Tuesday, December 29, 2020

Very Brief Blog: Medicare's 2018 Spending for CYP (PGx) Codes; Falling 2018-2019

The entry point for this blog is that I noticed a drop in national Part B spending for the most frequent pharmacogenetics codes (81225, 81226) between 2018 and 2019.  2018 Part B payments were $28M (81225, $10M, 81226, $18M) while 2019 Part B payments were $20M (81225, $8M, 81226 $12M)

Data analysis shows the change moving into 2019 was driven mostly by changing payments in Georgia, after the MAC transferred from "Cahaba MAC" to "Palmetto MAC," which introduced MolDx Z-codes and edits.


In CY2018, Medicare spent $33M on CPT codes 81225-81231 (CYP genes), of which $28M went to the top two codes, 81225/81226 which are CYP2C19 and CYP2D6.

I looked at the state and MAC distribution of CMS payments for these two codes.

For both CYP 2C19, $10M, and CYP 2D6, $17M, about 55-60% of payments went to providers in Georgia and Tennessee:

If we look at MAC level payments, these were dominated by the Cahaba MAC (over 50%), followed by either the Novitas MACs or the MolDx system MACs:

Cahaba, with 7% of the US Part B, paid a disproportionate number of PGx claims in 2018, over half of US claims for either 81225 or 81226.  

Transitions 2018, 2019

Georgia had a large portion of 81225/26 CYP payments in CY2018, but the dollar volume fell from $12.9M in CY2018 to only $1.5M in CY2019.   This was related to the transition from Cahaba PGx edits to Palmetto MolDx PGx edits.

CY2018, left.   CY2019, right. (Georgia only)


Rather than focus on PGx...

If instead we used all non-proprietary genetic CPT codes (811xx, 812xx, 813xx, 814xx), there were $736M dollars paid in CY2018, of which $128M were paid by Cahaba (18%).   In CY2018, $104M/$735M or 14% of all US non-proprietary genetics were paid to labs in Georgia, which has 3% of the FFS Medicare population. 

Using 2018 CMS cloud data for state files, and 2019 CMS excel data for state files, and using codes 811xx-814xx, we see $104M in payments in Georgia in CY2018, and $11.5M in CY2019.


Palmetto (MolDx) took over GA/TN in 2018,  but based on comparison of payments 2017-2018-2019, it appears that MolDx edits weren't fully implemented into well into 2018 (here). I've listed the payments under Cahaba, the 2017/partial 2018 contractor. Palmetto would have introduced a highly restrictive PGx policy at some point before the end of CY2018.  Palmetto updated its PGx policy in its MolDx regions in mid 2020.


When AMA and CMS transitioned from "stack codes" for mopath payments to specific CPT codes (such as for "CYP2C19") in 2013, there was a huge boom in PGx/CYP payments, which suddenly came to comprise half or more of CMS mopath payments around 2014.  

CMS rapidly rolled down LCDs that controlled PGx/CYP payments, and these payments fell drastically from 2014 to 2015 to 2016, as shown in the bar chart below.  (I've pulled this slide from an available lecture, and haven't revisited the original data today).

CYP (PGx) Pt B Spending Falls Sharply 2014-2015-2016


Saturday, December 26, 2020

Blog #5 of 5: 81407 as Weirdest Tier 2 Code (&) Strange Impact on 70/30 Billing Rule (&) Wacko CLIA-CMS Edits

Triggered by the December 22 release of the OIG report on lab spending for CY2019, I've written a set of 5 blogs.

  • #1 - OIG report; Focus on 81408 as biggest genomic code
  • #2 - Review of all Tier 2 payments (81400-81408)
  • #3 - Review of MolDx Tier 2 payments
  • #4 - Review of NGS MAC Tier 2 payments
  • #5 - 81407 as the weirdest Tier 2 code
click to enlarge

Blog #1 here, Blog #2 here, Blog #3 here, Blog #4 here, Blog #5 here.


Earlier this month, OIG announced that hitherto nearly unknown Tier 2 code 81408 had become the largest genomic CPT code in the Medicare system, with nearly $290M of payments in CY2019.  I've discussed aspects of this in several preceding blogs.   I've also looked at the Tier 2 system (codes 81400-81408) as a whole which are extraordinarily skewed to billing via the Novitas system (e.g. of $400M in Tier 2 codes in 2019, only a couple percent of all Tier 2 codes were paid via the MolDx and NGS MAC systems together, which are the vast majority of US population and states).   '

  • Tier 2 payments in 2019, at $400M, were almost as much as paid for all genomic codes just a few years ago (2017).  

In looking at Tier 2 codes, I noticed that the several labs that account for most Tier 2 code billing, tend to bill zero or very little of code Tier 2 Level 8, 81407 (here).   

You can see this in a basic national chart of Medicare Tier 2 code payments in CY2018.   Payments for 81406 were $22M, for 81408 $123M, but 81407 plummets to only $6M:

When I review of the genes earmarked by AMA CPT for Tier 2 codes 81406, 81407, and 81408, it's unclear when 81407 would work this way.  

(A partial explanation is that 81406 and 81408 have CMS MUE edits of 2, allow 2 payments per day, while 81407 has an MUE edit of 1.  However, that doesn't account for the plummeting utilization of 81407 compared to 81406, 81408, shown above.)

Exercise for the Reader

I'll leave it as an exercise for the reader to log on to CMS public records for recipients of payments in 2016, 2017, 2018, with a search on Code 81407.   

In contrast to the skyrocketing growth of 81408 in these three years (and 2019), 81407 was billed by a few labs that came and went from year to year without any pattern.   

Labs billing 81407 often had unusual names; I'm making this one up, but names like "ABC Laboratory, Boca Raton, FL" would not stand out.   

Sometimes, the only internet footprint of these labs is (A) registering an NPI number, (B) showing up in CMS records for Tier 2 codes with a few million dollars of payments, and then (C) a tiny bankruptcy notice in some court filing that Google detects.  Well, I'm not saying there's anything wrong with that.

Impact of Huge Lab Price Ranges on 70/30 Rule

In 1994, the OIG introduced the term "shell laboratory," remarking, " [it] conducts very little testing on the premises, even though it is Medicare certified....Despite the location of the actual testing, the local ``shell''  laboratory bills Medicare directly for these tests."  Correspondingly, Medicare statute at 1833(h)(5(A) allows referred test billing only if less than 30% of a lab's tests are referred out.  

However, in 1989/1990, when the 70/30 percentage rule was introduced by Congress (see OBRA 1989 §6111 and OBRA 1990 §4154), most lab tests had similar prices, e.g. $5, $10, $15.   

Today, lab tests run from $3-$4 (82962, glucose, 81002, urinalysis, 85611 prothrombin) up to over $5000.  

Let's stop at Medicare largest genetic test in 2019, 81408, at $2000.   It seems like a lab can refer out about 1,000 cases of 81408 for every 2,000 tests of 81002 it does in house (the 70/30 test percentage rule; see here).  
If there were #2,000 in-house $3 tests, that would garner $6,000, while the #1,000 81408 tests referred out would be payable at $2M.   A lab could seemingly could bill $2,006,000 while meeting the 70/30 rule at SSA 1834(h)(5).   This could lead to a sort of industry that outsources 81408 tests selectively.   (By the way, it would require only 500 patients, to reach $2,006,000, with each of the 500 patients getting 4 of the $3 test and 2 units, the payable limit, of the 81408 $2000 test).  

CMS paid for 146,000 81408 tests in 2019; it's not known how many were paid as reference tests (send out tests) from places like "ABC Laboratory" that vanish within the year.  

Referred lab tests (if scrupulously coded, which might not be the case) should have a -90 modifier, e.g. 81408-90.  

CMS doesn't publish data what proportion any particular lab code arrives with -90 modifiers, although those facts might be obtainable via a FOIA request.   

CLIA Edits for Tier 2 Codes: The Final Shock

I'm guessing that CMS processing lab codes on claims to require a CLIA number and category that is the same as those listed in the CMS CLIA/Code/Claim requirements files.  If so, the requirements are just batty.   

Most genetic codes require "220,310" which is "chemistry" or "immunology" status.  A few require 900 "cytogenetics."  

OMG.  The requirements for the Tier 2 codes are published by CMS as in the table below.  All can use 610 (histopathol) while several can also use 220, 310, 400, which are Dx Immunol, Chem, and Heme.  I'm not expert in this and it looks as if you would have to have CLIA certification 610 histopathology to bill whole gene sequencing code 81408.   I don't know that for sure and will check around, but it looks crazy.  See the chart below of what CLIA subtypes are assigned to what CPT codes.

This makes another point.  At least in principal, many CLIA classifications are predicated on CPT codes.  When CPT codes skew to grab-bag codes (like 81408 or 81479), it's not possible to tailor the CLIA classifications to the needs and qualificiations of a particular identified test.   We think of CPT codes as designed to handle provider-payer communications, but they also have a regulatory function in applying CLIA rules.

Blog #4 of 5: NGS MAC Manages 20% of Medicare, Pays 0.1% of Tier 2 Codes

Triggered by the December 22 release of the OIG report on lab spending for CY2019, I've written a set of 5 blogs.

  • #1 - OIG report; Focus on 81408 as biggest genomic code
  • #2 - Review of all Tier 2 payments (81400-81408)
  • #3 - Review of MolDx Tier 2 payments
  • #4 - Review of NGS MAC Tier 2 payments
  • #5 - 81407 as the weirdest Tier 2 code
click to enlarge

Blog #1 here, Blog #2 here, Blog #3 here, Blog #4 here, Blog #5 here.


Earlier this month OIG reported for Medicare that CPT code 81408, a hitherto almost unused Tier 2 code, was the highest-paid genomics code in 2019.  

This led me to study where it's billed (mostly states under Novitas, including Mississippi and Oklahoma), and its growth rate (81048 grew 580X in just three years 2016-2019).   

Then, I turned to Tier 2 billing as a whole (all the codes 81400-81408), here.   I had noted that billing for code 81408 plummeted 99% when Palmetto took over the southern states TN GA AL in across 2018/2019.  That led me to look at Tier 2 billing as a whole, and, in the MolDx system as a whole (here).  I discovered that only 2% of  all Tier 2 dollars 81400-81408 in 2018 were paid by the MolDx MACs, despite their paying a predominance of mopath codes overall.   

In all this, I've said that most of the billing went out via Novitas, which has no edits on Tier 2 codes per its public LCD.   (That's about $400M in Tier 2 codes in 2019, almost as much as the entire domain of mopath spending in 2016/2017.)   You may be asking, what about NGS MAC?   

Tier 2 Payments by NGS MACs, CY2018

NGS MAC has 10 states, three in the midwest under Jurisdiction 11 (MN, WI, IL) and 7 in the northeast under Jurisdiction JK.  

Together these are about 20% of fee for service Medicare.  

The cumulative billings for Tier 2 codes in these states for CY2018 are in this table:

NGS MAC has 20% of Medicare Population and 20% of States

RESULT:  NGS MAC has 20% of Medicare population but pays 0.1% of Tier 2 budget.

Take a moment to catch your breath.


So MolDx MACs had 2% of payments in 2018 ($4M/$190M), and NGS MACs had 0.1% of payments ($268K/$190M).   

I suspect the level of payments in MolDx MACs remained more or less similar in 2019 (circa $4M) and in the NGS MACs (circa $300,000) while the national Tier 2 spending (thus, mostly under Novitas and First Coast) grew from $190M to $400M.    Here's what Novitas says about Tier 2 edits:

Novitas LCD re Tier 2 Codes

Thursday, December 24, 2020

Blog #3 of 5: Inferring MolDx Edits on TIer 2 Codes: Amazing Data

Triggered by the December 22 release of the OIG report on lab spending for CY2019, I've written a set of 5 blogs.

  • #1 - OIG report; Focus on 81408 as biggest genomic code
  • #2 - Review of all Tier 2 payments (81400-81408)
  • #3 - Review of MolDx Tier 2 payments
  • #4 - Review of NGS MAC Tier 2 payments
  • #5 - 81407 as the weirdest Tier 2 code
click to enlarge

Blog #1 here, Blog #2 here, Blog #3 here, Blog #4 here, Blog #5 here.


This week, OIG released data that Tier 2 code 81408 was the highest-spending code of all genomic CPT codes (here).  This code rocketed in utilization by 580X (not 580%, 580X) from 2016 to 2019.  Moreover, most (or all) of that spending on 81408 was in handful of states with no edits on the code.

This led me to do a second study of Tier 2 spending more generally - all the codes from 81400 to 81408.  I published that earlier (here).   Spending nationawide rocketed from $190M in 2018 to almost $400M in 2019.  

Now, rather than looking at "national" data, I compare MolDx MACs to other MACs.  WE'LL LEARN:  If MolDx edits were applied nationally, all Tier 2 spending would fall by 96%.  Amazing but true.


Back Story

I had previously noted that the spending for Tier 2 codes was irregularly distributed and in a limited number of states, most especially for the $2000 code, 81408 (a code never paid in the NGS MAC or MolDx MACs).  

But after looking at national data for Tier 2 codes in 2018 vs 2019, I wondered how the MolDx states compared, and across all Tier 2 codes.

My Research on MolDx Rules

First: No consolidated publication.  As far as I know, there's no published general MolDx "policy" for Tier 2 codes.   

Second: Datamining the DEX Exchange.  You can go to the DEX Exchange website owned and run by MolDx, and look up individual Tier 2 codes by gene name (by gene names indentified inside each Tier 2 code by AMA definitions).   You could tally those up in a spreadsheet.  However, Palmetto includes a pretty strict disclaimer about using the data (other than reading it!) (here).   I looked up several genetics labs in the MolDx region, and the largest one had about 90% of its genes registered with MolDx were publicly listed on DEX as "N" nonpayable, so that suggested the edits were pretty numerous and stringent.  

Third: Inferring Edits from Payment Profiles.  But there's another approach.  I went to the CMS cloud database for CY2018 (here), and the State data files.  I downloaded all usage and spending data for all codes 81400-81408 (Tier 2 codes) for all 50 states.  

Then: I then subtracted the 24 states (and DC) that aren't part of MolDx.  

For this purpose, since MolDx took over TN/GA/AL in mid-2018, I left those out of those "newest" MolDx states, since MolDx didn't manage them for MolDx edits for the full year (see here).

Here is the result for MolDx Tier 2 payments in CY2018:

 MolDx Tier 2 Spending 2018

So we can see that MolDx authorized 28,742 across all Tier 2 services in Cy2018, dollars allowed being $4.6M.  

But, utilization dropped to around 1000 cases for 81405 and 81406, and payments from MolDx occurred in only 1-2 states.   There was NO payments by MolDx for either 81407 or 81408.

It's possible to look up each code, each state, each lab, behind the above table.  CMS public data will tell you every lab that billed, say, 81404, in what state, for how many patients.  For a deeper analysis of the ratios of distribution of these tiers of payments in 2018, here.

How does this MolDx utilization of Tier 2 codes - 28,742 services, $4M - compare to national utilization of Tier 2 codes?  

Pretty damn amazing.  Overall, MolDx pays a huge proportion of Mopath spending, as much as 80% as recently as 2017.  (NGS MAC, for example, pays almost nothing in Mopath codes, except for Cologuard in Wisconsin).   

But that MolDx domination of MoPath payments comes to a halt when Tier 2 codes are involved.

For the two highest Tier 2 codes, MolDx pays nothing.  For four other Tier 2 codes, MolDx pays only 2-8%, although its rules apply to about half the US population. 

Net-net, if MolDx managed Tier 2 codes nationally, spending would fall over 90%.   This is driven, of course, on the assumption that 81408 spending at $123M would fall to 0%, its MolDx rate.  Table:

MolDx pays 8% of services, but only in the low-cost tiers, so overall MolDx pays only 2% of Medicare Tier 2 dollars.  


I had had an inkling of this when I looked at 81408 data for CY2018, CY2019, comparing TN-GA under Cahaba edits versus under Palmetto MolDx edits:


Foonote 1.

Payments for all Tier 2 codes in 2018 were $37M in GA and $15M in TN.  This is out of keeping with MolDx edits, and I found some web source stating that MolDx wasn't applied until a few quarters after the February 2018 takeover of other Cahaba-state edits. 


Footnote 2.

CMS has released national payment data for CPT codes for 2018 and 2019.  That's how we know that 2019 Tier 2 payments were about $400M.   But 2018 data is available in cloud database format, and 2019 data is only publicly available in pretty awful individual state spreadsheet files.   My guess for 2019?   Well, 2018 there were $4M of MolDx Tier 2 payments and $186M of non-MolDx payments.  Since there are $390M of total payments in 2019, I'm betting MolDx payments were close to the same ($4M) and that neans that about $386M was in other states.  I'm not eager to cut and paste the data from 50 files just to prove that.  

CMS cloud data for 2018 lets you be even more granular than the state and MAC level, and drill down how different labs billed different patterns for the 81400-81408 code series.  This is pretty detailed stuff and I've put it on a side blog here.  For example, the 10 labs that did most of the billing for 81408 in 2017 (when total Tier 2 billing was $70M) did 70% of the billing for all Tier 2 genes in 2018, when billing was $190M.  

Footnote 3.

NGS MAC pays even less (!!) for Tier 2 services than MolDx does.  See the next blog, Blog #4.

Blog #2 of 5: Medicare Tier Two Molecular Code Spending: 2019 vs 2018, and Where

Triggered by the December 22 release of the OIG report on lab spending for CY2019, I've written a set of 5 blogs.

  • #1 - OIG report; Focus on 81408 as biggest genomic code
  • #2 - Review of all Tier 2 payments (81400-81408)
  • #3 - Review of MolDx Tier 2 payments
  • #4 - Review of NGS MAC Tier 2 payments
  • #5 - 81407 as the weirdest Tier 2 code
click to enlarge

Blog #1 here, Blog #2 here, Blog #3 here, Blog #4 here, Blog #5 here.


Earlier this week I did a report on the new OIG analysis of 2019 Medicare lab test spending - here.  

OIG noted that highest-spending MoPath code in 2019 was 81408 - the Tier Two, Level 9 code which can represent any of about a dozen rarely sequenced genes.  

MAC contractors like Novitas state, they have no edits on Tier 2 codes, while other MACs (including NGS MAC and MolDx MACs) pay Tier 2 codes rarely or never, especially not paying codes 81407 or 81408 (see blogs #3 and #4).

OIG Highlighted 81408; Let's Look at All Tier 2 Codes

Looking at national data, and ignoring particularly MACs, OIG emphasized that 81408 was the heaviest-paid of all genomic codes at $289M.  OIG didn't mention, but I've reported, this code came out of nowhere (volume up 580X in just 3 years, 2016-2019'; see table in Blog #1).  

Rather than viewing 81408 by itself, let's view national payments for the whole set of AMA Tier 2 codes 81400-81408.  Payment for this group doubled between 2018 and 2019, from $190M to $390M.   Tier 2 codes 81406, 81407, 81408 all grew between 2X and 3X in just one year.

  • This article focuses on national payments for Tier 2 codes.  For the mind-blowing story of how MolDx edits for Tier 2 codes differ from other MACs, see my sister article Blog #3 here.

What I Did:  National Data for Tier 2 Codes

I looked at the dollar volume and services volume for Tier 2 codes comparing 2018 and 2019.  See the next table.  

Lower Tier 2 Codes Shrank 2018-2019.  Interestingly, the services volume and dollar volume of the lower Tier 2 codes shrank between 2018 and 2019.   

For example, for 81400, there were 35,356 services in 2018 and 21,812 services in 2019.  Dollar volume shrank by ($872,755).   In fact, all the Tier 2 codes in the lower range 81400-81403 shrank coming in 2019.

Higher Tier 2 Codes Boomed 2018-2019.  At the same time the lower-level Tier 2 codes were shrinking, the higher-level and far more expensive Tier 2 codes grew A LOT from 2018 to 2019, rising from 62,280 uses to 146,014 uses, and dollars allowed for 81408 rising from $123M to $289M.  See table, click to enlarge.  The next tables contain multiples, the ratio of 2019 vs 2018 spending.  The lower Tier 2 codes have red multiples less than 1, and the top three Tier 2 codes double to triple.

DOJ-Supported Studies of Medicare Billing Data

In the past couple years, labs in GA, TN, OK seem to be disproportionately mentioned in DOJ-supported studies of Medicare billing data (here).  

In CY2018, these states had together 6% of the Medicare fee for service population, but they seemed disproportionately represented in 81408 billing (see pie chart in figure on this page).  I pulled these states separately across the range of Tier 2 codes.  Although holding only 6% of the population, these small states had up to 80% of Tier 2 services billed, e.g. for 81405.  

The table above had columns for GA, OK, TN.  In OK, the Tier 2 codes 81402, 81407, with N=1 in the MUE table below, were not billed.  But some labs have billing profiles for Tier 2 codes that mold to the MUE edits.

MUE Edits for Tier 2 Codes

While some MACs do not have edits on these Tier 2 codes and others do, there are National Medically Unlikely Edits which cap the volume of services per day per code per patient.  These range from 1 (81402, 81407) to 5 (81404).  For cryptic reasons, the third-highest Tier 2 code (81406) has an MUE of 2, 81407 has an MUE of 1, and 81408 has an MUE of 2 again.

If you multiply out the MUE edit allowance times the CLFS price, you get about $8600 allowed per day per patient under MUE edits.

The MAC for a state can alter the billing patterns  

We have a case study where the Palmetto MAC took over claims processing and edits from the former Cahaba MAC, for Jurisdiction J.  For example, billing for Tier 2 code "81408" was relatively high in Tennessee and Georgia in CY2018, but fell visibly when MolDx edits were fully implemented before the beginning of CY2019:

81408 Pmts Visibly Lower in 2019 than 2018

Let's take a pause there.  For more on how MolDx edits Tier 2 codes, see the sister article here.


Grand Finale:  MACs Mightily Impact Tier 2 Payments

Here I conclude my analysis of national use of all Tier 2 codes by making summary graphs.  I broke out the MACs into three cohorts.

  • In purple, the Novitas and FCSO MACs.
  • In Orange, the small and now-gone Cahaba MAC (TN, GA, AL).
  • In blue, all the remaining MACs (MolDx [Palmetto, WPS, Noridian, CGS], NGS MAC).
Practically all the spending for the Tier 2 codes 81400-81408 is for code 81408 by Novitas and Cahaba (which had no edits on these codes), piled up on the right of the bar chart.

Breathtaking.  From a payments viewpoint, we see two facts, unknown before now.

  • Fact #1:  Nearly all spending for all Tier 2 codes flowed out via the Novitas and Cahaba MACs (purple + orange).  Only a sliver of payments went out through other MACs, and only for the lower Tier 2 codes.  
  • Fact #2:  Total payments for all Tier 2 codes is massively skewed to 81408, the highest codes, often billed $4000 per patient (2 x 81408).  

In 2018+2019, Tier 2 spending (as above) was about $600M.



If you look at services rather than payments, the bar graph is more balanced, but it's comparing test numbers rather than prices (some tests are $60, some $2000).

The next two blogs, #3 and #4, will look at edits and payment patterns in MolDx MACs and then in NGS MACs.   The final blog, #5, will ask why 81407 (see bar chart just above) is so tiny compared to 81406 and 81408.



For lab-level detail on how different labs billing Tier 2 codes show different billing patterns, see my side blog here.


CMS Updates Pathology Correct Coding Manual for CY2021

 Medicare has a national Correct Coding Initiative whose rules and policies are binding for Medicare submissions.   (There is a Medicaid initiative as well, which as far as I know, is a perfect mirror of the "Medicare" documents.)  Medicare website here.

The program has three major parts.  One is code-to-code edits, for example, you can't use the code "incision" with the code "appendectomy" since an incision is part of an appendectomy.   The next part is Medically Unlikely Edits, which are volume limits per day per claim.  These represent admninistrative maximum or "stop loss" edits and LCD edits could be lower.   For example, you can't get more than one heart transplant per day or more than one unit of most genetic tests.   Finally, there are a set of Policy Manuals, which are basically arranged by CPT code series (for example, the lab and pathology series 80,000 is Chapter 10).

CMS updated the manuals on December 17, effective January 1, and the pathology manual is here.  See the updates, highlighted by CMS in red, for yourself.  I will highlight some but you are responsible for reading the actual manual yourself.

  • There are minor editorial updates such as referring to deleted codes.  
  • There is a new paragraph around tumor panel codes 81445, 81450, 81455.  
    • CPT codes 81445, 81450, and 81455 describe targeted genomic sequence analysis.  81445 applies to solid organ neoplasm type (5-50 genes) and 81450 applies to hematolymphoid neoplasm type (5-50 genes), while 81455 applies to the number of genes analyzed for either a solid or hematolymphoid neoplasm (51 or greater genes).  Providers/suppliers may not report 81455 with either 81445 or 81450.
    • The above red text seems to be an example of the CCI simply cutting and pasting existing, and here, longstanding, code definitions into the CCI.  It's cryptic why they find this necessary and for just these codes.  The rule about reporting 81455 with the smaller codes is already in the CCI edits and AMA manual. (FN1)
  • Older text, not new, states that the genomic procedures "simultaneously assay multiple genes or genetic regions" and "[do not] report testing for the same gene or genetic region by a different methodology." Adding, "A Tier 1 or Tier 2 molecular pathology procedure code should not, in general, be reported with a genomic sequencing procedure..."
  • There is a section regarding reporting molecular testing for infectious agents on the same day as non-molecular probe testing for infectious agents; see K:2.
I noticed in this year's  Physician Fee Schedule Annual Rulemaking, in sections regarding telehealth, CMS quoted, parsed, and hewed rigorously to the exact phraseology of some AMA CPT paragraphs and definitions (85FR50117ff).  On the other hand, when writing the Correct Coding Manual, CMS notes that its rules are binding on practitioners and may supersede CPT conventions.   So there's a lot to keep track of.


I recently studied CGP codes 81445, '50, '55 and found that most MACs simply do not pay 81455, while between 2018 and 2019, MolDx shifted some lab[s] in its jurisdictions from billing 81455 to billing 81479 for the same service.  Here.

Wednesday, December 23, 2020

Very Brief Blog: CMS COVID Data Shows Highest Risk for Black Beneficiaries

CMS has a webpage where it posts beneficiary data about COVID, based on claims data.  CMS notes the data is not designed for epidemiology purposes, for example, racial data is not adjusted for comorbidities.

The homepage for the data is here, the current 12 slide deck is here.

The chart below is normalized "per 100,000" beneficiaries.  Note that the rate of hospitalization for Black patients is about 4X the rate for whites.  

The rate of hospitalization per beneficiary at age 85+ is about triple that for beneficiaries age 65-74.  (Data shown is for information received by October 9).

Very Brief Blog: Noridian MAC Awarded Jurisdiction E Contract, Again

Early in 2020, I wrote a blog about the fact that the West Coast / Jurisdiction E Medicare contract was under a routine recompete (here).

On December 18, 2020, CMS updated its MAC webpage for Jurisdiction E, announcing that Noridian has been awarded the work.  It will continue "contracting in place," so to speak.  

The CMS webpage for Jurisdiction E is here;


See links on that page for additional data about the award.  JE is about 10% of the national Fee for Service Medicare A/B program (the states are about 14% of the US, but the Medicare Advantage population in California is quite high at between 40 and 45%).  It looks like the contract is expected to be 7 years, at about $80M/year.  

Very Brief Blog: CMS Offers Preview of Possible Future Quality Measures

CMS slowly rolls out changes to the complex system of hospital and physician quality measures through annual rulemaking, typically working a couple of years in advance.   See the spring hospital inpatient rulemaking, or the summer physician annual rulemaking, and so on.

On December 22, 2020, CMS released a lengthy PDF of possible new measures for future consideration.  This is a legally required process that essentially puts a review and comment period ahead of the next year's formal rulemaking for each provider category.

See the press release here.  Comments open through end of January.   See PDF of proposals here.

Other News - CMS and Value-Based Drug Pricing

In other CMS quality news, CMS finalized rules that liberalize "value based drug deals" where payments are tied to health quality - here.  CMS writes:  "The rule overhauls existing regulations that made structuring value-based payments difficult,.. Under these types of deals, payers negotiate prices with drugmakers based on outcomes and evidence-based measures like reduced hospitalizations or lab visits, and aren't accountable for the full price if those measures aren't met."

Very Brief Blog: A Very Brief Peek at Medicare Provisions in Budget Law

Many detailed analyses of the giant budget bill just passed by House and Senate will appear quickly.  It's 5593 pages (although only a svelte 7 mb, here.)

I'll just highlight a few headlines from Section "CC," healthcare law, at pages 4597 ff.

  • Page 4599, S. 102, Lengthy section on quality measures and quality measure funding ($20M/year).
  • Page 4631, S. 119, real-time benefits for patient prices.  Seems to require tools embedded in EHR's to help patients and doctors select services while knowing prices.  (Separately, of course, the bill contains a huge and complex section on surprising billing law.)
    • For some insights into this space, see a November 2020 sponsored article in MedCity by RxREVU on real time pricing - here; links to white paper.
  • Page 4642, S. 122, reduces copayments for colorectal cancer colonoscopy when it follows a positive FIT or Cologuard test or when it results in a biopsy.
  • Page 4644, S. 123, benefits regarding mental health diagnosis and therapy via telehealth.
  • Page 4647, S. 124, public-private partnership to fight fraud, including a third-party clearinghouse.  If Entity X commits healthcare fraud against one health plan, it would be quickly circulated to others.  This isn't entirely new but it sounds like it's being put on steroids.
    • See the HFPP 2018 report on "lab fraud" here.
  • Page 4656, S. 403, Medicare can direct-pay physician assistants (not only pay via a clinic or the name of the supervising physicians.)   
    • This is interesting law, a few words long, simply adds a date to make the current restriction apply only "for services before 1/1/2022."  That's about how long the amendment is.
  • Page 4771, S. 402.  Allows post-renal-transplant patients enrollment in Medicare Part B "solely for the purpose of immunosuppressive drugs."  

Tuesday, December 22, 2020

Blog #1 of 5: OIG Releases Annual Report on Lab Spending: CY2019

Triggered by the December 22 release of the OIG report on lab spending for CY2019, I've written a set of 5 blogs.

  • #1 - OIG report; Focus on 81408 as biggest genomic code
  • #2 - Review of all Tier 2 payments (81400-81408)
  • #3 - Review of MolDx Tier 2 payments
  • #4 - Review of NGS MAC Tier 2 payments
  • #5 - 81407 as the weirdest Tier 2 code

click to enlarge

Blog #1 here, Blog #2 here, Blog #3 here, Blog #4 here, Blog #5 here.


On December 22, 2019 OIG has released its annual Congressionally required support on lab spending - for CY2019.   It may seem quite recent that we had an OIG annual ;ab report, but that's because the CY2018 report was delayed to August 2020.   

For 2019, lab spending is up a little, from $7.6B to $7.7B.   

But the real boom is in genetic spending - rising from $475M in 2017 to $970M in 2018 to $1.36B in 2019.

  • Find the report  >> HERE.

Some of the rise from $475M to $1.4B in two years could be related to headlines of hundreds of millions of dollars of inappropriate charges:


Overall, total CLFS lab spending has gradually shifted from $7.0B in 2014 to $7.7B in 2020:

Limited Overall Budget Growth, 2014-2019

Genetic spending has risen much more sharply.

The top 15 tests, across all types, are as follows. Note that most dropped slightly from 2018 to 2019. Nerd fact: CPT code 81479, unlisted molecular test, should be in mid-chart at $202M, but isn't shown. Probably because "unlisted codes" aren't part of PAMA and the OIG report is on PAMA.

Note that the top genetic test is 81408, Tier 2, Level 9, which pays $2000 and has more than doubled from 2018 to 2019.   This led me to do some research on 81408 (which I've published in previous blogs the past couple months).


The 81408 story is so strange, it's hard to tell compactly, so I've assembled multiple data types onto one slide, at bottom.  

I have to mention, I'm transferring now to Part B data, which I have in detail. OIG data adds up both Part B and Hospital Outpatient data, but (you can show) for 81408 it makes no difference.  

81408 is the highest payment level in the Tier 2 genes, 81400-81408.  Each tier 2 gene code represents a procedure on any of one of  a list of gene names inside it.  81408 represents full sequencing of any one of about a dozen (rare) genes - you don't know which one was looked at.

First, we know that 81408 is the biggest genetic code in 2019.  What's it's utilization history?  

In a nutshell:  81408 exploded in scale 2016 to 2017 to 2018 to 2019.  81408 billing grew 580X from 2016 to 2019, and 270X from 2017 to 2019.  (See chart further below, "Muliples by Year.")

OK, that's odd.  Our next natural question is...

Who's billing for 81408?  

It turns out, it's never Labcorp, Quest, GeneDx, Myriad, Ambry, Invitae...

In fact, in CY2018, only about 10 labs got paid for 81408 (in more than 10 units).  (See 2018 lab-level data here).   Of the ten labs billing for 81408 in CY2017, by my count, 6 or 7 were implicated in the allegations of "Operation Double Helix" (DOJ here), 1 of them was bankrupt, and the 2 remaining had no public adverse notices that I could find.   I have no opinion on any of this; I'll let you decide if there seems anything unusual about this.  

In 2016 and 2017, the only labs paid (>10 claims) for 81408 were in TN and GA.  In 2018, of the top 13 labs billing 81408, and which accounted for 85% of far larger 81408 payments, 10 of the 13 were associated via Google with Operation Double Helix (allegations only).   

Typically, the 10 labs that billed 81408, billed it in pairs or in 2 units (2 x $2000). I learned that this is the allowable limit under national Medically Unlikely Edits (here).    

Labs billing for 81408 most typically billed many of the Tier 2 codes (81400-81408) but since 81408 and 81407 pay $2000 and $1000 respectively, most of the financial weight falls in the topmost Tier 2 codes, especially when billing 2x81408=$4000.   Another tidbit: If labs billed 81408 at all, in CY2018, it was usually 50-70-90% of their total Medicare payments.  But let's stop there; that's a topic I'll handle in my next blog, Blog #2, about Tier 2 billing as a whole.

LCD edits are regional so next, Where did labs bill?  

Only in a couple MAC systems.  Billing occurred mostly in the Novitas MAC (which stretches from New Jersey to Oklahoma) and in the former Cahaba MAC.   (See MAC MAP for 81408 below).  The Cahaba MAC was absorbed into Palmetto MAC in 2018 and MolDx edits were applied in 2H2018 to labs in GA and TN billing 81408.  

As far as I can tell, MolDx states never pay for 81408 and neither do the northern NGS MAC states.  (See blogs #3 and #4 on MolDx and NGS MAC handling of Tier 2 codes.)   Some commercial payers do not pay for 81408 (see here).  

So if you're keeping tally:  The LARGEST genetic code in Medicare in 2019, 81408, was not paid in most states, almost entirely paid in the Novitas MAC - whose LCD even today says "we have no edits on 81408" - and is not paid at all in some major private insurers either.  Plus, if you google-search the 10 labs paid for 81408 in 2018, you hit case after case related to DOJ.  (Note:  We don't know if any DOJ allegations will hold up in court or not.)   

Weirdest thing I've seen in almost 20 years around Medicare. 

click to enlarge

When Palmetto Took Over Cahaba MAC

The data shows: Palmetto crushed 81408 spending when they took over Tennessee and Georgia from Cahaba in 2H 2018 and put in MolDx rules.  81408 in Georgia in 2018 was $16M, in 2019, that plummets to $74,000.   81408 in Tennessee in 2018 was $7.7M, in 2019, that plummets to $4000.   (So MolDx stopped 81408 spending cold.  In fairness, for this issue, NGS MAC also stopped 81408 spending, with 99% less infrastructure than MolDx).    See blog #3 for a broader view of how MolDx handles all Tier 2 codes, not just 81408.
81408 exploded 2018/2019 (top of article), but plummeted in Palmetto MolDx states


Extra Facts

Bonus question.  Why the boom between 2017 and 2018?  Tier 2 codes date back to 2012 or so.   Answer?  maybe because the code was unpriced and manually processed with medical records before 2018.  On 1/1/2018, the code got a fixed fee schedule price of $2000, and an MUE of 2, and in some states, no edits.  Welcome to 2018.   

Definition of 81408.  List of 81408 gene names here.  They're generally very rare disorders.  The public DEX database from Palmetto shows that generally 81408 genes are not covered by MolDx - database here, example here.

Other explanations for 2019 genetic growth. Besides 81408, other codes rose substantially 2018-2019, such as 0037U, Foundation Medicine, rising well more than double, from 9900 to 22900 uses.   That added about +$45M from 2018 to 2019; whereas 81408 added nearly +$200M.  

OIG reports A+B; I have data for B.  OIG reports Part B lab + Part A hospital outpatient labs, getting about 150,000 tests or $290.4M.  If you look at CMS Part B lab data alone, it's $290M, suggesting +$400,000 in hospital 81408 cases or 200 cases of 81408 billing (about 0.14%). 

MUE could have saved $150M with a single digit.  Spending on 81408 in 2019 was about $290M; if the medically unlikely edit had been N=1 instead of N=2, about $150M would have been saved by that single digit on a CMS spreadsheet. 

No rocket science.  Nothing here is rocket science - it's all public data from the CMS website and basic Excel skills.  No machine learning needed.


Data for "> 10 cases billed" versus "All Data."   I said that 10 providers billed 81408 in CY2018.  In fact, that's providers named in the database with granular 81408 data (like charges and cases).  Cutting the data another way, CMS tells us that 74 total providers billed Medicare for 81408, but 64 of them must be for less than 10 cases.  Let's assume those 64 billed 5 each, or 320 cases; that's 320 cases of 81408 out of 62,479 cases, or 0.5%.  Outside the ten main labs billing 81408, then, 0.5% was also billed by labs with few cases, and 0.14% was billed by hospital-reference labs.  Ten labs accounted for >99% of 81408 billing, but not 100.0%.

The Novitas LCD L35396, states there are no edits on any Tier 2 codes:

Multiples by Year.

81408 billing grew 580X from 2016 to 2019, and 270X from 2017 to 2019.

Pled Guilty

Managers for two labs pled guilty in June 2020; the labs accounting for 8% of CY2018 81408 payments (here).  DOJ writes, "Once the amount of the bribe was calculated, Ark and Tamulski drafted and submitted sham invoices to the laboratories that backed into the agreed upon bribe amount and attempted to conceal the scheme through describing various services provided at hourly rates."  DOJ adds, and I emphasize, "The charge and allegations against the remaining defendants are merely accusations, and they are presumed innocent unless and until proven guilty."
Top Code 81408 - Dollars by State
(Here in table form, the data supporting the pie chart in the figure earlier in blog.)

In the bar chart below, 2018 payments in $$M; purple bars Novitas except for Florida; blue bars Cahaba (defunct MAC). No meaningful payments for 81408 in NGS MAC nor MolDx group of four MACs.

Monday, December 21, 2020

Thursday, December 17, 2020

Very Brief Blog: CMS Files Motion to Dismiss PhRMA's Injunction Against Drug Pricing Cuts for 2021

 A couple weeks ago, CMS announced a massive shift downwards in the prices it pays for Part B drugs (my blog here).   PhRMA and co-plaintiffs ask the court to stop it, on administrative and statutory, even constitutional, grounds.   

On December 15, CMS predictably filed a motion to dismiss.  Find it here.  

As is usual in these cases, the government offers a series of reasons why the case is void.  First, the claims are barred from judicial review, for two reasons.  One, it lacks a legitimate claim of harm (no claims have been filed yet).  Second, Congress barred judicial review of Innovation Center programs.   

If the plaintiffs did have standing, then the claims are without merit.  The new pricing program was within statutory authority.  No constitutional claim is valid.

Further, CMS argues, "plaintiffs have failed to establish irreparable harm" in their pleading, so a restraining order would be excessive at this time.


Wednesday, December 16, 2020

Which Two Things Are Enough Like the Other? NTAP and Product Similarity

CMS sometimes issues product-specific policies (e.g. the NCD on Provenge, the NCD on the Exact Sciences Cologuard test) and sometimes categorical policies (the NCD on transvascular valve implants, TAVR, which applies on a rolling basis to all similar valves the FDA approves).

Today at MedCityNews, Niall Brennan, former Chief Data Officer at CMS, has an interesting article on the definition of "similar" in New Technology Add-on Payments, the DRG supplement program (NTAP).  Essay here.  Brennan is CEO of the Health Care Cost Institute, HCCI.

I've always thought of these as product specific.  And usually that's an easy one: either there was only one product of the type being reviewed, or the product was a drug which by definition is usually unique.   I have a memory that CMS may have once or twice considered two products closely together, in parallel and consciously crosswalked to each other, but they were still two products with both being reviewed.

Brennan lays out the vision of some NTAP categories where they may be a profusion of products, besides the index product studied by CMS, or there may be disputes about whether some new product is similar enough to a benchmark product to be under the same NTAP.   

Can new products not reviewed by CMS tag along with an NTAP-eligible ICD10 procedure code that fits them?   I wouldn't be sure they could - but this is the point that Brennan's essay is about.  He writes;

Traditionally, a substantial similarity determination has been applied via new NTAP applications, and CMS has determined whether products are substantially similar and have told applicants whether they are eligible for coverage under the initial NTAP. Where things start to get problematic is if, instead of CMS determining substantial similarity, other vendors unilaterally assert substantial similarity in order to “piggyback” on an existing NTAP.

It's true that normally coding decisions are made by the hospital or other provider, if the real-world service is close enough to a CPT or ICD10 procedure code to fit.  It's a new viewpoint to note that the ICD10 procedure code could be locked to a special payment rule designed for one product, although the ICD10 procedure code itself is applicable to more than one product.

CMS Part B Data: Profiling Code 81479 Billing in CY2018

Summary:  This article briefly highlights use of unlisted molecular pathology code 81479 in 2018 versus 2019, with a deep dive into recently-released granular data for 2018.  CMS data is easy to find and review; this blog took less than an hour "A to Z" including finding the cloud data, arranging it in small Excel tables, and writing.  


National Data for 81479

From data released nationally by CMS in Excel spreadsheets (here), we know that CY2017 use of code 81479 (unlisted molecular test) was $116M, CY2018 use was $135M and CY2019 use was $202M.   The growth rate more than doubled from +16% to +50%, and the two-year growth in 81479 was +75%.

  • In CY2019, 81479 the third-highest MoPath code (behind 81408, #1 at $290M, and 81528 Cologuard, #2 at $245M).   
  • 81479 is an "unlisted" code that falls outside of PAMA pricing rules, both regarding data collection and regarding price setting.

By going through state-by-state Excel spreadsheets for CY2019, I ascertained that nearly all CY2019 use of 81479 was in MolDx states, with a few percent of use by Novitas in Pennsylvania.   For example, from 2018 to 2019, MolDx shifted its coding instructions for about 7000 cases of 81455 to 81479, adding to the 2019 growth in 81479 (see article on 81455 here). 

Here, I focus on detailed and easily-manipulated cloud databases released by CMS for CY2018 (here). This data is older than 2019 data, but far easier to use, since it's in cloud databases.


State Level Data for 81479 for 2018

CMS's state level data for CY2018 has 260,000 rows (each use of a CPT code in each state can be a row), but filtering for 81479, only 22 rows are left.  That is, 81479 was used in only 22 states.  

58% of the dollarized use was in California, 28% in Ohio (for Myriad Assurex Health, Genesight).  That's 86% of the 81479 use in 2 states out of our 50 states.  See table below (click to enlarge).

Nearly all 81479 billing in 2018 was in MolDx states - easily discovered in the CMS cloud database for 2018.   This is the same result I got by manually going through each one of the newer 2019 state level Excel files one by one for code 81479.  

Lab Level Data for 81479 for 2018

CMS's provider (lab) level data has a robust 10M lines of cloud data.  Filtering for 81479 gives only 58 rows, e.g. 58 labs billed at least 10 uses of 81479. (FN1)

  • The top payee for 81479 in CY2018 was Assurex Health, Ohio, 17,341 services, $38M (average allowed, $2,178).  
  • The next highest was CareDx, California, 7,571 services at $2841 for $22M.  
  • The third highest was Genomic Health, California, 5,184 services at $3121 for 416M.  
  • These three of the 58 lines total 56% of the 81479 payments.  
Below, I chart all of the labs paid at least $100,000 in CY2018 in this Part B data.  Like in the prior table, the few labs outside MolDx states are shown in red.  Click to enlarge.

Note that in "state" data for Ohio, there were 17,529 uses of 81479 among 5 providers.  While in "Assurex" data for Ohio, there are 17,341 uses of 81479. at $2178.  Some other 4 providers apparently divided 188 other uses of 81479 in Ohio.  Further down in the lab-level table, we find that Labcorp was paid for 178 uses of 81479 for $321 via its Dublin, OH location.  

Dublin wasn't the only place where Labcorp billed 81479.  Labcorp has two Research Triangle Park entries for 81479, under two different NPIs ('6001 and '8700), getting 3,563 81479 services at one, and 223 81479 services at the other.

The original spreadsheet data I used is in the cloud here. (FN2)


Caris - Mix of Molecular and Pathology Codes in 2018

Caris, which raised $310M in growth capital in fall 2020 (here) and ranked highly in recent data I studied for CY2018 use of CGP (tumor profiling) code 81455 (here).  

By pulling Caris lab data for CY2018 from the cloud, we find that 81455 (tumor, 51 or more genes) constituted 71% of its CY2018 payments from CMS; total CMS dollars allowed were $28M in 2018.  

Note that a significant percentage of Caris's CMS payments were 88-series codes, or tissue pathology codes rather than 81-series codes for genomics.

81479 Charge/Payment Ratios

Charge to payment ratios are pretty meaningless in Medicare, but sometimes fun to look at.  I sorted for the labs with Charge/Payment ratios at 5 or over (range, 5-12).  Most had less than $10,000 in Medicare payments.  The average payment was  $177.  All were in non-MolDx states.  (E.g., use of 81479 in non-MolDx states is rare, and pays poorly.)   

Though not a high-ratio case, Labcorp in NC was paid for 3,563 cases of 81479 for $1.2M, but that's only $327 a case.


Tennessee  switched from the Cahaba MAC to Palmetto MAC in 2018, but I don't believe MolDx rules were applied until mid-year or later.



109 providers billing 81479 are shown in the state data, but only 58 in the national data of labs.  The difference is that providers must be paid for 10 or more uses to show in the national data, providers paid for 9 or less uses are not shown.   Apparently, 51 providers were paid for 9 or less uses of 81479, and 58 were paid for 10 or more uses of 81479.


This week, Google Drive appears to convert uploaded Excel workbooks into Google Sheets automatically, although my settings tell it not to do that.  (?!)  I believe that if you click Download it will give the option of back-converting to Excel.