Data Models Old

Owned by David Milward
Last updated: Jul 31, 2017 by Adam Milward
3 min read

 

Data Models

A data model might describe or specify:

a database – e.g., the data held in a departmental system

a dataset for research – e.g., the data required for a clinical study

a data standard – e.g., an audit or reporting dataset

a data form – e.g., the data expected on an order form

a data message – e.g., the data included in an electronic order

a data report – e.g., the data presented in a pathology report

Data models are essential for efficient data management.   They provide specifications for developers to work to: in implementing a new query, a new dataflow, or a new system.  They can be used in the automatic generation of data schemas, and the automatic configuration of data tools. 

Data models are essential for good data governance.   They provide accounts of the data held in existing systems, the data flowing between systems, and data required for specific purposes.   They can be compared and audited automatically, allowing managers and data controllers to keep track of information flows in large organisations.

Model Creation

The model catalogue can be used to create a new data model.  For example, as a description of a planned dataset, or as a description of an existing database.   Data elements in this new model can be created by selecting data elements from other models in the catalogue; components of these models – classes of data elements – can be re-used in the same way; links recording the relationship will be created automatically.

Model Editing

The model catalogue can be used to update an existing data model.   The model structure, the data element definitions, the value domains, links to other models: all of these can be updated.   A system of version control applies to all models: previous versions of the model are stored, and links are maintained between data elements.  

The process of editing and revising a data model is faciliated also by a notion of publication status.   A new or working version cannot be accessed by other users, or referred to in other models, until it is formally published.

Data elements in successive versions of a model are linked, unless modified.   If a data element in another model is linked to a data element in an old version, then an indirect link will exist to the corresponding data element in the new version.   The automatic creation of a direct link would not be appropriate, as the intended relationship may no longer apply.

Two different data elements, in two different models, may be related.   For example, a data element in a database model may be based upon a data element in a data standard, as a specification that the database design should follow.   Alternatively, a data element in a research dataset may be based upon a data element in a database model, in the expectation that the data could then be extracted from an existing information system.

The model catalogue can be used to record this information, as a link between the data elements in the two different models. 

It can be used also to record relationships between groups of data elements, if a class of data elements in one model has been based upon a class of data elements in another.