Monday, September 9, 2019

Brief Blog: Understanding Concert Genetics' Two New White Papers on Genetic Coding Mess

Concert Genetics, a consultancy in Nashville, is currently hosting the annual Genetic Health Information Network Summit (GHIN Summit, September 9-11, 2019) - here.  Concert Genetics is well-known for its series of white paper on the U.S. genetic testing industry, volume, test types, and code mapping.

This week, they've also released two important new white papers on genetic coding.   I'll summarize briefly.

  • For Concert Genetics home page, here.
  • For white papers, here.  
    • A prior Concert Genetics white paper cited in Congressional testimony, here.
    • For some peer-reviewed papers Concert has written or collaborated on, here.
    • Patent, Systems and method for tracking...diagnostic testing products, here. (Cited in WP#1).
  • For blog, here.
  • For Variability in Coding white paper (#1), here.
  • For Coding Solution white paper (#2), here.

White Paper #1 - Variability

Working with a database of 35M claims and 2M genetic tests, Concert categorized by test type (for example, tumor testing, cancer risk testing, carrier testing.)  
  • Within each category, there was huge variability in coding and pricing.   
  • Those categories with the highest variability in coding had highest variability in pricing.

White Paper #2 - Coding Solution

We're all familiar with correct coding principles, some (brief ones) found in the CPT handbook, some in Medicare documents (Correct Coding Edits), some unwritten, some in payer articles, etc.   What Concert proposes is to have one uniform, systematic set of rules that are always applied as algorithms.  They are commonsensical (if there is a PLA code, use PLA code.  If there is a fit to a GSP code like Hereditary Breast and Ovarian Cancer, 81432, then apply that.   If there is a Tier 2 code, then apply that; and so on).   

It's different than today's coding because, while there is a dream that all human coders will code correctly and the same, here, a computer (or a human following iron-tight rules) will always reach the same result.   
Let's apply the idea to a dictionary.  We want to categorize the words in a dictionary.  First, categorize as English vs foreign.  Then, second, categorize as A-L or M-Z.  Then, third, categorize as one syllable, or two or more syllables, etc.   You'd have a rule set that could categorize any word, even words it hasn't seen before.  Same idea here, but for genetic coding.


One is is special rules.  Medicare MACs may have special coding rules they insist be applied (although sometimes this collides with payers; Myriad 2019 case here.)   Medicare has national special coding rules (Correct Coding Edits), although sometimes CCI rules are bizarre, as the industry has complained about some 2019 genetic coding rules therein (here, here).   MolDx has a "Test Panel" coding rule, which is dead-simple to state, but public data shows that Noridian is unable to consistently apply it.  Concert plans to have one uniform set of rules, limited in number, applied in specific order.



While it is common to hear of the 100,000 or 200,000 genetic tests, in 2017 Medicare data, 72% of volume was driven by only 10 codes.  Most of this was concentrated in the top 5 codes alone. 

While the top 5 codes did include 81479 (unlisted code) at 18%, if you went inside that unlisted code, most of the 81479 volume was driven by a few tests like Myriad Assurex Genesight, which is about half the 81479 volume.   (And almost all of the 81479 billing comes through the MolDx states, zero elsewhere).   In short, CMS data shows heavily concentrated use of a few blockbuster genomic tests.