Machine learning model for detecting masked hypertension in young adults.

Area of Science:

• Cardiology
• Medical Informatics
• Biostatistics

Background:

• Cardiovascular disease (CVD) is a leading cause of death globally, with hypertension (HT) contributing significantly.
• Masked hypertension (MHT), normal BP in clinic but high out-of-clinic, increases CVD risk and is often undiagnosed.
• Current diagnostic tools like ABPM and HBPM have accessibility and feasibility limitations.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting MHT using single-visit clinical data.
• To address the diagnostic challenges of MHT, particularly in resource-constrained environments.

Main Methods:

• Utilized data from the African-PREDICT study, including clinical, biomarker, body composition, and physical activity metrics from a young South African cohort.
• Employed ML models, including LASSO feature selection and extreme gradient boosting, for MHT prediction.
• Evaluated model performance using accuracy and ROC AUC scores.

Main Results:

• An ML model combining LASSO feature selection and extreme gradient boosting achieved 0.83 accuracy and 0.86 ROC AUC.
• The model primarily relied on four key features: systolic blood pressure, body weight, left ventricular mass at systole, and dehydroepiandrosterone sulfate levels.
• This predictive framework demonstrates potential for early MHT identification, reducing reliance on resource-intensive monitoring.

Conclusions:

• ML-based prediction of MHT offers a feasible approach for early detection and intervention.
• This strategy can help mitigate MHT progression and associated cardiovascular risks, particularly in underserved regions.
• Integrating computational techniques into clinical practice is crucial for addressing global health challenges like hypertension.