A Billing Code Pays the Claim. A Diagnosis Treats the Patient.

I have seen Charcot-Marie-Tooth disease (CMT) exactly once in my career. The patient arrived with complaints that pointed nowhere on their own: frequent falls, trouble writing, muscle cramps, a weakening grip, a little atrophy. What tied them together was the pattern, along with a family history of the same. Put those findings side by side and they form a phenotype that points toward CMT, a hereditary neuropathy, and from there toward the genetic screening that names the specific variant.

That case matters more now than it would have a few years ago. For most of my career, a CMT diagnosis changed little, because the only widely used treatment was over-the-counter pain relief. Today, targeted therapies are in development, and a patient reaches one only through a clinical diagnosis specific enough to identify the drug likely to work. With a vague label, the patient never becomes a candidate for that therapy, and no one realizes what was missed.

That distance, between a code specific enough to pay a claim and a diagnosis specific enough to treat a patient, is the clinical context gap. That gap sits inside most of the healthcare AI now arriving in the exam room, and as these tools push toward precision medicine, it turns from a nuisance into a patient safety problem.

While health tech is focused on billing codes, they’re missing the clinical specificity necessary to drive treatment decisions. The system loses data fidelity when technology extracts a medical code and strips away the clinical context from a patient encounter. While the code is important to bill for the encounter, essential clinical meaning is lost for downstream uses like interoperability, medication and therapy selection, and care coordination.

What Is the Clinical Context Gap?

The clinical context gap is the difference between a billing code chosen to process a claim and the structured clinical diagnosis a physician needs to treat a patient. A code classifies an encounter for reimbursement, while clinical context captures the clinical specificity necessary to guide treatment of an individual patient – whether for a rare disease like Charcot-Marie-Tooth, or something more routinely encountered.

Without a clinically specific diagnosis, you miss the relevant findings, phenotype, and history that guide the treatment decision. If your health technology only captures the billing code and misses the specifics of the disease, how is the next provider supposed to analyze the chart and continue where you left off?

Coding systems such as the International Classification of Diseases, Tenth Revision (ICD-10) were developed to classify encounters so that a claim can be paid. Trouble starts when we ask a billing category to stand in for clinical reality

CMT, for example, has a closely related hereditary neuropathy that presents almost identically but carries different genetics and treatment. However, its ICD-10 code reads only “hereditary motor and sensory neuropathy,” a label that will never help the physician differentiate between the two. The billing code “G60.0 hereditary motor and sensory neuropathy” not only fails to accurately describe the condition, it also fails to capture 61 other, more specific variants, with clinical phenotypes, testing and treatment protocols that would alter a clinician’s workflow. This is exactly why systems need to capture the clinical specificity using structured terminologies that capture the rich nuance, while retaining the correct coding for billing and reporting purposes.

What Ambient AI Misses

This problem could expand exponentially due to ambient AI documenting care. In much of today’s technology, the billing code comes first and the diagnosis is reverse-engineered to fit it. Clinically, that runs backward. The clinical picture should come first, and the code should follow from it. These tools infer a diagnosis from what was said aloud, yet they rarely prompt me for the finding that would refine or correct it.

A lot of clinical meaning never reaches a structured field at all. A 2025 study in the Journal of Medical Internet Research that analyzed 1.8 million Dutch primary care records found that only 13% of the clinical concepts in free-text notes had a structured counterpart, and a 2022 study in the Journal of the American Medical Informatics Association found 59.4% of chronic conditions captured consistently across encounter diagnoses.

A system built on a structured knowledge foundation works the other way, surfacing the findings that separate one diagnosis from a near neighbor. Each clinical diagnosis is identified as a unique clinical data entity, rather than a subsidiary of a generalized condition, enabling the precision needed to treat the patient’s specific clinical presentation.

Closing the Gap With Clinical Knowledge

Closing the clinical context gap means giving AI a clinical knowledge foundation that reflects how physicians reason. When a clinician documents toward a diagnosis, that foundation surfaces the distinguishing findings, prompts for what is missing, and records the result as structured, computable data. Structured, specific data is also the precondition for precision medicine, since a broad category cannot be tied to a genetic variant or a targeted pathway.

This is where an evidence-based clinical intelligence platform earns its place. Quippe®, built on the MEDCIN clinical data foundation and the Quippe Clinical Knowledge Graph, maps more than 430,000 clinical concepts and 100 million diagnostic concept relevancy links, laying out the relevant findings at the point of care and pairing them with a problem-oriented summary through Clinical Lens.

None of this is a case against ambient AI. The pressure to adopt it is real, and so is the promise of time saved. Saving me time, though, is different from helping me and my colleagues deliver better care, and the second part is far, far more important.

AI tools should only earn the trust clinicians are already extending them when they help capture the full clinical picture, rather than just a faster path to a billing code.