Fig. 6

Those who died during their index stay have a higher proportion of flags for life support features, features associated with frailty, and index stay complications. (A). heatmap visualization of the hierarchically clustered features used for the prediction tasks. The annotation legend on the left indicates those who died during their index stay (yellow) vs. those who survived the index stay (green). The heatmap also highlights groupings or clusters of features with a higher proportion of those who died during their index stay. Blue box—those whose indication for transplant was idiopathic pulmonary fibrosis; red box—a group of patients highlighted by black donor ethnicity; green box—a set of features associated with life support after transplant; grey boxes—2 groupings belonging to separate hierarchically clustered “clades” associated with recipient frailty. (B). Life support features that were statistically significant in patients who survived the index stay vs. those who died during the index stay, where stacked proportional bar plots were used to represent categorical features (green indicates “yes” and blue indicates “no” for the feature value) and violin plots were used to show continuous features. (C). Frailty features that were statistically significant in patients who survived the index stay vs. those who died during the index stay. For graphs in B and C a chi-square (categorical) or Wilcoxon-rank sum (continuous) test was applied with FDR correction < 0.05 for multiple comparisons.