Healthcare data managers have been hearing a lot about two interesting new approaches to building a data architecture: data fabric and data mesh.
With the increasing volumes, variety, and complexity of incoming healthcare data, and the pressures to enable data access while maintaining governance and security, healthcare data managers are looking for a reliable, scalable solution.
And they are increasingly aware that, no matter what approach they take, identity data management will have to be at its foundation.
What are data fabric and data mesh?
The similarity of the terms fabric and mesh seems to imply that these are variations on a single approach.
While both are trying to come to grips with the same complexities of healthcare data management, their paths to a solution are different. Data fabric is a technology, while data mesh is a process of organizational change.
Data fabric unites various data repositories with a unified integrated access layer available across the IT infrastructure. Rather than move data around, as current data warehouses or data lakes do, data fabric connects data with metadata. It automates many processes that are currently manual, increasing speed and reducing staff time.
Data fabric focuses on data integration.
Data mesh is an architectural framework for data management. Rather than uniting the data, as data fabric does, it distributes ownership of the data to various domains in the healthcare system, such as EMR, diagnostic imaging, and billing. Each domain owns and manages its own data, the data it has the greatest expertise in.
So the “mesh” is a network of interconnected domains, each creating high-quality data usable by the rest of the organization.
Data mesh focuses on data management.
Both are ways for users to get access to data from different sources in a common way, and try to solve the very real problems of healthcare data management.
Data mesh is more distributed, and views data produced by each domain as a product for data consumers throughout the organization, while data fabric is more centralized and automates the process with a variety of tools.
Since data fabric is a technological approach while data mesh is an organizational approach, they are not mutually exclusive, and healthcare organizations will likely adopt elements of both.
How do they improve healthcare data management?
Healthcare data managers must provide a wide range of users with quick and easy access to a wide range of data while adhering to governance and security rules and maintaining privacy.
The benefits of using either data fabric or data mesh include
- A connected data ecosystem beyond the EHR
- A comprehensive view of patient data
- Trusted analytics at scale that support deeper insights and improved decision-making
- More efficient data management
- Secure data sharing that supports improved collaboration
However, it’s important to understand that these benefits will not be achieved if either approach isn’t founded on trusted identity data management.
Why do they rely on trusted identity data management to succeed?
Healthcare data is about people, whether they are patients, providers, staff, or community members.
Regardless of how data is eventually used, each piece of data is originally generated in some specific domain. Billing systems generate billing data, CT scanners generate diagnostic image data, and physician visits generate unstructured interview data. To generate value, all this disparate data must be reliably linked to the person who generated them, or who it relates to.
Whether you keep data in its separate domains and provide access to it, as with data mesh, or integrate that data completely, as with data fabric, you need a way to provide reliable data management. You have to know who is who, or you can’t reliably tie all that data to the same person record and create a complete and trusted 360-degree view.
The process is extremely difficult. People change addresses, names, and other identifying aspects of themselves. Many different people have the same name and date of birth.
Failing to accomplish this first risks introducing significant errors and vagueness into the data on which your analytics and decision-making are based. Data can be duplicated or tied to the wrong individual. You also risk exposing private health information to the wrong person.
Without trusted identity data management, both data fabric and data mesh will unravel.
What do data and analytics officers need to consider?
No matter how sophisticated their data management or data integration, neither data fabric nor data mesh can bring success with poor-quality data, or with data that is not unambiguously tied to the right person.
Data governance is an important aspect of any solution. Because they differ in how centralized they are, data fabric and data mesh have differing requirements for effective data governance. The top-down nature of data fabric means that a central authority can set data policy guidelines. In the more federated data mesh, differing domain-specific policy rules and guidelines need to be reconciled in a negotiated way, including input from every domain.
Access rules and controls may also differ in how they are implemented, since data fabric data access is via APIs or SDKs, while data mesh access is through a controlled dataset.
Cloud transformation is a key initiative for healthcare organizations. Data fabric is entirely cloud and platform agnostic, meaning that integration across multiple cloud platforms is seamless. Current offerings from cloud platforms provide all the data-service tools needed to support a data mesh architecture. Both data fabric or data mesh approaches take full advantage of cloud infrastructure for data management and analytics services.
Whichever choice you make, either approach will rely on identity data management to ensure the correct references for the underlying data. The best identity data management system matches data across different tiers of data quality without eroding the integrity of identity management, or unnecessarily discarding data.
Once you have reliable identity data management in place, you can confidently move forward with the data architecture you have identified as consistent with your organizational capabilities and goals.
