Tuesday, December 15, 2020

CMS Part B Data: Profiling Tumor-Genomic-Profiling Billing in CY2018

I noted earlier this month that CMS had (finally) released provider-specific CPT code utilization data for CY2018 - here.   

In this blog, I look at the billing patterns for comprehensive genomic profiling (CGP; codes 81445, 81450, 81455) in CY2018.   One middle-ranked biller is Tempus, which caught my attention, and I pulled Tempus's billing for CY2018 also.

A major reason that the CGP codes caught my attention is the recent formation of a coalition to improve access to testing, the Access to CGP group - find it here, https://accesstocgp.com/


I went to the 2018 Part B cloud database (here) and searched on the following codes:

  • 81445, 5-50 tumor genes.  About $600.
  • 81450, 5-50 tumor genes, hematopoietic cancers,  About $600.
  • 81455, 51+ tumor genes.  About $3000.
  • 0037U, Foundation Medicine F1 (FDA) Test.  About $3500.
Here's the Medicare utilization for 2018:

  • For 81445, there were 3,173 uses totaling $1.9M, and the highest biller was Interpace (Pittsburgh) with 627 cases.
  • For 81450, there were 4,889 uses, for $3.7M, and the highest biller was Genoptix with 2,949 uses for $2.2M.
  • For 81455, there were 7,476, but from only 8 labs, totalling $22M.  The highest biller was Caris in Phoenix with 6,872 uses for $20M.  
  • Finally, Foundation Medicine 0037U received $33M for 9,884 FMI F1 cases.
In 2019, where we have national but not provider level data, these codes were used as follows: 81445, 7,424 cases for $4.4M; 81450, 12,196 cases for $9.2M, and 81455, only 847 cases for $2.3M.  (That's just about what would happen if you pulled out Caris's 6872 cases in 2018).   And 0037U rises in 2019 to 22,914 cases for $80M.

Understanding the Changes from 2018 to 2019

81445 and 81450, the 5-50 gene codes, both grew considerably from 2018 to 2019 (they're the blue and orange bars below), while the FMI code 0037U towers above the other codes in 2019.    Adding all four codes studied, usage grew from 25,422 in 2018 to 44,381 in 2019.

Curiously, the standard AMA CPT high-complexity tumor code, 81455, actually fell from 2018 to 2019. in fact, it fell by over 80% (gray bar).  What gives?   I think the answer is that most MACs don't pay for 81455 in 2019, while MolDx does pay for it, but under unlisted code 81479.  Thus, the drop in 81455 (51+ tumor genes) from 7476 uses to 847 uses in CMS 2019 data reflects the "shunting" of this service to unlisted code 81479 in 2019, at MolDx, about the only MAC that was paying for 81455 in 2018.  See my parallel blog on 81479 here.

MolDx and 81455, 51+ Genes

Note that the MolDx program (which paid the 7,476 cases of 81455 to Caris in CY2018) normally does not use code 81455, for reasons I've never understood, and probably pays most genomic profiling venders under unlisted code 81479 plus Z codes.  

In 2019, 2020 public meetings, I've heard MolDx staff say they've now stopped paying 81455 in the few cases where they might have paid it in 2017 or 2018.  Their LCD for large tumor gene panels clearly instructs labs to use 81479 (molecular unlisted) not 81455 (tumor genes 51+) when you study 51+ tumor genes.

Tempus 2018

Tempus shows up in the list for billers of 81445, with 39 cases for $23,000.   This goes along with something I noticed earlier this year, MoPath payments in any NGS MAC states are very tiny, with the sole exception of Cologuard, which is paid in NGS MAC state Wisconsin, but via an NCD, not local coverage.  

Turning to billing by Tempus in CY2018, in this Part B database, they were paid $1.9M, of which 47% or $905K was paid for code 81211, BRCA testing.  They billed in total two dozen codes, many paying less in total than $25,000.   

Tempus raised an additional $200M in 4Q2020, bringing net funding to $1B (here).

Find the Data By Individual Lab for 2018

See my Excel spreadsheet in the cloud here.  

I've also put screen shots below (click to enlarge), but it's going to be easier to read the cloud spreadsheet.



Readers interested in the tumor genome consortium A-CGP cited above may also be interested in a New York-based effort to document the clinical utility of whole genome sequencing - January 2015 Genomeweb article here.