FHIR Chat · icd10am to snomed map

it's not about having or wanting (or even needing tho :-) it's "wrong" Erich because ICD is a statistical classification (not a terminology and not an ontology, and it is mono hierarchical) and contains Other and Unspecified codes/categories... in fact if you look at the AIHW frequency data, you will find the top 30 most commonly/frequently used ICD-10-AM codes in any annual inpatient data collection (>7M cases) Of those, roughly (year on year) 10-12 codes are of the form "Other disease of the blah, Blah unspecified" etc... SNOMED does not "do" unspecifieds, or Others....maps have a direction for a reason - from the most granular/specific to the lumpier....(SCT clinical finding = 100K, ICD = 16K) I'm not sure what you want to achieve (or what your current task or interest is - but I don't think you can get there from here (if ICD is all you have))

Erich Schulz (Jun 28 2016 at 00:49):

year sorry I'm fully aware of the differences :-)

Erich Schulz (Jun 28 2016 at 00:50):

right now we have no coded problem list at all accessible to clinicians in my institution

Erich Schulz (Jun 28 2016 at 00:50):

absolutely nada

Erich Schulz (Jun 28 2016 at 00:51):

but we have a team of coders generating masses of ICD codes...

Erich Schulz (Jun 28 2016 at 00:52):

you maybe right there maybe zero extractable value from the ICD codes... but it seems worth a look

Erich Schulz (Jun 28 2016 at 00:53):

do you have a link to that "top 30 coded list" btw?

Erich Schulz (Jun 28 2016 at 00:58):

by way of background, a few colleagues and I are on a bit of an exploration to see if we can make better use of the data we have - what we do with the data will depend on what we find, audit, research and clinical care are all potential uses and this is merely one of many streams we're looking at... and we're looking to do specific data collection to fill in gaps - some of the gap filling will be triggered by what we have already.

Erich Schulz (Jun 28 2016 at 01:00):

and I have a particular research interest in "shareable knowledge artifacts" in the form of CQL

Donna Truran (Jun 28 2016 at 01:00):

table 4.12 (top 20 sorry, I mispoke, not the top 30) in AIHW Australian Hospital Statistics (pick a year, any year) http://www.aihw.gov.au/publication-detail/?id=60129550483

Erich Schulz (Jun 28 2016 at 01:00):

but without coded data its impossible to do any CDS

Michael Lawley (Jun 28 2016 at 01:07):

I would be very worried about doing CDS based on the ICD10AM codes coming from the coders because it is not *clinical* - it's also a bit late in the workflow - the codes get added after the patient is discharged and are against the episode of care.
Or do you have a different workflow going on and these are codes that are syntactically ICD10 but do not cary the proper meaning?

Donna Truran (Jun 28 2016 at 01:07):

yeah I know :-) which is why (I suspect) no-one does a lot of CDS....quite a few really good attempts, but AFAIK nothing that's been certified or officially sanctioned (by regualrtory authorities)... Happy to be corrected on this tho'.... there is quite a bit of 'performacne indicator' monitoring, and measures of compliance with care guidelines.. and I think that's where I've encountered the most effort being dedicated.... I suspect we lack enough coded data that serves as ground truth - and enough coded data for either training datasets or test datasets of enough volume and variety - to be sure of precision/recall, sensitivity/specificity