Medical Coding’s Intent is Sometimes Lost in Translation

I'm guessing that you've never heard of Semiotics. I hadn't until two years ago when I was reading a 1907 book about clinical microscopy of sputum which mentioned the “semiotic significance” of certain observable phenomena. Semiotics dates back to about 1875 when Charles Sanders Peirce coined the term to refer to the “formal doctrine of signs.”

To reduce this to plain language, it is the theory that words have meaning only when we associate a word (by sound or spelling) with a thought (concept) and a thing. Meaning exists only within the context of the triad. There is nothing that has meaning in the abstract (in the sense of a Platonic ideal) because different people and different cultures may form the association between the thing, the thought and the word differently or not at all. Cognitive scientists like George Lakoff have explored this same territory from the standpoint of how different cultures create categories of things that they perceive as being related in some way, sometimes with unexpected results, as the title of his book “Women, Fire and Dangerous Things” suggests.

Whew! What does this have to do with medical coding? It means that codes, any codes, do not intrinsically mean anything. Because the number of possible thoughts is infinite and the number of codes is small, no code can even come close to precisely standing for any specific thought. The developers of the original International Classification of Disease (ICD) understood this. They had a specific objective in mind and it was not to assign a specific code to every possible medical diagnosis or procedure. Their stated goal was to define a relatively small number of groups (categories) to which related disease processes could be assigned so that they could perform statistically valid comparisons between the incidences of disease from one country to another. They were concerned that if that introduced too much specificity into the coding scheme by creating a large number of categories that the individual categories might contain too few cases to allow statistically significant comparisons. The Systematized Nomenclature for Medicine (SNOMED) similarly started out as a way for pathologists to describe cancers in a way that would be suitable for staging. Since then these schemes have gone crazy. ICD-10 has mushroomed to over 150,000 codes from its original 15,000 and SNOMED has grown from 4 to 11 axes and now has hundreds of thousands of codes.

In spite of the proliferation of codes, it is still not possible to reliably reduce a patient's problems and conditions to a couple of codes. Why not? Because codes alone don't mean anything. They only designate separate categories, not precise singular things. There are not enough of them to find one that means exactly what you might want to say about your patient — so you are forced to approximate. All categorization involves approximation, as anyone that has tried to assign keywords to the articles in their reprint file knows.

The problem with approximation is that once done, there is no way to distinguish a precise code assignment from an approximate one. You must either trust them all and be wrong in an unknown percentage of cases, or trust none of them, which is probably the safer approach. The danger is always that once a code has been assigned, others will assume that it means all sorts of things that have no relationship to reality and then act accordingly — often to the detriment of the physician or the patient.

Next week, I will consider what might be a hidden agenda in the impending ICD-10 codes.

Find out more about Dan Essin and our other Practice Notes bloggers.