Re: [Gnumed-devel] [Fwd: Re: [GPCG] Embedded Flags]

[Karsten Hilbert]

> wrt to fast word search, how about pre-emptive indexing of clinical notes :
Sounds very helpful.

Once we actually *have* clinical notes.

> 3. add the clinical note id and lines of each token to a token link
> table.
That's the one thing I have always marvelled about: OTOH one
wants to put allergy information into an allergy table, BP in
a measurements table, etc. etc. A large chunk of free text is
in some generic SOAP-like tables, of course. So there's bits of the
clinical information spread out in various places where it
makes sense. That's good from a data handling point of view.
It's a nightmare from a fulltext search point of view.

Which prompts me to realize that we need to link all those
little "comment", "note", "description" fields into one larger
generic "clinical_free_text" table that has some "type" value
associated with each snippet. So, the allergies table would
pull the elements for it's "comment" from the
"clinical_free_text" table according to its foreign key
values. And those rows in the "clinical_free_text" may or may
not be marked with types (SOAP, at the crudest useful level)
or some generic type. Thanks for jogging my brain.

> Then finding 109 of 57,000 is as fast as doing a search on "implanon"
> in the token link table. 
Well, surely nice, but it ain't very likely that I'll need
this result in a split second. Plenty of time for that query
to finish.

>> Another aspect which interests me is the ability to embed flags in progress
>> notes for later identification of groups of patients for GP research.
This is useful in general. I can imagine writing free flow,
marking a word with a few keystrokes (2, really) and pressing
ALT-D to say "that's a diagnosis" which internally invokes
further processing. Or marking by some special sequence. This
has been brought up by some of the doctors on the German list
as well.

>> Of course for this to be useful it is worthwhile for younger GPs to develop
>> a habit of embedding strings which mark those conditions / events /
>> procedures / consents etc in which they might later have a research
>> interest.
IOW, to engage in witchcraft. In that case I'd prefer Syans
approach of tokenizing all non-trivial terms by brute-force.
Personal discipline as to vocabulary used goes a long way
anyways, which would be required in the mark-up case, too,
IMHO.

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