AnzarSeha

The clinical registry

By Dr Rida Akodad · Publication director · Updated 21 August 2026

A clinical registry is an organised, continuing collection of data on a population defined by a condition, an intervention or a device. It does not answer a question: it constitutes the resource from which several questions can be asked, including ones nobody has formulated yet.

Registry and study database: the founding distinction

Conflating the two is the error that condemns most registries to die after their first publication.

Study databaseRegistry
Starting pointA question askedA population defined
DurationThe length of the studyContinuous, with no planned end
Variable scopeWhat the planned analysis requiresA stable core, extensible
Success criterionA published answerA reusable resource
FreezeOnce, before analysisBy dated extraction; collection continues

The practical consequence is sharp: a registry is not frozen, snapshots are taken from it. Each study drawn from the registry freezes its extraction, not the registry itself. That is what lets one source feed several pieces of work without the second contradicting the first.

The design decisions that determine its lifespan

1. Define the population by an operational criterion

"Patients of the department" is not a criterion: it varies with organisation. "Every inpatient with a principal diagnosis coded C61 under ICD-10" is one, because it is reproducible by somebody else, and next year.

2. Separate the core from the extensions

A registry needs a small core of variables collected for every case, plus optional extensions per sub-population. A registry where everything is mandatory becomes impossible to complete; one with no mandatory core supports no cross-sectional analysis.

3. Plan versioning from day one

Definitions change: a classification is revised, a score's threshold moves. A registry that overwrites the old definition loses comparability across its own years. Each variable must carry the version of the definition it was collected under.

4. Decide who enters data, and when

This is the question that kills registries, and it is organisational rather than technical. Entry that depends on goodwill after a shift does not survive six months. The registries that last are those whose entry is anchored to an existing moment in the care pathway.

Governance

A registry is a collective resource, which raises questions a study database does not, and which must be settled before the first inclusion.

  • Who decides on access? A committee, not a person, under a written procedure.
  • Who may publish from the registry, and on what terms? Without a rule, the first team to publish appropriates a resource others built.
  • What is the inclusion policy? Consent, information, or a waiver under the applicable framework — the answer determines which uses are possible.
  • What is the retention period, and what becomes of the registry afterwards?
  • Who is the data controller? The institution, usually — the question must be written down, not assumed.

What makes a registry good

  • Case ascertainment, measured: what proportion of eligible cases is actually in it? A registry covering 60% of cases produces biased results in a way its size does not correct.
  • Stability of definitions, or failing that, explicit versioning.
  • Core completeness: fifteen variables filled everywhere beats eighty filled a third of the time.
  • Traceability: every value knows where it came from — see provenance.
  • Normalised coding: SNOMED CT and ICD-10 make the registry comparable with others.

Ascertainment deserves particular emphasis: it is the hardest property to obtain, the most rarely measured, and the one on which the validity of every descriptive analysis drawn from the registry depends.

How AnzarSeha addresses it

  • The registry is the single source: records are typed and versioned there, and each study builds on it instead of starting from an empty file.
  • A study drawn from the registry freezes its extraction; the registry keeps receiving records.
  • The dictionary carries mappings to standards, which keeps coding normalised.
  • Each study is isolated from the others by row-level security, closed by default.

Frequent questions

How many variables belong in a registry core?

Few enough to be filled systematically — in practice, a few dozen. The right test is not the count but the time: if the core takes more than a few minutes per case, it will not be completed exhaustively, and ascertainment is worth more than richness.

Does a registry need individual consent?

That depends on the applicable legal framework, the data type and the purpose, and cannot be answered in general. What is constant is that the question must be settled and documented before the first inclusion, not after.

Can a study database be turned into a registry?

Partly. Existing data joins the registry if its definitions are documented; ascertainment, however, cannot be recovered retroactively, since cases not included at the time stay excluded. The registry therefore really begins on the date its inclusion rule starts applying.

What is the difference between a registry and a health data warehouse?

A warehouse gathers data produced for care, with all its heterogeneity. A registry applies a population definition and a variable core decided in advance. The warehouse is broader, the registry more comparable.

Sources and standards