Skip to content
DEXPRADEXPRA home

Interoperability

HL7 101: Four Standards, One Name

Ask three engineers in a hospital what HL7 is and you will get three answers, all correct: pipe-delimited text over a socket, an XML clinical document, and a JSON REST API. They are three different standards, built decades apart, that happen to share a name — and none of the four generations has been retired. This paper is the orientation layer, written for engineers, data scientists and technical product people who have to work against real hospital interfaces without years of integration background behind them. It is organised generationally rather than alphabetically: what v2 is and why it still carries most of the traffic, what v3 got wrong and which parts of it survived, what CDA is for and why reports of its death are wrong on the evidence, where FHIR fits, and which piece of US regulation put it everywhere. Current to September 2026, and checked against ANSI/HL7 V2.9.1-2024, C-CDA Edition 5 and FHIR R4.

What’s inside

  • Four generations and none of them retired: which HL7 a vendor means, and why the 1987 one still carries the most traffic
  • A v2 message on the wire: segments, fields, components, the five delimiters, and the escape sequences a hand-rolled parser corrupts
  • The MSH dissected field by field, and the trigger event families you will actually meet — ADT, ORM and OM*, ORU, SIU, MDM, DFT, VXU, QBP
  • MLLP in three framing bytes, the acknowledgement contract in MSH-15 and MSH-16, and why none of it is secure
  • The optionality problem in the standard’s own words, Z-segments, and the conformance profiles that would fix it and are not used
  • v3 and the RIM: why the model-driven answer failed, and the parts of it still running in production
  • CDA and C-CDA through to Edition 5, FHIR placed among its siblings, and why a project starting today builds on R4 rather than R5
  • The regulation behind FHIR adoption, the seam where HL7 meets DICOM, nine expensive mistakes, a glossary and a full source list

PDF · 24 pages

Request this resource

Working on one of these problems?

We’re publishing more than we’re consulting right now. If you have something specific in interoperability, data platforms, or clinical AI, tell us what it is and we’ll tell you what we think.

Get in touch