Why you should resist the temptation of ICD-10 translation tools

Carl Natale
by Carl Natale

Grab a photograph. Any photo. Then take it to a photocopier, and copy it. Call that copy an ICD-9 code.

Then take that copy (the one called "ICD-9 code") and copy it on the same photocopier. That copy of a copy is called an ICD-10 code.

How does it look?

Not great. Making a copy a copy is less than desirable. You lose a lot of information that way.

Which helps explain what happened when University of Illinois Chicago researchers studied what happened when they applied a web-based ICD-10 converter to a bunch of ICD-9 codes. They lost information and potential reimbursements.

Why we shouldn't be surprised

Coding experts have been saying from the start that conversion and mapping tools are a bad way to generate ICD-10 codes. The crosswalks have very few direct translations. For the most part, they give users many codes to choose. Someone still needs to refer to the documentation and use judgment to choose the right code.

[See also: ICD-10 Myths: What you need to know about GEMs]

Canadian hospitals found this out when they tried mapping tools to convert ICD-9 codes to ICD-10-CA codes. They found it easier and more complete to learn the ICD-10 codes.

And the UIC study demonstrates that nicely.

Except not everyone wants to see that. Some are using the study to argue we should not switch to the ICD-10 code sets when they should be arguing we should not use automatic conversion tools. (I told you so)