Thursday, April 28, 2011

Certification Retrospective Part 2 – Method to our Madness?

Note: this blog post will continue discussing my experiences in CCHIT. The opinions are mine, and do not necessarily reflect the views of CCHIT or Siemens.

In Part 1, I talked about CCHIT giving me the privilege to work with great people over the years. But we didn’t volunteer for CCHIT just to meet people, but to do important work to advance the HIT industry. So what did we accomplish, and what were we thinking as we evolved?

One thing I learned is that no group can please all the people all the time. Sometimes, people wondered why we proposed some criteria and not others, or proposed the timeframes that we did. So we tried to explain these “whys” in a white paper Interoperability, Supplying the Building Blocks for a Patient-centered EHR in mid-2009 which still exists on the website of writer John Morrissey. It described the context for what we were doing and the “methodical march” to increasing levels of information interoperability that we proposed. I don’t think this was well understood by the public, who might have thought some of what we did was arbitrary. But here’s what was in our mind (well, I can only speak for myself – mine at least).

Key principles:
  • We aren’t starting with a blank sheet of paper. The existing state of adoption (immature as it might be) must be considered, so we started each year with an “environmental scan” on the state of the HIT industry.
  • Adequate time must be provided for EHR products to make the transition to a more desirable state. We aimed for 18 months lead time from our first “signal to the market” (new draft criteria) until certification testing of those criteria.

Given these principles, our methodical march followed the path below.
  • Data Capture precedes interoperability. You can’t share what you don’t have! When we started CCHIT, not all EHRs even captured all the data that other systems would want, so functionality criteria (ambulatory, inpatient, ED) needed to capture the data elements that could then be shared. Thus in the first certification year, 2006, there were no interoperability criteria required (though several criteria, such as Lab Results and e-prescribing, were roadmapped for later years).
  • Information Liquidity (a term we didn’t use, but which implicit in our criteria) is beneficial even prior to semantic interoperability. Thus, the first clinical summaries that CCHIT certified (Continuity of Care Documents, CCDs) in 2008 required human-readable sections for meds, allergies, problems, etc., but not structured data. We were later criticized for this, as if we didn’t care about structured data, but this was always intended as a stepping stone to later stages (as explained in the white paper). Even today (3 years later than the 2008 criteria), many voices espouse liquidity as a more important and realistic for the near-term than semantic interoperability (“don’t let the perfect be the enemy of the good”).  
  • Standardized Content (formats and vocabulary). Human-readability is fine for humans, but doesn’t help most EHR systems to “understand” (semantics) and process the information they receive from others. Because of the existing Tower of Babel syndrome most EHRs have been “confounded” in their ability to share information with other EHRs, and to “consume” (not just display) it. So CCHIT supported the goals of ONC and HITSP to progress toward standardization of vocabularies. But it realized that changing an EHR’s vocabulary overnight wasn’t realistic, and that mapping from one vocabulary to another is not simple or foolproof and must be done carefully for patient safety and other reasons. Thus a “glide path” from narrative to standard vocabularies was roadmapped.
  • Discrete Data Import (Consumption). When liquidity and standardized content are established, the EHRs that receive or retrieve clinical data can “do something” with the data, more than just file and display it. As early as 2007 criteria, we published certification criteria for ambulatory EHR discrete import of Laboratory Results that referenced the HITSP-endorsed HL7 2.5.1 Lab Results Reporting Implementation Guide, including a common subset of LOINC observation codes. Thus, many of us were disappointed when the ONC Final Rule for certification, published in 2010, lacked a specific HL7 standard and LOINC requirement, not going as far as what CCHIT had been certifying since 2007. There are 127 CCHIT Certified® ambulatory EHRs that already meet this requirement. For CCDs, the challenge of discrete data import is  greater because of the wider variety of data, and multiple applicable vocabularies. Rather than declaring the problem too big to tackle in one shot, we proposed chipping away a few sections at a time, starting in 2010 with discrete import of medications (using RxNorm as specified by HITSP), allergies (using RxNorm and UNII), Immunizations (using CVX) and demographics, then moving on to problems (using SNOMED-CT). These criteria are a matter of public record, and can be found on the CCHIT website: 2008 criteria. See 2008 Criterion IO-AM 11.06 for example, which originally roadmapped discrete data import for 2009, though the date changed to 2010 in the 2009 criteria (published but then withdrawn)***. You can decide based on the facts whether we were too aggressive, or too laggardly, or about right.

So what’s the point of this retrospective? I’ll sum up lessons learned in the next post. I’ll also speak more about how CCHIT’s interoperability criteria compared to ONC’s criteria.

I’ll also discuss some of the agonizing controversies that we faced, particularly in 2008 and 2009 over issues like standardized transport, discrete data import, data reconciliation turf (is it interoperability or functionality or both?) and the feasibility of requiring SNOMED-CT. We wrestled with each of these, and did our best. We might not have succeeded perfectly in all of them, and our criteria development work was superseded by ONC before we could resolve them all, but we learned much in the methodical march!


*** (Long Footnote)
The 2009 Ambulatory Certification criteria, along with the CCHIT 2009 certification program, were not launched, since they were in limbo awaiting decisions on ONC-ATCB certification. But eventually most of these criteria (without the roadmap) were relabeled and launched as the 2011 CCHIT Certified® ambulatory criteria. If the 2009 program had proceeded, Criterion IO-AM 11.06 would have required the following in 2010, adding much specificity beyond what had been proposed in the same criterion in 2008:
The system shall provide the ability to display CCD documents, using a subset of the HITSP C32 specification for the Medication and Immunization History module, file them as intact documents in the EHR, and import the discrete data from one or more of the entries in a structured form into the patient record. The system shall provide the ability to import as discrete data the following data elements from the C32 Medications Module: Free text SIG, free text product name, free text brand name, coded product name, coded brand name, product concentration, status of medication, quantity ordered, order date/time, order expiration date/time, prescription number, quantity dispensed, and fill number. (Note: fill number means the number of fills dispensed, not the number of fills prescribed). The “required if known” (R2) data elements will not always be known, and therefore will not be present in every instance of a C32 document, but if present, the system shall be able to import them

More was explained in the Comments column:

Medication and Immunization History-Coded Subset includes the following content modules of the HITSP C32: Person Information, Healthcare Provider, Medications-Prescriptions and Non-Prescription, Information Source, Comments
The intent for 10 Certification is not to require specific capability to use the imported data but only to show the system is capable of processing discrete data elements from a CCD. In future certification years, more functionality criteria to “use” the data may be added. Fulfillment of the 10 Certification criteria may be demonstrated in a variety of ways, including but not limited to any one of the following:
- Reconciliation functionality that allows CCD data to be compared with existing data, and selectively imported into a list
- A display of data captured from external sources, including the imported discrete data. The physical storage location (database table, file, etc.) is at the vendor’s discretion and does not need to be the same location as the active allergies.
- Creation of an CCD that includes the elements previously imported as discrete data
Vendors may innovate and differentiate their products by how well such data are used, but 10 Certification will not limit vendors to any particular approach


  1. Thanks for sharing, I will bookmark and be back again

    CISSP Certification

  2. Steffi, thanks so much for reading my blog and for your two kind comments. Best wishes as you work on your certification and education projects.