Wednesday, May 26, 2021

Very Brief Blog: CMS Publishes MAC Proposed Gapfill Rates for CY2021

CMS has some 50 codes in the CY2021 gapfill process.  These are codes, new in CY2020, that CMS could not price by its crosswalk method.  Therefore, CMS asks its MACs to submit their prices for the tests, now in CY2021.  CMS will take the median for setting its national fee schedule in 2022 and future years.

See the CMS lab website here:

And look for the file:  2021 CLFS Gapfill Preliminary Determinations (ZIP).

Regarding comments, CMS states,

  • Interim Medicare Administrative Contractor/Preliminary Gapfill payment recommendations have been posted below. We invite comments through July 26, 2021. Please submit comments to


Highest prices range up to $5475 and $5225 for 0212U and 0214U.   These are the Genomic Unity Whole Genome from Variantyx, and the Exome Plus test from Variantyx.   (For 2014U, the company asked for a code stack of 20+ codes over $10,000; MAC prices they got are pretty close to the off-the-shelf genome price for 81425.)

There are two codes priced in the $3000 range and 5 codes priced in the $2000 range.   

Several COVID tests are being priced by the gapfill method, including U0002 at $51 (regular COVID PCR) and U0003/U0004 at $75.   $75 is the CMS administrative price for high-throughput COVID, and the MACs are proposing just the same price.   The CMS administrative price would rule while during the public health emergency, and the gapfill price will replace the administrative price after the public health emergency.

MAC Variance

CMS picks the median price of the MACs, which for this purpose is set by states rather than MAC contractors.   All the MOLDX states always match, and they add up to more than half the states, so the median is set by whatever MOLDX picks.

That said, I did a very crude test by taking the 57 columns of the spreadsheet and calculating the SD and the proportional SD (SD/Median).   

  • For 25 codes, half the codes, the proportionate SD was less than 10%, so all the MACs basically agreed.   
  • For six codes, the SD divided by median (proportionate SD) was > 50% and up to 200%.[*] 
    • The far right of the graphic below. 
    • The largest variance was 0209U, PerkinElmer, cytogenomics of whole genome, NGS, which got pricing ranging from $787 to $5031 (whew, that's six-to-one.)  A fairly similar code, 0156U (NYGC whole genome cytogenomics), came in at a $1364 median but some MACs as low as $597.  

Errors Possible!

I discovered at least one code where the CMS median is clearly wrong.   At least in version 1, Code 0156U has median listed as $1364, but 80% of the states (prices) are clearly $597, so the median should be $597.   Odd.  This was the only line of the May 26 V1 with an incorrect median, and it was corrected on May 27 in a V2 spreadsheet.


Regulations and CMS policy require the MACs to submit a "rationale" but these are generally so bland as to be meaningless.  They appear on a separate tab.  In row after row, the price rationales are "price determined by 42 CFR 414 Section G" (the whole PAMA section in its entirety) or "research of tests and comparison to similar tests."  Some answers are more specific.  


Distinctive Code - AI Analysis of H&E Breast Cancer Slides

Code 0220U is for the PreciseDX test or test system, which images H&E breast cancer slides, adds AI analysis, and predicts which are very likely to have a low score on an Oncotype Dx test.  See press here, see company here.  It was priced very close to $706 by all MACs.

See earlier codes for cytology, screening by automated system, 88174/75, and 88361, IHC (including staining), computer-assisted.  


NGS MAC Departs from MOLDX Price, Twice

Comparing the columns for CT and KY, the NGS MAC matched the MOLDX MACs in all but two codes, the differences being,  0016M (a MAAA where NGS MAC is $700 high) and 0175U (a PGx where NGS MAC is $440 low.)  


There was no correlation between price (higher rank on the vertical axis) and the degree of MAC variance (farther right on horizontal access).  Arguably the higher-tier prices have less variance (clustering in upper left), but it could be chance (r = negative .3).

Yes, I know that if the SD is bigger than the median, the data is not gaussian, not parametric, and technically an SD is not valid.  A better non-parametric measure would be the high MAC price over the low MAC price ratio.  This goes from 1.0 to 6.5, but half the cases are less than about 1.1 or 1.2.