Stream: implementers
Topic: observation questions
Grahame Grieve (Mar 06 2016 at 08:12):
hey all - where do you think that posture goes in an observation?
Grahame Grieve (Mar 06 2016 at 12:38):
http://www.healthintersections.com.au/?p=2487 - comments welcome please. I think we should add the table at hte bottom to the specification , along with generated profiles. Additional suggestions for new rows in the table welcome
Grahame Grieve (Mar 06 2016 at 12:39):
Eric / Rob, note this list of aggregate data aspects:
Grahame Grieve (Mar 06 2016 at 12:40):
"average",
"maximum",
"minimum",
"standard deviation",
"variance",
"sum",
"median",
"count",
"quartile deviation",
"80th percentile",
"lower quartile",
"upper quartile",
"1st quintile",
"2nd quintile",
"3rd quintile",
"4th quintile"
Grahame Grieve (Mar 06 2016 at 12:41):
This is taken from openMHealth. we *really* need to sort out how that's going to work ASAP please
Josh Mandel (Mar 07 2016 at 00:35):
In my assessment, there is no good place to record a patient's posture in a post coordinated way in the current observation resource. Of course if you are measuring something for which pre coordinated codes exist, can you can capture a sitting blood pressure using a code like LOINC panel 34553-8
Grahame Grieve (Mar 07 2016 at 00:42):
posture is only relevant for some observations, though it's strongly relevant for them. Is it extension territory? should we propose a new element? Would it go in method?
Josh Mandel (Mar 07 2016 at 01:27):
I think it's not really a method. I mean, if you want to do the RIM-ish thing (and have nobody understand you) then you use Observation.related with a type of qualified-by. But it we want to support this nicely as a common case, it's adding something like bodyPosition (as an extension or a base element).
Grahame Grieve (Mar 07 2016 at 01:35):
so what does posture matter for?
- blood pressure
- height
?
Josh Mandel (Mar 07 2016 at 01:58):
Yeah, and even for height it's standard practice to use a 'length' code when taking a supine measurement. So really orthostatic BP is the prime use case I'm aware of.
Josh Mandel (Mar 07 2016 at 01:58):
Precordinated coded done seem so bad.
Grahame Grieve (Mar 07 2016 at 02:03):
that's troubling. What's the difference between 34553-8 and 55284-4?
Grahame Grieve (Mar 07 2016 at 02:05):
and if we said to use pre-coordinated LOINC values, would everyone have to know the relationship between 8455-8 , 8453-3 and 8454-1 ? (though since LOINC doesn't appear to provide a non-precoordinated term, it seems there's no option but to recommend both loinc and sct then?
Rob Hausam (Mar 07 2016 at 06:00):
34553-8 is a BP + heart rate panel with the the multiple body positions. I was going to mention heart rate as another observation where posture/body position may matter (but it doesn't always matter). 55284-4 is just a combined observation for systolic/diastolic BP which is discouraged from use (except as a section header). The recommendation instead is to use the separate BP observations of 8480-6 (systolic) and 8462-4 (diastolic).
Position should only infrequently matter for height (although there are specific LOINC height codes for lying and standing). I don't see any LOINC "length" codes for the entire body - the height code(s) would still be used, so I'm not sure what Josh is referring to with that.
Rob Hausam (Mar 07 2016 at 06:03):
I'm not quite sure what "non-precoordinated" term you would be thinking of that LOINC could provide, Grahame. If you are post-coordinating the position, then I think you would have to use LOINC and SNOMED CT (in the information model).
Grahame Grieve (Mar 07 2016 at 06:03):
so we recommend use of 8480-6 (systolic) and 8462-4 (diastolic) for the components (generally, even though the panel code says that 8480-6 is for standing, and I've never actually seen anyone have their blood pressure checked while standing. And what do you use for a removte monitory on a patient's arm where the patient gets to choose their own posture?
Grahame Grieve (Mar 07 2016 at 06:04):
what should we recommend for the overall observation? 55284-4?
Grahame Grieve (Mar 07 2016 at 06:05):
why does LOINC not say that 8480-6 is for standing in the code? is this LOINC being inconsistent?
Grahame Grieve (Mar 07 2016 at 06:05):
.. and why does Dan never respond to any questions.... so many questions....
Rob Hausam (Mar 07 2016 at 06:27):
the 34553-8 orthostatic BP panel actually includes 8460-8 "Systolic blood pressure--standing" (not 8480-6), so it is correct
when you are measuring orthostatic BP's, as is the panel, you would make a standing measurement (if the patient tolerates it), as well as sitting and supine
for the "overall" observation I think you could also use 55417-0 "Short blood pressure panel" (as only the systolic and diastolic components are required)?
Grahame Grieve (Mar 07 2016 at 06:28):
what good is a check digit when you end up with 8460-8 and 8480-6?
Grahame Grieve (Mar 07 2016 at 06:28):
and then use them for variants of each other :-( grrr
Rob Hausam (Mar 07 2016 at 06:29):
yeah - good question
Grahame Grieve (Mar 07 2016 at 06:29):
but you're saying that this example - http://hl7.org/fhir/observation-example-bloodpressure.html - is wrong? we better sort this out.
Grahame Grieve (Mar 07 2016 at 06:37):
I updated http://www.healthintersections.com.au/?p=2487 for this discussion
Rob Hausam (Mar 07 2016 at 06:37):
I don't know if the use of 55284-4 is actually "wrong", but its use is at least discouraged (unless you would want to call this use a "section header" in some sense, which is a probably a good bit of a stretch)
so maybe we should think about changing it to 55417-0?
Rob Hausam (Mar 07 2016 at 06:37):
ok
Grahame Grieve (Mar 07 2016 at 06:41):
well, do you want to make a task
Rob Hausam (Mar 07 2016 at 06:47):
I was also going to mention that posture/body position is
relevant for a good number of x-rays and other diagnostic radiology procedures - PA and lateral (standing), AP (normally supine) and left lateral decubitus chest x-rays are some examples
Grahame Grieve (Mar 07 2016 at 06:47):
where do they appear in the picture? diagnostic order? report code?
Rob Hausam (Mar 07 2016 at 06:51):
not quite sure what you mean
but the LOINC code 24637-1 "Chest X-ray AP left lateral-decubitus" has a component of "View AP L-lateral-decubitus", for example
Grahame Grieve (Mar 07 2016 at 06:52):
that wouldn't manifest in an Observation
Grahame Grieve (Mar 07 2016 at 07:00):
and it's rather overlaid with the idea of projection too, no? At least last time I had an image performed it was
Rob Hausam (Mar 07 2016 at 07:01):
still not sure that I'm following you
but it seems like DiagnosticOrder.item.code and DiagnosticReport.code would be the right place
Grahame Grieve (Mar 07 2016 at 07:01):
I'm just trying to rule it out of scope ;-)
Rob Hausam (Mar 07 2016 at 07:05):
ok, I see - I'm not sure you can do that, at least not easily
where it's actually needed it's pretty well entrenched - and generally pre-coordinated, I think
Grahame Grieve (Mar 07 2016 at 07:09):
out of scope for *Observation*
Rob Hausam (Mar 07 2016 at 07:18):
I still think you're going to have to capture it in some way in Observation (for cases like we've mentioned)
maybe that can be as a pre-coordinated LOINC (or other) code - and if not that, then as an extension?
Grahame Grieve (Mar 07 2016 at 08:54):
:-(
Eric Haas (Mar 07 2016 at 09:24):
The best solution is to creeate the appropriate loinc to include the relevent body positions - meanwhile use our own codes. no sense in having some LOINCs and some ans boolean elements
Eric Haas (Mar 07 2016 at 09:25):
and do i create a new stream for a new topic?
Grahame Grieve (Mar 07 2016 at 09:26):
don't create a new stream, but you can create a new topic just by saying what it is
Grahame Grieve (Mar 07 2016 at 09:26):
should we ask Dan V about this?
Eric Haas (Mar 07 2016 at 09:27):
don't ask tell I'd make a task but I', off to paris for a week and trying to finish up some argo stuff before I go.
Grahame Grieve (Mar 07 2016 at 09:28):
k
Josh Mandel (Mar 07 2016 at 14:36):
LOINC for body length is 8306-3 (vs 8302-2 for height)
Josh Mandel (Mar 07 2016 at 14:38):
I absolutely believe that a top level BP observation is a "section header": it creates a group structure between its parts. See https://github.com/argonautproject/implementation-program/wiki/Implementation-Sprint-7 for explanation / example and let me know if you disagree.
Rob Hausam (Mar 15 2016 at 03:02):
Josh, LOINC code 8306-3 is actually "Body height --lying" - there isn't actually a code with a component of "body length"
Andrew Ross (Mar 15 2016 at 20:59):
The Observation resource doesn't accept valueBoolean according to the docs because of the following:
The Boolean data type is rarely used for Observation.value[x] because most observations result values are never truly Boolean due to exceptional values such as "unknown". If needed, use valueCodeableConcept for a Boolean concept instead, and select codes from HL7 v2 Table 0136. These "yes/no" concepts can be mapped to the display name "true/false" or other mutually exclusive terms that may be needed.
However, I'm working with a system where the observation values really are booleans, and so we're stuck with hardcoding mappings between "Y" and true for particular code systems. Why not support Observation#valueBoolean for ease of implementation even if the use-case is rare?
Eric Haas (Mar 17 2016 at 01:23):
Create a tracker on Gforge with your use cases and we will reconsider.
Andrew Ross (Mar 17 2016 at 15:32):
ok, will do.
Last updated: Apr 12 2022 at 19:14 UTC
