Highly precise risk prediction model for new-onset hypertension using artificial intelligence techniques

Area of Science:

• Cardiovascular Health
• Public Health
• Machine Learning in Medicine

Background:

• Hypertension poses a significant public health challenge globally.
• Predicting hypertension risk is crucial for developing effective prevention strategies.
• Annual employee health checkups provide valuable longitudinal data.

Purpose of the Study:

• To develop and validate a novel risk prediction model for new-onset hypertension.
• To compare the performance of machine learning models against traditional statistical methods.
• To identify key predictors for hypertension development in a normotensive population.

Main Methods:

• Utilized longitudinal health checkup data from 18,258 individuals in Japan (2005-2016).
• Developed and compared prediction models using XGBoost, ensemble methods, and logistic regression.
• Employed a derivation set (75%) for model construction and a validation set (25%) for performance testing.

Main Results:

• Systolic blood pressure at Year (-1) during cardio-ankle vascular index measurement was the top predictor in the XGBoost model.
• Machine learning models (XGBoost: 0.877, Ensemble: 0.881) demonstrated high predictive accuracy (AUC) in the validation cohort.
• Logistic regression model achieved an AUC of 0.859.

Conclusions:

• Highly precise machine learning-based prediction models for future hypertension have been developed.
• These models can identify individuals at high risk within a general normotensive population.
• Facilitates early, non-pharmacological interventions to prevent hypertension onset.