Semantic Web Health Care and Life Sciences Interest Group Teleconference

04 Aug 2015

See also: IRC log


EricP, David_Booth, Lloyd_McKenzie, Paul_Knapp, Rob_Hausam, Tony_Mallia
David Booth and Paul Knapp
pknapp and dbooth


<trackbot> Date: 04 August 2015

Approve Minutes of previous meeting

july 28: http://wiki.hl7.org/index.php?title=ITS_RDF_Concall_Minutes_20150728

APPROVED: Minutes of July 28

computable semantic links from FHIR to RIM

david: first call last week

paul: No discussion on ARB last wk.

xsd:decimal precision

lloyd: no desire to change FHIR schema to say xsd:decimal . Grahame will add a warning to implementers to retain precision.
... What to do with RDF? Reasoners will only use declared types. want to be able to do math -- multiply, divide, numeric comparisons. But need to not lose precision.

<Zakim> ericP, you wanted to discuss closer ties to query in RDF labnd

lloyd: Probably need to have a separate attribute for the precision.

eric: we want standard tooling to preserve the precision. Having two components would do a good job of that: SPARQL could query as xsd:decimal, and a separate query could get the precision.

lloyd: if you're going to xquery, you'll risk losing precision, but you probably won't care. Need to be cautious about precision when we're reading data and writing it back out.
... Need more protection in RDF than in other spaces, because you won't round trip through xquery, but you might through a triplestore.

<pknapp> Lloyd: precision concern in FHIR is primaryily in storeage, not so much in XQuery

lloyd: I'll reply to Grahame's email and suggest adding guidance to the schema, and note challenges around xpath.

paul: Confusing precision and scale?
... Number of digits does not convey precision.
... Separate field that conveys precision makes sense.

paul, please chair while i'm trying to reconnect...

<ericP> Lloyd: typically humans pick units which permit one to capture the precision in decimals

<ericP> ... e.g. instead of 1230000^^whatsits, 1.23^^Mwhatsists

<ericP> Tony: we could define scale instead of precision

<ericP> pknapp: if we preserve scale, and in the FHIR side uses scientific notation or units which preserves precision, everything hangs together

david: proposal to use xsd:decimal to representation, plus an attribute to indicate the number of fractional digits.

lloyd: Cleanest would be to define an extension.

eric: If we say systolic BP is 150.0, then the quantity has a value property. it would either need an extension or something else. easist to have an extension.

david: In RDF, what would be meant by an extension?

<pknapp_> Tony: name the element 'scale'



david: propose we call it fhir:decimal.fractionDigits

paul: to make this work, this should be fhir:Quantity.decimal.fractionDigits

lloyd: there are some places where the units are carried separately
... such as currency

eric: we can apply this to decimal, but it won't be needed for some uses of xsd:decimal.

lloyd: where we have decimal where there are not units, are places we don't care about the precision. e.g., a search returns a match quality as a decimal.

paul: if someone requires decimal precision, then they either need to document that need and use xsd:decimal, carry the units separately, or use Quantity.

david: propose we call it fhir:decimal.fractionDigits

AGREED: use a separate attribute and call it fhir:decimal.fractionDigits

<pknapp> dbooth: side by side comparisons

Side-by-side comparison



<pknapp> tony: Grahames RDF ontology is accessable

https: //hl7-fhir.github.io/fhir.ttl

<pknapp> tony: called FHIR.turtle

<ericP> s/AGREED: david: use a separate attribute and call it fhir:decimal.fractionDigits/AGREED: use a separate attribute and call it fhir:decimal.fractionDigits/

<pknapp> tony: propose getting rid of the schema-only special primatives (code-primatibve, string-primative, etc)

tony: line94 use of fhir:uri-primitive will be changed to xsd:URI

<pknapp> uri-primative -> uri

lloyd: we would need fhir:code-primitive

tony: we can do that as restrictions on the value
... that will be a big change

Valueset Definition


tony: two ways: codes are instances, and terms align with SNOMED etc types. But we're dealing with two models: a model and a metamodel. An instance in the model becomes a type in the metamodel.

<pknapp> David: I have to leave to chair another meeting.

lloyd: can't do that.
... Code systems are doing two things.
... You can't list them like on lines 627-640, because codings have additional characteristics.
... We can only talk about unnamed codings.


Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.133 (CVS log)
$Date: 2008-01-18 18:48:51 $