Figure 3.

Schema of patient SPOKEsig generation. Diagnostic codes, prescribed medications, and laboratory test values are taken from the electronic health record (EHR, left side of image) and translated into codes readable by SPOKE. Integration of these concepts into SPOKE creates Propagated SPOKE Entry Vectors (PSEVs), which are then added together to create the SPOKEsig vector for each patient. The SPOKEsig of a patient represents a network of SPOKE nodes (right side of the figure; color represents node type), where each SPOKEsig value represents the relevance of the corresponding node in the network for that patient. Therefore, an element (or feature) in the SPOKEsig vector corresponds to a node in the SPOKE knowledge graph and the value depicts the weight of that node.