Figure 4.

Decision tree model for predicting COVID-19 risk of death. A decision tree prediction model was constructed to predict patient's risk of death given demographic and clinical attributes at initial evaluation. Shaded nodes are the parent/split nodes, and the non-shaded nodes are the leaf nodes of the DT model. The conditional recursive partitioning tree algorithm (ctree) was used for model building. The parameter p shown at the leaf nodes is the risk of death for patients satisfying the conditions of that tree branch.