Stream: cql
Topic: FHIR Normalization & QUICK
Clark C. Evans (Jul 13 2020 at 19:04):
Thank you for pointing me towards this January's connectathon as I'm looking for examples of CQL running over FHIR. I noticed that in the CMS 124 example normalization functions, such as Normalize Interval were introduced into the measure proper. I'm surprised to see this here, as I expected choice normalization to be done upstream. I'm wondering if this is the direction that QUICK is taking?
Bryn Rhodes (Jul 15 2020 at 05:39):
It's hard to do choice normalization upstream. We have tooling now that will construct model info files that are informed by profiles, so to the extent that those profiles do choice normalization, that will be reflected in the resulting model info. For an example of this, see the USCore/QICore model info included in the latest translators. Specifically, the evaluator tests here: https://github.com/DBCG/cql-evaluator/blob/fix-13-regression-tests/cli/src/test/resources/qicore/TestQICore.cql
Bryn Rhodes (Jul 15 2020 at 05:40):
But until there is agreement within the supported profiles about the possible representations, it's shifting the burden of between the implementation and the author. It's a straightforward pattern to do it in authoring, so that's the approach that's being pursued most commonly.
Clark C. Evans (Jul 16 2020 at 01:32):
Bryn, I can see why it may seem attractive to base quality measures on FHIR; on the other hand, the flexibility of FHIR seems to work against this approach. In the end, the quality model will still exist... it'll just be expressed at a lower level, with perhaps differing interpretations. Since I'm just playing, I'll stick with QI-Core => QDM. However, I'm curious if there is prose that describes why translation QI-Core to QDM was deemed lacking.
Last updated: Apr 12 2022 at 19:14 UTC
