Machine learning-based predictive modeling of depression in hypertensive populations.

Area of Science:

• Medical Informatics
• Public Health
• Computational Biology

Background:

• Hypertension is a prevalent condition in U.S. adults.
• Depression frequently co-occurs with hypertension, impacting patient outcomes.
• Predictive models are needed to identify at-risk individuals within this population.

Purpose of the Study:

• To develop and analyze machine learning (ML) models for predicting depression in U.S. adults with hypertension.
• To identify significant risk factors contributing to depression in this demographic.
• To evaluate the performance of various ML algorithms in depression prediction.

Main Methods:

• Utilized a cross-sectional dataset of 8,628 U.S. adults with hypertension from the National Health and Nutrition Examination Survey (2011-2020).
• Employed six ML classification methods (ANN, RF, AdaBoost, SGB, XGBoost, SVM) with 10-fold cross-validation.
• Applied feature selection and addressed data imbalance using random down-sampling; performance evaluated using AUC, accuracy, precision, sensitivity, specificity, and F1-score.

Main Results:

• Artificial Neural Network (ANN) achieved the highest Area Under the Curve (AUC) of 0.813 and specificity of 0.780.
• Support Vector Machine (SVM) demonstrated the highest accuracy (0.771), precision (0.969), sensitivity (0.774), and F1-score (0.860).
• Important predictive features included income-to-poverty ratio, triglyceride levels, white blood cell count, age, sleep disorders, arthritis, hemoglobin, marital status, and education.

Conclusions:

• Machine learning algorithms show comparable performance in predicting depression among hypertensive individuals.
• The developed models highlight the relative importance of various clinical and socioeconomic factors.
• Findings provide a foundation for further clinical research and targeted interventions for depression in hypertensive populations.