A substantial body of clinical information resides in electronic systems, represented
using existing coding schemes, terminologies and classifications. This information
may be of value to individual patient records or to population aggregations. Similarly,
there are many queries and decision support protocols that contain knowledge representation
based on existing terminologies.The volume and heterogeneous nature the existing data
means different approaches may be required to meet specific sets of requirements.
Factors that need to be considered include:
The volume and value of existing in the context of the anticipated uses of a future
SNOMED CT application :
Example: In the UK alone, there are over 50 million patients primary care electronic
records coded using one of the versions of the
Read Codes . Based on typical patterns of use this means there are several billion coded record
entries that may need to be taken into account in the migration process. A substantial
proportion of this data has continuing clinical value and thus despite the scale of
the task it is important to ensure that data is migrated accurately and efficiently.
- Data quality and consistency:
- Different users in different settings may select codes and terms in idiosyncratic
ways to reflect their needs. This may be acceptable locally but it creates an obstacle
to migration if the goal is consistent and comparable information at a regional, national
or global level.
- Different source code systems:
- Several different coding scheme versions are in use and each of these poses specific
challenges and offers a different profile of potential benefits.
- Different information systems:
- There are many system suppliers. As a result of system development and commercial
mergers and takeovers, many suppliers support more than one application in the same
domain. The challenge is to migrate from this diverse situation to a next generation
environment supporting standards such as
SNOMED CT .
- Different information models:
- In addition to differences in the use of codes, existing systems inevitably have a
variety of approaches to structuring clinical information. As a result, the process
of migrating data between systems is not simply a question of converting codes. The
underlying architecture of the source data also needs to be taken into account to
make optimal use of existing data without losing processable information or introducing