Table 2.

Depicted is the comparison of the 1-year mortality preinterventional TAVI predictive performance among the tested algorithms (ie, logistic regression, TreeNet/decision tree algorithm and CART, Classification And Regression Tree)

Best model within type	Average log-likelihood	Area under ROC curve	Misclassification rate
Logistic regression*	0.3000	0.7477	0.1056
TreeNet	0.3211	0.6176	0.1021
CART	0.3452	0.5568	0.4454
Random forests	0.6529	0.5238	0.1021

Predictive performance was measured using the following metrics: (a) minimal averaged log-likelihood, (b) area under the ROC curve and (c) misclassification (error) rate. The logistic regression model showed the best performance regarding 1-year mortality prediction.

*Best model across all model types with minimum average log-likelihood. Output for the best model follows.

CART, Classification and Regression Tree.