Stream: cql
Topic: How to use QUICK on Synthea data?
Clark C. Evans (Aug 07 2020 at 20:25):
I realize much of this work is still in development; but I'm curious how I could use the FHIR3 example data and measures (such as CMS124v7) in the January connectathon using the CQL Execution engine. I was able to convert the age.cql example to age.coffee; but I'm not sure how to go about using the pre-converted ELM or how to convert the bundles into coffee-styled elm.
Clark C. Evans (Aug 09 2020 at 15:17):
What would be great is if there were more detailed instructions on how to use one of the eCQMs with the cql-execution engine; in particular, how to use them against data created by Synthea.
Bryn Rhodes (Aug 11 2020 at 07:54):
Have you looked at the connectathon quick start? Those specifically illustrate how to load Synthea data sets into test servers and then evaluate measures using that data.
Clark C. Evans (Aug 11 2020 at 14:06):
@Bryn Rhodes Super, thanks (I'm not sure why I didn't see that). Does this have a command line interface or any regression tests?
Bryn Rhodes (Aug 11 2020 at 15:39):
The regression tests are the connectathon packages, we are working on getting that into an automated suite.
Clark C. Evans (Aug 11 2020 at 16:28):
@Bryn Rhodes So, I've updated both connectathon and cqf-ruler to the latest, and I'd note that the Wiki references an out-of-date blob for measure-EXM130-8.0.000; the one in the master repository is measure-EXM130-7.3.000. I'd also note that the R4 measure version, EXM130-7.2.000 fails with "Could not resolve membership operator" and "Could not resolve identifier timezone".
Other R4.01 measures also seem to fail when performed against their sample patient, measure-EXM124-9.0.000 errors out with Unable to resolve library reference: http://fhir.org/guides/dbcg/connectathon/Library/library-FHIRHelpers|4.0.1. Perhaps I've forgot another bundle I should be uploading?
It'd be great to have a way to run though these measure tests via command line.
Clark C. Evans (Aug 11 2020 at 17:35):
@Bryn Rhodes So, I was trying to run a very quick benchmark with a fairly old i7-4770 CPU @ 3.40GHz. I started the webserver after loading the EXM130 bundle, and then refreshed http://localhost:8080/cqf-ruler-r4/fhir/Measure/measure-EXM130-7.3.000/$evaluate-measure?patient=denom-EXM130&periodStart=2019-01-01&periodEnd=2019-12-31 twenty-five times for denom-EXM130 and twenty-five for numer-EXM130, discarding the first few warm up ones. What I got was an average computation time of 607.9ms. Is this in-line with your expectations, or is there something dramatically wrong with my setup?
Bryn Rhodes (Aug 11 2020 at 17:44):
We have a chunk of performance optimizations that we're working on getting in to the pipeline, I'll check where those are right now.
Bryn Rhodes (Aug 11 2020 at 17:44):
We're also actively working on connectathon prep, so there's a bit of flux in the connectathon repo right now.
Clark C. Evans (Aug 11 2020 at 17:45):
When is the next connectathon. I'm not an HL7 member though (currently unaffiliated).
Bryn Rhodes (Aug 11 2020 at 17:56):
The connectathon is September 9th through the 11th, and the price difference for members/non-members is very small. We would certainly welcome the participation.
Last updated: Apr 12 2022 at 19:14 UTC
