ICD-10 vs. SNOMED-CT: Part 2

There’s been some great debate since I posted my prior article, Let’s Abandon ICD-10! (and use SNOMED-CT instead), so I felt compelled to write a follow-up.

I completely agree with those who say that the specificity of documentation is a key challenge, regardless of which terminology we use for diagnostic coding. At the same time, ICD-10 forces clinicians to spend time away from patients documenting ICD-10-specific details, many of which have low relevance to modern day needs.

“It’s not just about capturing more details, it’s about capturing the right details.”

Medicine has changed since 1992 (when ICD-10 was released), and the details we need today in 2013 are far better served through SNOMED-CT than ICD-10. It’s not just about capturing more details, it’s about capturing the right details.

It also makes no sense to force doctors to document Problem Lists in SNOMED-CT and then ask them to document those same problems in a diagnosis section as ICD-10! The government has chosen “gold standard terminologies” to enable interoperability by reducing the need to translate data. Simultaneously they are creating interoperability problems within systems by picking two different terminologies for the same data in two different parts of the medical record. We can do better than this.

Many have noted that ICD-10 is a much better terminology for researchers than ICD-9 – absolutely true! However, as a researcher, I (and virtually every other researcher I know) would rather have data in SNOMED-CT than ICD-anything. Yes, ICD-10 is better than ICD-9, but SNOMED-CT is better still.

The argument is commonly made that a key value of ICD-10 lies in its use as a classification terminology. However, SNOMED-CT is a rich ontology with a multi-level hierarchy. As such it is highly capable for aggregating and classifying concepts — and does so in a more clinically relevant way. SNOMED-CT is superior to ICD-10 for this purpose.

It has been suggested that EHRs should just auto-translate SNOMED-CT to ICD-10 for the doctor. If ICD-10 and SNOMED-CT were easily and automatically interchangeable, then payors, researchers, or anyone else who felt the need for ICD-10 could simply take SNOMED-CT and run it through the translator — no need to burden doctors, EHRs, or anyone else with this work. Unfortunately, that’s not the case. There is not always a straightforward translation from SNOMED-CT to ICD-10, and therein lies the rub. This duplicate work and all associated training will be put on the backs of doctors. The result will be less clinical time at the patient’s bedside and longer waits for patients to see the doctor for care. Any small incremental benefit ICD-10 might offer over SNOMED-CT in rare cases must be weighed against this negative patient impact. In my opinion, it’s not even a close call. Providers should send payors granular SNOMED-CT data. If payors see value in converting that SNOMED-CT data into ICD-9, ICD-10, or TLA-whatever, they should feel free to translate as they see fit.

Even the ICD folks recognize these problems with ICD-10, and they are basing ICD-11 on SNOMED-CT. With ICD-11 due for release less than a year after the government mandated conversion date to ICD-10, the massive cost and negative patient impact of ICD-10 is not justifiable.

Fortunately, I’m not alone in this thinking. The American College of Physicians and the Texas Medical Association have made nearly identical recommendations to those in my first post on this topic — either to wait for ICD-11 or to use compositional SNOMED-CT for diagnosis coding.