Fig. 2

24-hour phenotype distribution and chord plot. In panel A, visualization of phenotypes using t-distributed stochastic neighbor embedding (t-SNE) technique with phenotypes shown in color from the consensus k-means clustering analysis visualizes distinction among four phenotypes. In panels B–E, each phenotype is highlighted separately and the ribbons connect to the different patterns of clinical variables and organ system dysfunctions on the top of the circle (inflammation = low temperature, high temperature, max CRP, max ferritin; organ failure = total OFI; pulmonary = pulmonary OFI, intubation; cardiovascular = high heart rate, low systolic blood pressure, cardiovascular OFI; renal = high creatinine, renal OFI; hepatic = hepatic OFI; hematologic = low hemoglobin, low platelets, hematologic OFI; neurologic = Low Glasgow Coma Score Scale, central nervous system OFI). The chords connect from an individual phenotype to a category if the group mean involvement of the variables differs from the overall mean for the entire cohort (see Table 1) specifically lower for low temperature, systolic blood pressure, hemoglobin, platelets, and Glasgow Coma Scale Score, but higher for all other variables