Machine Learning Approaches to Identify Influential Factors of the Comorbid Psychiatric Symptoms and Hypertension Among Rural Adults in Bangladesh: A Cross-Sectional Study.

Background and Aim: The burden of both psychiatric symptoms (anxiety and/or depression) and hypertension poses significant public health challenges in Bangladesh, especially in rural areas with limited healthcare access. Thus, this study aimed to identify the influential factors associated with comorbid psychiatric symptoms and hypertension in rural Bangladesh using machine learning (ML) algorithms. Methods: In this study, 1603 respondents were selected from a cross-sectional survey using a multistage random sampling method. Six commonly used ML algorithms were applied. The ML models' predictive performance was evaluated using standard validation metrics. Influential variables were ranked and explained using SHapley Additive exPlanations (SHAP) method. Results: The prevalence of comorbid psychiatric symptoms and hypertension was 8.7%. In predicting this outcome, the SVM model outperformed others across a variety of metrics: accuracy (0.748), RMSE (0.520), specificity (0.708), and Brier score (0.252). The model achieved a modest receiver operative characteristics (ROC) score of 0.613 (95% CI: 0.532-0.690) for predicting comorbidity. The top factor associated with this comorbid condition, as explained by the SHAP method, included respondents with cardiovascular disease (CVD), family history of hypertension, current smoking exposure, chronic disease, and current tobacco user. Conclusion: Approximately one out of ten people in rural areas experienced comorbid psychiatric symptoms and hypertension. The ML models highlighted several key associated factors, although their predictive performance was modest. The current situation highlights an urgent need for strengthened national and regional public health initiatives in rural Bangladesh.