Article published in Cap Today
In “Last Week Tonight with John Oliver,” the comedian makes a damning observation: How is the Domino’s Pizza delivery app better able to discern a customer’s location than the technology used by 911 dispatchers?
A similar comparison may be drawn between social media and health care information systems. While both hold massive amounts of personal data, websites like LinkedIn and Facebook make eerily accurate guesses about users’ relationships and preferences. Health care providers, on the other hand, sometimes lack essential knowledge about patients—a problem that may be addressed by applying to health care the same database technology that powers social media networking.
This technology, known as a graph database, uses graph theory to store information about relationships as data points. “The underlying premise of a graph database is there’s a lot of value in the knowledge of connections between data points and that value should be accessible in real time,” says Utpal Bhatt, vice president of marketing at Neo Technology. An early pioneer of graph technology, Neo Technology produces the Neo4j database, which is used by such companies as Walmart and eBay.
Like all data, Bhatt explains, health care data are highly interconnected. And graph database technology can provide a complete and real-time view of this interconnected information—the providers, health events, medications, and other relevant data linked to a patient.
Available in commercial form since the late 2000s, graph technology differs markedly from the relational databases health care organizations typically employ. “Relational databases store data in a tabular form, with foreign keys used to define how a record in one table relates to a record in another table,” explains Kevin Schmidt, director of product management at the health care data-management company NextGate. “Graph databases allow relationships to be modeled in a more flexible way.”
One benefit of this flexibility is enhanced performance and efficiency. Graph databases scale more naturally to large data sets and require shorter query times, presenting an advantage for anyone who wants to ask complicated questions of a health care data set. “Neo4j, for example, is a native graph database, which means . . . it actually stores data as a graph,” says Bhatt. “The relations between data points are treated as first-class objects [rather than as meta-data], which optimizes the database to answer questions about relationships.” Many Neo4j users are in health care analytics, Bhatt adds, citing as an example a customer who collected historical data from clinical trials to analyze the effects of pharmaceuticals on various patient populations.