Evaluation of coronary heart disease risk prediction based on simple physical examination parameters by machine...

Area of Science:

• Cardiovascular Medicine
• Biostatistics
• Machine Learning in Healthcare

Background:

• Coronary heart disease (CHD) poses a significant public health challenge.
• Accurate risk prediction is crucial for timely intervention and prevention strategies.
• Existing risk models may not fully leverage routinely collected clinical data.

Purpose of the Study:

• To develop and validate a CHD risk prediction model using readily available clinical indicators.
• To identify the most significant predictors of coronary heart disease.
• To create a clinically interpretable tool for cardiovascular risk assessment.

Main Methods:

• Utilized the Framingham Heart Study cohort (n=4,240) for model development.
• Employed advanced machine learning techniques including stacked ensembles (gradient boosting, random forest, XGBoost) and a logistic-regression meta-learner.
• Performed rigorous internal and external validation, incorporating feature selection and handling of data imbalances (SMOTEENN/SMOTETomek).

Main Results:

• Achieved high performance with an Area Under the Curve (AUC) of 0.977 and accuracy of 0.942 in internal validation.
• Demonstrated strong generalizability with AUC 0.929 and accuracy 0.885 in external validation.
• SHAP analysis identified systolic blood pressure, age, total cholesterol, and fasting glucose as key predictors.

Conclusions:

• A CHD risk model derived from routine clinical data exhibits excellent predictive performance.
• The model is robust and generalizes well to an independent cohort.
• This interpretable tool enhances cardiovascular risk assessment capabilities in clinical practice.