[Gnumed-devel] codes for narrative - how to store

[Karsten Hilbert]

Here's another question for the fellow GNUmedders:

Basically there's two approaches to associating narrative with
codes from coding systems (LOINC, ICD, ICPC, Ozcode, ...):

1) Have a table "coded_narrative". This table stores a
   text item, a code and a coding system.
   Advantages:
   - code-narrative association can be used to code narrative
     regardless of source table
   - codes are only stored once per narrative
   - code - narrative association lifetime is patient independant
   - once-coded narrative always reusable without recoding
   - avoids code-narrative association mismatch when narrative is changed
   Disadvantages:
   - cannot associate different codes to same narrative
     depending on patient (is this needed ?)
   - this could be mitigated by introducing a link stage
     table that allows for selection of which code-narrative
     association is to be *applied* to a patient's actual
     narrative
   - code-narrative association might get broken up while
     still needed somewhere else (eg no referential integrity
     link into patient narrative)

2) Have a table "lnk_code2narr". This table stores a link to a
   narrative item in the patient EMR, a code and a coding
   system.
   Advantages:
   - allows for patient-specific code-narrative association
   - cannot accidentally delete code-narrative association
     while in use (eg. referential integrity maintained)
   Disadvantages:
   - difficult to allow linking into several narrative source
     tables
   - codes per narrative are stored as many times as the
     narrative is coded, eg quite redundant
   - code-narrative association lost when narrative is deleted
   - prone to code-narrative mismatch after change of narrative
   - once-coded narrative may get lost for recoding if deleted

Problem with both approaches: They do not allow to specify
a *part* of a narrative row a certain code is to stand for.
Whether that is really needed remains to be seen. What *is*
possible, however, is to associate several codes with a given
narrative without regard for the fact that some may apply to
part of the narrative only.

I certainly favour approach #1.

What do other people think ?

Karsten
-- 
GPG key ID E4071346 @ wwwkeys.pgp.net
E167 67FD A291 2BEA 73BD  4537 78B9 A9F9 E407 1346