Figure 6. Overlaying predicted survival time plots for cardiovascular disease and mortality outcomes across the BIDMC broad QRS DDRTree.

A–I show the overlaid risk probabilities for the different outcomes. These were estimated for individual samples within the tree, using Cox proportional hazards models for fatal outcomes and Fine–Gray subdistribution hazards models for nonfatal outcomes accounting for the competing risk of death, with tree dimensions and pseudotime (position within the tree, relative to the tree core) as inputs. All‐cause mortality was treated as the competing event for other incident outcomes. ASCVD indicates atherosclerotic cardiovascular disease; BIDMC, Beth Medical Israel Deaconess Center; and DDRTree, dimensionality reduction via learning a tree.