Nonlaboratory-based risk assessment model for coronary heart disease screening: Model development and validation

Area of Science:

• Cardiovascular Disease Research
• Machine Learning in Healthcare
• Public Health Screening

Background:

• Identifying high-risk groups for coronary heart disease (CHD) is crucial for reducing mortality.
• Existing machine learning (ML) models often require parameters not readily available, limiting their widespread application.
• There is a need for simple, efficient ML models utilizing easily obtainable non-laboratory data for CHD risk prediction.

Purpose of the Study:

• To develop and validate a joint machine learning model for identifying individuals at high risk of CHD.
• The model utilizes easily obtainable non-laboratory parameters for enhanced accessibility and application.
• To assess the model's performance in terms of discrimination and calibration.

Main Methods:

• A prospective study utilized data from the Henan Rural Cohort Study (N=38,716) and external validation from the Dongfeng-Tongji cohort (N=17,958).
• A joint machine learning model was developed by combining logistic regression (LR), artificial neural network (ANN), random forest (RF), and gradient boosting machine (GBM).
• Readily accessible variables including demographics, medical/family history, lifestyle, and anthropometrics were used. Model performance was evaluated using Area Under the Receiver Operating Characteristic Curve (AUC) and Brier Score (BS).

Main Results:

• Key predictors identified were age, waist circumference, pulse pressure, heart rate, and family history of CHD, type 2 diabetes mellitus (T2DM), and dyslipidemia.
• Internal validation showed good discrimination (AUC=0.844) and acceptable calibration (BS=0.066).
• External validation demonstrated useful discrimination (AUC=0.792) and robust calibration (BS=0.069).

Conclusions:

• A novel, simple machine learning-based model using accessible variables accurately identifies individuals at high risk of CHD.
• The developed model demonstrates strong potential for large-scale CHD screening, particularly in resource-constrained settings.
• This approach offers a practical tool for early identification and management of individuals susceptible to coronary heart disease.