Table of Contents
Most people assume their medical record is a description of their health. A running account of what is wrong, what was treated, how it went.
It is closer to a claim. Almost every line in it exists partly to justify a payment, and that shapes what gets written, how it gets worded, and what quietly never gets updated.
This is not a conspiracy. It is just what happens when the same document has two jobs and only one of them comes with a deadline.
How the money shapes the page
In the US, insurers covering Medicare patients are paid more for covering sicker people. That design is deliberate and sensible, because otherwise insurers would quietly avoid anyone expensive.
How sick you are gets established by the diagnosis codes in your notes. More serious codes, larger payment. Which means there has always been a reason to make sure conditions are recorded, and never much of a reason to check whether they are still true.
An entire software industry grew up around the first half of that. Tools that read charts and find conditions that were missed. Almost nothing was built to find conditions that should no longer be there, because nobody was paying for that.
What that produces
A federal audit published in 2026 looked at 97 patient records where an insurer had billed for an acute stroke. All 97 failed review.
In 68 of them, the patient really had had a stroke. Documented, genuine, in the history. The billing described it as currently happening. An old stroke and an active one are worth different amounts, so the difference mattered to the payment even though nothing about the patient had been invented.
The regulator had actually published its detection method two and a half years earlier, including the database queries, so anyone could run the same check on themselves. The failure rate was still total.
The version of this you can actually see
You will not spot a coding decision in your own records. You can spot the symptom.
- Something from years ago described in the present tense, with nothing about how it has changed since.
- A problem list that has only ever grown across every visit you can remember.
- Vague entries. “Stable” on its own is a word that means very little to the next person reading it.
That last one is worth pushing on. “Diabetes, stable” and “diabetes, A1C 8.2, adjusting medication, review in three months” describe the same appointment. The second is a record. The first is a placeholder.
What to ask, and when
Not at the start of the appointment, when you are describing why you came. At the end, when the note is being written.
One question does most of the work: is that still current, or is it history? It takes five seconds to answer and it targets the exact distinction the chart keeps failing to make.
If you have moved practices, changed insurers, or had care split across hospitals, ask it about the older entries specifically. Those are the ones that have been carried forward the longest without anyone looking at them.
On the other side of it, the tooling is starting to change. Newer risk adjustment coding software checks whether a condition on file is still supported by the documentation rather than only hunting for conditions that are missing, which means it can flag something for removal instead of only ever adding. That is a narrow technical difference with a real effect on what ends up in your file. It is not widespread yet, which is why the one question above is still worth asking.