Tuesday, April 3, 2012

Value Set Registry Requirements

The Clinical Quality Workgroup "Essential Components Tiger Team" had its first meeting today.  The tiger team is focused on identifying essential components of high quality clinical quality measures, including but not limited to discussions of:
  • value sets, 
  • standard terminologies, and 
  • the technical & custodial requirements for 
    • creation, 
    • sharing and 
    • maintenance of these components.
On the "technical and custodial requirements", HL7 did extensive work in 2009 developed a set of business requirements addressing similar issues for a template registry instead of a value set registry. I just took a look at that document again, and did a mental search and replace using "Value Set" where ever the term   "Template" appeared.  It's actually an excellent source of requirements for a Value Set registry. 

One of the sources for the Template Registry Business requirements project was the HITSP TN903 Data Architecture Technical Note.  That technical note addressed key requirements for data elements, value sets and templates, including identifying key metadata used to access and manage these objects.  There's quite a bit of overlap in that metadata across each of the components.

I'm expecting that the "Essential Components Tiger Team" will examine these documents before their next meeting (next week).

This was the critical slides in our discussion today:

The critical components of this diagram are:

  • Discover and Reuse 
  • Public Domain
  • Governance and QA
Based on other slides in the deck (that we didn't get a chance to go over in the meeting), there are some critical problems with value set quality, and there are plenty of issues with reuse.  The idea that there would be a publicly accessible value set repository is a good one.  PHIN VADS is already one such registry supported by the CDC.  The value sets in VADS are available for anyone to use, which brings me to the next bullet: Public Domain.  The model for measures today is such that they may be licensed materials, and so the issue of reuse could be problematic NOT from a technology standpoint, but from one of policy.  That's not within the scope of this  HIT standards workgroup, but would have to be addressed by the HIT Policy committee or one of its workgroups.

The issue of Governance and QA is also important.  The OpenEHR community has a pretty well executed process for governance that could be adopted with respect to open review and feedback processes on value sets (as well as some really great tools in that space).  Building a national repository of value sets is a critical first step in being able to establish governance over value sets.  Most of the other difficult issues are again related to policy.

In terms of standards: I've already recommended using the IHE Sharing Value Sets profile for accessing Value Sets for Query Health, and I'd also recommend it as a format for exchange of known value sets for use in quality measures. SVS is an implementation of just a very small part of the HL7 Common Terminology Services 2.0 DSTU.  That standard is appropriate to describe the higher level functionality needed to enable Discovery and Reuse of value sets, as well as some of the necessary management capabilities.  Some features (e.g., the annotation capability found in the registry business requirements document) are yet missing from CTS 2.

Across these collections of existing standards, there's a great deal to be examined and implemented.   It's not so much that we need the standards, as that we need them implemented in a national infrastructure.

  -- Keith

1 comment:

  1. Keith,

    I've been working with the British Medical Journal clinical content team to develop standardized, evidence based definitions of various disease states-- the denominator in patient registries for the large n chronic conditions is our first target. We are implementing these definitions in the national health system of the Cayman Islands, but I'm also implementing them in my other life in the US as an "analytics guy". Would you be interested in collaborating or introducing me to the ONC Tiger Team? I'd rather not duplicate efforts if ONC is also working on similar initiatives, but I can't tell for sure from my limited insight to ONC if the overlap is significant. Dana Womack suggested I contact you. You can find my full profile on LinkedIn.

    Best,
    Dale

    ReplyDelete