How do you handle “unable to determine” query responses?

By Brian Murphy

Unable to determine.

Love it? Hate it? Or do you understand why it must (usually) be offered as a query option?

The latest ACDIS-AHIMA guidance, Guidelines for Achieving a Compliant Query Practice, imply that “unable to determine” should typically (but crucially, not always) be offered to the physician in a traditional multiple-choice query. From the brief:

     “Unable to Determine” requires specific consideration to determine if needed as a multiple choice option.

     “Unable to determine” is defined as the provider being clinically unable to determine if a diagnosis or further clarity can be provided in the documentation. This terminology does not equate to an “unable to rule out” option and does not represent an uncertain diagnosis (e.g., possible, probable, unlikely).

     Unable to determine must always be offered for present on admission and yes/no queries.

Despite this extant guidance some CDI departments are not fans of “unable to determine.” It gives a provider an easy out, short-circuits critical thinking, and leads to an inaccurate depiction of risk of mortality and lesser reimbursement.

But sometimes physicians just will not know. Determining whether a patient is septic is not easy (as I always caveat, I am not a physician nor a coder).

The early symptoms of sepsis, such as fever, rapid heartbeat, and difficulty breathing, can be non-specific and resemble common infections. Sepsis symptoms can overlap with other medical conditions, making it difficult to distinguish sepsis from other infections or inflammatory disorders. There is also a lack of specific diagnostic tests: Currently, there is no single definitive test to diagnose sepsis. Diagnosis relies on a combination of clinical assessments, such as vital signs, laboratory tests, and medical history. And there is still debate on the efficacy of sepsis-2 vs. sepsis-3 criteria.

All of this can lead to delayed diagnosis or misdiagnosis. Which makes “unable to determine” not just an easy out, but an understandable provider response, especially if the CDI professional renders the query early before all the clinical evidence is in.

So it seems we will always need an option for uncertainty. At least, for now.

But I wonder: Will an AI be able and willing to report “unable to determine? I doubt it; I suspect it either will draw inferences and make deductions from a much wider and deeper data set, and then render a diagnosis that appears definitive.

Or, it may hallucinate and render a false answer. Because AI like chat GPT is designed to provider answers, not wait on a lab test or diagnostic interpretation that has not yet arrived. At least in its current form.

The FY 2023 ICD-10-CM Official Guidelines for Coding and Reporting define a provider as, “physician or any qualified healthcare practitioner who is legally accountable for establishing the patient’s diagnosis.”

What if the “practitioner” is a machine? Is the machine legally accountable? Its manufacturer?

I’d love to hear your thoughts, as well as how you handle the dreaded phrase “unable to determine.”

Reference

Guidelines for Achieving a Compliant Query Practice—2022 Update: https://acdis.org/resources/guidelines-achieving-compliant-query-practice%E2%80%942022-update