OHDSI Home | Forums | Wiki | Github

Creating HEDIS measures as cohort definitions in ATLAS

(Evan Minty) #21

Thanks for directing me towards this thread Gowtham.

I agree with this workgroup wholeheartedly. As a benefit for the wider project, I think that having an existing library of quality measures would provide considerable incentive for an organization to adopt the OMOP CDM for their data assets.

There’s a number of measures out there for consideration, and part of the fun would be compiling candidates for implementation. See for example, NICE (https://www.nice.org.uk/standards-and-indicators/quality-standards-topic-library)

I don’t think I can make the Georgia F2F, but I’m happy to help make this happen.


(Gowtham Rao) #22


Good evening - would it be possible to create a workgroup that is interested in using Atlas to calculate Quality measures?

(Siqin Ye) #23

Hi all,

just joining this conversation-- I work with Karthik from Columbia where I also over see a lot of our reporting, and we are in the process of building an OMOP instance one of the goals for which is to address quality reporting needs.

In general (having handled a lot of ACO reporting to CMS), the issue ends to be that many measures require specific documentation that is not standardized concepts. For the HEDIS BMI measure for instance, the eligibility and BMI are readily computable but the follow-up for abnormal BMI is difficult to compute (CMS is also recommending against using the previously recommended GPRO codes for these measures now a days). Measures like HTN control, DM control, or cervical cancer screening are readily computable though (since the data elements are already structured and well-defined in OMOP; though outside data is still an issue for many cancer screening tests).

Anyway if there is interest in a working group I would be very interested to join-- I have a strong interest in this work both from ops and research perspective.

Nice to meet everyone!


(Evan Minty) #24

Hi all -

Wondering if any of you are heading to the OHDSI symposium. That might be a good venue to coordinate a face to face meeting and possibly strike up a working group in this area. I’m happy to help try to coordinate this. Let me know either on this thread or via PM if you’re interested.


(Gowtham Rao) #25

Yes - @Evan_Minty happy to connect at the symposium.

Regarding Work Group – I think that is a really good idea and we have a WG http://www.ohdsi.org/web/wiki/doku.php?id=projects:workgroups:quality-measures but we have not met yet. The reason is we tried to build a measure at the OHDSI face to face spring 2017 in Atlanta - and found the exercise very difficult.

What we learnt is – that we had two options. Build measures to the detail specification of the measure stewards manual, OR to build measures the OHDSI way. E.g. the HEDIS measure specification manual would be 1,000s of pages, with some measure specification going for 100s of pages - this is difficult, mostly because of the detail. Alternative is to “approximate” it using the OHDSI, by leveraging the OMOP CDM standard vocabularies for example with the hierarchy’s. When we went thru the exercise we concluded that it is possible to build the measures using OMOP CDM, but the level of details makes it difficult and this task becomes not a scientific task, but a technical task that a voluntary community is not interested in pursuing. The general consensus was that it maybe best to do commission this as a paid service.

There were also some additional considerations. e.g. some of the measure specifications required rules like checking for a procedure in the same visit as a procedure. The OHDSI tools like Atlas did not support these rules at that time (but now do! https://github.com/OHDSI/Atlas/issues/454) .

We need a lot of new functionality like above before beginning a workgroup that implements complex specifications e.g. “Gaps in care”. i.e. an eligible Diabetic person who has a gap in HbA1c testing, then we need two cohort (denominator and numerator – and difference between them). HEDIS measure specifications will require the ability to union, intersect or subtract cohorts from each other.

(Evan Minty) #26

Interesting, thanks.
It strikes me that for ‘reporting’ , where specificity is paramount, we would struggle based on your comments.

But the development of measures that can give insight into care gaps, in a defacto standard data model, would still be of considerable value to any organization considering adoption. The main constraint being expressed here are limitations in ATLAS, as the requirements you’re mentioning could certainly be achieved with a series DB queries (I guess that’s the source of your concern with voluntary labour supply - this isn’t my strength either, but would be a great exercise to push one’s knowledge of the CDM).

I’ll fire some PM’s to the folks on this thread to see if we can find a way to sit down in a couple of weeks.

(Rajeev Dixit) #27

We are interested in mapping HEDIS measures to Atlas concept_sets. Is there any further progress on this? We are ready to contribute if others are interested.