Figure 6.

Cluster characterization reveals clinical variables associated with clusters

(A) Clinical variables associated with patient clusters in the training cohort and at least one validation cohort (p = 0.001). Continuous variables were categorized, and Fisher’s tests were applied to each factor level. The strength of the association was used to construct a word cloud with larger words having a stronger association with a cluster in the training cohort.

(B) Age distribution of patients within clusters. Boxes represent the median and IQR with whiskers extended to the lowest/highest datum within 1.5 IQR of the lower/upper quartile. Clusters are ordered along the x axis in relation to the increasing rate of associated 28-day mortality. Analysis was performed on clusters which comprised at least 10 patients in each cohort.