Figure 2.

Bayesian network analysis. Hypothesized model (knowledge-based DAG) (A) and automatically learned model with imposed constraints (e.g., cancer and MTX) (BIC-based DAG) (B). The BIC-based DAG is obtained through logical and clinical associations and by selecting the model (DAG) with the best (i.e., lowest) BIC. Legend: DAG: directed acyclic graphic; BIC: Bayesian Information Criterion. Abbreviations: BMI, body mass index; MTX, metastasis; ALB, albumin; NUTR, nutritional support; BTcP, breakthrough cancer pain; RADIO, radiotherapy; HRA, hospital readmission. (** = p < 0.05; * = p < 0.10; • = Not Significant).