Stream: implementers
Topic: New to FHIR - how to capture data quality metrics / tests
Jessica Malenfant (Mar 23 2018 at 01:14):
Hi all - I'm pretty new to FHIR and have spent a lot of time recently looking for information on if and how we might capture data quality tests and results using FHIR (as both a transport mechanism and format to capture the tests and results (I think that'd be a FHIR resource?). Not specifically clinical quality measures, but more like frequency distribution of values in a field in an RDBMS (not from an EHR directly, but as part of projects interested in the secondary use of EHR data) - examples: distribution of Race values in a data set and data completeness.
I've seen a few resources related to medical device measures but wanted to see if anyone in the community had any thoughts or experience in this area? Thanks very much!
Lloyd McKenzie (Mar 23 2018 at 02:07):
It's not something I've heard of anyone doing, but I think the Observation resource should be a pretty good starting point.
Jessica Malenfant (Mar 23 2018 at 17:13):
Thanks - I hadn't considered the Observation resource and have been reading more on the Operation Measure - Evaluate Measure and the Resource Measure and wasn't sure how appropriate it would be to use resources that don't quite capture / define the data structure of the tests we'd use to describe a data source.
Jose Costa Teixeira (Mar 23 2018 at 18:01):
why not measureReport?
Jose Costa Teixeira (Mar 23 2018 at 18:02):
seems more adequate esp if you are doing data profiling / DQ measurement.
Jose Costa Teixeira (Mar 23 2018 at 18:04):
I've worked with Data Quality rules and measurements (completeness, accuracy, etc) but not in healthcare. But they are a technical measurement, so I would not think of Observation first. measure and measureReport seem to be a nicer fit.
Jessica Malenfant (Mar 23 2018 at 18:11):
Thanks @Jose Costa Teixeira thanks - I somehow missed the MeasureReport yet (again, I'm very new to FHIR), but at 1st glance you're right that it looks like it might be a better fit. So Measure to capture the test (e.g. completeness test) and MeasureReport to capture the result ( e.g. # of missing values for a specific field in a table).
Bryn Rhodes (Mar 23 2018 at 18:14):
+1, I was going to suggest that as well, especially with the addition of subject, this would be a good fit for a structure measure.
Jose Costa Teixeira (Mar 23 2018 at 18:16):
yes, Measure for the test and MeasureReport for the result would be my choice. Curious to see how this ends up looking like..
Jessica Malenfant (Mar 23 2018 at 19:00):
Thanks! @Bryn Rhodes - when you say "addition of subject" can you clarify that a bit more? would that be part of the https://www.hl7.org/fhir/measure-definitions.html in the measure.identifier.type? Appreciate your help on this!
Bryn Rhodes (Mar 23 2018 at 19:03):
In STU3, the Measure resource was focused on quality measures that had patients as subjects, but we've expanded that in STU4 with the addition of the subject elements to Measure and MeasureReport to support measures that focus on other areas. For example, a structure measure might be "number of beds", and the subject in that case would be the "location". For the use case you're describing, I would think the subject would be "device".
Jessica Malenfant (Mar 23 2018 at 19:39):
@Bryn Rhodes thanks that makes sense and should be useful. Can you point me to the STU4 version? I can't seem to find it.
Bryn Rhodes (Mar 23 2018 at 20:54):
@Jessica Malenfant , current build is here: http://build.fhir.org, Measure and MeasureReport
Last updated: Apr 12 2022 at 19:14 UTC
