The Healthcare Interoperability Universe: FHIR, Governance, and Transparency

By Pieter De Leenheer, Ph.D., CTO, 1upHealth
LinkedIn: Pieter De Leenheer
LinkedIn: 1upHealth Inc.

The healthcare industry is generating data at an astronomical scale, from clinical and claims to genetic and lifestyle data via wearables and other IoT devices. This zettabyte scale reservoir of largely untapped and unstructured data opens a plethora of promising applications in areas such as care management, population health, and prior authorization. However, to realize its true value, all this data needs to move better across the four trillion-dollar system, and that starts with data interoperability and transparency, which are deeply intertwined through data governance and data intelligence.

Making Interoperability Possible

Computable interoperability refers to the technological capability for systems to exchange data seamlessly where both sender and receiver understand the meaning (read: the semantics) of the data based on agreed-upon ontologies. The Fast Healthcare Interoperability Resources (FHIR) standard has emerged as a pivotal ontology for exchanging health data within the cloud. Developed by Health Level Seven International (HL7), FHIR leverages open standards, such as JSON, OAuth, and REST, and other ontologies, such as SNOMED and ICD, making it an extensible and widely applicable interoperability tool. FHIR’s adoption is supported by various stakeholders, including bipartisan federal legislation, international health communities, and venture capital-backed tech companies.

FHIR is designed to be a modular and open-source standard, encouraging global adoption and continual refinement. It allows for local customization to meet specific needs, which is critical given the diverse regulatory and operational needs in healthcare. The ongoing development of FHIR includes the addition of new resource types to its ontology and the establishment of implementation guides by communities like DaVinci and CARIN, which advocate for the standard’s extension. The recently published rule on CMS Interoperability and Prior Authorization (a.k.a. CMS-0057-F) is expected to drive more FHIR network effects. Building on the technological foundation of the May 2020 CMS Interoperability and Patient Access final rule, this new rule intends to improve patient, provider, and payer access to interoperable patient data and reduce the burden of prior authorization processes.

The Power of Data Governance

Data governance is critical for interoperability as it constitutes the orchestration of activities to underwrite the meaning, quality, compliance, security, and availability of data. Every enterprise has a data governance operating model that defines how it delivers high-quality health data to customers and beneficiaries, as well as how the organization runs itself. Federated data governance, such as that outlined in the latest CMS rule, would do this in an open-ended, distributed, and often inter-organizational setting. A single data element (or FHIR resource type) may trigger myriad internal or public data policies regarding data privacy, ethics, and interoperability, depending on its context of application. An operating model running data governance at the scale of US healthcare should be able to exchange trillions of data elements between millions of stakeholders, and should balance between innovation (i.e., maximizing flow) and risk mitigation (i.e., constraining flow).

The Case for Health Data Intelligence

Data governance activities intend to underwrite data veracity, or, trustworthiness and truthfulness. Therefore, transparency provided by auditable metadata is key. Data quality, compliance, security, and availability are continuously monitored and recorded in the form of metadata. This metadata is often called “data intelligence” as it explains events pertaining to a data product’s lifecycle, and helps to further remediate and fine-tune data governance activities to avoid or resolve issues faster in the next release. Events in the data’s lifecycle can include changes in the schema, data volume anomalies, data quality trends, and opportunities for (compute and store) cost optimization.

Data intelligence is an essential part of the data product fact sheet. Like Carfax for cars, it provides a standardized and transparent way to see the history, purpose, and service-level objectives of a data product. It drives trust and, with it, adoption and reusability. Just like software components, data products have to be curated, cataloged, and continuously improved so they are easy to discover, access, and apply. This requires a comprehensive data product development lifecycle, which includes the following key phases:

  • Discoverability: high-precision search and retrieval of relevant data products in terms of active metadata
  • Cataloging: curation of data products, including selecting, onboarding, certifying, and publishing relevant data products
  • Applicability: frictionless and secure access to selected data products with proper governance controls
  • Monitorability: subscribing to data change events while the data product is in use

Once healthcare organizations notice the adoption of their foundational health data product catalog, they can start to assess the extent to which their interoperability investments pay off. They can drill deeper into which foundational data products drive the most adoption and why and then use that as a basis to expand their investments into more advanced data product capabilities.

Take, for example, clinical and claims data products that each on their own have specific value for medical and underwriting use cases, respectively. Once combined in an aggregated data product, they can open up higher orders of value for more complex use cases such as population health and prior authorization. This aggregation exercise might also uncover patient entity resolution challenges and necessitate data acquisition strategies At this stage the data product catalog – which is initially populated with foundational data products that depend on internal data flows – is extended with aggregate data products that recursively rely on other data products, effectively generating a mesh of data products, or “data mesh”.

Meshing it all Together

Conceiving data as a product extends beyond facilitating data interoperability. The modular programmability enabled by recombining data products across various domains – clinical, lifestyle, and financial – can uncover correlations to inform preventative treatments and optimize healthcare delivery. This integration enables a deeper understanding of the relationship between patient behaviors and tactics, risk factors, and health outcomes, which are essential for advancing personalized medicine and value-based care models.

Current interoperability standards like FHIR are not yet capable of encoding transparency into health data products, which is needed to make them easily discovered and meshed into aggregate data products at scale. That is why FHIR-compliant APIs and CMS-regulated data exchange protocols require robust and transparent quality and governance controls. As FHIR experts innovate and refine the standard to further support healthcare’s needs, we will see more progress toward unlocking the true potential that health data contains.