Key result
The SAINT machine learning model incorporating multimodal clinical data achieved a higher AUC (0.732) than models using only traditional Framingham Risk Score features (0.656) for predicting cardiovascular disease, while analysis revealed women are under-diagnosed by up to 2-fold compared to men when using non-sex-specific criteria.
Why the study?
The lack of sex-specific cardiovascular disease criteria contributes to the under-diagnosis of women compared to men, and machine learning may provide more nuanced cardiovascular disease risk prediction than the Framingham Risk Score.
Do machine learning models incorporating multimodal data improve cardiovascular risk prediction and identify sex-specific under-diagnosis compared to the Framingham Risk Score in the UK Biobank cohort?
Observational (n=20,542)
Do machine learning models incorporating multimodal data improve cardiovascular risk prediction and identify sex-specific under-diagnosis compared to the Framingham Risk Score in the UK Biobank cohort?
Absolute Event Rate: 0.732% vs 0.656%
Machine learning models using multimodal UK Biobank data highlight significant under-diagnosis of cardiovascular conditions in women and demonstrate the need for sex-specific diagnostic criteria.
May support sex-specific CVD diagnostic thresholds; leaves open prospective validation of multimodal ML models.
The lack of sex-specific cardiovascular disease criteria contributes to the under-diagnosis of women compared to men. For more than half a century, the Framingham Risk Score has been the gold standard to estimate an individual’s risk of developing cardiovascular disease based on age, sex, cholesterol levels, blood pressure, diabetes, and smoking. Now, machine learning can offer a much more nuanced insight into predicting the risk of cardiovascular disease. The UK Biobank is a large database that includes traditional risk factors as well as tests related to the cardiovascular system: magnetic resonance imaging, pulse wave analysis, electrocardiograms, and carotid ultrasounds. Here we leverage 20,542 datasets from the UK Biobank to build more accurate cardiovascular risk models than the Framingham Risk Score, and quantify the under-diagnosis of women compared to men. Strikingly, for first-degree atrioventricular block and dilated cardiomyopathy, two conditions with non-sex-specific diagnostic criteria, our study shows that women are under-diagnosed 2x and 1.4x more than men. Similarly, our results demonstrate the need for sex-specific criteria in essential primary hypertension and hypertrophic cardiomyopathy. Our feature importance analysis reveals that, out of the top 10 features across three sex and four disease categories, traditional Framingham factors made up between 40-50%, electrocardiogram 30-33%, pulse wave analysis 13-23%, and magnetic resonance imaging and carotid ultrasound 0-10%. Improving the Framingham Risk Score by leveraging big data and machine learning allows us to incorporate a wider range of biomedical data and prediction features, enhance personalization and accuracy, and continuously integrate new data and knowledge, with the ultimate goal to improve accurate prediction, early detection, and early intervention in cardiovascular disease management. Our analysis pipeline and trained classifiers are freely available at https://github.com/LivingMatterLab/CardiovascularDiseaseClassification
No takes yet. Share an insight, caveat, or question.
Pierre et al. (2023) conducted an observational in Cardiovascular disease (n=20,542). Machine learning models (SAINT, XGBoost, MLP) including multimodal clinical features vs. Models using Framingham Risk Score features only was evaluated on Area under the curve (AUC) for predicting any cardiovascular disease (both sexes). The SAINT machine learning model incorporating multimodal clinical data achieved a higher AUC (0.732) than models using only traditional Framingham Risk Score features (0.656) for predicting cardiovascular disease, while analysis revealed women are under-diagnosed by up to 2-fold compared to men when using non-sex-specific criteria.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: