Fig 1.
Overview of experimental setup of proposed approach.
Table 1.
CVD outcomes statistics according to definition in current study and the comparator study definition by Alaa et al. [29].
Table 2.
Performance of all tested classifiers including baseline models.
Fig 2.
AUROC of logistic regression with L1 regularization and XGBoost.
Fig 3.
AUROC curves of baseline models on imputed data.
Fig 4.
AUROC curves of baseline models on unimputed data.
Fig 5.
Performance of best logistic regression model depending on number of features.
AUROC performance of best performing logistic regression model with L1 regularization (continuous blue line) compared to number of features utilized in each iterative feature elimination step (orange line), dotted blue horizontal line showing intersection of 25 features with iterative feature elimination step, allowing for extrapolation to model performance.
Table 3.
Performance of best logistic regression model depending on number of features.
Fig 6.
AUROC of logistic regression with L1 regularization and XGBoost when trained on Whites and tested on non-Whites.
Table 4.
Model performance when trained on Whites and tested on non-Whites.
Table 5.
Relative regression feature weights of 25 most informative risk factors from best logistic regression model.
Table 6.
Categorization of the 25 most informative risk factors into categories from the best logistic regression model.