Health Data Standards

A patient's care is recorded in many places: a family doctor's notes, a hospital laboratory system, a pharmacy database, a ward chart, a national registry.

9 frameworks, from 1918 to today · Health Informatics · Medicine & Health

still a working view ended · each leads to its entry below

Overview

A patient's care is recorded in many places: a family doctor's notes, a hospital laboratory system, a pharmacy database, a ward chart, a national registry. Health data standards are the agreed rules that let those records mean the same thing and move between systems without losing their sense1. They come in several kinds. Classifications sort diseases and procedures into numbered categories for counting and billing2. Terminologies capture clinical detail in coded form—not just "diabetes" but which kind, in which part of the body, with what complications3. Messaging standards define the structure of a message sent from one computer system to another4. Identifier systems give laboratory tests and observations stable codes5. Document standards specify how a whole clinical document is packaged6. Device standards let bedside machines talk to information systems7. And resource-based standards model health data as modular, web-friendly objects8.

What is at stake is interoperability: the ability of different systems to exchange data and use it9. Without standards, every hospital builds its own tower of records, and information dies at the boundary. With them, a laboratory result can follow a patient across a city, a country, or a lifetime. The history of health data standards is the story of how the field moved from paper forms and statistical lists toward a global, web-based architecture—and of the compromises and competing philosophies along the way.

Paper Records, Hospital Standardization, and the Birth of the Medical Record Profession

In the early twentieth century, hospitals in North America and Europe began to be judged on how well they documented care. The American College of Surgeons, founded in 1913, launched a hospital standardization program in 1918: hospitals seeking approval had to keep complete, organized medical records. This created a new occupation. The Association of Record Librarians of North America was founded in 1928 (it later became the American Health Information Management Association), and its members—mostly women—developed rules for what a record should contain, how it should be indexed, and how diagnoses should be coded. The problem they solved was basic: a hospital could not audit its own care, or defend itself in court, without records that were complete and retrievable. Their solution was a profession with its own standards of practice, and the medical record became an object of deliberate design rather than an afterthought.

Statistical Classification: The ICD

The other great strand came from public health and epidemiology. International statistical classification of causes of death began in the nineteenth century; the International Statistical Institute adopted a classification in 1893, and the first International Classification of Diseases (ICD) was issued by the World Health Organization's predecessor in 1948. The ICD assigns each disease a code so that deaths and hospital discharges can be counted and compared across places and years10. Its logic is statistical: categories are broad enough to aggregate, and revisions every decade or so reflect changing medical knowledge. The ICD became the backbone of morbidity and mortality reporting worldwide, and later revisions—ICD-9, ICD-10, and the current ICD-11—extended it to procedures, functioning, and traditional medicine. Its limitation is equally important: designed for counting, it cannot express the fine clinical detail a treating doctor needs.

Compositional Clinical Terminology: SNOMED

That limitation drove the development of clinical terminologies. SNOMED—Systematized Nomenclature of Medicine—grew from the Systematized Nomenclature of Pathology, published by the College of American Pathologists in 1965, which organized terms along axes such as topography and morphology. Successive versions added clinical terms, and in 2000 SNOMED RT merged with the UK's Clinical Terms Version 3 to produce SNOMED CT, now maintained by SNOMED International. SNOMED CT is compositional: concepts are defined by their relationships to other concepts, so a clinician can build a precise expression—say, a fracture of a specific bone with a specific type—from simpler parts11. The idea is that a computer can reason with the result, not merely store it. SNOMED CT is now used in dozens of countries, though full implementation remains uneven.

Messaging: HL7 Version 2

Meanwhile, hospital computers needed to talk to each other. In 1987 a group of hospital system vendors and users in the United States formed Health Level Seven (HL7), named for the seventh layer of the OSI networking model—the application layer, where health data live. HL7 Version 2, first published in 1989, defined messages for admissions, orders, results, and billing, built from segments of pipe-delimited fields. Its great virtue was pragmatism: it was relatively simple to implement, and it spread globally. Its weakness was that it was flexible to the point of ambiguity—different sites interpreted fields differently, so "HL7-compliant" did not guarantee that two systems could actually understand each other. Version 2 remains in wide use today, a testament to the value of a workable standard over a perfect one.

Naming Observations: LOINC

A laboratory result is only useful if the receiving system knows what was measured. In the 1990s, Clement McDonald and colleagues at the Regenstrief Institute in Indianapolis developed LOINC—Logical Observation Identifiers Names and Codes—first released in 1994. LOINC gives each laboratory test and clinical observation a stable code built from parts: the component measured, the property, the timing, the specimen, and so on. A glucose measurement in serum has a different LOINC code from a glucose measurement in urine, and both are distinct from a glucose measurement in a different unit. LOINC solved the problem of semantic drift—the same test called by different names at different hospitals—and is now the international standard for identifying observations.

Devices at the Bedside: ISO/IEEE 11073

Medical devices—infusion pumps, ventilators, patient monitors—generate a constant stream of data. The ISO/IEEE 11073 family of standards, developed from the 1990s onward by the IEEE and the International Organization for Standardization, defines how these devices communicate with each other and with information systems. The standards specify not only the transport of data but the nomenclature of measurements and device states, so that a monitor's reading of heart rate can be understood by a clinical information system without custom integration7. Adoption has been slower than for messaging standards, partly because device vendors had little incentive to open their systems, but the family remains the reference point for device interoperability.

The Model-Driven Turn: HL7 Version 3 and the Reference Information Model

By the late 1990s, HL7's leadership concluded that Version 2's flexibility was a trap. The answer was Version 3, built on a formal Reference Information Model (RIM): a small set of object classes—entities, roles, acts, participations—from which all messages would be derived. The RIM was intended to make messages unambiguous and machine-processable. It was an ambitious, model-driven approach, and it produced a generation of specifications, including the Clinical Document Architecture (CDA), which defines how to structure a clinical document such as a discharge summary so that it is both human-readable and machine-processable6. CDA found real use in national programs, notably in the United States and Europe, for exchanging summaries of care. But Version 3 was complex, expensive to implement, and slow to win adoption; many implementers stayed with Version 2. The episode taught the field that formal rigor alone does not guarantee uptake.

FHIR and the Resource-Based Present

The response came from a new generation. In 2011, Grahame Grieve, an Australian health informatics engineer, proposed a simpler approach: model health data as resources—discrete, reusable objects such as a Patient, an Observation, or a MedicationRequest—each with a defined structure and a web API. This became FHIR, Fast Healthcare Interoperability Resources, developed under HL7 and first published as a draft standard in 2014. FHIR borrowed the best of the web: RESTful interfaces, JSON and XML formats, and modular resources that could be combined as needed8. It also learned from the past by reusing established terminologies—SNOMED CT, LOINC, ICD—rather than inventing new ones. FHIR has been adopted with unusual speed by vendors, national programs, and app developers, and it is now the focus of global interoperability efforts, including the United States' interoperability rules and the European Union's health data space.

Where the Field Stands

Today the main approaches coexist rather than replace one another. HL7 Version 2 still carries most hospital messaging; CDA remains in use for document exchange; FHIR is the growth area, extending from clinical data to genomics, public health, and research. SNOMED CT, LOINC, and ICD remain the semantic backbone, and the ISO/IEEE 11073 family continues to define device communication12. The open questions are familiar in new form: how to map between terminologies without losing meaning; how to govern standards so that they serve patients across countries with different resources and legal systems; how to make FHIR implementations genuinely interoperable rather than merely conformant; and how to keep standards current as medicine and technology change. The history suggests that the standards which endure are those that balance rigor with usability—and that the work of making health information travel is never finished.

References

  1. Clinical data standards, Wikipedia
  2. Medical classification, Wikipedia
  3. SNOMED CT, Wikipedia
  4. HL7, Wikipedia
  5. LOINC, Wikipedia
  6. Clinical Document Architecture, Wikipedia
  7. ISO/IEEE 11073, Wikipedia
  8. Fast Healthcare Interoperability Resources, Wikipedia
  9. Interoperability, Wikipedia
  10. International Classification of Diseases, Wikipedia
  11. Systematized Nomenclature of Medicine, Wikipedia
  12. IEEE 11073 service-oriented device connectivity, Wikipedia

The frameworks of Health Data Standards

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1918-Present
  2. 1948-Present

    International Classification of Diseases

    How can diseases be counted and compared across places?

    Fly to its tower in Nousfarer

  3. 1965-Present

    SNOMED CT

    How can clinical detail be coded so computers can reason with it?

    Fly to its tower in Nousfarer

  4. 1989-Present

    HL7 Version 2 Messaging

    How can hospital computer systems exchange messages?

    Fly to its tower in Nousfarer

  5. 1990-Present

    ISO/IEEE 11073

    How can bedside devices talk to information systems?

    Fly to its tower in Nousfarer

  6. 1994-Present

    LOINC

    How can a laboratory test be identified unambiguously?

    Fly to its tower in Nousfarer

  7. 1999-2015

    Model-Driven Standards (HL7 Version 3 and RIM)

    How can health messages be made formally unambiguous?

    Fly to its tower in Nousfarer

  8. 2000-Present

    Clinical Document Architecture (CDA)

    How should a whole clinical document be structured?

    Fly to its tower in Nousfarer

  9. 2014-Present

    Fast Healthcare Interoperability Resources

    How can health data be modelled as modular web resources?

    Fly to its tower in Nousfarer

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

Hospital Standardization and Medical …International Classification of Disea…SNOMED CTHL7 Version 2 MessagingISO/IEEE 11073LOINCModel-Driven Standards (HL7 Version 3…Clinical Document Architecture (CDA)Fast Healthcare Interoperability Reso…

Beyond the field

Where Health Data Standards's ideas came from in other fields, and what they led to there: the strongest 12 of 38 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Health Data Standards is a planet in Nousfarer

Its city has a tower for each of these 9 frameworks. Land, walk in, and climb one concept at a time.

Land on Health Data Standards

Related fields

Fields its ideas are linked to

More in Health Informatics

See where Health Data Standards sits on the Map of Human Knowledge