 7 ) 125 ( 9 ) Memory impairment, n (%) 0.025 No 1830 (95) 550 (97) 1280 (94) Yes 100 ( 5 ) 19 ( 3 ) 81 ( 6 ) Drinking behavior n (%) 0.19 No 1638 (85) 473 (83) 1165 (86) Yes 292 ( 15 ) 96 ( 17 ) 196 ( 14 ) Health insurance, n (%) 1 No 169 ( 9 ) 50 ( 9 ) 119 ( 9 ) Yes 1761 (91) 519 (91) 1242 (91) Smoking, behavior n (%) 0.448 No 1815 (94) 531 (93) 1284 (94) Yes 115 ( 6 ) 38 ( 7 ) 77 ( 6 ) Systolic Blood Pressure, Median (Q1,Q3) 122.5 (110, 138.5) 124.5 (111.5, 141) 122 (109, 138) 0.014 Diastolic Blood Pressure, Median (Q1,Q3) 73 (65.5, 81) 73.5 (66, 82) 72.5 (65.5, 81) 0.137 Pulse, Median (Q1,Q3) 74 (68, 81) 74.5 (68.5, 80.5) 74 (67.5, 81) 0.724 Left grip strength, Median (Q1,Q3) 22 (17.2, 27) 23.3 (18.5, 28.5) 21.8 (17, 26.5) < 0.001 Right strength, Median (Q1,Q3) 23.5 (18.9, 28.27) 24 (20, 29.2) 23 (18.1, 28) < 0.001 Pulmonary function, Median (Q1,Q3) 260 (190, 320) 270 (200, 330) 250 (190, 310) 0.017 Sleep duration, Median (Q1,Q3) 5.5 ( 4 , 7 ) 6 ( 5 , 8 ) 5 ( 4 , 7 ) < 0.001 Age, Median (Q1,Q3) 59 (51, 67) 58 (50, 66) 60 (51, 67) 0.149 To minimize bias caused by missing data, this study employed multiple imputation using the "MICE" (Multivariate Imputation by Chained Equations) package( 14 ). The predictive mean matching (PMM) method was used to generate five imputed datasets, with 50 iterations performed to ensure the stability of the imputation results. Subsequently, Z-score standardization was applied to all features to mitigate the influence of large-scale variables on model training, thereby improving model accuracy and stability. To address class imbalance in the original dataset—where 1,361 patients (70.5%) were diagnosed with depression and 569 (29.5%) were not—the Synthetic Minority Over-sampling Technique (SMOTE) was applied to the training set. Using five nearest neighbors, SMOTE generated synthetic samples with 100% oversampling of the minority class and 200% undersampling of the majority class to achieve balance. By interpolating between existing minority instances, SMOTE mitigated class bias and improved the model's sensitivity to non-depressed cases, thereby enhancing its generalizability and clinical relevance( 15 ). For data partitioning, all samples were randomly divided into training and testing sets in a 7:3 ratio. The training set was used for data preprocessing, feature selection, and model development, while the testing set was reserved for performance evaluation and model tuning. Feature selection was performed using univariate logistic regression, multivariate logistic regression (see Supplementary Table S1 ), and the Least Absolute Shrinkage and Selection Operator (LASSO) regression. Based on previous research and clinical interpretability, seven key predictors were ultimately selected: hope for the future, self-rated health, retirement status, multisite pain, history of falls, life satisfaction, and sleep duration(Fig. 2 A. B ). For model development and comparison, ten machine learning algorithms were implemented, including Neural Network, Extreme Gradient Boosting (XGBoost), Categorical Boosting (CatBoost), Logistic Regression, Random Forest, Gradient Boosting Machine (GBM), Support Vector Machine (SVM), Light Gradient Boosting Machine (LightGBM), k-Nearest Neighbors (KNN), and Adaptive Boosting (AdaBoost). Hyperparameters for each model were optimized using a grid search strategy( 16 ).. Model training was conducted on the training set using 10-fold cross-validation combined with resampling techniques to reduce overfitting. Model performance was subsequently evaluated on both the internal testing set and the temporally separated external validation set derived from the 2011 wave of CHARLS. This approach enabled assessment of the models’ robustness and generalizability over time. Model performance was assessed from three key perspectives: discrimination, calibration, and clinical utility. Discrimination was evaluated using the area under the receiver operating characteristic curve(AUC) and the F1 score, Precision-Recall curves. Calibration was assessed through calibration plots and Brier scores to determine the agreement between predicted probabilities and actual outcomes. Clinical utility was evaluated using Decision Curve Analysis (DCA), which measures the net clinical benefit across a range of risk thresholds. In addition, confusion matrix metrics—including sensitivity, specificity, accuracy, and precision—were comprehensively analyzed to further evaluate model performance. 2.4 Machine Learning Algorithms This study employed R version 4.4.3 to develop ten widely used classification models: XGBoost, CatBoost, LightGBM, Random Forest, GBM, Neural Network, Logistic Regression, SVM, KNN, and AdaBoost. Hyperparameter tuning for each model was conducted using grid search, combined with 10-fold cross-validation repeated five times to assess model stability and generalization performance. Each model offers distinct advantages: Logistic Regression is straightforward and interpretable, suitable for linear classification problems.( 17 ) SVM performs robustly in high-dimensional spaces and is effective for nonlinear data. GBM and XGBoost enhance performance by integrating weak learners, offering strong nonlinear modeling capabilities( 18 ). Neural Networks can automatically extract features, making them adept at complex pattern recognition tasks( 19 ), Random Forest reduces overfitting risk by aggregating multiple decision trees( 20 ). KNN classifies based on instances without requiring a training phase( 21 ). AdaBoost iteratively improves weak classifiers' performance( 22 ). LightGBM and CatBoost are particularly efficient in handling large-scale data and categorical features, with high training efficiency( 23 ). performance for specific predictive tasks, aiding in the selection of the optimal model. Grid search, a hyperparameter optimization technique, exhaustively searches through a predefined set of hyperparameter combinations to identify the configuration yielding the best model performance. This method systematically evaluates each combination using cross-validation, ensuring a thorough exploration of the hyperparameter space( 16 ), The hyperparameter tuning methods for the models are as follows: Support Vector Machine (SVM), GBM, Neural Network, Random Forest, XGBoost, KNN, and AdaBoost were tuned using grid search within predefined parameter ranges; whereas CatBoost and LightGBM were trained using parameters defined by our own specifications. In this study, all models underwent hyperparameter tuning using 10-fold cross-validation repeated five times (“10-fold CV with 5 repeats”) combined with grid search. The optimal hyperparameters identified are as follows: SVM sigma = 0.001, C = 0.5 GBM : n.trees = 100, interaction.depth = 3, shrinkage = 0.01, n.minobsinnode = 5 Neural Network : size = 5, decay = 0.6 Random Forest : mtry = 5 XGBoost : nrounds = 10, max_depth = 5, eta = 0.001, gamma = 0.5, colsample_bytree = 0.5, min_child_weight = 1, subsample = 0.6 KNN : kmax = 12, distance = 1, kernel = "optimal" AdaBoost : mfinal = 2, maxdepth = 2, coeflearn = "Zhu" LightGBM : learning_rate = 1.0, min_data = 1, num_threads = 2, force_col_wise = TRUE, CatBoost : iterations = 100, depth = 5, learning_rate = 0.03, border_count = 32 2.5 SHAP Explainability Analysis We applied the Shapley Additive Explanations (SHAP) method( 24 ),to quantify each feature’s contribution to our model predictions. Positive SHAP values indicate factors that increase predicted risk, while negative values denote protective effects. After examining seven key predictors via their SHAP distributions, we implemented an interactive Shiny application so clinicians can explore these feature attributions and obtain case-specific explanations on demand. 2.6 Individual Treatment Effect Estimation To investigate heterogeneity in intervention effects, we applied a counterfactual framework and uplift modeling to estimate individual treatment effects (ITEs) on depression risk. Binary treatment assignments were defined based on modifiable psychosocial variables. Life satisfaction was selected as the intervention variable due to its modifiability and clinical relevance. Participants reporting "Somewhat satisfied," "Very satisfied," or "Completely satisfied" were classified as the high life satisfaction (treated) group, whereas those reporting "Not very satisfied" or "Not at all satisfied" were classified as the low life satisfaction (control) group. ITEs were estimated using a two-model approach with random forest classifiers. Separate outcome models were trained for the treated and control groups using the same covariates as in the primary prediction model. For each individual, the difference between predicted depression risk under treatment and control conditions provided the ITE. To identify individuals most likely to benefit from improvements in life satisfaction, we extracted the top 5% of ITE scores (highest estimated benefit) as the high-benefit subgroup. Descriptive statistics (means and standard deviations) were calculated for baseline numeric variables and compared between this subgroup and the overall population. Differences in means were used to rank variables contributing to treatment heterogeneity. The clinical utility of individualized predictions was evaluated using uplift modeling and Qini curves. Individuals were ranked by their ITE estimates, and cumulative incremental benefit was plotted against the cumulative population fraction. The Qini coefficient (area under the Qini curve, uplift AUC) summarized model discriminative performance, with higher values indicating better identification of individuals likely to benefit from interventions. Subgroup-level treatment benefits were further quantified by reporting cumulative uplift within specific strata, such as the top 5% and top 20% of predicted beneficiaries. 2.7. Statistical analysis Continuous variables are reported as mean ± SD (range) or median (IQR) and compared using Student’s t-test for normally distributed data) or the Wilcoxon rank-sum test for nonparametric data. Categorical variables are expressed as counts (percentages) and compared using Pearson’s chi-square test) or Fisher’s exact test when expected cell counts are small. Two-sided P < 0.05 was considered statistically significant. All analyses were performed in R version 4.4.0 (released April 24, 2024). 3. Results 3.1 Patient Characteristics From the 2015 wave of the CHARLS (n = 21,038), 11,006 women who reported headache symptoms were identified. After excluding individuals with missing or implausible values and those not meeting inclusion criteria, a total of 1,930 middle-aged and older women were included in the final analysis. Among them, 1,361 (70.5%) had CESD-10 scores ≥ 10 and were classified as having depressive symptoms. The median age of the cohort was 59 years. In terms of sociodemographic characteristics, 94% were retired and 6% remained employed; 82% were married and 8% were unmarried; 69% resided in urban areas and 31% in rural or township areas. Educational attainment was distributed as follows: primary school or less (63%), secondary school (23%), high school (10%), and tertiary education or above (4%) (Table 1 ). Significant differences between depressed and non-depressed participants were observed across multiple domains, including psychosocial factors, socioeconomic status, health behaviors, functional health, comorbidity burden, physical performance, and physiological indicators. For temporal external validation, a separate sample of 1,518 women from the 2011 CHARLS wave, selected using analogous inclusion and exclusion criteria, was used as a time-series test set. The 2011 CHARLS cohort and the 2015 cohort differed in life satisfaction (n (%)), multisite pain, educational attainment, self-perceived health status (n (%)), comorbidities, and age (see Supplementary Table S2), thereby meeting the criteria for external validation. 3.2 Multi-Model Comparative Analysis We systematically evaluated the predictive performance of ten machine learning models—XGBoost, CatBoost, LightGBM, Random Forest, GBM, Neural Network, Logistic Regression, SVM, KNN, and AdaBoost—across training, internal test, and temporal validation datasets. Comprehensive metrics including AUC, precision-recall AUC (PR-AUC), Brier score, accuracy, sensitivity, specificity, precision, and F1 score were used for robust assessment (Table 2 ; Fig. 2 ).Among all models, GBM consistently demonstrated superior overall performance and clinical applicability. It achieved AUCs of 0.823, 0.724, and 0.785 and PR-AUCs of 0.818, 0.855, and 0.901 in the training, test, and temporal validation sets, respectively, highlighting its excellent discrimination and strong capability in handling imbalanced data. GBM’s Brier scores were stably low (around 0.17–0.18), reflecting good calibration. Notably, GBM balanced sensitivity (0.812 training; 0.73 test; 0.778 validation) and specificity (0.71 training; 0.631 test; 0.668 validation) effectively, with F1 scores consistently above 0.77, indicating reliable positive predictive performance. Table 2 Detailed performance metrics of multiple machine learning models for predicting depression risk in patients with headache across training, testing, and validation datasets. Training set Model Threshold Accuracy Sensitivity Specificity Precision F1 Brierscore PR-AUC Logistic 0.525887 0.671 0.652 0.691 0.688 0.67 0.21484 0.713 SVM 0.542945 0.698 0.669 0.729 0.72 0.694 0.200852 0.746 GBM 0.492458 0.762 0.812 0.71 0.745 0.777 0.173917 0.818 NeuralNetwork 0.548343 0.68 0.599 0.763 0.726 0.657 0.2075 0.742 RandomForest 0.493 1 1 1 1 1 0.027376 1 Xgboost 0.512044 0.722 0.737 0.707 0.724 0.731 0.211511 0.79 KNN 0.545801 0.83 0.826 0.835 0.839 0.833 0.137648 0.921 Adaboost 0.783073 0.664 0.488 0.848 0.77 0.597 0.240072 0.707 LightGBM 0.476093 0.75 0.787 0.711 0.74 0.763 0.175359 0.798 CatBoost 0.620233 0.737 0.794 0.678 0.72 0.755 0.24324 0.803 Test set Model Threshold Accuracy Sensitivity Specificity Precision F1 Brierscore PR-AUC Logistic 0.447533 0.724 0.793 0.557 0.812 0.803 0.205221 0.855 SVM 0.509945 0.689 0.723 0.608 0.817 0.767 0.203391 0.837 GBM 0.687687 0.701 0.73 0.631 0.827 0.776 0.177467 0.855 NeuralNetwork 0.551785 0.63 0.573 0.767 0.856 0.686 0.208239 0.853 RandomForest 0.805 0.585 0.488 0.818 0.867 0.625 0.184902 0.836 Xgboost 0.552936 0.683 0.692 0.659 0.831 0.755 0.208459 0.85 KNN 0.613803 0.605 0.587 0.648 0.801 0.678 0.21559 0.805 Adaboost 0.5 0.65 0.683 0.568 0.793 0.734 0.230642 0.776 LightGBM 0.753478 0.645 0.622 0.699 0.833 0.712 0.196562 0.817 CatBoost 0.639369 0.716 0.763 0.602 0.823 0.792 0.199979 0.86 temporal validation set Model Threshold Accuracy Sensitivity Specificity Precision F1 Brierscore PR-AUC Logistic 0.510353 0.706 0.686 0.761 0.886 0.773 0.205221 0.904 SVM 0.496642 0.735 0.747 0.702 0.872 0.805 0.203391 0.893 GBM 0.661602 0.748 0.778 0.668 0.864 0.819 0.177467 0.901 NeuralNetwork 0.474907 0.719 0.719 0.72 0.874 0.789 0.208239 0.904 RandomForest 0.711 0.688 0.676 0.72 0.867 0.76 0.184902 0.88 Xgboost 0.536153 0.727 0.749 0.668 0.859 0.8 0.208459 0.894 KNN 0.651937 0.632 0.606 0.702 0.846 0.706 0.21559 0.851 Adaboost 0.5 0.668 0.644 0.734 0.867 0.739 0.230642 0.838 LightGBM 0.753478 0.672 0.67 0.678 0.849 0.749 0.196562 0.827 CatBoost 0.626773 0.742 0.773 0.661 0.86 0.814 0.199979 0.894 Random Forest, despite near-perfect metrics in the training set (accuracy and F1 of 1.0), showed significant performance degradation on the test and validation sets, indicative of overfitting and limiting its clinical utility. XGBoost and LightGBM also demonstrated robust discrimination with AUCs and PR-AUCs slightly below GBM, but exhibited somewhat higher Brier scores and less consistent calibration across datasets. CatBoost showed comparable discrimination (AUCs ~ 0.74) and PR-AUCs (up to 0.894) but had relatively poorer calibration (Brier score ~ 0.20), potentially restricting its applicability in clinical settings. The superior PR-AUC values of GBM, especially in the temporal validation cohort (0.901), underscore its strength in predicting true positives in a context where class imbalance exists. This is critical for clinical scenarios where accurate identification of at-risk patients is paramount. Overall, GBM’s combination of high discrimination, stable calibration, and balanced sensitivity and specificity across datasets positions it as the most reliable and clinically valuable predictive model for individualized risk stratification 3.3 Model Interpretation and Application Figure 3 (K.L.M) presents the results of the feature importance analysis and SHAP summary plots for the GBM model, offering a visual interpretation of how each predictor contributes to the model's output. In these plots, higher SHAP values represent a stronger positive impact on the prediction of depression risk, whereas lower or negative SHAP values indicate weaker or mitigating effects. The SHAP summary plot employs a color gradient from purple (low feature value) to yellow (high feature value), effectively illustrating both the distribution and directional influence of individual feature values on the model's predictions. Among all features, hope for the future (hope) exhibited the widest SHAP value distribution, suggesting it had the most substantial influence on model predictions. Notably, lower levels of hope (indicated by purple) were associated with higher SHAP values, implying that hopelessness is a strong risk factor for depression. Other key features—sleep duration, life satisfaction (satlife), and self-rated health (srh)—also showed significant impact, where lower values were generally associated with higher predicted risk of depression. Additionally, multisite pain and history of falls were positively associated with predicted depression risk, indicating that individuals experiencing frequent pain or with a history of falling were more likely to be classified as depressed by the model. While retirement status (retire) had a smaller overall contribution, its SHAP distribution revealed a slight tendency for non-retired individuals to be more prone to depression. Based on these key features, we developed an interactive web-based calculator using Shiny, allowing healthcare professionals to conveniently estimate individual depression risk: https://wenxiaohui.shinyapps.io/gbmforshinny/ 3.4 Individual Treatment Effect (ITE) Estimation and Stratified Response Analysis Based on prior subgroup comparison (Fig. 3 N.O), life satisfaction emerged as the most distinctive categorical variable differentiating high-risk individuals from the overall population, with substantially lower proportions of high life satisfaction levels (satlife_4 and satlife_5) observed in the high-ITE subgroup. Motivated by this finding, we further estimated the ITEs using an uplift modeling approach, treating life satisfaction as a modifiable proxy intervention (≥ 3 vs < 3). The ITE distribution (Fig. 3 P) revealed significant inter-individual heterogeneity in predicted treatment responses. A large proportion of individuals exhibited negative ITEs (i.e., reduced depression probability under higher life satisfaction), suggesting a generally beneficial effect. Notably, approximately 5% of the population demonstrated highly negative ITEs, indicating strong potential benefit from life satisfaction improvement. These individuals represent key targets for preventive intervention. This finding is further supported by the Qini curve(see Supplementary Picture S2), According to the Qini curve, the model achieved an uplift AUC of 0.558, slightly above the random level. The cumulative incremental response for the top 5% of high-score individuals was 15, and for the top 20% it was 37, indicating that the model can to some extent identify potential beneficiaries and maximize the effect of targeted interventions. 4. DISSCUION This study systematically developed and validated ten mainstream machine learning algorithms using the CHARLS database to predict depression risk among middle-aged and older women with headaches. Across the training, internal test, and temporal validation datasets, the GBM model demonstrated superior and stable predictive performance, achieving AUCs of 0.823, 0.724, and 0.785(Fig. 2 B.C.D), respectively, alongside consistent Brier scores around 0.17, indicating excellent discrimination and calibration. In contrast, although Random Forest performed exceptionally well on the training set, its marked decline in validation sets suggested overfitting, limiting its clinical utility. CatBoost, despite competitive metrics, showed poorer probability calibration, reducing its practical applicability. The GBM model’s balanced performance in accuracy, sensitivity, specificity, and F1 score highlights its rationale as the optimal model with significant clinical potential results (Table 2 ). Given its strong discrimination and calibration capabilities, GBM was ultimately selected as the optimal model and deployed through a Shiny-based web application to facilitate personalized risk estimation. Compared with traditional psychometric tools, machine learning models allow for the integration of multidimensional variables and the exploration of nonlinear and interactive effects, thereby improving predictive accuracy. Moreover, decision curve analysis (Fig. 2 J) indicated that the GBM model could guide the determination of optimal clinical intervention thresholds, enhancing the utility of community-based screening programs. Using univariate, multivariate logistic regression, and Lasso regression, we identified seven core predictors of depression: self-rated health, sense of hope, life satisfaction, history of falls, retirement status, multisite pain, and Insomnia. All of these variables have been previously linked to depression in existing literature( 8 , 9 , 25 – 27 ), and their interactive effects further enhanced the interpretability of our model. For instance, poor self-rated health is a consistent predictor of depressive symptoms( 28 ). Hope for the future and life satisfaction, as components of subjective well-being, are protective psychological resources among older adults. A Swedish study found that retirement generally improves depressive symptoms and enhances well-being, although residual occupational stress may exacerbate symptoms in certain individuals( 29 , 30 ). Additionally, the interaction between sleep disturbances and work-related stress is known to aggravate depressive outcomes( 31 ), while chronic multisite pain can increase depression risk through mechanisms such as hormonal dysregulation and impaired sleep quality. In recent years, the association between headache and depression has attracted increasing academic and clinical attention. Epidemiological studies have shown that individuals with migraine are approximately 38% more likely to experience depression than the general population( 32 ). In patients with headache disorders, female individuals exhibit a higher suicide rate( 33 ). Adults suffering from migraine in combination with depression and/or anxiety tend to have significantly higher healthcare utilization and associated costs, reflecting a substantial burden on both patients and healthcare systems( 34 ). Despite such evidence, studies specifically focusing on middle-aged and older women with headache—particularly those utilizing machine learning techniques for risk modeling—remain scarce. In the domain of predictive modeling, Zheng et al. developed a depression risk prediction model among older adults using CHARLS data and found that the random forest algorithm yielded the best performance; however, the study lacked external validation( 34 ). Shaojie Duan and colleagues expanded the migraine attack prediction model for the Chinese population by integrating traditional Chinese medicine (TCM) theory, and explored the correlations among depression, headache, and metabolic syndrome based on TCM principles( 35 ). Furthermore, Amoozegar’s study demonstrated that screening tools such as PHQ-9 and HADS perform well in migraine patients; however, optimal cut-off points for depression screening vary according to specific clinical objectives. Moreover, the high prevalence of depression among patients in headache clinics and the frequent undertreatment emphasize the necessity of scientifically validated predictive and screening models. These models are vital for the early identification of depression comorbidity in headache sufferers, enabling timely intervention and reducing disease burden( 36 ). One of the most notable findings is the central role of life satisfaction as a modifiable, high-impact predictor. While prior research has consistently linked low life satisfaction to increased depression risk in older adults, few studies have translated this association into actionable, personalized interventions( 37 – 39 ). Our ITE results demonstrated that individuals with lower life satisfaction would derive the greatest benefit from targeted psychosocial interventions, suggesting that improving life satisfaction should be prioritized as a key preventive strategy. ( 40 ). This finding aligns with literature supporting life satisfaction as a core dimension of subjective well-being and a potential protective factor against mental health decline. The model’s strong and consistent discrimination in both the internal test set and an independent temporal validation cohort highlights its generalizability, despite differences in baseline characteristics and population profiles (see Supplementary Table S2). Importantly, our study not only demonstrated statistical robustness but also addressed the practical gap between model development and clinical application by deploying an interactive Shiny-based web calculator. This tool enables real-time, individualized risk assessment using easily accessible inputs, thereby supporting integration into primary care and community health screening. Compared with existing depression prediction models, which often rely solely on sociodemographic or symptom-based screening tools, our approach offers three advantages: ( 1 ) the integration of machine learning for optimized variable selection and performance, ( 2 ) external validation to enhance credibility and applicability, and ( 3 ) the novel application of ITE analysis for precision targeting of interventions. These features strengthen the model’s translational potential in both preventive and therapeutic contexts. Nonetheless, this study has several limitations. First, although the temporal validation set represents an independent time point, it originates from the same data source as the training set, which may introduce temporal bias. Second, the retrospective nature of the study limits data reliability, as most predictors were self-reported rather than objectively measured. Third, while the model has undergone internal and temporal validation, prospective studies are still warranted to confirm its clinical applicability and real-world utility. Additionally, the ITE analysis, although useful for identifying patient subgroups with differential predicted risk, is limited by potential unmeasured confounding and model assumptions. Therefore, the interpretation of ITE results should be cautious, and further validation in independent cohorts and prospective trials is necessary to fully establish its clinical relevance. In conclusion, our study advances the field by integrating machine learning and ITE analysis to develop a clinically interpretable and actionable depression risk prediction model for middle-aged and older women with headache. By highlighting life satisfaction as a strategic target for precision prevention, this work bridges the gap between statistical modeling and practical mental health care, offering a pathway toward more personalized and effective intervention strategies in this high-risk population. 5. Conclusion This study developed and externally validated a machine learning–based prediction model for depression risk in middle-aged and older women with headache, integrating both population-level prediction and individualized intervention analysis via ITE. The model demonstrated strong discrimination and calibration across datasets, underscoring its generalizability and clinical utility. Our findings highlight life satisfaction as a pivotal and modifiable predictor, with ITE analysis revealing that individuals with low life satisfaction are likely to experience the greatest preventive benefit from targeted psychosocial interventions. By translating the model into an interactive web-based tool, we provide a practical solution for clinicians and public health practitioners to implement personalized screening and precision prevention strategies. Future research should focus on prospective validation in diverse populations and experimental evaluation of targeted life satisfaction–enhancing interventions. Such efforts will be essential to confirm causal effects and optimize intervention strategies, ultimately contributing to improved mental health outcomes and reduced suicide risk in this high-vulnerability population. Abbreviations • marry Marital status • rural Residence • srh Self-rated health • hope Hope for the future • satlife Life satisfaction • falldown History of falls • retire Retirement status • mulpain Multisite pain • edu Education level • disability Disability • hibpe Hypertension • diabe Diabetes • cancre Cancer • arthre Arthritis • lunge Chronic lung disease • hearte Cardiovascular diseases • stroke Stroke • dyslipe Dyslipidemia • livere Liver disease • kidneye Kidney disease • digeste Gastrointestinal disorders • asthmae Asthma • memrye Memory-related disorders • drinkev Alcohol use • smoken Smoking status • ins Health insurance coverage • systo Systolic blood pressure • diasto Diastolic blood pressure • pulse Pulse rate • lgrip Grip strength of left hand • rgrip Grip strength of right hand • sleep Sleep duration • puff Lung function Declarations 9.1Ethics approval and consent to participate Our study uses the China Health and Retirement Longitudinal Study (CHARLS) public dataset and does not require additional Institutional Review Board approval as the primary data collection was approved by the Biomedical Ethics Review Committee of Peking University (IRB00001052–11015). Written informed consent was obtained from all participants by the principal investigators of the survey. The methods in this study were carried out in accordance with the Declaration of Helsinki and relevant guidelines and regulations. 9.2Consent for publication Not applicable 9.3Availability of data and materials The datasets analyzed during the current study include the 2011 and 2015 waves of the China Health and Retirement Longitudinal Study (CHARLS). Access to CHARLS data is publicly available upon application at http://charls.pku.edu.cn. The datasets used and/or analyzed during the current study are also available from the corresponding author upon reasonable request. Competing Interests The authors declare that they have no competing interests. 10.hors' Contributions Xiaohui Wen : Conceptualization, Methodology, Software, Formal Analysis, Visualization. Zhijun Lin: Investigation, Data Curation, Validation. Yihui Qian: Writing – Original Draft, Writing – Review & Editing. Lin Wang: Data Curation, Validation, Software. Jingtong Zhou: Data Curation, Resources, Formal Analysis. Tingting Wen: Methodology, Data Transformation, Visualization. Qianying Cao: Conceptualization, Data Reduction, Quality Control. Zhou Liu, M.D.: Conceptualization (lead), Funding Acquisition (lead). All authors: Read and approved the final manuscript; contributed to critical revisions. 11.Funding This work was supported by the following funding sources: Guangdong Provincial Basic and Applied Basic Research Fund (Grant No. 2024A1515220002) Clinical + Basic Research Project of Guangdong Medical University (Grant No. 4SG23284G) Guangdong Provincial Medical Association Clinical Research Fund - Healthcare Special (Grant No. 2024HY-A6006) Guangdong Medical University Affiliated Hospital High-level Talent Research Launch Fund (Grant No. GCC2022011) 2023 Special Project of the Songshan Lake Medical-Engineering Integration Innovation Center of Guangdong Medical University (Grant No. 4SG22307P) 2023 Guangdong Provincial Administration of Traditional Chinese Medicine Research Project (Project No. 20232100) References Global regional, national burden of neurological disorders. 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet Neurol. 2019;18(5):459–80. Vetvik KG, MacGregor EA. Sex differences in the epidemiology, clinical features, and pathophysiology of migraine. Lancet Neurol. 2017;16(1):76–87. Kuan AS, Wang YF, Chen SP, Chuang YF, Wang SJ. Sex differences in pain, suicidal ideation, and suicide attempts in patients with migraine. Headache. 2025. Bolay H, Ozge A, Saginc P, Orekici G, Uludüz D, Yalın O, et al. Gender influences headache characteristics with increasing age in migraine patients. Cephalalgia. 2015;35(9):792–800. Buse DC, Manack AN, Fanning KM, Serrano D, Reed ML, Turkel CC, et al. Chronic migraine prevalence, disability, and sociodemographic factors: results from the American Migraine Prevalence and Prevention Study. Headache. 2012;52(10):1456–70. Slavich GM, Irwin MR. From stress to inflammation and major depressive disorder: a social signal transduction theory of depression. Psychol Bull. 2014;140(3):774–815. Pan A, Sun Q, Okereke OI, Rexrode KM, Hu FB. Depression and risk of stroke morbidity and mortality: a meta-analysis and systematic review. JAMA. 2011;306(11):1241–9. Vieira WF, Coelho DRA, Litwiler ST, McEachern KM, Clancy JA, Morales-Quezada L, et al. Neuropathic pain, mood, and stress-related disorders: A literature review of comorbidity and co-pathogenesis. Neurosci Biobehav Rev. 2024;161:105673. Rist PM, Schürks M, Buring JE, Kurth T. Migraine, headache, and the risk of depression: Prospective cohort study. Cephalalgia. 2013;33(12):1017–25. Beam AL, Kohane IS. Big Data and Machine Learning in Health Care. JAMA. 2018;319(13):1317–8. Zhao Y, Hu Y, Smith JP, Strauss J, Yang G. Cohort profile: the China Health and Retirement Longitudinal Study (CHARLS). Int J Epidemiol. 2014;43(1):61–8. Irwin M, Artin KH, Oxman MN. Screening for depression in the older adult: criterion validity of the 10-item Center for Epidemiological Studies Depression Scale (CES-D). Arch Intern Med. 1999;159(15):1701–4. Andresen EM, Malmgren JA, Carter WB, Patrick DL. Screening for depression in well older adults: evaluation of a short form of the CES-D (Center for Epidemiologic Studies Depression Scale). Am J Prev Med. 1994;10(2):77–84. Zhang Z. Multiple imputation with multivariate imputation by chained equation (MICE) package. Ann Transl Med. 2016;4(2):30. Zeng ZQ, Wu Q, Liao BS, Gao J. Tien Tzu Hsueh Pao/Acta Electronica Sinica. 2009;37(11):2489–95. A Classfication Method For Imbalance Data Set Based on Kernel SMOTE. Ali YA, Awwad EM, Al-Razgan M, Maarouf A. Hyperparameter Search for Machine Learning Algorithms for Optimizing the Computational Complexity. Processes. 2023;11(2):349. Cox DR. The Regression Analysis of Binary Sequences. J Roy Stat Soc: Ser B (Methodol). 2018;20(2):215–32. Friedman JH. Greedy Function Approximation: A Gradient Boosting Machine. Ann Stat. 2001;29(5):1189–232. Shao F, Shen Z. How can artificial neural networks approximate the brain? Front Psychol. 2022;13:970214. Breiman L. Random forests, machine learning 45. J Clin Microbiol. 2001;2:199–228. Cover TM, Hart PE. Nearest neighbor pattern classification. IEEE Transinftheory. 1953;13(1):21–7. Freund Y, Schapire R. A decision-theoretic generalization of on-line learning and an application to boosting. J Comput Syst Sci. 1997. Meng Q, editor. editor LightGBM: A Highly Efficient Gradient Boosting Decision Tree. Neural Information Processing Systems; 2017. Lundberg SM, Erion G, Chen H, Degrave A, Lee SI. Explainable AI for Trees: From Local Explanations to Global Understanding. 2019. Benca RM, Peterson MJ. Insomnia and depression. Sleep Med. 2008;9(Suppl 1):S3–9. Goldzweig G, Baider L, Andritsch E, Pfeffer R, Rottenberg Y. A Dialogue of Depression and Hope: Elderly Patients Diagnosed with Cancer and Their Spousal Caregivers. J Cancer Educ. 2017;32(3):549–55. Huang R, Li S, Hu J, Ren R, Ma C, Peng Y, et al. Adverse childhood experiences and falls in older adults: The mediating role of depression. J Affect Disord. 2024;365:87–94. Maier A, Riedel-Heller SG, Pabst A, Luppa M. Risk factors and protective factors of depression in older people 65+. A systematic review. PLoS ONE. 2021;16(5):e0251326. Weber E, Hülür G. Co-development of Couples' Life Satisfaction in Transition to Retirement: A Longitudinal Dyadic Perspective. J Gerontol B Psychol Sci Soc Sci. 2021;76(8):1542–54. Åhlin JK, Peristera P, Westerlund H, Magnusson Hanson LL. Psychosocial working characteristics before retirement and depressive symptoms across the retirement transition: a longitudinal latent class analysis. Scand J Work Environ Health. 2020;46(5):488–97. Magnusson Hanson LL, Peristera P, Chungkham HS, Westerlund H. Psychosocial work characteristics, sleep disturbances and risk of subsequent depressive symptoms: a study of time-varying effect modification. J Sleep Res. 2017;26(3):266–76. Mollan SP, Subramanian A, Perrins M, Nirantharakumar K, Adderley NJ, Sinclair AJ. Depression and anxiety in women with idiopathic intracranial hypertension compared to migraine: A matched controlled cohort study. Headache. 2023;63(2):290–8. Vuralli D, Ayata C, Bolay H. Cognitive dysfunction and migraine. J Headache Pain. 2018;19(1):109. Zheng Y, Zhang C, Liu Y. Risk prediction models of depression in older adults with chronic diseases. J Affect Disord. 2024;359:182–8. Duan S, Xia H, Zheng T, Li G, Ren Z, Ding W, et al. Development and validation of non-invasive prediction models for migraine in Chinese adults. J Headache Pain. 2023;24(1):148. Amoozegar F, Patten SB, Becker WJ, Bulloch AGM, Fiest KM, Davenport WJ, et al. The prevalence of depression and the accuracy of depression screening tools in migraine patients. Gen Hosp Psychiatry. 2017;48:25–31. Cheng Y, Wei Y, Tang SL. Does Helping Others Always Benefit Health? Longitudinal Evidence on the Relationship between Helping Behavior and Depression: The Mediating Role of Life Satisfaction and the Moderating Effect of IADL. Depress Anxiety. 2024;2024:2304723. Tong Y, Xie F, Wen X, Li Y, Yuan M, Zhang X, et al. Longitudinal Association between Bullying Victimization and Depressive Symptoms in Chinese Early Adolescents: The Effect of Life Satisfaction. Depress Anxiety. 2024;2024:6671415. Abdin PVA, Roystonn E, Devi K, Wang F, Shafie P. Tracking the Prevalence of Depression Among Older Adults in Singapore: Results From the Second Wave of the Well-Being of Singapore Elderly Study. Depress Anxiety. 2025;2025:9071391. Künzel SR, Sekhon JS, Bickel PJ, Yu B. Metalearners for estimating heterogeneous treatment effects using machine learning. Proc Natl Acad Sci U S A. 2019;116(10):4156–65. Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterial.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7517256","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":533878078,"identity":"0600f32f-4eb0-4606-a569-c503bcdf3390","order_by":0,"name":"Xiaohui Wen","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xiaohui","middleName":"","lastName":"Wen","suffix":""},{"id":533878079,"identity":"d9647e19-f51a-471c-a7fa-4532106f9cda","order_by":1,"name":"Zhijun Lin","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Zhijun","middleName":"","lastName":"Lin","suffix":""},{"id":533878080,"identity":"ca85ab67-cb8e-4998-93c4-8b5cbffcef9f","order_by":2,"name":"Yihui Qian","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yihui","middleName":"","lastName":"Qian","suffix":""},{"id":533878081,"identity":"245d07de-855b-4ef3-919f-f9a9de86b1cc","order_by":3,"name":"Lin Wang","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Lin","middleName":"","lastName":"Wang","suffix":""},{"id":533878082,"identity":"db6de458-b8c1-4afb-a423-bf1fb8478517","order_by":4,"name":"Jingtong Zhou","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jingtong","middleName":"","lastName":"Zhou","suffix":""},{"id":533878083,"identity":"f60e9f15-6d9b-4455-a92e-a9008d59a91f","order_by":5,"name":"Tingting Wen","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Tingting","middleName":"","lastName":"Wen","suffix":""},{"id":533878084,"identity":"d15c79eb-24a8-4959-bf4e-e9574cbd824d","order_by":6,"name":"Qianying Cao","email":"","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":false,"prefix":"","firstName":"Qianying","middleName":"","lastName":"Cao","suffix":""},{"id":533878085,"identity":"e3f6f7ac-b66e-4891-b8a4-30d48b2b0c95","order_by":7,"name":"Zhou Liu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzElEQVRIiWNgGAWjYBACPmYGBiBikGNgSADxmQlrYYNqMSZBC1RZYgPxWth5jD8XVBxOn++enSbBUGGd2MB+9gABh/EYGM84czh345m32yQYzqQnNvDkJRDUkszbBtQyI3ebBGPb4cQGCR4DgloOA7WkG4K1/CNOi2EzUEuCvARISwNRWtiKmXnOpBtu4Hm72SLhWLpxG08Ofi38/Ic3f+apsJaXb8/deONDjbVsP/sZ/FrgwOAAAyQ22YhTDwTyDUQrHQWjYBSMgpEGAKcBPFonC0i2AAAAAElFTkSuQmCC","orcid":"","institution":"Affiliated Hospital of Guangdong Medical University","correspondingAuthor":true,"prefix":"","firstName":"Zhou","middleName":"","lastName":"Liu","suffix":""}],"badges":[],"createdAt":"2025-09-02 11:38:22","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7517256/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7517256/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":94409612,"identity":"c06bf690-3844-43f4-95ea-eb20bfab15cc","added_by":"auto","created_at":"2025-10-27 14:04:13","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1181968,"visible":true,"origin":"","legend":"","description":"","filename":"ManuscriptFileWHBMC.docx","url":"https://assets-eu.researchsquare.com/files/rs-7517256/v1/dea3ac4cad5634e357989e27.docx"},{"id":94410039,"identity":"f27b2ccd-fea1-483e-89d4-ead34bf2762c","added_by":"auto","created_at":"2025-10-27 14:04:25","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9526,"visible":true,"origin":"","legend":"","description":"","filename":"4cb6a9d8837e4cf4a73b78834835bb54.json","url":"https://assets-eu.researchsquare.com/files/rs-7517256/v1/77b9ef424346d26d4e22e7b1.json"},{"id":94409347,"identity":"1e961f3f-f366-408b-a994-e79f7b671d8e","added_by":"auto","created_at":"2025-10-27 14:04:05","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":294258,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-7517256/v1/a8ed9462a117cad8e45dee56.docx"},{"id":94410188,"identity":"58a30b6e-7bf8-4bd9-8dac-1d54b78462ce","added_by":"auto","created_at":"2025-10-27 14:04:34","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":175758,"visible":true,"origin":"","legend":"","description":"","filename":"4cb6a9d8837e4cf4a73b78834835bb541enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7517256/v1/7f8240bf549335cf8ced063e.xml"},{"id":94409483,"identity":"1ed58889-c43a-4bce-a7a8-1fac276c49c8","added_by":"auto","created_at":"2025-10-27 14:04:09","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":46567,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7517256/v1/8ab853612bdafc7b663ce9c2.png"},{"id":94408836,"identity":"eb1b0c49-07b5-4ae3-bfef-26f17ce4a345","added_by":"auto","created_at":"2025-10-27 14:03:53","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":114777,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7517256/v1/696ffaff4483c0ccaae09fd4.png"},{"id":94409746,"identity":"44a81e4c-0f98-40fc-96a3-cb7ad5458773","added_by":"auto","created_at":"2025-10-27 14:04:16","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":49595,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7517256/v1/5a7b09d320bbbdfa44066375.png"},{"id":94410266,"identity":"382e6490-bf51-4086-97b5-1f5799cf5ef7","added_by":"auto","created_at":"2025-10-27 14:04:37","extension":"xml","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":172577,"visible":true,"origin":"","legend":"","description":"","filename":"4cb6a9d8837e4cf4a73b78834835bb541structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7517256/v1/60aa201621264b8522b84d01.xml"},{"id":94410324,"identity":"eb99aadb-7f28-49b0-a035-83a11e3388e7","added_by":"auto","created_at":"2025-10-27 14:04:40","extension":"html","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":183076,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7517256/v1/4fc9f23ee65da63f26cfe587.html"},{"id":94408838,"identity":"0c3cdf63-0f6b-4883-a062-4a487da79ba7","added_by":"auto","created_at":"2025-10-27 14:03:53","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":173074,"visible":true,"origin":"","legend":"\u003cp\u003eResearch design and technical roadmap\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7517256/v1/e3cda38bdd52fa67f47a0406.png"},{"id":94410452,"identity":"3fa1f7c1-3688-44ee-b1c6-c6522bde3c78","added_by":"auto","created_at":"2025-10-27 14:04:43","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":612840,"visible":true,"origin":"","legend":"\u003cp\u003eModel Construction\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7517256/v1/d537afc7a2026e88107afa93.png"},{"id":94408607,"identity":"6bdd18a6-da3f-4b06-a316-d7564a082f7b","added_by":"auto","created_at":"2025-10-27 14:03:41","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":267813,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7517256/v1/5a0756964971e0d2c3acc72d.png"},{"id":106399458,"identity":"4b1120d6-e255-4ec4-8352-7d7f492c80cd","added_by":"auto","created_at":"2026-04-08 08:29:59","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2575095,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7517256/v1/2141455b-9447-4015-b6be-84fb996d2038.pdf"},{"id":94410508,"identity":"2e7ca0c1-c380-488d-b9cc-df772971fa2c","added_by":"auto","created_at":"2025-10-27 14:04:44","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":294258,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-7517256/v1/68d103074111854daf694d5a.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Predictive Model for Headache-Related Depression in Middle- Aged and Older Women Incorporating Individual Treatment Effects","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eHeadache is one of the most prevalent neurological symptoms worldwide, with a notably higher burden in women than in men. Epidemiological studies indicate that active headache disorders affect approximately 52.0% of the global population, with a prevalence of 57.8% in females compared to 44.4% in males (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). The lifetime prevalence of headache remains consistently higher among women. In young women, hormonal fluctuations and menstrual cycles often influence headache frequency and severity.(\u003cspan additionalcitationids=\"CR3 CR4 CR5\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). Depression has emerged as a major global public health concern, substantially impairing psychological well-being and quality of life. It has been linked to increased risks of suicidal behavior, stroke, cognitive decline, and neurodegenerative diseases such as Alzheimer\u0026rsquo;s disease(\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e), A bidirectional relationship between chronic pain and depression is well documented, and women with chronic headache are particularly vulnerable to depressive symptoms, which may further exacerbate neurological and psychiatric outcomes.(\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e). Despite this, research specifically examining depression risk in middle-aged and older women with headache remains limited, However, research specifically focusing on depression risk in middle-aged and older women with headache remains limited. .(\u003cspan additionalcitationids=\"CR9\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e). In this study, we leveraged data from the 2015 wave of the China Health and Retirement Longitudinal Study (CHARLS) to develop an ML-based prediction model for depression risk in middle-aged and older women with headache. We systematically compared ten widely used ML algorithms\u0026mdash;including XGBoost, CatBoost, random forest, and neural networks\u0026mdash;using ten-fold cross-validation to ensure robustness and mitigate overfitting. The model\u0026rsquo;s generalizability was assessed through temporal validation with the 2011 CHARLS cohort. To enhance clinical applicability and promote precision prevention, we incorporated Individual Treatment Effect (ITE) estimation into the analytical pipeline, enabling the identification of subgroups most likely to benefit from targeted interventions and the exploration of heterogeneity in predicted depression risk across diverse patient profiles.\u003c/p\u003e"},{"header":"2. Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Data Sources and Study Population\u003c/h2\u003e\u003cp\u003eThis study utilized data from a large-scale waves of the CHARLS (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e) for model development and validation. The training and internal validation datasets were derived from the 2015 wave of CHARLS, a nationally representative longitudinal survey of middle-aged and older Chinese adults. Female participants aged 45 years or older who reported chronic headache symptoms were included. For temporal external validation, data from the 2011 wave of CHARLS were used. Inclusion criteria were: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) female gender; (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) age\u0026thinsp;\u0026ge;\u0026thinsp;45 years; (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) self-reported presence of headache symptoms; and (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) complete data on all predictor variables. Exclusion criteria included: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) a history of severe neurological or psychiatric disorders; and (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) missing values in key variables.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 Data Collection Dimensions\u003c/h2\u003e\u003cdiv id=\"Sec5\" class=\"Section3\"\u003e\u003ch2\u003e2.2.1 Depression and Headache\u003c/h2\u003e\u003cp\u003eIn this study, the presence of a significant risk of depression was used as the target outcome variable. The definition and standardization of depressive status were based on the depression screening tools applied in each respective dataset. In the CHARLS dataset, depressive symptoms were assessed using the validated 10-item Center for Epidemiologic Studies Depression Scale (CES-D 10)(\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). The CES-D 10 consists of 10 items, each scored on a 4-point Likert scale from 0 (rarely or none of the time) to 3 (most or all of the time), with total scores ranging from 0 to 30. Higher scores indicate more severe depressive symptoms. Based on previous studies(\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e), a score of \u0026ge;\u0026thinsp;10 was used to define clinically significant depressive symptoms. For the sake of clarity, the term \"depression\" is used throughout this study to refer to depressive symptoms of varying severity. As for headache assessment, the CHARLS dataset contains only a single self-reported item regarding headache symptoms and does not further distinguish between specific headache subtypes such as chronic headache, migraine, or tension-type headache. Therefore, in this study, headache was treated as a broad clinical concept. Given prior epidemiological evidence indicating that chronic headache, migraine, and tension-type headache are the most prevalent forms among middle-aged and older adults(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e), it is reasonable to assume that individuals in CHARLS who reported headache symptoms potentially represent a heterogeneous group comprising various headache subtypes.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section3\"\u003e\u003ch2\u003e2.2.1 Demographic Characteristics\u003c/h2\u003e\u003cp\u003eThis study assessed a range of sociodemographic variables, including age; sex (male/female); marital status (married/unmarried); place of residence (urban/rural); educational attainment (no formal education, primary school, junior high school, high school or above); and retirement status (yes/no). Subjective health status and life satisfaction were evaluated using a five-point Likert scale ranging from \u0026ldquo;very good\u0026rdquo; to \u0026ldquo;very poor.\u0026rdquo; Multisite pain was defined as the presence of pain in more than five body regions, based on the sample median. Hopefulness was categorized by its frequency over the past week: \u0026ldquo;\u0026lt;1 day,\u0026rdquo; \u0026ldquo;1\u0026ndash;2 days,\u0026rdquo; \u0026ldquo;3\u0026ndash;4 days,\u0026rdquo; or \u0026ldquo;5\u0026ndash;7 days.\u0026rdquo; Behavioral factors included smoking status, alcohol consumption, and history of falls (all dichotomous), as well as average nighttime sleep duration over the past month, recorded in hours.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section3\"\u003e\u003ch2\u003e2.2.2 Health Status\u003c/h2\u003e\u003cp\u003eBased on previous literature and clinical expertise, we included several health-related variables potentially associated with depression. These included self-reported history of chronic conditions such as hypertension, cancer, dyslipidemia, heart disease, stroke, arthritis, psychiatric disorders or rheumatism, liver disease, kidney disease, gastrointestinal disorders, and asthma. Additionally, self-rated health and life satisfaction were assessed as subjective perceptions of overall health and life quality, categorized as \"very good,\" \"good,\" \"fair,\" \"poor,\" or \"very poor.\" Multisite pain was defined as reporting pain in more than five body regions and treated as a binary variable. For physical examination data, the mean value of repeated measurements was used.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e2.3 Preprocessing and Model Construction\u003c/h2\u003e\u003cp\u003eAs illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, data preprocessing and model development followed a structured workflow. To prevent model overfitting due to multicollinearity and to minimize noise interference, we performed variable screening using Pearson correlation heatmaps and variance inflation factor (VIF) analysis (see Supplementary Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). Variables with high pairwise correlations (|r| \u0026gt;0.8) and severe multicollinearity (VIF\u0026thinsp;\u0026gt;\u0026thinsp;5) were excluded to reduce computational complexity and eliminate potential confounding interactions.\u003c/p\u003e\u003cp\u003eThe exclusion criteria for variables included:(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) variables with more than 20% missing data,\u003c/p\u003e\u003cp\u003e(\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) absolute Pearson correlation coefficients greater than 0.8, (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) VIF values exceeding 5, and\u003c/p\u003e\u003cp\u003e(\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) variables deemed not significantly associated with depression risk based on previous studies and clinical expertise. Ultimately, 37 variables were retained for further analysis, encompassing sociodemographic characteristics, health status, medical history, lifestyle behaviors, and physiological measurements. These variables included: Age, Marital status(marry), Residence༈rural༉, Self-rated health༈srh༉, Hope for the future༈hope༉, Llife satisfaction༈satlife༉, History of falls༈falldown༉, Retirement status༈retire༉, Multisite pain༈mulpain༉,Education level༈edu༉, Disability, Hypertension༈hibpe༉, Diabetes༈diabe༉, Cancer༈cancre༉, Arthritis༈arthre༉, Chronic lung disease༈lunge༉, Cardiovascular diseases༈hearte༉, Stroke, dyslipidemia༈dyslipe༉,Liver disease༈livere༉, Kidney disease༈kidneye༉,Gastrointestinal disorders༈digeste༉, Asthma༈asthmae༉, Memory-related disorders༈memrye༉, Alcohol use༈drinkev༉,Smoking status༈smoken༉Health insurance coverage(ins), Systolic blood pressure(systo), Diastolic blood pressure(diasto), Pulse rate(pulse), Grip strength of left(lgrip) and Right hand(rgrip), Sleep duration(sleep), and Lung function(puff) (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 table of the raw dataset\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\u003eVariables\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTotal (n\u0026thinsp;=\u0026thinsp;1930)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eNon-depression\u003c/p\u003e\u003cp\u003e(n\u0026thinsp;=\u0026thinsp;569)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDepression\u003c/p\u003e\u003cp\u003e(n\u0026thinsp;=\u0026thinsp;1361)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003ep\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFeel hopeful about the Future, n (%)\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\u003eRarely or never (\u0026lt;\u0026thinsp;1 day)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e776 (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e164 (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e612 (45)\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\u003eOccasional (1\u0026ndash;2 days)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e300 (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e66 (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e234 (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e)\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\u003eSometimes (3\u0026ndash;4 days)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e334 (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e78 (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e256 (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e)\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\u003eAlmost always (5\u0026ndash;7 days)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e520 (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e261 (46)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e259 (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e)\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\u003eRetirement status n (%)\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\u003e1820 (94)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e516 (91)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1304 (96)\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\u003e110 (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e53 (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e57 (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e)\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 falls, n (%)\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\u003e1326 (69)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e440 (77)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e886 (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\u003e604 (\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e129 (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e475 (\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e)\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, n (%)\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\u003e0.015\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eUnmarried\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e347 (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e83 (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e264 (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e)\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\u003e1583 (82)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e486 (85)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1097 (81)\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\u003eLife satisfaction, n (%)\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\u003eNot at all satisfied\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e92 (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e9 (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e83 (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e)\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\u003eNot very satisfied\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e252 (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e34 (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e218 (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e)\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\u003eSomewhat satisfied\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e940 (49)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e263 (46)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e677 (50)\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\u003eVery satisfied\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e539 (\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e221 (\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e318 (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e)\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\u003eCompletely satisfied\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e107 (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e42 (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e65 (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e)\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\u003eMultisite pain, n (%)\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\u003e612 (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e222 (\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e390 (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e)\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\u003e1318 (68)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e347 (61)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e971 (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\u003ePlace of residence, n (%)\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\u003erural\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e592 (\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e219 (\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e373 (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e)\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\u003eUrban\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1338 (69)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e350 (62)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e988 (73)\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\u003eEducational attainment, n (%)\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\u003eElementary school or below\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1212 (63)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e321 (56)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e891 (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\u003eMiddle school\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e452 (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e146 (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e306 (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e)\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\u003e188 (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e65 (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e123 (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e)\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\u003eAbove high school\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e78 (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e37 (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e41 (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e)\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\u003eDisability, n (%)\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\u003e1070 (55)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e363 (64)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e707 (52)\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\u003e860 (45)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e206 (\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e654 (48)\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\u003eSelf perceived health status, n (%)\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\u003eVery good\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e298 (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e50 (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e248 (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e)\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\u003ePoor\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e718 (\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e167 (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e551 (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e)\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\u003eFair\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e802 (42)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e298 (52)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e504 (\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e)\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\u003eGood\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e70 (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e33 (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e37 (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e)\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\u003eVery good\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e42 (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e21 (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e21 (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e)\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\u003eHypertension n (%)\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\u003e0.859\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\u003e1166 (60)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e346 (61)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e820 (60)\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\u003e764 (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e223 (\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e541 (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e)\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\u003eDiabetes mellitus, n (%)\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\u003e0.524\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\u003e1645 (85)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e490 (86)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1155 (85)\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\u003e285 (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e79 (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e206 (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e)\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\u003eCancer, n (%)\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\u003e0.835\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\u003e1883 (98)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e554 (97)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1329 (98)\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\u003e47 (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e15 (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e32 (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e)\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\u003eLunge disease, n (%)\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\u003e0.224\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\u003e1560 (81)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e470 (83)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1090 (80)\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\u003e370 (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e99 (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e271 (\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e)\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\u003eHeart disease, n (%)\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\u003e0.128\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\u003e1331 (69)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e407 (72)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e924 (68)\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\u003e599 (\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e162 (\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e437 (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e)\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 stroke n (%)\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\u003e0.875\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\u003e1855 (96)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e548 (96)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1307 (96)\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\u003e75 (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e21 (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e54 (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e)\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\u003eArthritis, n (%)\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\u003e610 (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e214 (\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e396 (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e)\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\u003e1320 (68)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e355 (62)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e965 (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\u003eDyslipidemia, n (%)\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\u003e0.853\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\u003e1465 (76)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e434 (76)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1031 (76)\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\u003e465 (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e135 (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e330 (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e)\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\u003eLiver disease, n (%)\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\u003e0.915\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\u003e1743 (90)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e515 (91)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1228 (90)\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\u003e187 (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e54 (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e133 (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e)\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\u003eKidney disease, n (%)\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\u003e0.005\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\u003e1635 (85)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e503 (88)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1132 (83)\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\u003e295 (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e66 (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e229 (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e)\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\u003eDigestive disease, n (%)\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\u003e0.004\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\u003e926 (48)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e302 (53)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e624 (46)\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\u003e1004 (52)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e267 (47)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e737 (54)\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\u003eAsthma, n (%)\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\u003e0.113\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\u003e1766 (92)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e530 (93)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1236 (91)\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\u003e164 (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e39 (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e125 (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e)\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\u003eMemory impairment, n (%)\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\u003e0.025\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\u003e1830 (95)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e550 (97)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1280 (94)\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\u003e100 (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e19 (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e81 (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e)\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\u003eDrinking behavior n (%)\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\u003e0.19\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\u003e1638 (85)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e473 (83)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1165 (86)\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\u003e292 (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e96 (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e196 (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e)\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\u003eHealth insurance, n (%)\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\u003e1\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\u003e169 (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e50 (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e119 (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e)\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\u003e1761 (91)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e519 (91)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1242 (91)\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\u003eSmoking, behavior n (%)\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\u003e0.448\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\u003e1815 (94)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e531 (93)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1284 (94)\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\u003e115 (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e38 (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e77 (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e)\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\u003eSystolic Blood Pressure, Median (Q1,Q3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e122.5 (110, 138.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e124.5 (111.5, 141)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e122 (109, 138)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.014\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiastolic Blood Pressure, Median (Q1,Q3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e73 (65.5, 81)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e73.5 (66, 82)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e72.5 (65.5, 81)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.137\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePulse, Median (Q1,Q3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e74 (68, 81)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e74.5 (68.5, 80.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e74 (67.5, 81)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.724\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLeft grip strength, Median (Q1,Q3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e22 (17.2, 27)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e23.3 (18.5, 28.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e21.8 (17, 26.5)\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\u003eRight strength, Median (Q1,Q3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e23.5 (18.9, 28.27)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e24 (20, 29.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e23 (18.1, 28)\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\u003ePulmonary function, Median (Q1,Q3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e260 (190, 320)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e270 (200, 330)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e250 (190, 310)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.017\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSleep duration, Median (Q1,Q3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e5.5 (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6 (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e5 (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e)\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\u003eAge, Median (Q1,Q3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e59 (51, 67)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e58 (50, 66)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e60 (51, 67)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.149\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eTo minimize bias caused by missing data, this study employed multiple imputation using the \"MICE\" (Multivariate Imputation by Chained Equations) package(\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). The predictive mean matching (PMM) method was used to generate five imputed datasets, with 50 iterations performed to ensure the stability of the imputation results. Subsequently, Z-score standardization was applied to all features to mitigate the influence of large-scale variables on model training, thereby improving model accuracy and stability.\u003c/p\u003e\u003cp\u003eTo address class imbalance in the original dataset\u0026mdash;where 1,361 patients (70.5%) were diagnosed with depression and 569 (29.5%) were not\u0026mdash;the Synthetic Minority Over-sampling Technique (SMOTE) was applied to the training set. Using five nearest neighbors, SMOTE generated synthetic samples with 100% oversampling of the minority class and 200% undersampling of the majority class to achieve balance. By interpolating between existing minority instances, SMOTE mitigated class bias and improved the model's sensitivity to non-depressed cases, thereby enhancing its generalizability and clinical relevance(\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eFor data partitioning, all samples were randomly divided into training and testing sets in a 7:3 ratio. The training set was used for data preprocessing, feature selection, and model development, while the testing set was reserved for performance evaluation and model tuning. Feature selection was performed using univariate logistic regression, multivariate logistic regression (see Supplementary Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e), and the Least Absolute Shrinkage and Selection Operator (LASSO) regression. Based on previous research and clinical interpretability, seven key predictors were ultimately selected: hope for the future, self-rated health, retirement status, multisite pain, history of falls, life satisfaction, and sleep duration(Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA.\u003cb\u003eB\u003c/b\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFor model development and comparison, ten machine learning algorithms were implemented, including Neural Network, Extreme Gradient Boosting (XGBoost), Categorical Boosting (CatBoost), Logistic Regression, Random Forest, Gradient Boosting Machine (GBM), Support Vector Machine (SVM), Light Gradient Boosting Machine (LightGBM), k-Nearest Neighbors (KNN), and Adaptive Boosting (AdaBoost). Hyperparameters for each model were optimized using a grid search strategy(\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e).. Model training was conducted on the training set using 10-fold cross-validation combined with resampling techniques to reduce overfitting. Model performance was subsequently evaluated on both the internal testing set and the temporally separated external validation set derived from the 2011 wave of CHARLS. This approach enabled assessment of the models\u0026rsquo; robustness and generalizability over time.\u003c/p\u003e\u003cp\u003eModel performance was assessed from three key perspectives: discrimination, calibration, and clinical utility. Discrimination was evaluated using the area under the receiver operating characteristic curve(AUC) and the F1 score, Precision-Recall curves. Calibration was assessed through calibration plots and Brier scores to determine the agreement between predicted probabilities and actual outcomes. Clinical utility was evaluated using Decision Curve Analysis (DCA), which measures the net clinical benefit across a range of risk thresholds. In addition, confusion matrix metrics\u0026mdash;including sensitivity, specificity, accuracy, and precision\u0026mdash;were comprehensively analyzed to further evaluate model performance.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e2.4 Machine Learning Algorithms\u003c/h2\u003e\u003cp\u003eThis study employed R version 4.4.3 to develop ten widely used classification models: XGBoost, CatBoost, LightGBM, Random Forest, GBM, Neural Network, Logistic Regression, SVM, KNN, and AdaBoost. Hyperparameter tuning for each model was conducted using grid search, combined with 10-fold cross-validation repeated five times to assess model stability and generalization performance.\u003c/p\u003e\u003cp\u003eEach model offers distinct advantages:\u003c/p\u003e\u003cp\u003e\u003cb\u003eLogistic Regression\u003c/b\u003e is straightforward and interpretable, suitable for linear classification problems.(\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e)\u003c/p\u003e\u003cp\u003e\u003cb\u003eSVM\u003c/b\u003e performs robustly in high-dimensional spaces and is effective for nonlinear data.\u003c/p\u003e\u003cp\u003e\u003cb\u003eGBM\u003c/b\u003e and \u003cb\u003eXGBoost\u003c/b\u003e enhance performance by integrating weak learners, offering strong nonlinear modeling capabilities(\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cb\u003eNeural Networks\u003c/b\u003e can automatically extract features, making them adept at complex pattern recognition tasks(\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e),\u003c/p\u003e\u003cp\u003e\u003cb\u003eRandom Forest\u003c/b\u003e reduces overfitting risk by aggregating multiple decision trees(\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cb\u003eKNN\u003c/b\u003e classifies based on instances without requiring a training phase(\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cb\u003eAdaBoost\u003c/b\u003e iteratively improves weak classifiers' performance(\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cb\u003eLightGBM\u003c/b\u003e and \u003cb\u003eCatBoost\u003c/b\u003e are particularly efficient in handling large-scale data and categorical features, with high training efficiency(\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eperformance for specific predictive tasks, aiding in the selection of the optimal model.\u003c/p\u003e\u003cp\u003eGrid search, a hyperparameter optimization technique, exhaustively searches through a predefined set of hyperparameter combinations to identify the configuration yielding the best model performance. This method systematically evaluates each combination using cross-validation, ensuring a thorough exploration of the hyperparameter space(\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e), The hyperparameter tuning methods for the models are as follows: Support Vector Machine (SVM), GBM, Neural Network, Random Forest, XGBoost, KNN, and AdaBoost were tuned using grid search within predefined parameter ranges; whereas CatBoost and LightGBM were trained using parameters defined by our own specifications. In this study, all models underwent hyperparameter tuning using 10-fold cross-validation repeated five times (\u0026ldquo;10-fold CV with 5 repeats\u0026rdquo;) combined with grid search. The optimal hyperparameters identified are as follows:\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eSVM\u003c/strong\u003e\u003cp\u003esigma\u0026thinsp;=\u0026thinsp;0.001, C\u0026thinsp;=\u0026thinsp;0.5\u003c/p\u003e\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eGBM\u003c/b\u003e: n.trees\u0026thinsp;=\u0026thinsp;100, interaction.depth\u0026thinsp;=\u0026thinsp;3, shrinkage\u0026thinsp;=\u0026thinsp;0.01, n.minobsinnode\u0026thinsp;=\u0026thinsp;5\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eNeural Network\u003c/b\u003e: size\u0026thinsp;=\u0026thinsp;5, decay\u0026thinsp;=\u0026thinsp;0.6\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eRandom Forest\u003c/b\u003e: mtry\u0026thinsp;=\u0026thinsp;5\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eXGBoost\u003c/b\u003e: nrounds\u0026thinsp;=\u0026thinsp;10, max_depth\u0026thinsp;=\u0026thinsp;5, eta\u0026thinsp;=\u0026thinsp;0.001, gamma\u0026thinsp;=\u0026thinsp;0.5, colsample_bytree\u0026thinsp;=\u0026thinsp;0.5, min_child_weight\u0026thinsp;=\u0026thinsp;1, subsample\u0026thinsp;=\u0026thinsp;0.6\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eKNN\u003c/b\u003e: kmax\u0026thinsp;=\u0026thinsp;12, distance\u0026thinsp;=\u0026thinsp;1, kernel = \"optimal\"\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eAdaBoost\u003c/b\u003e: mfinal\u0026thinsp;=\u0026thinsp;2, maxdepth\u0026thinsp;=\u0026thinsp;2, coeflearn = \"Zhu\"\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eLightGBM\u003c/b\u003e: learning_rate\u0026thinsp;=\u0026thinsp;1.0, min_data\u0026thinsp;=\u0026thinsp;1, num_threads\u0026thinsp;=\u0026thinsp;2, force_col_wise\u0026thinsp;=\u0026thinsp;TRUE,\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eCatBoost\u003c/b\u003e: iterations\u0026thinsp;=\u0026thinsp;100, depth\u0026thinsp;=\u0026thinsp;5, learning_rate\u0026thinsp;=\u0026thinsp;0.03, border_count\u0026thinsp;=\u0026thinsp;32\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e2.5 SHAP Explainability Analysis\u003c/h2\u003e\u003cp\u003eWe applied the Shapley Additive Explanations (SHAP) method(\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e),to quantify each feature\u0026rsquo;s contribution to our model predictions. Positive SHAP values indicate factors that increase predicted risk, while negative values denote protective effects. After examining seven key predictors via their SHAP distributions, we implemented an interactive Shiny application so clinicians can explore these feature attributions and obtain case-specific explanations on demand.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e2.6 Individual Treatment Effect Estimation\u003c/h2\u003e\u003cp\u003eTo investigate heterogeneity in intervention effects, we applied a counterfactual framework and uplift modeling to estimate individual treatment effects (ITEs) on depression risk. Binary treatment assignments were defined based on modifiable psychosocial variables. Life satisfaction was selected as the intervention variable due to its modifiability and clinical relevance. Participants reporting \"Somewhat satisfied,\" \"Very satisfied,\" or \"Completely satisfied\" were classified as the high life satisfaction (treated) group, whereas those reporting \"Not very satisfied\" or \"Not at all satisfied\" were classified as the low life satisfaction (control) group.\u003c/p\u003e\u003cp\u003eITEs were estimated using a two-model approach with random forest classifiers. Separate outcome models were trained for the treated and control groups using the same covariates as in the primary prediction model. For each individual, the difference between predicted depression risk under treatment and control conditions provided the ITE.\u003c/p\u003e\u003cp\u003eTo identify individuals most likely to benefit from improvements in life satisfaction, we extracted the top 5% of ITE scores (highest estimated benefit) as the high-benefit subgroup. Descriptive statistics (means and standard deviations) were calculated for baseline numeric variables and compared between this subgroup and the overall population. Differences in means were used to rank variables contributing to treatment heterogeneity.\u003c/p\u003e\u003cp\u003eThe clinical utility of individualized predictions was evaluated using uplift modeling and Qini curves. Individuals were ranked by their ITE estimates, and cumulative incremental benefit was plotted against the cumulative population fraction. The Qini coefficient (area under the Qini curve, uplift AUC) summarized model discriminative performance, with higher values indicating better identification of individuals likely to benefit from interventions. Subgroup-level treatment benefits were further quantified by reporting cumulative uplift within specific strata, such as the top 5% and top 20% of predicted beneficiaries.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e2.7. Statistical analysis\u003c/h2\u003e\u003cp\u003eContinuous variables are reported as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD (range) or median (IQR) and compared using Student\u0026rsquo;s t-test for normally distributed data) or the Wilcoxon rank-sum test for nonparametric data. Categorical variables are expressed as counts (percentages) and compared using Pearson\u0026rsquo;s chi-square test) or Fisher\u0026rsquo;s exact test when expected cell counts are small. Two-sided P\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant. All analyses were performed in R version 4.4.0 (released April 24, 2024).\u003c/p\u003e\u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Patient Characteristics\u003c/h2\u003e\u003cp\u003eFrom the 2015 wave of the CHARLS (n\u0026thinsp;=\u0026thinsp;21,038), 11,006 women who reported headache symptoms were identified. After excluding individuals with missing or implausible values and those not meeting inclusion criteria, a total of 1,930 middle-aged and older women were included in the final analysis. Among them, 1,361 (70.5%) had CESD-10 scores\u0026thinsp;\u0026ge;\u0026thinsp;10 and were classified as having depressive symptoms. The median age of the cohort was 59 years. In terms of sociodemographic characteristics, 94% were retired and 6% remained employed; 82% were married and 8% were unmarried; 69% resided in urban areas and 31% in rural or township areas. Educational attainment was distributed as follows: primary school or less (63%), secondary school (23%), high school (10%), and tertiary education or above (4%) (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Significant differences between depressed and non-depressed participants were observed across multiple domains, including psychosocial factors, socioeconomic status, health behaviors, functional health, comorbidity burden, physical performance, and physiological indicators.\u003c/p\u003e\u003cp\u003eFor temporal external validation, a separate sample of 1,518 women from the 2011 CHARLS wave, selected using analogous inclusion and exclusion criteria, was used as a time-series test set. The 2011 CHARLS cohort and the 2015 cohort differed in life satisfaction (n (%)), multisite pain, educational attainment, self-perceived health status (n (%)), comorbidities, and age (see Supplementary Table S2), thereby meeting the criteria for external validation.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Multi-Model Comparative Analysis\u003c/h2\u003e\u003cp\u003eWe systematically evaluated the predictive performance of ten machine learning models\u0026mdash;XGBoost, CatBoost, LightGBM, Random Forest, GBM, Neural Network, Logistic Regression, SVM, KNN, and AdaBoost\u0026mdash;across training, internal test, and temporal validation datasets. Comprehensive metrics including AUC, precision-recall AUC (PR-AUC), Brier score, accuracy, sensitivity, specificity, precision, and F1 score were used for robust assessment (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e; Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).Among all models, GBM consistently demonstrated superior overall performance and clinical applicability. It achieved AUCs of 0.823, 0.724, and 0.785 and PR-AUCs of 0.818, 0.855, and 0.901 in the training, test, and temporal validation sets, respectively, highlighting its excellent discrimination and strong capability in handling imbalanced data. GBM\u0026rsquo;s Brier scores were stably low (around 0.17\u0026ndash;0.18), reflecting good calibration. Notably, GBM balanced sensitivity (0.812 training; 0.73 test; 0.778 validation) and specificity (0.71 training; 0.631 test; 0.668 validation) effectively, with F1 scores consistently above 0.77, indicating reliable positive predictive performance.\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\u003eDetailed performance metrics of multiple machine learning models for predicting depression risk in patients with headache across training, testing, and validation datasets.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"9\"\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\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTraining set\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eThreshold\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSensitivity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eSpecificity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eF1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eBrierscore\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003ePR-AUC\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLogistic\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.525887\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.671\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.652\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.691\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.688\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.21484\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.713\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSVM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.542945\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.698\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.669\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.729\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.72\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.694\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.200852\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.746\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGBM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.492458\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.762\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.812\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.71\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.745\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.777\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.173917\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.818\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNeuralNetwork\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.548343\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.68\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.599\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.763\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.726\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.657\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.2075\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.742\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRandomForest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.493\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.027376\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e1\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\u003e0.512044\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.722\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.737\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.707\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.724\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.731\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.211511\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.79\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eKNN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.545801\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.83\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.826\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.835\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.839\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.833\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.137648\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.921\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAdaboost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.783073\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.664\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.488\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.848\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.77\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.597\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.240072\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.707\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLightGBM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.476093\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.75\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.787\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.711\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.74\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.763\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.175359\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.798\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCatBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.620233\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.737\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.794\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.678\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.72\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.755\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.24324\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.803\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eTest set\u003c/b\u003e\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\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eThreshold\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSensitivity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eSpecificity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eF1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eBrierscore\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003ePR-AUC\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLogistic\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.447533\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.724\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.793\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.557\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.812\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.803\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.205221\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.855\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSVM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.509945\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.689\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.723\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.608\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.817\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.767\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.203391\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.837\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGBM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.687687\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.701\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.73\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.631\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.827\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.776\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.177467\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.855\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNeuralNetwork\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.551785\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.63\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.573\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.767\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.856\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.686\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.208239\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.853\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRandomForest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.805\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.585\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.488\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.818\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.867\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.625\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.184902\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.836\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\u003e0.552936\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.683\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.692\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.659\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.831\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.755\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.208459\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.85\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eKNN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.613803\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.605\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.587\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.648\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.801\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.678\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.21559\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.805\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAdaboost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.65\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.683\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.568\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.793\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.734\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.230642\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.776\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLightGBM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.753478\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.645\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.622\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.699\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.833\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.712\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.196562\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.817\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCatBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.639369\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.716\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.763\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.602\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.823\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.792\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.199979\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.86\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003etemporal validation set\u003c/b\u003e\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\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eThreshold\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSensitivity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eSpecificity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eF1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eBrierscore\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003ePR-AUC\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLogistic\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.510353\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.706\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.686\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.761\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.886\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.773\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.205221\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.904\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSVM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.496642\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.735\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.747\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.702\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.872\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.805\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.203391\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.893\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGBM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.661602\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.748\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.778\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.668\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.864\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.819\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.177467\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.901\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNeuralNetwork\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.474907\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.719\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.719\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.72\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.874\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.789\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.208239\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.904\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRandomForest\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.711\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.688\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.676\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.72\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.867\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.76\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.184902\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.88\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\u003e0.536153\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.727\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.749\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.668\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.859\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.208459\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.894\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eKNN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.651937\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.632\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.606\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.702\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.846\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.706\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.21559\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.851\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAdaboost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.668\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.644\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.734\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.867\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.739\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.230642\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.838\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLightGBM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.753478\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.672\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.678\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.849\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.749\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.196562\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.827\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCatBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0.626773\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.742\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.773\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.661\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.86\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.814\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e0.199979\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.894\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eRandom Forest, despite near-perfect metrics in the training set (accuracy and F1 of 1.0), showed significant performance degradation on the test and validation sets, indicative of overfitting and limiting its clinical utility. XGBoost and LightGBM also demonstrated robust discrimination with AUCs and PR-AUCs slightly below GBM, but exhibited somewhat higher Brier scores and less consistent calibration across datasets. CatBoost showed comparable discrimination (AUCs\u0026thinsp;~\u0026thinsp;0.74) and PR-AUCs (up to 0.894) but had relatively poorer calibration (Brier score\u0026thinsp;~\u0026thinsp;0.20), potentially restricting its applicability in clinical settings.\u003c/p\u003e\u003cp\u003eThe superior PR-AUC values of GBM, especially in the temporal validation cohort (0.901), underscore its strength in predicting true positives in a context where class imbalance exists. This is critical for clinical scenarios where accurate identification of at-risk patients is paramount. Overall, GBM\u0026rsquo;s combination of high discrimination, stable calibration, and balanced sensitivity and specificity across datasets positions it as the most reliable and clinically valuable predictive model for individualized risk stratification\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Model Interpretation and Application\u003c/h2\u003e\u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e\u003cb\u003e(K.L.M)\u003c/b\u003e presents the results of the feature importance analysis and SHAP summary plots for the GBM model, offering a visual interpretation of how each predictor contributes to the model's output. In these plots, higher SHAP values represent a stronger positive impact on the prediction of depression risk, whereas lower or negative SHAP values indicate weaker or mitigating effects. The SHAP summary plot employs a color gradient from purple (low feature value) to yellow (high feature value), effectively illustrating both the distribution and directional influence of individual feature values on the model's predictions.\u003c/p\u003e\u003cp\u003eAmong all features, hope for the future (hope) exhibited the widest SHAP value distribution, suggesting it had the most substantial influence on model predictions. Notably, lower levels of hope (indicated by purple) were associated with higher SHAP values, implying that hopelessness is a strong risk factor for depression. Other key features\u0026mdash;sleep duration, life satisfaction (satlife), and self-rated health (srh)\u0026mdash;also showed significant impact, where lower values were generally associated with higher predicted risk of depression. Additionally, multisite pain and history of falls were positively associated with predicted depression risk, indicating that individuals experiencing frequent pain or with a history of falling were more likely to be classified as depressed by the model. While retirement status (retire) had a smaller overall contribution, its SHAP distribution revealed a slight tendency for non-retired individuals to be more prone to depression. Based on these key features, we developed an interactive web-based calculator using Shiny, allowing healthcare professionals to conveniently estimate individual depression risk:\u003c/p\u003e\u003cp\u003e\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://wenxiaohui.shinyapps.io/gbmforshinny/\u003c/span\u003e\u003cspan address=\"https://wenxiaohui.shinyapps.io/gbmforshinny/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003e3.4 Individual Treatment Effect (ITE) Estimation and Stratified Response Analysis\u003c/h2\u003e\u003cp\u003eBased on prior subgroup comparison (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eN.O), \u003cem\u003elife satisfaction\u003c/em\u003e emerged as the most distinctive categorical variable differentiating high-risk individuals from the overall population, with substantially lower proportions of high life satisfaction levels (satlife_4 and satlife_5) observed in the high-ITE subgroup. Motivated by this finding, we further estimated the ITEs using an uplift modeling approach, treating life satisfaction as a modifiable proxy intervention (\u0026ge;\u0026thinsp;3 vs\u0026thinsp;\u0026lt;\u0026thinsp;3). The ITE distribution (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eP) revealed significant inter-individual heterogeneity in predicted treatment responses. A large proportion of individuals exhibited negative ITEs (i.e., reduced depression probability under higher life satisfaction), suggesting a generally beneficial effect. Notably, approximately 5% of the population demonstrated highly negative ITEs, indicating strong potential benefit from life satisfaction improvement. These individuals represent key targets for preventive intervention. This finding is further supported by the Qini curve(see Supplementary Picture S2), According to the Qini curve, the model achieved an uplift AUC of 0.558, slightly above the random level. The cumulative incremental response for the top 5% of high-score individuals was 15, and for the top 20% it was 37, indicating that the model can to some extent identify potential beneficiaries and maximize the effect of targeted interventions.\u003c/p\u003e\u003c/div\u003e"},{"header":"4. DISSCUION","content":"\u003cp\u003eThis study systematically developed and validated ten mainstream machine learning algorithms using the CHARLS database to predict depression risk among middle-aged and older women with headaches. Across the training, internal test, and temporal validation datasets, the GBM model demonstrated superior and stable predictive performance, achieving AUCs of 0.823, 0.724, and 0.785(Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB.C.D), respectively, alongside consistent Brier scores around 0.17, indicating excellent discrimination and calibration. In contrast, although Random Forest performed exceptionally well on the training set, its marked decline in validation sets suggested overfitting, limiting its clinical utility. CatBoost, despite competitive metrics, showed poorer probability calibration, reducing its practical applicability. The GBM model\u0026rsquo;s balanced performance in accuracy, sensitivity, specificity, and F1 score highlights its rationale as the optimal model with significant clinical potential results (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Given its strong discrimination and calibration capabilities, GBM was ultimately selected as the optimal model and deployed through a Shiny-based web application to facilitate personalized risk estimation. Compared with traditional psychometric tools, machine learning models allow for the integration of multidimensional variables and the exploration of nonlinear and interactive effects, thereby improving predictive accuracy. Moreover, decision curve analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eJ) indicated that the GBM model could guide the determination of optimal clinical intervention thresholds, enhancing the utility of community-based screening programs.\u003c/p\u003e\u003cp\u003eUsing univariate, multivariate logistic regression, and Lasso regression, we identified seven core predictors of depression: self-rated health, sense of hope, life satisfaction, history of falls, retirement status, multisite pain, and Insomnia. All of these variables have been previously linked to depression in existing literature(\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan additionalcitationids=\"CR26\" citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e), and their interactive effects further enhanced the interpretability of our model. For instance, poor self-rated health is a consistent predictor of depressive symptoms(\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e). Hope for the future and life satisfaction, as components of subjective well-being, are protective psychological resources among older adults. A Swedish study found that retirement generally improves depressive symptoms and enhances well-being, although residual occupational stress may exacerbate symptoms in certain individuals(\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e). Additionally, the interaction between sleep disturbances and work-related stress is known to aggravate depressive outcomes(\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e), while chronic multisite pain can increase depression risk through mechanisms such as hormonal dysregulation and impaired sleep quality.\u003c/p\u003e\u003cp\u003eIn recent years, the association between headache and depression has attracted increasing academic and clinical attention. Epidemiological studies have shown that individuals with migraine are approximately 38% more likely to experience depression than the general population(\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e). In patients with headache disorders, female individuals exhibit a higher suicide rate(\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e). Adults suffering from migraine in combination with depression and/or anxiety tend to have significantly higher healthcare utilization and associated costs, reflecting a substantial burden on both patients and healthcare systems(\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e). Despite such evidence, studies specifically focusing on middle-aged and older women with headache\u0026mdash;particularly those utilizing machine learning techniques for risk modeling\u0026mdash;remain scarce. In the domain of predictive modeling, Zheng et al. developed a depression risk prediction model among older adults using CHARLS data and found that the random forest algorithm yielded the best performance; however, the study lacked external validation(\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e). Shaojie Duan and colleagues expanded the migraine attack prediction model for the Chinese population by integrating traditional Chinese medicine (TCM) theory, and explored the correlations among depression, headache, and metabolic syndrome based on TCM principles(\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e). Furthermore, Amoozegar\u0026rsquo;s study demonstrated that screening tools such as PHQ-9 and HADS perform well in migraine patients; however, optimal cut-off points for depression screening vary according to specific clinical objectives. Moreover, the high prevalence of depression among patients in headache clinics and the frequent undertreatment emphasize the necessity of scientifically validated predictive and screening models. These models are vital for the early identification of depression comorbidity in headache sufferers, enabling timely intervention and reducing disease burden(\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e). One of the most notable findings is the central role of life satisfaction as a modifiable, high-impact predictor. While prior research has consistently linked low life satisfaction to increased depression risk in older adults, few studies have translated this association into actionable, personalized interventions(\u003cspan additionalcitationids=\"CR38\" citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e). Our ITE results demonstrated that individuals with lower life satisfaction would derive the greatest benefit from targeted psychosocial interventions, suggesting that improving life satisfaction should be prioritized as a key preventive strategy. (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e). This finding aligns with literature supporting life satisfaction as a core dimension of subjective well-being and a potential protective factor against mental health decline.\u003c/p\u003e\u003cp\u003eThe model\u0026rsquo;s strong and consistent discrimination in both the internal test set and an independent temporal validation cohort highlights its generalizability, despite differences in baseline characteristics and population profiles (see Supplementary Table S2). Importantly, our study not only demonstrated statistical robustness but also addressed the practical gap between model development and clinical application by deploying an interactive Shiny-based web calculator. This tool enables real-time, individualized risk assessment using easily accessible inputs, thereby supporting integration into primary care and community health screening.\u003c/p\u003e\u003cp\u003eCompared with existing depression prediction models, which often rely solely on sociodemographic or symptom-based screening tools, our approach offers three advantages: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) the integration of machine learning for optimized variable selection and performance, (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) external validation to enhance credibility and applicability, and (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) the novel application of ITE analysis for precision targeting of interventions. These features strengthen the model\u0026rsquo;s translational potential in both preventive and therapeutic contexts. Nonetheless, this study has several limitations. First, although the temporal validation set represents an independent time point, it originates from the same data source as the training set, which may introduce temporal bias. Second, the retrospective nature of the study limits data reliability, as most predictors were self-reported rather than objectively measured. Third, while the model has undergone internal and temporal validation, prospective studies are still warranted to confirm its clinical applicability and real-world utility. Additionally, the ITE analysis, although useful for identifying patient subgroups with differential predicted risk, is limited by potential unmeasured confounding and model assumptions. Therefore, the interpretation of ITE results should be cautious, and further validation in independent cohorts and prospective trials is necessary to fully establish its clinical relevance. In conclusion, our study advances the field by integrating machine learning and ITE analysis to develop a clinically interpretable and actionable depression risk prediction model for middle-aged and older women with headache. By highlighting life satisfaction as a strategic target for precision prevention, this work bridges the gap between statistical modeling and practical mental health care, offering a pathway toward more personalized and effective intervention strategies in this high-risk population.\u003c/p\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eThis study developed and externally validated a machine learning\u0026ndash;based prediction model for depression risk in middle-aged and older women with headache, integrating both population-level prediction and individualized intervention analysis via ITE. The model demonstrated strong discrimination and calibration across datasets, underscoring its generalizability and clinical utility. Our findings highlight life satisfaction as a pivotal and modifiable predictor, with ITE analysis revealing that individuals with low life satisfaction are likely to experience the greatest preventive benefit from targeted psychosocial interventions. By translating the model into an interactive web-based tool, we provide a practical solution for clinicians and public health practitioners to implement personalized screening and precision prevention strategies. Future research should focus on prospective validation in diverse populations and experimental evaluation of targeted life satisfaction\u0026ndash;enhancing interventions. Such efforts will be essential to confirm causal effects and optimize intervention strategies, ultimately contributing to improved mental health outcomes and reduced suicide risk in this high-vulnerability population.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; marry\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eMarital status\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; rural\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eResidence\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; srh\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eSelf-rated health\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; hope\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eHope for the future\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; satlife\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eLife satisfaction\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; falldown\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eHistory of falls\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; retire\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eRetirement status\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; mulpain\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eMultisite pain\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; edu\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eEducation level\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; disability\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eDisability\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; hibpe\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eHypertension\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; diabe\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eDiabetes\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; cancre\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eCancer\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; arthre\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eArthritis\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; lunge\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eChronic lung disease\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; hearte\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eCardiovascular diseases\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; stroke\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eStroke\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; dyslipe\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eDyslipidemia\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; livere\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eLiver disease\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; kidneye\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eKidney disease\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; digeste\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eGastrointestinal disorders\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; asthmae\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eAsthma\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; memrye\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eMemory-related disorders\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; drinkev\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eAlcohol use\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; smoken\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eSmoking status\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; ins\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eHealth insurance coverage\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; systo\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eSystolic blood pressure\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; diasto\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eDiastolic blood pressure\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; pulse\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003ePulse rate\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; lgrip\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eGrip strength of left hand\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; rgrip\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eGrip strength of right hand\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; sleep\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eSleep duration\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003e\u0026bull; puff\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eLung function\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003e\u003cem\u003e9.1Ethics approval and consent to participate\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOur study uses the China Health and Retirement Longitudinal Study (CHARLS) public dataset and does not require additional Institutional Review Board approval as the primary data collection was approved by the Biomedical Ethics Review Committee of Peking University (IRB00001052\u0026ndash;11015). Written informed consent was obtained from all participants by the principal investigators of the survey. The methods in this study were carried out in accordance with the Declaration of Helsinki and relevant guidelines and regulations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003e9.2Consent for publication\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003e9.3Availability of data and materials\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets analyzed during the current study include the 2011 and 2015 waves of the China Health and Retirement Longitudinal Study (CHARLS). Access to CHARLS data is publicly available upon application at http://charls.pku.edu.cn. The datasets used and/or analyzed during the current study are also available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eCompeting Interests\u003c/em\u003e\u003c/strong\u003e\u003cem\u003e\u003cbr\u003e\u0026nbsp;The authors declare that they have no competing interests.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e10.hors\u0026apos; Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eXiaohui Wen\u003c/strong\u003e\u003cstrong\u003e:\u003c/strong\u003e Conceptualization, Methodology, Software, Formal Analysis, Visualization.\u003cbr\u003e\u003cstrong\u003eZhijun\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;Lin:\u003c/strong\u003e Investigation, Data Curation, Validation.\u003cbr\u003e\u003cstrong\u003eYihui\u003c/strong\u003e\u003cstrong\u003eQian:\u003c/strong\u003eWriting \u0026ndash; Original Draft, Writing \u0026ndash; Review \u0026amp; Editing.\u003cbr\u003e\u003cstrong\u003eLin Wang:\u003c/strong\u003e Data Curation, Validation, Software.\u003cbr\u003e\u003cstrong\u003eJingtong Zhou:\u003c/strong\u003e Data Curation, Resources, Formal Analysis.\u003cbr\u003e\u003cstrong\u003eTingting Wen:\u003c/strong\u003e Methodology, Data Transformation, Visualization.\u003cbr\u003e\u003cstrong\u003eQianying\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;Cao:\u003c/strong\u003e Conceptualization, Data Reduction, Quality Control.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eZhou Liu, M.D.:\u003c/strong\u003e Conceptualization (lead), Funding Acquisition (lead).\u003cbr\u003e\u003cstrong\u003eAll authors:\u003c/strong\u003e Read and approved the final manuscript; contributed to critical revisions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e11.Funding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the following funding sources:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eGuangdong Provincial Basic and Applied Basic Research Fund (Grant No. 2024A1515220002)\u003c/li\u003e\n \u003cli\u003eClinical + Basic Research Project of Guangdong Medical University (Grant No. 4SG23284G)\u003c/li\u003e\n \u003cli\u003eGuangdong Provincial Medical Association Clinical Research Fund - Healthcare Special (Grant No. 2024HY-A6006)\u003c/li\u003e\n \u003cli\u003eGuangdong Medical University Affiliated Hospital High-level Talent Research Launch Fund (Grant No. GCC2022011)\u003c/li\u003e\n \u003cli\u003e2023 Special Project of the Songshan Lake Medical-Engineering Integration Innovation Center of Guangdong Medical University (Grant No. 4SG22307P)\u003c/li\u003e\n \u003cli\u003e2023 Guangdong Provincial Administration of Traditional Chinese Medicine Research Project (Project No. 20232100)\u003c/li\u003e\n\u003c/ul\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eGlobal regional, national burden of neurological disorders. 1990\u0026ndash;2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet Neurol. 2019;18(5):459\u0026ndash;80.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVetvik KG, MacGregor EA. Sex differences in the epidemiology, clinical features, and pathophysiology of migraine. Lancet Neurol. 2017;16(1):76\u0026ndash;87.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKuan AS, Wang YF, Chen SP, Chuang YF, Wang SJ. Sex differences in pain, suicidal ideation, and suicide attempts in patients with migraine. Headache. 2025.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBolay H, Ozge A, Saginc P, Orekici G, Ulud\u0026uuml;z D, Yalın O, et al. Gender influences headache characteristics with increasing age in migraine patients. Cephalalgia. 2015;35(9):792\u0026ndash;800.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBuse DC, Manack AN, Fanning KM, Serrano D, Reed ML, Turkel CC, et al. Chronic migraine prevalence, disability, and sociodemographic factors: results from the American Migraine Prevalence and Prevention Study. Headache. 2012;52(10):1456\u0026ndash;70.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSlavich GM, Irwin MR. From stress to inflammation and major depressive disorder: a social signal transduction theory of depression. Psychol Bull. 2014;140(3):774\u0026ndash;815.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePan A, Sun Q, Okereke OI, Rexrode KM, Hu FB. Depression and risk of stroke morbidity and mortality: a meta-analysis and systematic review. JAMA. 2011;306(11):1241\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVieira WF, Coelho DRA, Litwiler ST, McEachern KM, Clancy JA, Morales-Quezada L, et al. Neuropathic pain, mood, and stress-related disorders: A literature review of comorbidity and co-pathogenesis. Neurosci Biobehav Rev. 2024;161:105673.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRist PM, Sch\u0026uuml;rks M, Buring JE, Kurth T. Migraine, headache, and the risk of depression: Prospective cohort study. Cephalalgia. 2013;33(12):1017\u0026ndash;25.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBeam AL, Kohane IS. Big Data and Machine Learning in Health Care. JAMA. 2018;319(13):1317\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhao Y, Hu Y, Smith JP, Strauss J, Yang G. Cohort profile: the China Health and Retirement Longitudinal Study (CHARLS). Int J Epidemiol. 2014;43(1):61\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eIrwin M, Artin KH, Oxman MN. Screening for depression in the older adult: criterion validity of the 10-item Center for Epidemiological Studies Depression Scale (CES-D). Arch Intern Med. 1999;159(15):1701\u0026ndash;4.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAndresen EM, Malmgren JA, Carter WB, Patrick DL. Screening for depression in well older adults: evaluation of a short form of the CES-D (Center for Epidemiologic Studies Depression Scale). Am J Prev Med. 1994;10(2):77\u0026ndash;84.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang Z. Multiple imputation with multivariate imputation by chained equation (MICE) package. Ann Transl Med. 2016;4(2):30.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZeng ZQ, Wu Q, Liao BS, Gao J. Tien Tzu Hsueh Pao/Acta Electronica Sinica. 2009;37(11):2489\u0026ndash;95. A Classfication Method For Imbalance Data Set Based on Kernel SMOTE.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAli YA, Awwad EM, Al-Razgan M, Maarouf A. Hyperparameter Search for Machine Learning Algorithms for Optimizing the Computational Complexity. Processes. 2023;11(2):349.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCox DR. The Regression Analysis of Binary Sequences. J Roy Stat Soc: Ser B (Methodol). 2018;20(2):215\u0026ndash;32.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFriedman JH. Greedy Function Approximation: A Gradient Boosting Machine. Ann Stat. 2001;29(5):1189\u0026ndash;232.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShao F, Shen Z. How can artificial neural networks approximate the brain? Front Psychol. 2022;13:970214.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBreiman L. Random forests, machine learning 45. J Clin Microbiol. 2001;2:199\u0026ndash;228.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCover TM, Hart PE. Nearest neighbor pattern classification. IEEE Transinftheory. 1953;13(1):21\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFreund Y, Schapire R. A decision-theoretic generalization of on-line learning and an application to boosting. J Comput Syst Sci. 1997.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMeng Q, editor. editor LightGBM: A Highly Efficient Gradient Boosting Decision Tree. Neural Information Processing Systems; 2017.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLundberg SM, Erion G, Chen H, Degrave A, Lee SI. Explainable AI for Trees: From Local Explanations to Global Understanding. 2019.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBenca RM, Peterson MJ. Insomnia and depression. Sleep Med. 2008;9(Suppl 1):S3\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGoldzweig G, Baider L, Andritsch E, Pfeffer R, Rottenberg Y. A Dialogue of Depression and Hope: Elderly Patients Diagnosed with Cancer and Their Spousal Caregivers. J Cancer Educ. 2017;32(3):549\u0026ndash;55.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHuang R, Li S, Hu J, Ren R, Ma C, Peng Y, et al. Adverse childhood experiences and falls in older adults: The mediating role of depression. J Affect Disord. 2024;365:87\u0026ndash;94.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMaier A, Riedel-Heller SG, Pabst A, Luppa M. Risk factors and protective factors of depression in older people 65+. A systematic review. PLoS ONE. 2021;16(5):e0251326.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWeber E, H\u0026uuml;l\u0026uuml;r G. Co-development of Couples' Life Satisfaction in Transition to Retirement: A Longitudinal Dyadic Perspective. J Gerontol B Psychol Sci Soc Sci. 2021;76(8):1542\u0026ndash;54.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e\u0026Aring;hlin JK, Peristera P, Westerlund H, Magnusson Hanson LL. Psychosocial working characteristics before retirement and depressive symptoms across the retirement transition: a longitudinal latent class analysis. Scand J Work Environ Health. 2020;46(5):488\u0026ndash;97.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMagnusson Hanson LL, Peristera P, Chungkham HS, Westerlund H. Psychosocial work characteristics, sleep disturbances and risk of subsequent depressive symptoms: a study of time-varying effect modification. J Sleep Res. 2017;26(3):266\u0026ndash;76.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMollan SP, Subramanian A, Perrins M, Nirantharakumar K, Adderley NJ, Sinclair AJ. Depression and anxiety in women with idiopathic intracranial hypertension compared to migraine: A matched controlled cohort study. Headache. 2023;63(2):290\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVuralli D, Ayata C, Bolay H. Cognitive dysfunction and migraine. J Headache Pain. 2018;19(1):109.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZheng Y, Zhang C, Liu Y. Risk prediction models of depression in older adults with chronic diseases. J Affect Disord. 2024;359:182\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDuan S, Xia H, Zheng T, Li G, Ren Z, Ding W, et al. Development and validation of non-invasive prediction models for migraine in Chinese adults. J Headache Pain. 2023;24(1):148.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAmoozegar F, Patten SB, Becker WJ, Bulloch AGM, Fiest KM, Davenport WJ, et al. The prevalence of depression and the accuracy of depression screening tools in migraine patients. Gen Hosp Psychiatry. 2017;48:25\u0026ndash;31.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCheng Y, Wei Y, Tang SL. Does Helping Others Always Benefit Health? Longitudinal Evidence on the Relationship between Helping Behavior and Depression: The Mediating Role of Life Satisfaction and the Moderating Effect of IADL. Depress Anxiety. 2024;2024:2304723.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTong Y, Xie F, Wen X, Li Y, Yuan M, Zhang X, et al. Longitudinal Association between Bullying Victimization and Depressive Symptoms in Chinese Early Adolescents: The Effect of Life Satisfaction. Depress Anxiety. 2024;2024:6671415.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAbdin PVA, Roystonn E, Devi K, Wang F, Shafie P. Tracking the Prevalence of Depression Among Older Adults in Singapore: Results From the Second Wave of the Well-Being of Singapore Elderly Study. Depress Anxiety. 2025;2025:9071391.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eK\u0026uuml;nzel SR, Sekhon JS, Bickel PJ, Yu B. Metalearners for estimating heterogeneous treatment effects using machine learning. Proc Natl Acad Sci U S A. 2019;116(10):4156\u0026ndash;65.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Headache, Depression, Middle-aged and Older women, Risk Prediction Model, Machine Learning, Shiny Application, Mental Health, ITE","lastPublishedDoi":"10.21203/rs.3.rs-7517256/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7517256/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eDepression commonly co-occurs with headache, especially among middle-aged and older women, increasing symptom burden and healthcare utilization. To enable early identification and targeted intervention, we developed a machine learning\u0026ndash;based risk prediction model and deployed an interactive Shiny web application. Data from 1,930 individuals aged\u0026thinsp;\u0026ge;\u0026thinsp;45 years with self-reported headache in the 2015 wave of the China Health and Retirement Longitudinal Study were analyzed and randomly split into training and testing subsets. Depressive symptoms were assessed using a validated instrument, and seven predictors spanning clinical and lifestyle domains were selected. Ten machine learning algorithms were compared using discrimination, calibration, and decision curve analysis, with temporal validation on the 2011 CHARLS wave. The gradient boosting machine model achieved the best performance (area under the curve 0.823 training, 0.724 test, 0.785 temporal validation) and favorable calibration. Key predictors included hope, sleep duration, life satisfaction, and self-rated health. Individualized treatment effect analysis identified approximately 5% of participants most likely to benefit from interventions enhancing life satisfaction, and the Qini curve confirmed heterogeneity in treatment effects (uplift area under the curve 55.8). Targeting interventions based on predicted risk can achieve greater benefits than random allocation, supporting precision mental health strategies. This model and its Shiny tool facilitate early identification and tailored intervention for high-risk middle-aged and older women with headache, warranting prospective validation.\u003c/p\u003e","manuscriptTitle":"A Predictive Model for Headache-Related Depression in Middle- Aged and Older Women Incorporating Individual Treatment Effects","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-26 13:25:10","doi":"10.21203/rs.3.rs-7517256/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"e34b00f6-33e2-42d1-b71b-3a13034d1fe4","owner":[],"postedDate":"October 26th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-08T08:28:41+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-26 13:25:10","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7517256","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7517256","identity":"rs-7517256","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}