Development and Internal Validation of Machine-Learning Models for Short-Term Prediction of Day-to-Day Home Blood...

Area of Science:

• Cardiovascular Health
• Biomedical Engineering
• Machine Learning in Healthcare

Background:

• Day-to-day blood pressure variability (BPV) is linked to cardiovascular risk.
• Environmental factors influence daily BPV.
• Predicting short-term BPV increases from personal sensor data is not well understood.

Purpose of the Study:

• Develop and validate a machine learning model.
• Predict short-term increases in day-to-day BPV.
• Utilize personal sensor data on behavioral and environmental exposure.

Main Methods:

• 30-day monitoring study in community-dwelling adults.
• Home BP measured twice daily; sensors recorded environmental data and physical activity.
• Machine learning models trained and validated using cross-validation; XGBoost model interpreted with Shapley values.

Main Results:

• Primary outcome: episode of increased systolic day-to-day BPV (rolling 5-day CV ≥11.0%).
• Candidate predictors derived from preceding 5-day exposure.
• Performance evaluated using AUC, calibration, Brier score, and decision-curve analysis.

Conclusions:

• A framework for predicting short-term BPV increases from personal environmental and behavioral data is established.
• Supports future personalized interventions.
• Targets modifiable environmental and behavioral factors for cardiovascular risk management.