Predicting Cardiovascular Diseases Using Neural Networks: Validation With the SCORE2 Risk Assessment Tool

Abstract

This study introduces an approach to predicting cardiovascular diseases (CVDs) by leveraging neural networks and evaluates its performance against the widely recognized SCORE2 risk assessment tool. Given the significant global health impact of CVDs, our research aims to assess the effectiveness of deep learning‐based models compared with traditional methods. Our methodology involves developing and deploying a neural network model capable of learning intricate patterns and relationships from a comprehensive dataset. Leveraging a large population cohort with extended follow‐up enables precise and long‐term risk estimation. We utilize the T‐paired test to compare risk predictions between our neural network model and the SCORE2 tool. Our results indicate a notable accuracy of 0.77 for our neural network‐based model in predicting CVDs, with the T‐paired test revealing no significant variations in risk levels between the two methods. These findings underscore the effectiveness of neural networks as a robust tool for CVD risk prediction and advocate for further exploration and integration of these technologies to enhance cardiovascular risk assessment, thereby advancing predictive modeling in healthcare.