Convert your FHIR JSON -> XML and back here. The CDA Book is sometimes listed for Kindle here and it is also SHIPPING from Amazon! See here for Errata.

Thursday, February 18, 2010

Subsets and Value Sets

The HITFACA Blog reports:
On February 23, 2010, the Vocabulary Task Force established by the Clinical Operations Workgroup of the Health IT Standards Committee will hold a public hearing on “Vocabulary Subsets and Value Sets” as facilitators of meaningful use of electronic health records (EHRs).

And then provides a list of questions which I've responded to below:
  1. Who should determine subsets and/or value sets that are needed?
    It depends.  Subsets or value sets are needed for implementation guides and for much broader use cases such as Laboratory ordering and Results.  Consensus standards organizations should be responsible for determining which subsets or value sets are needed for their implementation guides.  Broader use cases may be driven by various initiatives at regional or national levels.  An organization responsible for harmonization of standards similar to ANSI/HITSP should also have a role in identifying value sets.
  2. Who should produce subsets and/or value sets?
    Consensus based Standards bodies should produce and MAINTAIN them.  Production seems easy, but a value set or subset that has no maintenance process has no life.  
  3. Who should review and approve subsets and/or value sets?
    It depends upon what they are used for.  Primarily the concensus groups of the producer organizations, but in some cases, such as value sets used for quality measures, review and approval could also include organizations like NCQA. 
  4. How should subsets and/or value sets be described, i.e., what is the minimum set of metadata needed?
    See HITSP TN903: Data Architecture Technical Note
  5. In what format(s) and via what mechanisms should subsets and/or value sets be distributed?
    Value sets should be available in a standard format, such as the Rich Release format used by NLM for RxNORM and UMLS.
  6. How and how frequently should subsets and/or value sets be updated, and how should updates be coordinated?
    It depends on their use.  Updates for fairly static value sets should be reviewed at least every five years (ANSI rules uses this figure for reaffirmation of Standards).  Value sets for clinical use should be reviewed and updated at least annually.  Some value sets and subsets may need to be updated quarterly, montly or even weekly (e.g., medications).  Updates may be delivered as a subset containing only the changes in more frequently updated value sets.
  7. What support services would promote and facilitate their use?
    Value sets should be available from a Web Service, such as that described in the HITSP T66 Retrieve Value Set Transaction.
  8. What best practices/lessons learned have you learned, or what problems have you learned to avoid, regarding vocabulary subset and value set creation, maintenance, dissemination, and support services?
    Building a value set requires a commitment to ongoing maintenance of it.  Dissemination should support both manual download automated retrieval and update.  Support services require that there be a feedback mechanism (such as an e-mail list service) to comment on it.  Public input is absolutely necessary in the creation and maintenance of a value set.  Quick response may be needed for clinical value sets to address issues like H1N1 or new medications or treatment options.
  9. Do you have other advice or comments on convenience subsets and/or value sets and their relationship to meaningful use?
    Isn't this enough...
  10. What must the federal government do or not do with regard to the above, and/or what role should the federal government play?
    The Federal Government should have a role in the coordination of value set deployment activities.  Presently the CDC, NLM and AHRQ (USHIK) all have some role in the development or deployment of value sets, which includes overlapping distribution, delivery and maintenance responsibilities.  Duplication of these efforts is not useful.  It would be better if there was a single coordinated effort, which could include participation from all of these bodies.
    NLM has appropriate infrastructures for manual download, licensing and deployment.  CDC has appropriate infrastructures for some development of public health oriented value sets.  USHIK has appropriate infrastructures for delivery of knowledge about value sets (e.g., metadata).  To my knowledge, none of these provide for automated computer update of value sets using simple web services such as those described in the HITSP T66 Retrieve Value Set Transaction, but I believe CDC is closest to having that capability.