A cost-effective, machine learning-driven approach for screening arterial functional aging in a large-scale Chinese...

Area of Science:

• Cardiovascular Health
• Machine Learning in Medicine
• Public Health Screening

Background:

• Vascular aging, a key indicator of cardiovascular risk, often goes undetected in early stages.
• Current screening methods can be resource-intensive and may not identify high-risk individuals proactively.
• Developing accessible tools for early detection of vascular aging is crucial for preventative healthcare.

Purpose of the Study:

• To develop and validate a cost-free machine learning model for predicting elevated arterial stiffness (EAS).
• To identify high-risk populations for vascular aging using readily available questionnaire and physical data.
• To optimize healthcare resource allocation through early identification of individuals susceptible to arterial aging.

Main Methods:

• Utilized a dataset of 77,134 adults including questionnaire responses and physical measurements.
• Employed feature selection techniques (LASSO, RFE-LGE) to identify key predictive variables.
• Trained and evaluated four machine learning algorithms, including XGBoost, for EAS prediction using a 70/30 train-test split.

Main Results:

• A model was constructed using 14 accessible features, including systolic blood pressure, age, and waist circumference.
• The Extreme Gradient Boosting (XGBoost) model demonstrated superior performance with Area Under the Curve (AUC) values of 0.8722 (training) and 0.8710 (test).
• Key predictors identified were systolic blood pressure, age, waist circumference, hypertension history, and sex.

Conclusions:

• The XGBoost model effectively predicts the prior probability of elevated arterial stiffness in the general population.
• Integration into primary care can significantly reduce healthcare costs associated with vascular aging.
• This accessible screening tool facilitates enhanced management of arterial aging and preventative cardiovascular care.