Medicare doesn’t negotiate what it pays for a lipid panel. It asks labs what private payers paid them, lines every price up from cheapest to dearest, weights each by how many tests were run at it, and takes the point where half the tests sit on each side. Here is that machine, drawn, with the real reports.
Primary source:CMS, “Preliminary CY 2027 Medicare Clinical Laboratory Fee Schedule Payment Rates,” September 21, 2026: “The weighted median is calculated by arraying the distribution of all private payor rates, weighted by the volume for each payor and each laboratory.” Data: the first PAMA round’s raw private-payer file (collected Jan–Jun 2016, reported 2017) and 2026 CLFS rates, via MIMI Labs. Prices grouped into log-spaced bins (~13% wide). Weighted medians recomputed by us from the raw rows.
weighted median (what the formula picks)median report (the typical contract)Medicare pays in 2026
Weighted median–
Median report–
Medicare 2026–
Top 3 price bins hold–
Top panel: share of all tests performed at each price. Bottom panel: share of price reports (one report = one lab’s rate with one payer). Same horizontal axis. Prices below $0.50 or above $1,000 are off the chart and counted in the note under it.
The 80/20 of the weighted median
Look at the two panels together. The reports spread wide — thousands of contracts at every price. The tests pile into a few narrow spikes. Those spikes are a handful of very large contracts — almost certainly the big national labs, though the file doesn’t say who. A few price points carry most of the volume, so they choose the median. That’s why the red line lands left of the grey one. Weighting by volume hands the price to whoever runs the most tests at the lowest rate.
Why this matters for 2027
The first round’s reporters were mostly independent and physician-office labs: of 1,942, just 21 were hospital labs. The 2026 round drew 6,411 labs, 875 of them hospital labs (CMS fact sheet). More reporters means more contracts in the bottom panel. Whether it moves the red line depends on where their volume sits, not how many of them there are. CMS’s answer is in: about 16% lower on average, cuts capped at 15% a year from 2027 to 2029.
If your tool triggers an order — a CDS nudge, an RPM panel, a screening outreach — the unit price under it is set by this machine. The companion explorer runs it across 942 tests.
Where this is thin
2016 prices. This is the only per-report file public so far. The 2027 rates come from 2025 data; CMS says it will release raw data for codes reported by 10+ entities for public comment.
No lab identifier. We can’t see which lab sits behind a spike, or apply CMS’s duplicate rule. Our median counts every row.
Obvious junk. The lipid-panel file has reports at five-figure prices and at a few cents. CMS now drops records above 100× the current rate; the first round didn’t publish that rule. The weighted median barely notices them, which is one thing it does well.
Binning. Bars are ~13%-wide price bins, so the chart shape is approximate. The three headline numbers are computed from the raw rows.