Machine learning to predict cardiovascular risk
Problem
Accurately estimating a patient's cardiovascular event risk is difficult, and traditional scales may be outperformed by data-driven methods.
Why it matters
Cardiovascular events are major causes of death, and better risk stratification targets prevention effectively.
What the AI did
Fifteen machine learning methods were used to estimate 5-year cardiovascular risk in 38,527 patients of the Spanish ESCARVAL RISK cohort, compared against the SCORE and REGICOR scales using AUC, C-index and diagnostic metrics.
Results
Best method: quadratic discriminant analysis, AUC 0.7086; Naive Bayes and neural networks next (values truncated in abstract)
Impact
Machine learning offers a data-driven alternative to traditional risk scales, though predictive capacity appears modest.
Record
| Venue | International Journal of Clinical Practice |
|---|---|
| Published | 2019 |
| Work type | article |
| Authors | José A. Quesada, Adriana López‐Pineda, Vicente Francisco Gil-Guillén, Ramón Durazo-Arvizú, Domingo Orozco‐Beltrán, Angela López-Domenech, Concepción Carratalá‐Munuera |
| Citations | 104 (OpenAlex, at import time) |
| References | 31 |
| Model or method | 15 machine learning methods; best was quadratic discriminant analysis, followed by Naive Bayes and neural networks |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1111/ijcp.13389 |
| Open access | bronze |
| Keywords | cardiovascular risk, machine learning, AUC, cohort study, risk prediction |
Importance rating
significant — thousands to millions affected
Cardiovascular events are major causes of death; study on 38,527-patient cohort shows most ML methods outperform SCORE and REGICOR risk scales.