Predicting cardiovascular events using three stage Discriminant Function is much more accurate than Framingham or...

Area of Science:

• Cardiology
• Biostatistics
• Predictive Analytics

Background:

• Current cardiovascular risk prediction models (Framingham, QRISK) have limited accuracy (<70%) and high false positive rates.
• Existing methods apply predictor combinations only once, potentially missing complex risk interactions.

Purpose of the Study:

• To develop and validate a more accurate method for predicting cardiovascular incidents (CVI).
• To improve the true positive rate and reduce misclassification errors in cardiovascular risk assessment.

Main Methods:

• Utilized Principal Components Analysis to identify four independent determinants of cardiovascular risk in a British dataset.
• Employed a multi-stage Discriminant Function analysis with repeated application to residuals for enhanced prediction accuracy.

Main Results:

• The novel approach correctly predicted 94% of CVIs more than 20 years in advance.
• Achieved a low misclassification rate (2.8 errors per correct prediction) and 92% accuracy when validated with the Jacknife method.

Conclusions:

• Repeated application of Discriminant Function analysis to residuals offers superior cardiovascular risk prediction compared to single-application models.
• This enhanced method provides a more accurate and potentially simpler way to communicate individual cardiovascular risk to patients.