Figure 1.

Comparison of landmarking approach. (A) The Concordance index. (a) ESKD: The Kruskal–Wallis test comparing the ESKD concordance index of the models for each landmark time was non-significant. Each boxplot was created using the model’s performance on the five-fold cross-validation test dataset. (b) Death: The Kruskal–Wallis test comparing the death concordance index of the models for each landmark time was non-significant. Each boxplot was created using the model’s performance on the test dataset of the five-fold cross-validation. (B) Integrated Brier score. (a) ESKD: The Kruskal-Wallis test comparing the ESKD integrated Brier score of the models for each landmark time was non-significant. Each boxplot was created using the model’s performance on the test dataset of the five-fold cross-validation. (b) Death: The Kruskal–Wallis test comparing the death-integrated Brier score of the models for each landmark time was non-significant. Each boxplot was created using the model’s performance on the test dataset of the five-fold cross-validation.