Prediction Algorithms for Blood Pressure Based on Pulse Wave Velocity Using Health Checkup Data in Healthy Korean...

Area of Science:

• Cardiovascular Physiology
• Biomedical Engineering
• Health Informatics

Background:

• Pulse transit time and pulse wave velocity (PWV) are indicators related to blood pressure (BP).
• Previous research on predicting BP using wearable devices was limited in scale and scope.
• The relative importance of various clinical variables for BP prediction remained unclear.

Purpose of the Study:

• To predict systolic and diastolic blood pressure using PWV.
• To assess the relative importance of clinical variables in BP prediction models.
• To develop a feasible and efficient model for BP estimation using wearable technology.

Main Methods:

• A study involving 1362 healthy men over 18 years old.
• Systolic and diastolic blood pressure were estimated using multiple linear regression.
• Models were stratified by age (under 60 and 60+) with repeated partitioning to mitigate bias.

Main Results:

• Age-stratified models (under 60 and 60+) outperformed non-stratified models.
• Models with three variables (PWV, BMI, age) showed comparable performance to models with 17 variables.
• The final model using PWV, BMI, and age met established medical instrumentation criteria with prediction errors similar to mercury sphygmomanometers.

Conclusions:

• Age stratification at 60 years improved BP prediction accuracy.
• A parsimonious model incorporating PWV, BMI, and age demonstrated high predictive performance.
• The minimal variable model is efficient and feasible for practical BP prediction.