Assessment of cardiovascular risk based on a data-driven knowledge discovery approach

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Existing risk assessment models for acute myocardial infarction (AMI) lack integration with current clinical interventions and often use hard-to-obtain risk factors.
• There is a need for updated, simple, and interpretable risk models for AMI patients that are practical for clinical use.

Purpose of the Study:

• To develop and validate a simple, interpretable, data-driven risk assessment model for patients hospitalized with acute myocardial infarction.
• To integrate existing clinical risk tools with data-driven knowledge discovery using routinely collected hospitalization data.

Main Methods:

• A decision tree model was developed using a reduced set of six binary risk factors.
• The model was validated on a recent dataset of 11,113 patients provided by the Portuguese Society of Cardiology, which originally contained 77 risk factors.

Main Results:

• The proposed decision tree model achieved a sensitivity of 80.42%, specificity of 77.25%, and accuracy of 78.80%.
• The model demonstrates the effectiveness of using a simplified set of risk factors for accurate risk prediction in AMI patients.

Conclusions:

• A simple, interpretable decision tree model using six binary risk factors effectively assesses short-term risk in acute myocardial infarction patients.
• This approach facilitates the integration of risk assessment tools into clinical practice, enhancing confidence and utility.