Depression is a pervasive mental health challenge worldwide, but its intersection with chronic diseases remains an area of intense study, particularly in low- and middle-income countries. A groundbreaking study conducted in Bangladesh has now shed light on the prevalence of probable depression among individuals living with chronic illnesses, revealing alarming rates and uncovering crucial socio-behavioral and clinical factors that exacerbate mental health risks in this vulnerable population.
The research, published in the esteemed journal BMC Psychiatry, represents one of the first comprehensive investigations leveraging advanced machine learning techniques to predict probable depression in chronic disease patients in Bangladesh. Chronic conditions such as diabetes, hypertension, and heart disease often complicate patients’ mental well-being, yet quantifying and predicting associated depression has been challenging, primarily due to limited data and sociocultural underreporting.
In a meticulously designed cross-sectional study, researchers enrolled 1,222 adults diagnosed with various chronic diseases from multiple healthcare centers across Bangladesh over a six-month period in 2024. This extensive recruitment effort ensured broad demographic representation, encompassing a diversity of urban and rural settings essential for understanding geographic disparities in mental health outcomes.
Data collection involved structured interviews capturing a rich spectrum of domains, including sociodemographic characteristics, lifestyle behaviors such as tobacco and alcohol use, physical activity, sleep patterns, family medical history, and indicators of unmet mental healthcare needs. Depression was assessed using the validated Bangla version of the Patient Health Questionnaire-9 (PHQ-9), a globally recognized tool for detecting probable depressive disorders.
What distinguishes this study is its dual analytical approach: employing both traditional multivariable logistic regression and cutting-edge machine learning (ML) algorithms to analyze the dataset. This strategy allowed for a robust assessment of predictive factors and the development of computational models capable of identifying patients at elevated risk of depression with greater precision than conventional statistical techniques.
The results revealed a striking prevalence rate of 29.7% for probable depression among the chronic disease cohort, signaling a substantial burden of comorbid psychiatric distress that could impede effective disease management. Notably, the adjusted regression models identified multiple risk factors significantly associated with depression, including unemployment, residence in urban areas, consumption of smokeless tobacco and alcohol, substance use, physical inactivity, shorter nighttime sleep duration under seven hours, a family history of chronic illnesses, and crucially, unmet mental healthcare needs.
Machine learning analyses unveiled further nuances in risk prediction. Among six ML algorithms evaluated, CatBoost emerged as the superior model, achieving an impressive accuracy of 71.1% and an area under the receiver operating characteristic curve (AUC) of 0.76, benchmarks that underscore its efficacy in distinguishing depressed from non-depressed patients. Interpretability of machine learning outputs was enhanced through SHapley Additive exPlanations (SHAP) and feature importance techniques, which consistently highlighted residence, employment status, family medical history, and mental healthcare access as the most influential predictors.
Urban residence as a key risk factor challenges conventional assumptions that rural isolation predisposes to depression, suggesting that the stresses inherent in urban environments, including socioeconomic pressures and lifestyle factors, may disproportionately affect mental health among chronic disease sufferers. Unemployment’s association with depression further implicates economic stability as a determinant of psychological resilience.
The identification of substance use behaviors and physical inactivity as modifiable correlates opens potential avenues for integrated intervention strategies targeting both physical and mental health. Short sleep duration’s linkage to depressive symptoms corroborates growing evidence that sleep hygiene is critical for emotional regulation, especially in medically vulnerable populations.
A particularly concerning finding is that many patients faced gaps in mental healthcare fulfillment, signifying barriers to accessing psychological support that could mitigate depression’s impact. This deficiency underscores the urgent need for healthcare systems in Bangladesh to prioritize mental health services within chronic disease management frameworks.
The successful application of machine learning in this context demonstrates the transformative potential of data science in public health. Predictive models like CatBoost can facilitate targeted pre-screening, enabling healthcare providers to identify and intervene early with individuals at higher risk of depression, optimizing resource allocation and enhancing patient outcomes.
This study’s comprehensive approach, integrating epidemiological rigor with technological innovation, provides a blueprint for similar investigations globally, particularly in settings where mental health remains stigmatized and under-addressed. It also calls attention to the complex interplay of socioeconomic, behavioral, and clinical determinants that shape mental health trajectories among those burdened with chronic illness.
Future research should expand longitudinally to assess how depression evolves in chronic disease populations and examine the effectiveness of ML-driven interventions in real-world clinical settings. Moreover, culturally tailored programs addressing the identified risk factors could drastically diminish depression prevalence and improve quality of life for millions.
By illuminating the mental health challenges faced by chronic disease patients through machine learning lenses, this research sets a new standard for integrated care approaches, driving forward both scientific understanding and practical solutions in global mental health.
Subject of Research: Prevalence, risk factors, and machine learning-based prediction of probable depression among individuals with chronic diseases in Bangladesh
Article Title: Prevalence, associated factors, and machine learning-based prediction of probable depression among individuals with chronic diseases in Bangladesh
Article References:
Das, P., Hasan, M.E., Arif, M. et al. Prevalence, associated factors, and machine learning-based prediction of probable depression among individuals with chronic diseases in Bangladesh. BMC Psychiatry 25, 1093 (2025). https://doi.org/10.1186/s12888-025-07542-4
Image Credits: AI Generated
DOI: 10.1186/s12888-025-07542-4
Keywords: Depression, chronic diseases, machine learning, Bangladesh, CatBoost, PHQ-9, mental health prediction, epidemiology, risk factors, socio-behavioral determinants
