Is a computer-generated nudge a query? According to extant guidance, yes… but do we need new guidance?

By Brian Murphy

Is a nudge, a best practice alert, or some other machine prompted question to a physician, a query?

I’m risking a very contentious topic here. But it’s one that comes up a lot in conversation.

According to the latest ACDIS-AHIMA query brief, the answer is yes. A query is a query is a query, whether done by nurse, coder, MD, man or machine. Per the brief:

“All queries must meet the same compliant standards regardless of how or when they are generated, including those autogenerated by artificial intelligence (AI) and computer-assisted coding (CAC), whether in real-time computer-assisted physician documentation (CAPD) or after the episode of care is complete. Any technology-generated documentation query must follow the query compliance guidance discussed above.”

But not everyone agrees, and new tech doesn’t always conform.

This was a problem beginning to arise during the end of my time as ACDIS director in late 2021/early 2022. I was seeing a blurry line with computer assisted coding and NLP applications. Today artificial intelligence has exploded onto the scene, including solutions powered by generative AI that can engage providers in dialogue. Ambient AI that listens, and doesn’t just transcribe but summarizes and offers next steps.

Why can’t AI prompt a provider for hyponatremia when the sodium level warrants, and ignore the formal query process?** Should we cede the query space to AI? Here’s a few reasons for pause.

• Denials are being issued on the basis of leading queries.
• The liability for misdiagnosis falls on the provider, not the tool he or she uses.
• AI will always give you an answer and never say, “I don’t know.”
• AI still hallucinates answers and must be fact checked … but it keeps getting better.

The reason query guidance is needed in the first place is that coders aren’t allowed to assume. Harried physicians historically had a tendency to ask “just tell me what to write.” CMS punted queries to professional associations and ACDIS and AHIMA stepped in with practice briefs.

I’m proud of the work both have done in this space. And not just because I was part of it, but because it was and is a thoughtful, multidisciplinary effort. Does that mean perfect? No. But the line to perfect starts on the left, and look—no one is in it. CMS doesn’t seem to know what to do about AI, either—see the recent Medicare Advantage rule, where it punted the issue entirely.

What is fair to ask is: Has the time come for something new, for new tech? It’s a question worth asking.

I won’t pretend to have the answers nor would I have the temerity to offer any. This is for professional organizations and possibly CMS to decide.

Final thought: If the big tech companies want their product to avoid the query process and prompt, they should be held jointly liable for erroneous diagnoses. My opinion only… but with great power comes great responsibility.

Further commentary

**This is a complex issue and I recommend reading ACDIS/AHIMA compliant clinical documentation integrity technology standards paper below. It distinguishes between a “documentation alert” and a “clinical/critical alert,” and states that the latter is not subject to query compliance guidelines (and gives hyponatremia as an example). You would not for example expect a sepsis detection tool to offer up a multiple-choice query to a provider alerted to a critically ill patient.

References

• Guidelines for Achieving a Compliant Query Practice—2022 Update: https://acdis.org/resources/guidelines-achieving-compliant-query-practice%E2%80%942022-update
• ACDIS/AHIMA compliant clinical documentation integrity technology standards: https://acdis.org/resources/acdisahima-compliant-clinical-documentation-integrity-technology-standards