Figure 1. Classification tree used to define cut-offs for component score and identify profiles of risk of death in the derivation cohort across the four PIRO components.

Each node split decision in the tree was chosen from the possible cut-offs for all components, maximizing the within-node homogeneity.