The challenges of HL7 to FHIR data integration
by Mike Hayball, Chief Technology Officer
Much of the vision and initiatives that form the foundations of the 10-year-plan for the NHS are underpinned by the reality that to better connect services to improve patient care, our healthcare data first needs to be better connected.
Across the course of a patient’s care journey they will accumulate and create data over numerous different systems, and finding meaningful ways to integrate that data to use and share it remains a central concern. Without the fundamental ability to connect healthcare data, adopting the technologies and innovate to overcome the most significant healthcare challenges will be impossible.
As we develop the neighbourhood care model, we will also need to integrate non-NHS healthcare data also, which further complicates the picture.
Although the industry is steadily moving toward modern standards like HL7 FHIR (Fast Healthcare Interoperability Resources), the legacy of older systems continues to pose significant challenges.
Despite the promise of data standardisation and the availability of tools that offer interpretation, however, much of the work in integrating data remains complex, error-prone, and difficult to automate.
At Feedback, our integration team collaborates closely with individual care settings to tailor data integration solutions that ensure Bleepa® can seamlessly access comprehensive patient information, breaking down isolated data silos.
As our Head of Integration, Thomas Spankie, puts it:
“You can’t get away from doing the dirty work of data transformation. You can move the problem around, but invariably you end up with a human doing the heavy lifting at some point. It’s a moving target, and there’s no silver bullet,”
Thomas Spankie, Head of Integration,
Feedback Medical
The challenges
- Semantic mismatches
HL7 v2 messages are often terse, cryptic, and highly context dependent. They were designed for rapid message exchange in local systems, not necessarily for semantic clarity. FHIR, by contrast, emphasizes readable, self-describing data structures. Bridging this gap involves more than simple field mapping; it requires a deep understanding of both data models to ensure that the meaning and intent of the data are preserved.
- Customisation and variability
No two HL7 implementations are truly the same. Hospitals and vendors often customize HL7 messages to fit their unique workflows. As a result, integration teams must contend with idiosyncratic fields, local code sets, and inconsistent usage patterns. FHIR attempts to offer a consistent model, but mapping to it from these inconsistent sources often requires bespoke transformation logic.
- Inadequate tools
There are tools that promise to automate parts of this process—Mirth Connect, interface engines, FHIR converters—but the reliability of these tools remains dependent on the people who configure them. Automation can assist, but it cannot replace the need for domain knowledge and context-sensitive decisions. Most of the time, it takes a human to interpret ambiguous or incomplete data and decide how best to represent it in FHIR.
- Evolving standards
FHIR itself is still maturing. While it is widely adopted and supported, the standard is continually evolving, with frequent updates to resource definitions, implementation guides, and profiles. Keeping up with these changes is a constant challenge, especially when working in environments with tight regulatory or production constraints like healthcare.
- Validation and testing
Even after data has been transformed, ensuring that the result is correct, complete, and clinically safe requires rigorous validation. This step often falls to integration engineers or clinical informaticists who must verify not only that the data looks right but that it behaves correctly when used downstream—for example, in clinical decision support, reporting, or billing.
No simple solution
HL7 to FHIR integration is not a plug-and-play process. It’s an ongoing effort that requires a blend of technical expertise, domain knowledge, and human judgment. While tools can help streamline parts of the workflow, they cannot eliminate the ultimate need for human oversight. The burden of getting it right ultimately rests with people—not just code.
The reality is that healthcare data is inherently complex, and there is no simple solution.
Achieving meaningful integration requires not only collaboration but also appropriate expertise and a clear understanding of how to handle and interpret this data effectively. Only by combining these elements can we properly integrate healthcare data to embrace technology innovations that will improve patient care.