Cardiovascular Risk Estimation in Colombia Using Artificial Intelligence Techniques.

Area of Science:

• Cardiovascular disease research
• Machine learning applications in healthcare
• Biostatistics and predictive modeling

Background:

• Cardiovascular disease risk estimation in Colombia lacks specific ML-based insights.
• Existing models may not be optimally adapted for the Colombian population.
• Artificial intelligence offers novel approaches for risk prediction.

Purpose of the Study:

• To evaluate the potential of five machine learning techniques for cardiovascular risk prediction in a Colombian cohort.
• To compare the discriminatory ability of ML models against the traditional Framingham model.
• To explore innovative AI-driven strategies for enhanced cardiovascular risk assessment.

Main Methods:

• A cohort of 847 disease-free patients was followed for 10 years.
• Five ML techniques (neural networks, decision trees, SVM, random forests, Gaussian Bayesian networks) were employed.
• 5-fold cross-validation and AUC-ROC analysis were used for model evaluation.

Main Results:

• The neural network model achieved the highest discriminatory ability (AUC-ROC 0.69).
• All evaluated ML models demonstrated superior performance compared to the Framingham model (AUC-ROC 0.53).
• ML techniques showed promising results for cardiovascular event prediction in the study population.

Conclusions:

• Flexible machine learning approaches are viable for improving cardiovascular risk prediction in Colombia.
• ML-based risk prediction may offer greater discrimination than established models like Framingham.
• Further prospective studies are recommended to validate these findings before widespread implementation.