Skip to main content
Fig. 1 | BMC Medical Informatics and Decision Making

Fig. 1

From: Interpretable time-aware and co-occurrence-aware network for medical prediction

Fig. 1

The data structure of EHR based on different methods. a Original EHR data structure. b EHR data structure based on time relation. c EHR data structure based on co-occurrence relation. d Data relation under our TCoN model. The data form b arranges codes in a random order, but different sequences have different effects on results. For example, the sequence ‘heart disease —> influenza —> coronary’ has closer relation between ‘heart disease’ and ‘influenza’ than the sequence ‘heart disease —> coronary —> influenza’. The data form c can make every two codes have the equal relation, but if ‘heart disease’, ‘atrial fibrillation’ and ‘diabetes’ are in three different visits, the equal relation will fail as there are different time intervals among them. The data form d is the combination. it describes both the equal code relation in the same visit and the time relation in different visits

Back to article page