Coding for meaningful data and the limitations of “correct coding”

April 10th, 2017 / By Rhonda Butler

When coders say “coding is an art,” they don’t mean art—they don’t imagine themselves creating little medical record haiku. Describing coding as an art is a way of saying that coding is a messy and often difficult business. It’s messy because it involves understanding and applying an enormous number of things that generically can be called “coding rules.”

Coding rules are expressed in different forms and maintained by different organizations. There are coding guidelines for diagnosis and procedure codes, maintained by the Cooperating Parties. Then there are diagnosis code instructional notes, developed by the CDC, and procedure code definitions developed by CMS. And finally there is Coding Clinic advice, published by the American Hospital Association.

These things collectively define what is called “correct coding.” The rules that comprise “correct coding” multiply as exceptions are found to the existing rules, and then exceptions to the exceptions—rinse and repeat. Clarifications of existing rules decrease confusion in some areas but unintentionally cause confusion in others. Any new area of confusion needs further clarification—“in this situation, do x, in another situation do y.” And the result is a set of rules that becomes fractally detailed over the years, with the initial, broad rules covering most situations, and subsequent rules covering the less common situations, until it seems we will write rules to infinity, chasing down that “last” ambiguity.

I am not fond of the term “correct coding,” with its implication that there is only one acceptably coded solution for every medical record. It is well documented that several coders may code the same medical record differently and all of them are considered “correct.” There is even the “same record, different day” phenomenon, where the same person coding the same record multiple times get a slightly different but still “correct” list of codes each time. Rather than using the term “correct coding” to describe their goal, the coding community could adopt language that acknowledges both the purpose and the limitations of coding. For instance, we could speak of coding in terms of whether or not it is “defensible” rather than whether or not it is “correct.”

Would it make a difference if coders, coding auditors, and educators worked to create, evaluate and teach “defensible coding” rather than “correct coding”? I think it would. Words matter, because they subtly influence the thinking of the people using them. Using the right word instead of its second cousin can help us clarify our attitudes and focus more effectively on the goal—meaningful coded data, wherever possible.

I added the phrase “wherever possible” because it acknowledges that a classification system has limits. By design, a diagnosis or procedure classification makes a limited number of distinctions, and then it stops. The “leftovers” are classified in non-specific codes that use words like “other” and “unspecified” in the diagnosis code title, and non-specific root operations like Repair for procedures. In a world where “defensible coding for meaningful data” is the stated goal, we could focus our time and energy on creating good data where good data is an achievable goal. As it is now, too much time is spent debating the merits of non-specific code A vs. non-specific code B, when both codes are defensible and the data in that area will have limited utility regardless. As a community, if we are clear enough in our own minds to say, “In this situation the data is going to be clear as mud regardless,” then perhaps we will be confident enough to say, “Either A or B is fine, we don’t need one more rule that tells us arbitrarily to code A instead of B.”

I know better than most what a huge change in habits I am advocating. It means that if, after following their usual process for finding the most accurate code available, coders find only non-specific code options, they don’t immediately write to Coding Clinic. They use their best judgment instead. And it means that if, after following their usual process for reviewing a coded record, a coding auditor finds that the coder chose code A and the coding auditor would have chosen code B, and both are defensible, nothing happens—the coding auditor moves on to the next record.

In reporting on the selection process for Supreme Court justices, recent news articles have discussed the extent to which the Constitution is open to interpretation. This question should be asked of coding rules as well as the Bill of Rights. To me, coding rules should contribute to meaningful coded data. If they don’t, they are adding complexity to the coding process without substantially improving the results. A fractally detailed set of rules that doesn’t give us meaningful coded data is an unnecessary burden on an already burdened industry.

Rhonda Butler is a clinical research manager with 3M Health Information Systems.