What happens when you comment on a Medicare coverage policy

Draft Local Coverage Determinations open for comment right now, and the public record of every comment period that has already closed — what was argued, and what the contractor wrote back. Every line links to its CMS source.

Generated 2026-08-31 · coverage database vintage 2026-08-27 · comment record vintage 2026-08-31

5 policies are accepting comment. After a window closes, 42 CFR 426.320 limits review to a Medicare beneficiary with a denied service and a treating physician’s documentation — not the manufacturer.

Contractors answer every comment — 3,747 of 3,768. But most responses do not announce a decision: of 708 distinct substantive responses, 192 state a change and 42 state a refusal, while 446 restate the policy or explain its basis.

A comment period is also rarely crowded. 99 of 339 drafts drew three comments or fewer.

These counts are of what contractors published. They are not a success rate: no comparison exists against drafts that were revised for reasons unrelated to any comment.

The 60 most recent of 339 on the record, newest first. Quotes are the contractor’s own words about what it did.

Every finding below is checkable against the CMS pages linked above. They are here because each one was missed by an obvious shortcut, and each changes what a commenter would write.

1. Identical length is not identical text

The three Genome-Wide Molecular Methodologies drafts carry indication sections of 5,123 characters each and analysis sections of 5,934 each. Comparing those counts says the drafts are the same. Comparing the text says otherwise.

Palmetto and Noridian are identical. CGS differs, and every difference is a cross-referenced policy number. A letter citing L38047 to CGS cites a policy CGS does not have.

2. The constraint is not in the draft you are reading

Those criteria bind by incorporation: the draft says non-NGS testing “must also fulfill criteria outlined in L39919”. So the rules that decide whether a test is covered live partly in three other policies — and those are different documents at each contractor.

Assessing one code against one draft means reading six documents per contractor. That is invisible to anyone reading a single page, and it is most of the work.

3. A two-sentence ask, granted four times

The Association for Molecular Pathology wrote to four contractors about solid organ allograft rejection. Among a short letter’s asks: use dd-cfDNA rather than cfDNA, because the general term is not specific to the donor-derived form.

One ask, four contractors, four changes — published on both sides. The cheapest asks are the ones that get granted, and almost nobody makes them.

Each limit below is a measurement, not a disclaimer. The number that establishes it is given so you can decide whether to believe the rest.

Counts above are of what contractors published. None is a causal claim: drafts are revised for reasons unrelated to any comment, and no comparison against that exists.