Improving clinical models based on knowledge extracted from current datasets: a new approach.

Area of Science:

• Cardiology
• Medical Informatics
• Data Science

Background:

• Cardiovascular diseases (CVD) are a leading global cause of mortality.
• Current CVD risk prediction models have limitations in accuracy and applicability.
• Effective CVD prevention strategies are crucial for public health.

Purpose of the Study:

• To develop an improved model for cardiovascular disease risk prediction.
• To enhance the accuracy of predicting new cardiovascular events.
• To ensure clinical interpretability in risk assessment models.

Main Methods:

• Utilized a decision tree scheme for clinical interpretability.
• Developed an innovative optimization strategy to adjust decision tree thresholds.
• Validated the approach using a real-world dataset from the National Registry on Acute Coronary Syndromes.

Main Results:

• The new approach achieved a sensitivity of 80.52%, specificity of 74.19%, and accuracy of 77.27%.
• Demonstrated a significant improvement of approximately 26% in accuracy compared to the original risk score.
• The decision tree model provided clinically interpretable risk predictions.

Conclusions:

• The proposed decision tree-based approach offers a significant improvement in cardiovascular disease risk prediction.
• Optimized decision tree thresholds using recent clinical data enhance predictive performance.
• This method provides a more accurate and interpretable tool for clinical decision-making in CVD prevention.