Next-visit prediction and prevention of hypertension using large-scale routine health checkup data

Area of Science:

• * Computational medicine and predictive analytics.
• * Health informatics and data science.

Background:

• * Hypertension is a major global health concern.
• * Early risk prediction is crucial for preventative interventions.
• * Routine health checkup data offers a rich source for predictive modeling.

Purpose of the Study:

• * To develop and evaluate machine learning models for predicting future hypertension risk.
• * To leverage large-scale, high-dimensional health checkup data for enhanced prediction.
• * To compare model performance against baseline methods and identify key predictive factors.

Main Methods:

• * Utilization of a large-scale dataset from MJ Health Research Foundation, Taiwan.
• * Implementation of 5-fold cross-validation for robust model training.
• * Ensemble prediction using voted results from 5 trained models.
• * Feature selection and comparison with the Framingham risk score.

Main Results:

• * Achieved 69.59% precision, 77.90% recall, and 73.51% F1-score.
• * Models outperformed a baseline using only the last visit's blood pressure.
• * More frequent health checkups correlated with better prediction accuracy.
• * Models trained with selected factors surpassed the Framingham risk score.
• * Virtual weight loss scenarios indicated potential for personalized intervention suggestions.

Conclusions:

• * Machine learning models can effectively predict future hypertension risk from routine health data.
• * Incorporating historical and selected health factors improves predictive performance.
• * The models show potential for guiding preventative strategies, such as weight management.