Thanks for the reply.
I suppose I failed to distinguish between EHRs, the record systems, and EHRs, the standardized XML representations of the record systems. While standards such
as CCR and CDA are certainly well structured and defined (as best as can be done in medicine), the systems that actually host and process electronic health record information are not so (necessarily). Additionally, while a system may be certified for Continuity
of Care Document support (by implementing either CCR or CDA) or Meaningful Use compliance, the system does not necessarily have to make the use of these features easy, nor provide a comprehensive implementation.
>That standard CCR spec is enforced through product testing as per the CCHITT (Sp?) testing documents – every EHR product must be certified through testing of the
XML against the spec, and must follow it exactly. There is a suite of certification tests that is produced (I suppose by HHS) and which each and every EHR system must pass. Without certification, it can’t be used by the docs – uncle won’t pay them to buy
it and use it. So ONLY certified EHR products that produce spec’d EHR readable XML data are available to health providers.
I located an old Meaningful Use Stage 1 specification (May 2011) for Syndromic Surveillance specifies that the HL7 2.x message that must be produced to pass
the test is valid if the required fields (specified in the HL7 standard) are populated. If the same requirements were imposed on CDA certification, basic information like lab result values would not need to be provided. HL7 2.x, the standard used for message
transmission between electronic health systems in Meaningful Use certification, is another well-structured specification that creates massive headaches in implementation due to the number of optional fields. Having been through the certification process for
Meaningful Use Stage 1, and part of Stage 2, it’s been my experience that the requirements for certification are quite lax.
>The CCR specs document rigorously defines the XML forms and symbols. It’s a 100 page dense pdf, and just browsing quickly over it, you will find XML used in a
Lisp-like way to structure object-attribute-value paths for finding a “logical” diagnosis. The constraints on the provider/user force answers to questions in a sequential way, guided in part by previous answers from higher up the tree.
Rich, are you sure that this is the CCR standard? It seems quite out of scope for that specification, and the competing standard (CDA), which has been overtaking CCR, provides
nothing as complex as clinical decision support. I only retain copies of the CDA standard and am not inclined to pay $80 for a copy of the CCR standard. Both of these standards, as far as I am aware, are only concerned with structuring information for exchange
between systems. Perhaps I’m confused about what you’re discussing when you mention that the systems (presumably the CCR) places “constrains on the provider/user [to] force answers to questions in a sequential way”. It is worth noting that the CCR specification
has nothing to do with the internal structure of EMR/EHR systems, nor the naming of internal database columns, nor the data types and constraints upon those columns. It merely specifies an XML schema that must be complied to when producing electronic copies
of patient records.
>The original motivating reason behind all this EHR orthodoxy was to minimize health costs by standardizing all the paperwork, which as you know, takes a lot more
doctor, nurse and tech time than treating the patient. You can’t easily data mine a database where every provider puts in his own terminology – and medicine is rigidly categorized, defined and named by the long history of health care development. So compared
to most fields, I would expect it to be easier, not harder, than the usual vocabulary acquisition problem. Not so to date.
I agree that it would be easier to reuse standard vocabularies, though this is not necessarily a requirement in EHRs as of yet. Additionally, as I pointed
out earlier, implementation quality is inconsistent; while CDA or CCR may act as useful mediums for representation, the EMR/EHR systems that produce those records are far from standard in their implementations and perhaps even less useful for data mining.
This brings me back to my original point: EHRs (the systems, not the CDA/CCR standards) are hardly ontological. These systems are, for the most part, massive relational or graphical databases with business logic on top. The most widely used products don’t
make use of any sort of ontology modeling language, and while one might consider a highly normalized database to represent an ontology (to an extent), they can only truly be considered schemas. While some may use ontological subsystems, like SNOMED-CT, for
decision support and vocabulary selection, the actual patient record components of the EHR systems are most often merely schemas.
I agree that there is much room for the use of ontologies in the medical community, but I don’t think EMR/EHR systems are useful guides to industry utilization
of ontologies nor their effectiveness.
Booz | Allen | Hamilton
308 Sentinel Drive
Annapolis Junction, MD
Booz Allen: 301.419.5962