Thirteen years of Medicare rate history, read correctly every time
Ask Localis what CPT code 99213 paid under Medicare in CY2015, and the number that comes back needs to match what CMS actually published that quarter, not a modern assumption stretched backward. Billing teams pull rates like this for appeals, contract audits, and old claims review, and a wrong historical rate is worse than no rate at all.
That's a harder promise than it sounds. CMS ships the Physician Fee Schedule as spreadsheets, not as a versioned API, and it doesn't treat the layout of those files as a stable contract. Across the thirteen years Localis has ingested this data, CMS has renamed columns, reordered them, and abbreviated the same field differently from one release to the next. The contractor column is a good example: it ran as "Carrier" through 2017, spent several years afterward as "Medicare Administrative Contractor," and now carries a parenthetical "(MAC)" — the same column, the same meaning, three different labels depending on which year's file you open.
None of that drift is a defect in CMS's data. It's what happens when a federal agency updates a spreadsheet template every so often without asking the people building tools on top of it. But it means a lookup that assumes this year's layout also describes an older file will eventually read the wrong column and compute a wrong answer without raising any error at all — silently, which is the worst way to be wrong.
Localis reads every CMS column by what it's labeled, checked against everything we know CMS has called that field across all thirteen years of releases, rather than trusting where the column happens to sit in a given file. A label CMS retired years ago is still recognized in the file that used it; a label CMS only started using recently doesn't need to exist for an older lookup to work. The same care applies to what a column's values mean, not just what it's called — a field can keep an identical name from one release to the next while the way its number is encoded quietly changes, and treating "found the column" as "understood the column" is exactly how a plausible-looking wrong number ends up published.
The part that matters most to a reader: when a column genuinely can't be found, or a file's format doesn't match anything we understand, the ingest stops and nothing publishes. We'd rather fail loudly on a CMS format we haven't seen before than ship a guessed rate that looks right and isn't. That's why a 2015 lookup and a current lookup carry the same confidence — both return either a real, sourced number or nothing.
Try it yourself: look up 99213's rate history across every release we carry, or use search to find any code.