Cardiovascular risk prediction: from classical statistical methods to machine learning approaches

Area of Science:

• Cardiology
• Medical Informatics
• Data Science

Background:

• Cardiovascular risk prediction scores are vital for primary and secondary prevention.
• Traditional scores often assume linear relationships, limiting accuracy.
• Machine learning (ML) offers potential for improved prediction by modeling non-linear data.

Purpose of the Study:

• To review and compare classical statistical and ML-based cardiovascular risk scores.
• To highlight the strengths and limitations of each approach for clinical application.
• To provide physicians with a critical understanding of available risk prediction tools.

Main Methods:

• Literature review of classical statistical and ML-based cardiovascular risk scores.
• Comparative analysis of methodologies, accuracy, and clinical utility.
• Discussion of non-linearity in cardiovascular risk factor modeling.

Main Results:

• Classical scores have limitations due to linear assumptions.
• ML techniques demonstrate superior ability to capture complex, non-linear patterns in cardiovascular data.
• Both approaches have distinct advantages and drawbacks for risk stratification.

Conclusions:

• ML-based scores show promise for enhancing cardiovascular risk prediction accuracy.
• Understanding the differences between classical and ML scores is crucial for effective clinical implementation.
• Further research is needed to optimize ML applications in cardiovascular disease prevention.