Re: [Gnumed-devel] Measurement workflows - result shading

[Karsten Hilbert]

On Sun, Jul 14, 2013 at 04:47:28PM +0000, Jim Busser wrote:

> We can do without a legend, if the shading is made self-evident by one (maybe 
> both) of the following:
> 
> - a character indicator appended (after two spaces) to the right of the 
> result in the shaded cell
>       see 
> http://lists.gnu.org/archive/html/gnumed-devel/2013-07/pngDRyvVBnnRD.png
> 
> - information contained in the tooltip

In that case we only need one color. A second may be a nice
convenience.

> Several potential scenarios exist, and it may be a good idea to agree first 
> on a limit to how many colour states.
> 
> Might a colour for each of the following be excessive, and how many of these 
> states should be coalesced?
> 
> 1. result is above upper limit of *target* range
> 
> 2. result is above upper limit of *normal* range in 3 situations:
>       with no target defined
>       within the target range (think: INR 2.6)
>       below the lower limit of the target range (INR 1.6)
> 
> 3. result is within the normal range in 3 situations
>       with no target defined
>       within the target range (think: INR < 1.5 peri-procedurally)
>       below the lower limit of the target range
> 
> 4. result lacks numeric normal/target data, but is coded / flagged by the lab
>       4a. with an indicator
>       4b. with a comment, but no indicator
> 
> - plus the mirror (inverses) of #1-3 ...
> 
> At the highest level of abstraction, the questions include
> 
> 1) whether, in the case of results that are computationally evaluable as 
> "high" or "low", to have distinct colours for
> 
>       numerically above target (supra-therapeutic, else *unacceptably* 
> elevated)
>       numerically below target (infra-therapeutic, else *unacceptably* 
> depressed)
> 
>       numerically above normal
>       numerically below normal
> 
>       … i.e. 4 shades
>       … or whether to combine together the two "aboves", and the two "belows"
>       … and whether -- if combining -- "targets" (when present) overrule 
> "normals"
> 
> 2) whether to map, and auto-interpret, a range of potential lab supplied 
> indicators such as
> 
>       + -
>       high low
>       h l
> 
> as meaning "high" or "low" except I think this could be unreliable depending 
> on the source labs and the language being used, and therefore clinically 
> risky.
> 
> 3) whether, in the case of the remaining results, which are *not* readily 
> evaluable as "high" or "low", but are nonetheless characterized by any of
> 
>       lab-supplied indicator
>       lab-supplied comment
>       clinician-supplied (or revised) indicator
>       clinician-supplied (or revised) comment
>       clinician signed-as Abnormal
>       clinician signed-as Relevant
> 
> to assign yet some *other* colour, or to dispense with colour for these.

This is entirely too complicated for the average user.

I suggest 2 colors at most (well, 3, if you count "neutral"):

1) below target, or below normal if no target

2) above target, or above normal if no target

After all, the very fact that a target is defined means that
the normal is not applicable, by definition. We could do
with one color (out-of-target or out-of-normal if no target
defined) if we combine that with the already existing
abnormality indicator.

I don't think auto-interpreting + and - is risky.

Karsten
-- 
GPG key ID E4071346 @ gpg-keyserver.de
E167 67FD A291 2BEA 73BD  4537 78B9 A9F9 E407 1346