Open Access

ARTICLE

The Plateau Dilemma: Identifying Key Factors of Depression Risk among Middle-Aged and Older Chinese with Chronic Diseases

Zhe He¹, Yaning Zhang^2,*

1 Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, 430022, China
2 School of Economics, Sichuan University, Chengdu, 610065, China

* Corresponding Authors: Yaning Zhang. Email: ,

(This article belongs to the Special Issue: Depression Across the Lifespan: Perspectives on Prevention, Intervention, and Holistic Care)

International Journal of Mental Health Promotion 2025, 27(11), 1747-1768. https://doi.org/10.32604/ijmhp.2025.070491

Received 17 July 2025; Accepted 26 September 2025; Issue published 28 November 2025

Abstract

Background: Depression represents a significant global mental health burden, particularly among middle-aged and older Chinese with chronic diseases in high-altitude regions, where harsh environmental conditions and limited social support exacerbate mental health disparities. This paper aims to develop an interpretable machine learning prediction framework to identify the key factors of depression in this vulnerable population, thereby proposing targeted intervention measures. Methods: Utilizing data from the China Health and Retirement Longitudinal Study in 2020, this paper screened out and analyzed 2431 samples. Subsequently, Recursive Feature Elimination and Least Absolute Shrinkage and Selection Operator were applied to screen predictors from 32 alternative variables. Furthermore, through hyperparameter tuning and 5-fold cross-validation, 8 machine learning models were constructed, namely, Random Forest, Extreme Gradient Boosting, Light Gradient Boosting Machine, Gradient Boosting Machine, K-Nearest Neighbor, Naive Bayes Classifier, Support Vector Machine, and Logistic Regression. Finally, the SHAP algorithm was applied to analyze the interpretability of the best-performing model, quantifying nonlinear relationships and threshold effects. Results: Among the respondents, the prevalence of depression was approximately 46.89%. After feature engineering screening, 8 variables were retained for inclusion in the prediction model. Furthermore, the Gradient Boosting Machine performed optimally in terms of comprehensive performance, with an Area Under Receiver Operating Characteristic Curve (AUC) of 0.845, an Accuracy of 0.714, a Sensitivity of 0.655, a Precision of 0.711, a Specificity of 0.766, and an F1 of 0.682. In addition, Life satisfaction, PM2.5, Self-rated health, and Education were identified as the top 4 key factors. Meanwhile, the influence of these variables on depression showed nonlinear and threshold effects. Conclusion: This research highlights the value of machine learning in mental health. Based on the identified key factors, this paper proposed a series of policy measures to improve the health pattern of the middle-aged and elderly populations facing the dual challenges of chronic disease and environmental adversity.

---

Keywords

---