Scholarlink

Objective <p>Depressive symptoms are significant and deleterious complication of cardiovascular and cerebrovascular diseases (CCVD), profoundly impairing patients’ quality of life. This study aimed to develop and validate an online screening model to estimate the risk of depressive symptoms among patients with CCVD.</p> Methods <p>This study utilized data from the 2020 China Health and Retirement Longitudinal Study (CHARLS), including 4,086 patients with CCVD. A total of 31 behavioral, health-related, psychological, and sociodemographic indicators were evaluated. Feature selection was performed using least absolute shrinkage and selection operator (LASSO) regression, Random Forest (RF), and the Boruta algorithm. Six machine learning models, Logistic Regression (LR), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Decision Tree, RF, and XGBoost, were developed and compared. The best-performing model was used to construct a clinically interpretable nomogram, which was deployed as a real-time, web-based risk screening tool via a Shiny application. Temporal validation was conducted using CHARLS 2015 data.</p> Results <p>Among the 4,086 CCVD patients included, 1,990 (48.70%) exhibited depressive symptoms. Feature selection identified 13 key variables. The Logistic Regression model demonstrated superior performance among the tested algorithms, with an AUC of 0.767, and showed moderate discriminative ability (C-index = 0.776, 95% CI: 0.761–0.790) and good calibration (Hosmer–Lemeshow test, <i>P</i> = 0.51). The model was implemented into a freely accessible online calculator for real-time risk assessment.</p> Conclusion <p>Using machine learning approaches, this study developed and validated a screening model with moderate discriminative ability for estimating the risk of depressive symptoms in CCVD patients. The online Shiny-based calculator serves as a practical screening tool for clinicians, supporting early identification and intervention in high-risk populations. Further prospective validation is warranted to assess its real-world utility.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A machine-learning-derived online screening tool for depressive symptoms in patients with cardiovascular and cerebrovascular diseases (CCVD): a cross-sectional study with temporal validation from CHARLS

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月26日

查看全文
  • Chun-Juan Zhang,
  • Min-Xia Liu,
  • Mei-Zhen Wu,
  • Xi-Cheng Zhou,
  • Xiao-Dong Ma

摘要

Objective

Depressive symptoms are significant and deleterious complication of cardiovascular and cerebrovascular diseases (CCVD), profoundly impairing patients’ quality of life. This study aimed to develop and validate an online screening model to estimate the risk of depressive symptoms among patients with CCVD.

Methods

This study utilized data from the 2020 China Health and Retirement Longitudinal Study (CHARLS), including 4,086 patients with CCVD. A total of 31 behavioral, health-related, psychological, and sociodemographic indicators were evaluated. Feature selection was performed using least absolute shrinkage and selection operator (LASSO) regression, Random Forest (RF), and the Boruta algorithm. Six machine learning models, Logistic Regression (LR), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Decision Tree, RF, and XGBoost, were developed and compared. The best-performing model was used to construct a clinically interpretable nomogram, which was deployed as a real-time, web-based risk screening tool via a Shiny application. Temporal validation was conducted using CHARLS 2015 data.

Results

Among the 4,086 CCVD patients included, 1,990 (48.70%) exhibited depressive symptoms. Feature selection identified 13 key variables. The Logistic Regression model demonstrated superior performance among the tested algorithms, with an AUC of 0.767, and showed moderate discriminative ability (C-index = 0.776, 95% CI: 0.761–0.790) and good calibration (Hosmer–Lemeshow test, P = 0.51). The model was implemented into a freely accessible online calculator for real-time risk assessment.

Conclusion

Using machine learning approaches, this study developed and validated a screening model with moderate discriminative ability for estimating the risk of depressive symptoms in CCVD patients. The online Shiny-based calculator serves as a practical screening tool for clinicians, supporting early identification and intervention in high-risk populations. Further prospective validation is warranted to assess its real-world utility.