Personalization algorithms applied to cardiovascular disease risk assessment.

Area of Science:

• Cardiology
• Medical Informatics
• Health Services Research

Background:

• Cardiovascular disease (CVD) is a leading global cause of mortality.
• Current clinical guidelines utilize risk assessment tools for patient CVD risk stratification.
• Existing tools may exhibit variable performance across different patient populations.

Purpose of the Study:

• To explore personalized CVD risk assessment strategies.
• To develop and validate patient grouping methodologies for enhanced risk prediction.
• To compare the performance of novel personalization methods against existing CVD risk assessment tools.

Main Methods:

• Developed two patient personalization methods: a clustering approach and a similarity measures approach.
• Validated these methodologies using data from 460 Portuguese Acute Coronary Syndrome with non-ST segment elevation (ACS-NSTEMI) patients.
• Evaluated the sensitivity, specificity, and geometric mean of the proposed methods.

Main Results:

• The similarity measures approach demonstrated superior performance compared to the clustering method.
• The similarity measures approach achieved a sensitivity of 77.7%, specificity of 63.2%, and geometric mean of 69.7%.
• These results represent an improvement over the performance of current CVD risk assessment tools (78.5%, 53.2%, 64.4% respectively).

Conclusions:

• Personalized CVD risk assessment, particularly using similarity measures, can optimize patient stratification.
• The similarity measures approach offers enhanced accuracy for CVD risk prediction in ACS-NSTEMI patients.
• This study highlights the potential of tailored risk assessment to improve healthcare strategies for cardiovascular events.