Hubris

Coding Systems

Up - Home

Vamp coding
OXMIS & Read

The report on classification of data in General Practice

What else could we do?

Read 3 in 1998 MIG paper

Firstly, although there is still some dispute about whether we should code medical information, and which code system we should use, for the foreseeable future we are stuck with doing so.

No coding system is satisfactory to all users, indeed most are unsatisfactory to all users, and even if one was ideal and fully worked out for whoever has control of its development (most unlikely to be the end-users) it would not be a complete solution for everyone.

However, each coding system contains elements which are highly satisfactory and useful, and we must be ready to accept codes from other people

The paradigm of the Web offers one hint at the solution. Web browsers should be strict in what they do, following the standards precisely, but permissive in what they accept. So should our systems.

Therefore the best approach is to design a system which can store codes from an arbitrary number of coding systems, and search on all or any of them. At first look it appears trivially simple to do so, and indeed on further inspection the difficulties appear to relate to copyrights, egos and attempts to coerce the users rather than to the technical difficulty of the solution.

the NHS Data Dictionery "defines data standards to support the collection, sharing and exchange of data throughout the NHS".

Each existing code system consists of a table of codes, often with subcodes, and associated with a text phrase. For use in finding or selecting a code many of the codes will have a keyword ("Termkey" in Read) which serves as a shortcut, however this is not part of the code, and can be omitted or replaced at will.

Structure of a table for storing codes regardless of the coding system they came from

Accession
number

Shortcut

Code
System

Code

Subcode1

Subcode2

Rubric

Velocity

1

Read25

2

Read24

3

Read31

4

ICD9

5

ICD10

6

OXMIS

7

SNOWMED

8

Local

• HC99R1 - the Read 4 character set
• HC99R2 - Read versions 1 and 2
• 9999R3 - regrettably, an arbitrary description for Read version 3

Those might be reasonably used instead of the descriptive strings above.
Description of fields

The table would in my Practice be about 6000 lines long by now, assuming that only codes which have been used from the 78000 lines of Read 2 5 byte version which we use had been copied in. It would also have many ICD 9 codes from hospital notes, and many OXMIS codes from patients notes which have been transferred from practices using VAMP systems.
And all without a single code conversion.

In use the selected dictionary of codes would be the first place the software would automatically look when one wished to code a new item in a patient's notes. This in no way prevents one using the supplied Read Code table which would still be there (unless one's system is supplied with an ICD code table, for which the same argument applies) and in fact if one wished the entire set of codes could be copied in.

However for most users the selection of a new code would be a fairly rare event.

Each coding system is hierarchical in some fashion, and the software would as it already does have to allow users to go up or down a level in the hierarchy. This would be more complex, as the code for navigating the hierarchy would have to be written for each hierarchy (it has already been written of course, but the system would have to include each algorithm) but this is what computers are for.

The Use of Code Dictionaries

They should be a lexicon, a list of phrases actually observed from medical records, thus reflecting the phrases we use and recognise. The effect is to define a reduced subset of English, and to make best use of it this must be the reduced subset of English which is used in medical schools when students are taught to describe patients and diseases (and administrative events and all the rest of the items it has been felt necessary to code).

There is no sign of this occurring