SNOMED Documentation Search


 Other Documents
Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 2 Current »

Current Version - Under Revision

The form in which records are represented may have a substantial impact on the efficiency, accuracy and completeness of retrieval. The forms that best suit retrieval may differ from the forms that are required to meet the principles of clinically safe and legally valid .

Storing information as entered

This option leaves information in the form entered in the with no additions to assist future retrieval. The application must do all the work needed to locate the required records and compute subsumption and when a request is made to retrieve data.

Using an Expression Repository

An innovative approach to the issues raised by literal storage of is to implement an repository. Each unique used in the system is stored in a referenced database table and assigned an internal unique (e.g. a UUID). When an is used in a clinical record entry the unique id is used to reference the in the repository.

The key advantages of this approach of this approach are:

  • The

    can have a fixed size whereas a

    in of variable and indeterminate size. This significantly improves storage and index efficiency.
  • The

    repository can also be used to store

    representations of each

    and to relate these to the original

    . This optimizes performance for

    normalization during retrieval.
  • The

    repository could also be processed by a

    and a

    table of all the

    used in the application could then be generated.

    retrieval would then be highly optimized by using the

    to test a single join between each predicate and the candidate

    .

Related Links

Minimizing postcoordination

One possible approach to optimization of retrieval is to the original stored information into an equivalent representation with the minimum number of components.

The objective of this approach is to allow the generation of simple indices for the representation. It is then possible to undertake most retrievals using the |is a| to compute whether in the record are of the used to specify retrieval. Where is required, the minimum number of additional tests are required to confirm that a in the record meets the specified retrieval criteria.

One difficulty with this approach is that there may be more than one representation that requires the same degree of . This is discussed in more detail and illustrated in Transforming expressions to normal forms .

If this approach is adopted additional rules need to be applied to determine the choice between alternatives with a similar number of components.

Example:

In the hypothetical example illustrated in , the "red steel pedal bicycle", for which no representation exists, could be represented as:

"red pedal bicycle" + |make of| = |steel|

or

"steel pedal bicycle" + " color " = "red"

Both are equally close to the objective of minimizing . A rule is needed to determine which of these is preferred. There is no obvious right or wrong solution to this but a simple rule that places the attributes in an order will, if applied consistently, allow all representations to be reduced to a single minimized form.

Maximizing postcoordination

An alternative approach is to expand any in the record to their fullest possible forms. This general type of is illustrated in Transforming expressions to normal forms .

This approach requires a richer record structure but has the advantage that there are three possible end-points to , each of which ensures that any computably equivalent representations of will expand to an identical form. The three end-points are summarized here:

  • Short

    :
    • This is the most parsimonious of the three options.
    • A

      is represented as the combination of:
      • with its most proximate

        supertypes;
      • The recorded

        values and/or

        that distinguish it from its most proximate

        supertypes.
  • Long

    :
    • This option is more verbose as it includes some redundancy.
    • A

      is represented as the combination of:
      • with its most proximate

        supertypes;
      • All of its recorded

        values and/or

        , irrespective of whether they are shared by its most proximate

        supertypes.
  • Exhaustive

    form:
    • This option is extremely verbose.
    • A

      is represented as a combination of:
      • with all of its

      • All of its recorded

        values and/or

        , irrespective of whether they are shared by its most proximate

        supertypes.

If the retrieval criteria are expressed in a similar form, a relatively simple query can interrogate the record for all entries with a matching set of and specified characteristics.


Feedback
  • No labels