Value-Based Healthcare Translational Data Analytics using the Problem Oriented Medical Record Graph Representation

Translational data analytics play an important role in developing, applying and evaluating health outcomes to improve the effectiveness, safety, and patient experiences of care. Through this integrative approach of linking the bedside clinical practice with the benchside biomedical research relevant information of different varieties from multiple sources can provide a more accurate patterns contributing to health outcomes. In this article we are introducing an exploratory analytics approach to generate knowledge graphs from patient cases described via the SOAP note at the bedside and expand this graph through building attributes like medications to what the bench can provide on research related to these attributes. This exploratory approach has been applied to the problem of identifying polypharmacy knowledge graphs. However, we are proposing to used the generated knowledge graphs using an autoencoder to decode neighborhoods of polypharmacy and label nodes like drugs as candidate for deprescribion or not. The autoencoder is our next phase of research.