In Canada, the implementation of ICD-10 caused an approximate 50% drop in productivity in rates of coding in the health profession. Reports show that a year after implementation, productivity only returned to 80% of the original ICD-9 baseline. Now, more than a decade since Canada adopted the new coding system, they still have yet to reach original productivity levels.

This reduction in productivity is understandable. ICD-10 is vastly more complex than ICD-9, and people sometimes have difficulty adapting to change. It’s naturally going to cause fear in the industry. The 3-page bill proposed to the US Congress by Rep. Ted Poe (R-TX), HR 2126, would allay those fears for a while by delaying U.S. implementation of ICD-10, which has already occurred several times. As noted in the referenced article, however, neither committee has decided to hear the bill, so yes, it seems highly likely that ICD-10 implementation will occur in the United States on October 1, as (currently) scheduled.

But tell me: do we really need to know that someone was “bitten by a turtle” (W59.21XA) rather than “struck by a turtle” (W59.22XA)? And if the person was bitten on the toe by that turtle, do we really need to know whether it was the right big toe (S90.471A) or the left big toe (S90.472A)? What if the incident happened, for some reason, while the person was waterskiing (Y93.17)? Do we need to know that? It is definitely helpful to know something about laterality, which side was the injury was on, and about the person’s activity when the injury occurred. It is also nice to know how many people a turtle bit this year, as well as how many waterskiing accidents.

In another (less funny but more likely) example, someone driving a pickup strikes a lamppost while texting. The person’s head strikes the dashboard, causing a contusion (S00.83XA), while the truck’s airbag expands and strikes them on the right side of the chest, causing a contusion there (S20.211A). That it happened in a pickup matters (V57.5XXA), as does the fact that they were texting at the time (Y93.C2). This is all highly useful information, but does it have to be done as a pre coordinate highly complex classification system? The format of the ICD-10 codes, and for that matter the ICD-9 codes, is quite old fashioned in a post-coordinate world.

A taxonomy — and it would be a large one — would better provide the desired data and in a much more flexible form than the ICD-10. Just a single laterality rule for left and right would remove a goodly portion of the listed codes.  If it is the interest of the Centers for Medicare and Medicaid Services (CMS) to gather more data about health and health care in the USA, then why not apply some proven, inexpensive, easy-to-implement algorithms rather than cause a Y2K-style panic throughout the entire healthcare industry?

In example one above, we could code for “turtle,” “toe,” “bite,” and even water or waterskiing. If any of those items was missing in the source data, i.e., the electronic medical record (EMR), we could still do a good job of determining the cause of the bite and collating the data for later reporting and retrieval. It would be possible to mine the data more effectively than using the classification system, since the data would be disaggregated and available. For example, all kinds of water-related accidents could be retrieved, as could all turtle-related injuries, bite-related injuries, etc.

Going one step further, one could even link the data as RDF triples for a full semantic enrichment and ontology approach, which would surface all kinds of fascinating relationships. One could then visualize the data in various presentations for quick understanding of how much danger there is the general populace regarding turtles bites and waterskiing.

Access Integrity has built a system to review EMRs and instantly provide a suggested list of ICD-10 codes, just like they have done for the ICD-9. For this, ICD-10 is certainly a good move forward. We still maintain that a human should review and make the final selection of the codes submitted for billing, due to the complexity of the classification system, the ambiguities in the EMR as written by the healthcare provider, and of course the liabilities in a litigious world. For those of us in the information business, more data is always better than less, so even if ICD-10 is imperfect, and it is, there is no doubt that it’s a step in the right direction.

Marjorie M.K. Hlava
President, Access Innovations