Application of physiological network mapping in the prediction of survival in critically ill patients with acute liver failure

Fig. 1. Network map of clinical and laboratory variables showing correlations and k-clique percolation clusters of patients with acute liver failure who survived their ICU stay (optimized k—clique size = 3). The size of the nodes is automatically generated and does not represent any measure of influence or magnitude. Also, the colours of the nodes and edges were used to help differentiate the clusters from one another and have no analytical translation beyond this. Cl chloride, AST aspartate transaminase, ALT alanine aminotransferase, GCS Glasgow Coma Score, Bil total bilirubin, ALP alkaline phosphatase, Cr serum creatinine, Na serum sodium, Glu blood glucose, HR heart rate, RespR respiratory rate, Temp temperature.

Fig. 2. Network map of clinical and laboratory variables showing correlations and k-clique percolation communities of patients with acute liver failure who did not survive their ICU stay (Optimized k—clique size = 3). Cl chloride, AST aspartate transaminase, ALT alanine aminotransferase, GCS Glasgow Coma Score, Bil total bilirubin, ALP alkaline phosphatase, Cr serum creatinine, Na serum sodium, Glu blood glucose, HR heart rate, RespR respiratory rate, Temp temperature.

Fig. 3. Kaplan–Meier graphs of patients with acute liver failure admitted to the intensive care unit who survived and those who did not survive after 28 days, according to SOFA cut-off scores (A) and the cut-offs of the composite scores from the combination of SOFA with the parenclitic deviations along the pH-bicarbonate (pH-CO3) (B), pH-creatinine (pH-Cr) (C), lactate-heart rate (lactate-HR) (D), lactate-glucose (lactate-Glu) (E), and oxygen saturation-respiratory rate (SpO2-RespR) (F) axes.