All Things Data

The nitty gritty on value sets? They’re more than alphabet soup

In my previous post, I introduced the concepts of measure logic, value sets and attribution. Today we’re going to do a deeper dive into value sets. We’re going to look at how three different specifications (UDS, HEDIS, and Meaningful Use) define value sets, and we’ll assess the pros and cons of each. We’ll close with an overview of some of the current challenges related to value sets.

Of the several hundred measures that Azara offers, we’ve probably spent the most time on UDS, HEDIS and Meaningful Use. Those of you in community health know UDS well. For the uninitiated, UDS is a reporting requirement for federally qualified health centers (FQHCs) that includes just over a dozen clinical quality measures (CQMs). The UDS manual is a narrative form specification that expresses both the measure logic and value sets. The approach makes the spec easier to read, but it is a somewhat unstructured way to organize value sets. It also does not provide complete value sets for all clinical concepts, especially medications. However, in 2015 the UDS manual approved use of the Value Set Authority Center (VSAC) to define value sets (page 88 of the 2015 UDS manual). See below for more about the VSAC.

HEDIS is an annual reporting specification released by NCQA and is primarily used by health plans. However, NCQA also offers a Technical Specification for Physician Measurement that can be used by non-health plans, which Azara uses as the spec for our HEDIS measures. HEDIS specifications have been published yearly since 1991, which may be why they are well documented and contain clear, structured value sets. The HEDIS specifications are broken into narrative Word documents that define the measure logic, and into a (read-only) Excel spreadsheet that provides value sets.

While HEDIS specifications implement more structure than UDS specifications, the Meaningful Use eCQMs implement the most structure, both in the measure logic and the value sets. The Meaningful Use eCQMs take advantage of the National Library of Medicine’s (NLM) VSAC to organize and publish value sets. It is an extremely robust system that is worthy of explanation in a separate blog post. VSAC usage causes the value set definitions to appear somewhat disconnected from the measure logic, and understanding the way everything ties together has a bit of a learning curve. However, the VSAC is a great resource for implementing Meaningful Use eCQMs and for facilitating independent research into clinical value sets.

Each specification handles value sets differently, and many don’t consider them at all. When we come across a measure requirement that doesn’t include any value sets, we often refer to the VSAC. As we’ve mentioned before, a good spec reduces ambiguity. Well documented, discrete, value sets make for a good spec, but, as always, there is a trade off between robustness and ease of use.

One of the biggest challenges associated with value sets is “value set harmonization.” Spend some time with the VSAC and you’ll notice many clinical concepts have multiple value sets with different codes. Outside the VSAC, the codes that UDS considers for diabetes may not align with what HEDIS considers for diabetes. This is the reality of decentralized value set definition: if multiple organizations are tasked with defining the codes for a given clinical concept, the value sets will vary. That said, much effort has been put into matching these value sets, and into highlighting differences where they exist.

Another value sets issue relates to timing. Work flows that record data change over time, codes are added and removed from terminologies regularly, and value sets are constantly reevaluated and modified. This poses a challenge to reporting systems like DRVS. What if ICD-9-CM codes are removed from a diabetes value set because of the upgrade to ICD-10-CM? Should DRVS continue to consider historical ICD-9-CM codes, or should we expect that source system data is upgraded to reflect the latest version of the terminology? (Note, this scenario hasn’t actually happened; we are still looking at ICD-9-CM codes). If we need to check historical ICD-9-CM codes, then must all newly created diagnosis value sets include both ICD-9-CM and ICD-10-CM codes? Maintaining a full historical list of all valid diagnosis codes is definitely a burden for measure authors.

Value set definition is moving in the right direction, and there are some great tools available for reporting systems. We at Azara are excited to see the progress being made in this area of CQM definition, and we look forward to continued improvement.

Here are some helpful resources:

Eric Gunther is an engineer at Azara Healthcare. 

Copyright 2018 © Azara Healthcare