The explosion of connected devices for health applications enables previously unimaginable potential for advancing sports medicine, not to mention overall health-system function, precision care, and individual wellness. Technologies ranging from smart watches, fitness-tracking wearables, and biosensors to medical-grade equipment for remote glucose, blood-pressure, and heart-rate monitoring all supply enormous new streams of critical data enabling a host of health-related wonders.
But your old-school databases cannot make the most of it.
Data and its flow are both central to the task of gleaning actionable insight from all these new technologies. But there’s a third factor that has proven equally vital: data relationships.
Graph technology is the future for leveraging all three.
The importance of data relationships
Traditional relational database (RDB) technologies do not account for the importance of data relationships. If you have ever opened a spreadsheet, you are likely familiar with the way information is stored in traditional databases. Data is gathered in tables and organized as a series of rows, and each row can contain one or more columns. Any relational information between columns within tables requires building shared pieces of information (keys); and connecting tables requires building a bridge (join) between them. Relationships amongst data are inferred, but not stored. Thus, exploring those relationships is complicated, laborious, and thwarted by technical limitations.
The volume and pace of information from new connected devices chokes traditional data management systems reliant upon arbitrary relational models. They cannot function with real-time, high-volume, sensitive, variably structured, and interconnected data ingested at velocity from a plethora of diverse sources.
A different approach to data
Graph technology is different. Graph databases consist of nodes (representing people, places, and things, for example) and edges (representing the relationships between nodes) — all stored as data points. The nodes and their relationships (edges) become collective data entities in a graph, and their interplay doesn’t require laborious changes to the database to be stored or explored.
Graph databases are schemaless — making them inherently flexible — which is great for complex, connected, and dynamic systems. Semistructured data from wearables and IoT devices, for example, can be streamed into a graph database along with stores of traditional structured health data, allowing their multifaceted relationships to manifest without friction.
This enables data to be represented as a network of relationships in action. These relationships can be traversed quickly, in real time, for pattern detection and discovery and a host of powerful new ML/AI applications. Many popular graph database algorithms deal with paths, centrality, and clustering.
- Traverse algorithms, for example, enable expansive analysis of routes and paths between nodes.
- Centrality algorithms measure the influence and importance of myriad data relationships.
- Clustering algorithms help identify groups that share common traits and identify their relationships to other groups.
The current market leader in graph databases is NEO4J, with other providers supplying a bevy of alternatives including Microsoft (Cosmos DB), Amazon (Neptune), ArangoDB, GraphDB, TigerGraph, Stardog, and OrientDB among others.
Promising developments and tough hurdles
Graph databases have been around for quite some time — underpinning the functionality of social platforms like Twitter, Facebook, or LinkedIn. But their power extends beyond trending hashtags, influencer marketing, and professional networking. Graph technology has seen a surge in adoption over the past few years across enterprises as digitization has increased the volume and complexity of data in every industry.
Gartner’s review of the top data and analytics trends last year noted that graphs will form the foundation for many of the data and analytics automation and modern AI capabilities needed by organizations, reporting that “as many as 50% of Gartner client inquiries around the topic of AI involve a discussion around the use of graph technology.” The firm estimates that by 2025, graph technologies will be used in 80% of data and analytic innovations, up from only 10% in 2021.
In the innovative consumer sports and wellness technology ecosystem, graph technology helps “see the data forest for the trees,” and acquire striking insight. It is especially useful for surfacing previously unidentified or little-understood relationships, such as individual behavioral patterns or coincidental patterns of change, that impact progress, performance, or risk.
But despite the surge in interest, graph technology utilization is still relatively rare in traditional healthcare organizations like hospitals, where master data management is still grounded in old-school database technologies and infrastructure. The acquisition of expertise for modeling highly connected data to define the structure or ontology of graphs, populating graphs from existing data sources, ensuring quality and consistency of data, and managing unique security and compliance requirements (HIPAA, the Cures Act, etc.) make for a tall order in the IT-talent-starved healthcare industry.
That has to change. The utility of graph technologies is too great. The healthcare sector’s capacity to ingest more and richer data from an expanding pool of sources is critical, as is the ability to extract meaning from all that information. By discerning data relationships, graph technologies can unlock insights that bridge amazingly personalized health monitoring to amazingly optimized healthcare delivery.
Leveraging the new health data landscape means embracing the types of data-forward technologies underpinning today’s popular sports and wellness innovations, which will ultimately produce better medicine and more effective healthcare writ large.