MedFuse 2018

Healthcare data is not like general-ledger data. General ledger data fits well into a table and the data format of credits and debits is consistent across many industries. In contrast, data from an MRI, an eye exam and a blood test are all different. We find that healthcare data does not fit will into tabular formats. We call this characteristic “high variability” data. It is just one of the reasons that companies are moving away from the relational model to a more flexible model of storing and retrieving healthcare data. Another reason is the need to connect this health data with other data such as claims that provide insight into the frequency and severity of health problems, a person’s genetic information that allows for more precision in treatment. In addition, wouldn’t it be valuable to connect this endogenous data with exogenous data of an individual that is increasingly available via wearables? By connecting all this information and mining this connected health information for the data as well as structure, is it possible to improve healthcare? We think so! In this presentation we will cover an innovative way of storage and retrieval of data in a graph model, visualizing information and ways to gain insights by querying and processing this data, where every fact is broken down into nodes and relationships. We will discuss the rise of the graph industry and the adoption by healthcare companies. We will discuss many of the classic use-cases of graph technologies including master data management, complex rule processing, and fraud detection. We will also discuss some emerging topics such as causal graphs, the combination of deep learning and graphs and possible ways for graphs to enhance the ability to explain machine-driven decisions.