Re: [Gnumed-devel] state of test results handling

[Karsten Hilbert]

On Sat, Mar 29, 2008 at 12:04:20PM -0700, James Busser wrote:

>       sort order (or vertical presentation order) among the results that are 
> to be viewed...

Currently, results are sorted by
- timestamp
  - episode
    - unified name

> many of us are very accustomed to seeing tests presented 
> with both internal relationships (tests within a battery
> like hematologic tests Hb WBC platelets) appearing together

Over here are mostly represented without internal
relationships other than "ordered together" (not
test-battery related). Battery parts often *appear* next to
each other on display, however, as that is their incidental
ordering provided by the labs -- which is, incidentally
again, preserved locally.

But ordering, of course, makes sense and is, in fact, on the
TODO for one of the next iterations. It'll be one of the
available sorters alluded to in a previous mail.

> and --- in my case --- above 
> the "chemistry" with immunology and other special testing and 
> microbiology and pathology lower down.
That is, in fact, another quality of sorting which will be
provided eventually. This is what Liz' test panels are about
-- the local user can define content and order of any panels
them deem sensible and order/filter by that.

> Is there some way to establish and 
> manage this?
Not yet.

> I do notice, within the tooltip or right-click, the fields 
> "Grouped under" and "Type comment"
That alludes to our grouping of tests of different labs
under the same common name, say,

Lab A test name: "CRPxact (our super-exact version".
Lab B test name: "el cheapo CRP (we do it for less)".

Labs are wont to profile themselves where possible. Doctors
couldn't care lass at the point of care.

Importers would duly create two different test types from
the above data.

Locally both of the above would be associated under a
unified test name "CRP" and displayed in one very row.

>       test names in plain text, instead of  bold
Done. However, test names will then not be visually
different from test results (aparat from results now being
monospace) so I set the column 0 background colour to be the
same as the row/column label background colour thus making
column 0 appear to be a label. Make sense ?

>       right-justify the results, within the cells
They are, aren't they ?

>       for non-numeric results, present the first 4 characters?
Ah, excellent point. Make that 8, perhaps ? If more than 8
display 5 + ... (to indicate there's more) else display up
to 8.

>       pad the right end of each result with 3 "space" characters
>       3rd space is populated with first character of abnormality_indicator
>               (dunno if this will get confused by varied indicators, like * 
> or H or h 
> or upward-pointing arrow)
Current logic:

- pad result with 6 spaces
  - I have observed "+++" as an indicator
  - collapsing +++ and ++ and + to 1 char falsifies data (acceptably ?)
- fill those spaces with " (3-chars-of-indicator)" or 6 spaces if no indicator

- this will align result values given ...

>       maybe needs to be monospace font to maintain vertical visual alignment

... this is (now) done, too.

>       make technically abnormal results red

Which logic do you want me to apply here ? We have several
sources of abnormality state:

- test_result.abnormality_indicator as provided by test_org
        - we already explicitely display that

- your (the current users) result-signing decision

- the intended reviewers (may or may not be you) result-signing decision

- anyone elses result-signing decision

- neither of the above available but availability of either of
  - normality: min/max/range
  - clinical target: min/max/range
    (this might be more clinical relevance than technical abnormality)

How to apply in which order ?

>       suggest clinically significant results do not affect formatting, but 
> only the filterability
Currently, layout is not affected but the result is rendered
in bold and firebrick red -- depending on your and the
reviewers (in that order) idea of significance. You really
want no indication whatsoever of clinical significance ?

I'd have figured significance to be of greater concern than
abnormality since the latter just may have been flagged
solely by a machine somehwere upstream while the former must
have undergone *some* sort of deliberate (hopefully appraisal)
process taking into account degree of abnormality and
clinical context. Machines flag all kinds of things.

> - if it would be possible for a single view to display a mix of  
> already-signed plus unsigned results, is it desired that any distinction 
> be made visually?
I do believe so but this is as yet unresolved as to how to
represent. I am inclined to either color code or add a
one-character indicator to the result. A "?" perhaps ?
Ideally a crossed out checkmark but we can't rely on fonts
to contain one even if it'd be defined in the unicode range.

Any of these might work:

http://www.fileformat.info/info/unicode/char/2300/index.htm
http://www.fileformat.info/info/unicode/char/270D/index.htm
http://www.fileformat.info/info/unicode/char/237B/index.htm

> or just rely on persistence of the unreviewed results 
> line item in the inbox?
No, we need some sort of immediate indication of sign-off
state right from the results viewer, too.

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