A Clinical Interpretable Approach Applied to Cardiovascular Risk Assessment.

Area of Science:

• Computational intelligence
• Medical informatics
• Cardiology

Background:

• Clinical Decision Support Systems (CDSS) require interpretable predictive models.
• Developing accurate and understandable models is crucial for clinical adoption.
• Existing approaches often rely on prior knowledge or purely data-driven methods.

Purpose of the Study:

• To develop a data-driven, interpretable predictive model for assessing 30-day mortality risk in acute coronary syndrome (ACS) patients.
• To extract meaningful rules from clinical data for patient risk stratification.
• To enhance the utility of CDSS through accurate and understandable predictions.

Main Methods:

• A data-driven supervised clustering approach was employed.
• The model learned distinct patient groups based on similar characteristics.
• Validation utilized the largest Portuguese ACS patient dataset (13902 patients).

Main Results:

• The developed model achieved acceptable performance with a Geometric Mean (GM) of 72%.
• Clinical interpretability was maintained through derived decision rules.
• The approach demonstrated potential for effective patient risk stratification.

Conclusions:

• Data-driven computational intelligence techniques can create interpretable predictive models for clinical decision support.
• This method shows promise in stratifying patient risk for conditions like ACS.
• Further research is needed to enhance model performance and applicability.