A Cardiovascular Disease Prediction Model Based on Routine Physical Examination Indicators Using Machine Learning...

Area of Science:

• Public Health
• Medical Informatics
• Epidemiology

Background:

• Cardiovascular diseases (CVD) are a leading global cause of premature death.
• Early detection of high-risk populations is crucial for CVD prevention.
• This study focuses on the Xinjiang rural population, a region with unique health considerations.

Purpose of the Study:

• To develop a machine learning (ML) model for predicting CVD risk.
• To identify key indicators for CVD prediction using routine physical examination data.
• To establish a model suitable for the Xinjiang rural population.

Main Methods:

• Utilized data from two-stage surveys (2010-2017 and 2016-2021) with 12,692 participants.
• Employed feature selection techniques including Lasso regression, FLR, and RF.
• Compared prediction models: L1-LR, RF, SVM, and AdaBoost for CVD risk assessment.

Main Results:

• Identified key predictors: age, systolic blood pressure, lipid profile indices, triglyceride-glucose index, BMI, and BAI.
• The L1-LR model demonstrated superior prediction performance in discrimination and calibration.
• The cumulative incidence of CVD was 9.27% after a 4.94-year follow-up.

Conclusions:

• The L1-LR based prediction model shows the best performance for the Xinjiang rural population.
• Routine physical examination indicators can effectively predict CVD risk in this demographic.
• This model aids in early detection and prevention strategies for CVD in the region.