Why does one number matter?

Consider one patient: 50 years old, male, Black, with a serum creatinine of 1.5 mg/dL.

That creatinine is a measurement. It does not change. What changes is the equation the lab runs it through.

Under CKD-EPI 2009, which includes a coefficient for Black race, this patient’s estimated GFR is 62.0. Under CKD-EPI 2021, which does not, the same blood gives 56.4.

One patient, both equations. The gap is small; the boundary it crosses is not.

The difference is 5.7 units. On its own that is unremarkable — well inside the noise you would accept from a different assay or a different day.

But the number does not stay a number. It lands in a table of stage boundaries that somebody drew, and this patient lands on different sides of one of them: stage G2 under the 2009 equation, stage G3a under the 2021 one.

Nothing about the patient changed. Their stage did.

That is the whole shape of what follows. A change this small is invisible at the point it happens — it shows up as a slightly different float in a slightly different column. It becomes visible only downstream, at the places where somebody drew a line and attached a decision to it. The rest of this workshop follows that number to those places.

Every figure on this page is computed

Nothing above is typed by hand. Each value is read from findings.json, emitted by 01_act1.ipynb when it runs against the real cohort — so the page cannot drift from the notebook without the build failing.