In case you had to guess, the answer is “no.” Recently, DrBicuspid.com ran a story about the EZCode system and SNODENT titled “Diagnostic dental codes: Are we there yet?” Unfortunately, the answer to that question is also “no.”
The emergence of not one, but two, dental diagnostic vocabularies is troubling. First we have essentially none, then suddenly two. This reminds me of a recent quote by Doug Fridsma, the Director of the Office of Standards and Interoperability in the Office of the National Coordinator for Health IT. At an AMIA 2012 Annual Symposium panel discussion, he remarked: “Standards are like toothbrushes. Everyone has one, but nobody wants to use someone else’s.” I appreciated the dental analogy, but duplicate standards are no laughing matter. Dentistry is currently wasting a huge opportunity to create a novel, forward-looking approach to representing dental diagnoses. Unfortunately, we seem to be stuck somewhere between the Stone Age and the 19th century.
To understand why, we need to look back a few years. The ADA started working on the first incarnation of SNODENT, its diagnostic vocabulary, sometime in the 90s. The project took a while and SNODENT was supposed to be released in January 2000 (see “SNODENT to provide inclusive means of transmitting dental information,” ADA News, 30:9; 5/3/1999 [unfortunately, not available online]). When the ADA News asked then ADA President Dr. Tim Rose: “Will you be using SNODENT in your office?” he confidently answered: “I sure will.” Fast-forward to 2012: Yes, we are still waiting for SNODENT, now going into its second incarnation. (Actually,can you be reincarnated when you have never been born? Sounds like a Zen koan.)
In part, the EZCodes diagnostic vocabulary, a project of Harvard University School of Dental Medicine‘s Dr. Elsbeth Kalenderian, emerged as a reaction to this “Waiting for Godot” scenario. Normally, few people would care about competing dental diagnostic vocabularies, were it not for two important factors. One is the HITECH Act. Sometime in the future, the Department of Health and Human Services will anoint one dental diagnostic vocabulary as “the” standard for interoperability of diagnostic information in dentistry. The second is that the ADA is realizing quite a bit of non-dues revenue from licensing the CDT. (According to a recent conversation with an ADA staffer, a CDT license is about $11/year per customer of an electronic practice management system and $1,000/year per institutional site license. With over 90% of all dentists using a computer in their office, you can do the math.) So, it stands to reason that licensing a diagnostic vocabulary similarly might generate another nice chunk of cash for the ADA every year. As a result, Harvard and the ADA have been at war over their respective diagnostic vocabularies for quite some time. (Of course, if you only have the slightest inkling about the organizational psychology of both entities, you know this had to end up in a mudwrestling match. But, that is another story.)
So, what about the comparative merits of SNODENT and EZCodes? The DrBiscuspid article provides some basic information: SNODENT has about 7,000 terms, EZCodes about 1,300. EZCodes was developed by a working group of the Consortium for Oral Health Research and Informatics, mainly by merging several existing dental diagnostic vocabularies. SNODENT was developed through a somewhat opaque process that, to my knowledge, was never really published. Both vocabularies are currently free for researchers after signing a licensing agreement.
So, how well do these codes work? According to the DrBicuspid article, the EZCodes are currently being piloted in 17 dental schools and institutions located in the U.S. and Europe. There is a 2011 paper on the evaluation of the Z Codes, a major component of the EZCodes. SNODENT is rumored to be evaluated in a few dental schools, but a search for “SNODENT” in PubMed today only turned up the same three papers that have been there since 2006. The only other reference to a comparison between the two vocabularies alleges that “the ADA sees EZCodes as an ‘interface terminology’ useful for capturing health problems but not as a replacement for SNODENT in storing information in EHRs.”
So, what should we make of all of this? I fear that neither effort at developing a dental diagnostic vocabulary will produce a very satisfying result in the long term, unless some radical changes are made. Even worse, the tug-of-war and duplicate work consumes precious resources that dentistry, as a profession, can ill afford to waste. Here are a few relevant observations:
- The world is, in general, moving away from top-down, bureaucratic approaches to developing and implementing standards. Why? Because they don’t work. The healthcare landscape is littered with ailing, moribund or just plain dead standardization efforts that consumed a lot of time and energy, and are essentially not used in practice. The more promising approaches are smaller, nimbler and less bureaucratic, and engage the communities who care about and use the product from the very beginning.
- Dentistry has a very successful and broadly used coding system, the Current Dental Terminology (CDT). The CDT has about 710 codes. Clinicians know most of the ones they use frequently by heart. Clearly, knowing codes by heart gets harder the larger a code set is – difficult with 1,300 codes and fairly impossible with 7,000. However, that is not an unsolvable problem. The entry interface for the codes in the electronic patient record simply must be smart enough to make choosing the right code easy for the user. This is a significant, but solvable, human-computer interaction design challenge.
- Speaking of design: One adage in the design community is “fail early, fail often.” Bringing something as complex as a new diagnostic vocabulary online rarely works with a big-bang approach. It makes much more sense to focus on smaller pieces of the puzzle, get the bugs out, and then move on to developing the next bigger increment. (One clue for this is hidden in the Z Codes evaluation paper cited above: Over a period of one year, UCSF used only 93 [63%] of 147 Z codes.) Unfortunately, developing a dental diagnostic vocabulary in an incremental, iterative approach would require a level of collaboration, shared vision and coordination between dentists, informaticians, vocabulary and terminology specialists, and the dental IT industry, that is unlikely to materialize.
- Coming back to the statement above that “the ADA sees EZCodes as an ‘interface terminology’ useful for capturing health problems but not as a replacement for SNODENT in storing information in EHRs,” we need to clear up a misunderstanding. Separating work on various aspects of a vocabulary makes no sense. As Kent Spackman states in an authoritative paper on terminologies, ideally, interface and other terminologies should be derived from a common reference terminology because this “may allow different terminological efforts to focus on separate parts of the problem and to cooperate in solving the overall problem.” Given what we are witnessing, wishful thinking indeed!
- Unfortunately, both vocabulary development efforts decided to stick with outdated models of representing classifications and terminologies. Over the long term, those approaches will be about as efficient and effective as the horse and buggy are for transportation today. The formalism for representing “things” for the foreseeable future are “ontologies,” which even JADA discovered in a 2010 editorial. For a number of reasons, ontologies are way more powerful for representing diagnoses, treatments and other concepts in healthcare than traditional approaches. (One thing they do very elegantly is to combine the terminology, information and inferencing models described by A. Rector in The interface between information, terminology, and inference models.) So, at this point, ontologies are the way to go in architecting vocabularies. The good thing is that you can largely reuse the work spent on creating vocabularies when you build ontologies, so not all past effort is wasted.
- Developing and maintaining large vocabularies requires a lot of time and money. Very few organizations have the wherewithal to support this process. Here, again, we can take a lesson from the ontology world. Many ontologies are developed in a completely open process by their community of users. While development still must be organized and regulated, costs and effort are spread over a much larger number of individuals, groups and institutions. This has two benefits: (1) everyone who needs the ontology uses it and (2) no one has to ask how much it costs. One example: the Gene Ontology, one of the most successful ontologies ever created.
So, what will the future hold for dental diagnostic vocabularies? Given the current path, most likely mediocrity, tension, conflict, widespread dissatisfaction and little benefit. Not a pretty picture.
– Titus Schleyer, DMD, PhD
Assoc. Professor and Director, Center for Dental Informatics