Discussion flowing from Redwood MedNet on Friday. Rumor that ICD-11 is either going to be replaced by a version of SNOMED or going to be incorporated into a future version of SNOMED. Possible impllication: the US should wait for ICD-11 because coding will be so much easier then.
I don’t buy the implication at all, for two reasons. We can’t continue to limp along on ICD-9 until ICD-11 is fully cooked and goes through the processes that enabled ICD-10, such as adoption in other countries.
My other concern is the fundamental difference between a classification system and a concept enumeration. Classification systems are designed to always be able to express an answer (perhaps better, perhaps worse.) The question might be “what did this patient die of, what was the regional distribution of various kinds of lung cancer” or “given the problems and course of treatment, under which of the clauses in our contract or regulation do you want to be paid?”
In order to ensure that the question can always be answered, classification systems include “not otherwise specified” (NOS) categories to catch the loose ends. When too many diverse loose ends are lumped together in the same NOS category, the answers to the questions become less discriminating and conclusions (or payments) derived from them more random or subject to “optimization” by plausible choices among loosely defined categories.
For keepers of systems of concepts “not otherwise specified” is anathema. As soon as you add a concept that was previously included in NOS, the definition of NOS has changed and you can’t compare data across encoded at different times. They prefer not to add concepts until the medical version of the Council of Elders has convened to determine where the new concept fits in the ontology and how it maps to other concepts. This is very important for research and advanced applications such as clinical decision support based on coded data.
What might the rumor mean?
One interpretation of the rumor might be “just send the data (in SNOMED concepts) and let the receiver decide on the classification. This is like when I ask my wife “what color do you want the bathroom” and she says “well powder blue goes better with the tiles but aqua goes better with the towels.” I got a bunch of data but woe be onto me if I were to use it to take responsibility for picking the paint. This shifting of roles is unlikely to be acceptable any time soon.
Another interpretation might be that IHTSDO would map any given version of SNOMED to a specific version of ICD-11 and distribute ICD-11 along with SNOMED. Careful attention to the exact version of ICD-11 that was in use could permit using the SNOMED ontologies to manipulate data expressed in ICD-11. That would be a valuable addition to medical informatics and has some benefits in terms of funding translation of both code sets into various languages.
I think the latter interpretation sounds feasible. It still wouldn’t justify holding on to that collection of big, vague buckets called ICD-9 for another 3-5 years.