The CPT Editorial Panel created the Digital Medicine Coding Committee in 2024. The committee’s charge is to respond to requests from the Panel and to support the Panel on coding issues that involve or may include a digital medicine and/or an artificial intelligence component by providing or obtaining expertise or advice on a specific subject matter. At all times, the Panel shall have the sole authority to create, revise and update codes, descriptions and applicable guidelines for appropriate CPT coding.
The CPT Editorial Panel continues to evolve its thinking on artificial intelligence and strives to better accommodate the evolving role of advanced digital technologies in health care.
In response, the Panel’s Digital Medicine Coding Committee (DMCC) has been working to develop a coding solution. The potential new category of CPT code would be a mechanism to codify algorithmic analysis of clinically relevant patient data (e.g., biophysical signals, imaging data, lab results) to produce clinically meaningful output or conclusions that impact patient care.
This framework is still under development and is tentatively titled Clinically Meaningful Algorithmic Analyses (CMAA).
The DMCC is actively seeking stakeholder feedback to ensure the structure is clinically relevant, scalable and aligned with current health care delivery and payment models.
Each stakeholder listening session will allow for stakeholder comments. We cannot guarantee that everyone will have time to speak during these open comment sections, but it is our intent to hear all perspectives. Attendees must register for each virtual meeting and participants will be held to the CPT confidentiality policy (PDF).
Medicare’s Epilepsy Panel Problem: A Naming Analysis of Labs Billing 81419, 81408, and 81443
Executive Summary
Medicare Part B has experienced a dramatic surge in payments for CPT 81419 (epilepsy gene panel) — from 41 payments nationwide in 2022 to 30,377 payments and $74 million just two years later. This pattern mirrors the earlier wave of abuse around 81408 (rare gene sequencing) that prompted OIG investigations and $888M in improper payments.
A close review of lab names billing these codes reveals a distinctive profile: generic or inflated names, heavy LLC usage, no locality identifiers, and lookalike branding— all consistent with shell labs designed to rapidly capture Medicare dollars.
For comparison, we analyzed 973 labs billing 81443 (thyroid stimulating hormone, $16), a common and low-risk test. These names show a markedly different profile: they include major national labs, hospital-affiliated labs, and regional providers with established reputations.
1. Background
Medicare Part B pays for clinical laboratory tests on a fee-for-service basis, with few real-time edits or prior authorization requirements. This “pay and chase” model has historically been vulnerable to sudden surges in utilization when a code is unguarded — as seen with 81408 between 2018–2021.
The new concern is CPT 81419, an epilepsy gene panel. Despite being rarely used until recently, its utilization skyrocketed in 2023 and 2024, concentrated in Novitas and FCSO MAC regions
DIHP 81419 blog
.
2. Methodology
81419 dataset: 26 unique labs billing this code in 2023.
81408 dataset: 74 unique labs billing this code in 2020, at the peak of its abuse cycle.
81443 dataset: 973 unique labs billing this routine $16 test in 2023, used as a baseline for “normal” lab naming patterns.
Each list was analyzed for:
Legal form (LLC vs. Inc.)
Use of prestige or marketing words (“Elite,” “Express,” “Global”)
Use of genomics/biotech buzzwords (“Gene,” “Bio,” “Diagnostics”)
Geographic or ownership identifiers
3. Findings
A. 81419 Epilepsy Panel Labs (2023)
58% are LLCs, suggesting easy setup/dissolution.
High prevalence of prestige branding: “Elite,” “Global,” “Vanguard” — names that project importance without revealing a real-world presence.
Lookalike risk: “Elite Bio Reference Laboratory” echoes BioReference (a major national lab).
Almost no locality: None reference a city, region, or hospital system, making them placeless.
Representative examples:
Yeva Diagnostics
Global Gene Labs, LLC
Elite Bio Reference Laboratory, LLC
Ultra Gene, LLC
B. 81408 Rare Gene Sequencing Labs (2020)
Even higher LLC rate (74%), confirming “pop-up shell” behavior.
Plainer names: Fewer biotech buzzwords, more generic business titles (Best Choice Laboratory LLC, Oakwood Lab Services LLC).
Known fraud history: OIG has documented massive improper payments from this code cycle, confirming that this pattern represents high risk.
C. 81443 TSH Labs (2023)
Diverse but legitimate mix: Quest, LabCorp, hospital labs, regional systems, academic centers.
Higher Inc. rate (41%) and lower LLC rate (30%) — suggesting long-standing corporate structures.
Locality common: Many names reference geography or parent institutions (e.g., University Medical Center Lab, Midwest Clinical Laboratory).
Minimal marketing jargon: Almost no “Elite” or “Express” (0.2% combined), very low use of “Gene” or “Bio” (2–3%).
4. Comparative Table
Feature
81408 (2020)
81419 (2023)
81443 (Routine)
Number of Labs
74
26
973
LLC usage
74%
58%
30%
Prestige words
Low
High
Very Low
Genomics buzzwords
Moderate (13%)
High (19%)
Rare (2%)
Geographic identity
Rare
Rare
Common
Known fraud risk
Confirmed by OIG
Emerging risk
Low
5. Interpretation
The naming pattern itself is a strong indicator of risk:
81419 labs look more like marketing constructs than medical institutions.
Their profile closely parallels the labs that fueled the 81408 fraud wave.
In contrast, 81443 labs resemble real-world providers with geographic anchoring and brand equity, providing a baseline for what “normal” looks like.
6. Policy Implications
Early warning system: CMS could use naming analytics as a heuristic signal. Labs with formulaic, placeless names — especially LLCs — should be flagged for additional review.
Targeted prepayment edits: As with 81408, Novitas and FCSO regions appear to be the locus of uncontrolled growth and should receive earlier interventions.
Transparency and reporting: Regular publication of top billing labs per CPT code could let watchdog groups and researchers catch anomalous patterns months (not years) earlier.
Conclusion
The explosive growth of 81419 billing — funneled through a small number of suspiciously named labs — represents a replay of the 81408 fraud scenario. Medicare cannot afford to wait for another OIG report after hundreds of millions are spent. Naming analysis, combined with real-time data monitoring, can provide a low-cost, high-yield early detection tool to protect the Part B trust fund.
We have studied the odd names of labs that billed 81419 (epilepsy gene panel) to Medicare in 2023. There was a lot of known Medicare fraud in 2020 caused by labs that billed code 81408 (full sequence, rare gene, $2000) to Medicare. I include the names of labs that billed 81419 in 2023, and labs that billed 81408 in 2020.
Now, I am also including labs that billed a common and mundane code to CMS - 81443, Thyroid Stiumulating Hormone $16. Using some of the insights you saw above, but also new insights you may develop, discuss anomalous names of 81419 labs, of 81408 labs, and, finally, put these in a setting of relatively unselected labs billing 81443. (Of course, a few of the labs billing 81443 could be fraudulent, not impossible.)
###
Here’s a structured comparison of the naming patterns across three groups of Medicare labs — the ones billing 81419 (epilepsy panel), 81408 (rare gene full seq), and 81443 (routine TSH).
1. 81419 Epilepsy Gene Panel (2023)
Highly “engineered” names: 58% include “LLC,” suggesting rapid shell formation for billing.
Marketing-style branding: Frequent use of “Elite,” “Express,” and “Global,” which signal prestige rather than location or science.
Genomics jargon:“Gene,” “Bio,” “Diagnostics” are common — giving a veneer of modernity.
Lookalike risk: Names like Elite Bio Reference Laboratoryecho national players like BioReference, which might confuse payers. See also: GeneTx, which someone might confuse for "GeneDx," a very large publicly held lab.
Pattern: Concentrated in Novitas/FCSO states, similar to prior 81408 fraud.
This cluster feels like it was designed for Medicare billing rather than clinical branding — often placeless, ownerless, and likely ephemeral.
2. 81408 Rare Gene Full-Seq (2020)
Even higher LLC usage:74% of labs are LLCs — even more than 81419, reinforcing the “set up & dissolve” pattern.
Fewer prestige words: Most names are plain but corporate-sounding (Your Choice Laboratory LLC, Oakwood Lab Services LLC), projecting neutrality.
Somewhat less genomics jargon: Only 13% mention “Gene,” vs. nearly 20% of 81419 labs.
Key observation: 81408 labs were numerous (74 unique labs) and widely dispersed, but many were later linked to OIG findings of massive improper payments. [See OIG Report about the "billion dollar code."]
3. 81443 TSH (Routine, $16 Test)
Much more mainstream profile: 973 unique labs, including Quest Diagnostics, LabCorp, regional hospitals, and university systems.
Lower LLC rate (30% compared to 74% for 81408), higher Inc. rate (41%) — reflecting large established corporations.
Rare “prestige words”:Almost no “Elite” or “Express” — only 0.2% use them.
Clear locality/ownership signals: Many names include city, region, or hospital name, suggesting real brick-and-mortar providers.
4. Putting It Together
Feature
81408 (2020)
81419 (2023)
81443 (Routine)
LLC usage
74% – very high
58% – still high
30% – lower
Prestige words
Low
High (“Elite,” “Express,” “Global”)
Nearly absent
Genomics jargon
Moderate (13%)
High (19%)
Rare (2%)
Locality in name
Rare
Rare
Common
Association with fraud
Known OIG target
Rapid rise, suspicious pattern
Mostly mainstream
Interpretation
81419 is the “new 81408.” The naming style shows slightly more polish, more “science branding,” and an obvious intent to look legitimate to payers.
81408 was rawer: mostly cookie-cutter LLCs, still enough to fool Medicare into paying $888M before OIG stepped in.
81443 shows what normal looks like: established labs with geographic anchors, brand equity, and no need to sound like a biotech marketing deck.
I would not take this too seriously but it's provided "FYI" only as an example of AI writing with science policy documents. Not as a real "expert" judgment.
AI picked up on topics like CMS staff overload and the contrast with an accelerated NCD in 2014 and whether that could occur again.
Based on the content of the letter and CMS precedent, the chances of favorable CMS action look fairly strong, with a few nuanced considerations that experts would flag for monitoring:
Reasons CMS Is Likely to Act Favorably
1. Strong Regulatory Trigger
ColoSense is FDA-approved via PMA and even had Breakthrough Device designation, which CMS generally views positively.
CMS has a history of aligning coverage decisions with FDA approval for CRC screening tests — notably, Cologuard’s NCD was updated within 60 days of FDA approval in 2014.
2. Clear Clinical Validation Evidence
The pivotal CRC-PREVENT trial is large (8,900+ participants) and published in JAMA, which meets CMS expectations for peer-reviewed evidence.
Performance metrics (≈93–94% sensitivity for CRC, ≈45–56% for advanced adenomas, ≈88% specificity) are comparable to or slightly better than Cologuard, which CMS already covers.
3. Category-Wide Policy Trend
CMS has been moving away from single-test NCDs toward category-based coverage (e.g., NCD 90.2 for NGS, 2021 blood-based CRC test NCD).
This request aligns with CMS’s stated aim to reduce administrative burden by avoiding repeated reconsiderations for each new technology.
4. Public Health Framing
Letter ties request to Cancer Moonshot priorities and COVID-related screening backlogs — aligning with [2024] White House and CMS equity initiatives.
Potential Red Flags / Risks
1. CMS Resource Constraints
NCD reconsiderations compete with many priorities. CMS may delay due to staffing bandwidth rather than scientific objections.
2. Frequency / Utilization Management
CMS might scrutinize the three-year screening interval, especially given concerns about over-utilization and cost. They could request additional modeling or health-economic analysis.
3. Category-Wide Language
CMS may hesitate to immediately grant category-wide coverage for all future sRNA tests, preferring to first set precedent with ColoSense specifically, then broaden coverage later.
4. Sensitivity for Advanced Adenomas
ColoSense shows ~45–56% sensitivity for AAs — comparable to Cologuard but not perfect. CMS could ask whether this represents a “meaningful clinical improvement” over FIT.
That said, CMS did not require Cologuard to show superiority over FIT, so this is likely not disqualifying.
Bottom Line
Probability of favorable action:High (80% of higher) given FDA approval, strong evidence, and CMS precedent.
Biggest risk:Timeline, not outcome — CMS could take 6–12 months to issue a Proposed Decision Memo rather than fast-tracking as they did for Cologuard in 2014.
Likely compromise: CMS may first issue coverage for ColoSense specifically (test-specific coverage) before moving to category-wide sRNA coverage. [BQ: Not likely]
Here’s a side-by-side, graduate-level comparison of the FDA validation packages—what each kit is cleared to do, how it was analytically/clinically validated, and how “safety” (i.e., risk of incorrect classification and controls/QC to mitigate it) is handled.
Scope & Intended Use (what each kit may legally report)
Level 2/3 variants with evidence/potential evidence of clinical significance; TMB score
Level 2/3 variants across a 46-gene panel; no TMB
20240 Illumina FDA 0801 P23001…
2025 FDA 0702 TFS TMO Oncomine …
Analytical Validation—design choices and headline performance
Study design highlights
Specimens: Both PMAs validate on intended-use FFPE tissue (multiple tumor types), with DNA and/or RNA workflows aligned to the claimed variant classes. Controls (positive and NTC) are required each run, and runs/samples that fail QC are invalidated by rule.
20240 Illumina FDA 0801 P23001…
2025 FDA 0702 TFS TMO Oncomine …
Comparators: Both use validated NGS orthogonal methods as the primary accuracy comparator for claimed biomarkers; tumor-profiling concordance studies span representative variants across genes and tumors.
20240 Illumina FDA 0801 P23001…
2025 FDA 0702 TFS TMO Oncomine …
Error control & calling:
Illumina: UMI-based error correction on DNA; BWA-MEM/STAR alignment; variant quality and likelihood metrics; TMB is calculated from filtered small-variant calls with germline-exclusion heuristics; explicit analyte-level QC (e.g., fusion score, splice score). Performance not established for tumor-profiling small variants <5% VAF.
20240 Illumina FDA 0801 P23001…
ODxET: Variant caller supports read-based and molecular-family–based evidence; variant-class-specific thresholds (SNV/indel/CNV/fusion/splice) and tumor-cellularity-aware CNV logic; reportability driven by an ADF (assay definition file).
2025 FDA 0702 TFS TMO Oncomine …
Accuracy (agreement vs orthogonal methods)
Claim
Illumina TSO-Comprehensive
ODxET
Primary CDx accuracy
NTRK1/2/3 fusions (solid tumors): PPA 96.6% (114/118; 95% CI 91.5–99.1); NPA 94.5% (273/289; 95% CI 91.2–96.8). Study noted a putative plate cross-contamination cluster driving several TSO-positive/comparator-negative discordants; invalids occurred in both methods; ~17% TSO invalids in the mixed cohort (larotrectinib trial + supplemental).
Established across small DNA variants (with ≥5% VAF), fusions, splice variant; gene-level analyte QC and control pass/fail gating (DNA control requires 23/24 specified variants; RNA control 12/13).
Three study sets spanning SNV/indel/CNV/fusion/splice; variant-class-specific PPA/NPA reported vs NGS comparators; additional focused accuracy on ERBB2 CNV (breast), ALK & RET fusions (NSCLC/thyroid), EGFR SNVs & exon 19 deletions (NSCLC).
Run/Sample/Analyte QC (Illumina): tiered gating (run-level PF % and Q30; sample-level median insert size, median exon coverage, RNA on-target reads & expression-control detection; analyte-level thresholds including fusion score, splice score, small-variant AQ/LQ). Failure at any tier blocks reporting for the run/sample/analyte. Controls carry strict pass counts (DNA control 23/24 variants; RNA 12/13).
20240 Illumina FDA 0801 P23001…
Run/Sample/NTC/Purification QC (ODxET): templating control minimum reads and read length; DNA mapped reads ≥100k, RNA mapped fusion reads ≥6,000 and expression controls detected; NTC zero-call rules; purification quant constraints (0.5–512 ng/µL). Variant calls must also satisfy class-specific calling criteria (read count, allele fraction, strand, hotspot coverage).
2025 FDA 0702 TFS TMO Oncomine …
Interpretation: Both PMAs front-load risk control into automated QC + tight control design that halts reporting if quality is inadequate—this is the main “safety” mechanism for an IVD NGS panel.
Notable design differences with regulatory impact
Breadth vs depth
Illumina offers very broad content (517 DNA genes; 24 fusion genes + one splice; TMB), two CDx claims (NTRK/RET), and extensive tumor-profiling output. Accuracy for very low VAF tumor-profiling variants (<5%) is not established, which is typical for large-panel CGP PMAs.
20240 Illumina FDA 0801 P23001…
ODxET is narrower but adds CNV calling under IVD claims and focuses its CDx on EGFR exon 20 insertions; it provides a full tumor-profiling menu across 46 genes.
2025 FDA 0702 TFS TMO Oncomine …
Change management (post-market updates)
ODxET includes an FDA-approved Predetermined Change Control Plan (PCCP) to update its ADF (variants, clinical categorization) and certain software elements without a new submission, contingent on predefined V&V, documentation, and regression testing—important for keeping the tumor-profiling knowledge current.
2025 FDA 0702 TFS TMO Oncomine …
Illumina provides a Knowledge Base that is updated periodically for tumor profiling and installed on the instrument; the PMA positions this as part of the LRM workflow rather than a PCCP. (Operationally useful but not a PCCP.)
20240 Illumina FDA 0801 P23001…
Systems engineering
Illumina separates extraction/library/enrichment from sequencing on NextSeq 550Dx with UMI error correction on DNA libraries.
20240 Illumina FDA 0801 P23001…
ODxET leverages the Genexus Dx end-to-end automation (purification → library → templating → sequencing → analysis) with tight run planning/traceability and templating-control gates—reduces operator variability by design.
2025 FDA 0702 TFS TMO Oncomine …
“Safety” in the PMAs (how risk is addressed)
No device-related adverse events were reported; both PMAs frame risk as false positives/false negatives leading to inappropriate therapy or missed therapy, plus delayed results if runs/samples fail. Mitigation is via: strict controls (positive/NTC), multi-layer QC gating, invalidation rules, and restricted, indication-specific CDx interpretation algorithms.
20240 Illumina FDA 0801 P23001…
2025 FDA 0702 TFS TMO Oncomine …
Practical takeaways for molecular pathology & regulatory teams
If you need TMB under IVD claims or a broad CGP report alongside two pan-tumor/NSCLC CDx claims, Illumina TSO-Comprehensive is the match; just note the ≥5% VAF floor for established tumor-profiling small-variant accuracy and enforce the analyte-level QC gates (fusion/splice scores, control pass counts).
20240 Illumina FDA 0801 P23001…
If you need an IVD with CNV calling, rapid “sample-to-answer” automation, and an FDA-approved PCCP to streamline future variant/reporting updates—and a focused CDx for EGFR exon 20 insertions—ODxET is compelling; its CDx accuracy vs orthogonal methods is excellent, and prevalence-adjusted NPA/PPV provide confidence at real-world EGFR-ex20 prevalence.
If you want, I can drop these into a 1-page table you can paste into a validation plan or TA memo.
###
###
Post Market Letter
Here’s a focused, side-by-side on post-approval (postmarket) requirements for the two PMA’d kits, highlighting what’s routine for all PMAs vs. what’s device-specific (e.g., ODxET’s CNB/FNA commitments).
What both PMAs must do (standard conditions)
Annual Reports (21 CFR 814.84) with required content and timing, plus units sold/distributed (to contextualize adverse events) under 21 CFR 814.82(a)(9).
Change control: PMA supplement / 30-day notice obligations per 21 CFR 814.39 for any change affecting safety/effectiveness (Illumina), and the same for Thermo Fisher plus alignment to its approved PCCPs (see below).
MDR (21 CFR 803), recalls/corrections (21 CFR 806.10), UDI requirements, and QS regulation (21 CFR 820) expectations are called out explicitly in both letters.
24-month shelf life noted. (Not a study commitment, but establishes the approved dating.)
4-month shelf life; protocol used is approved for future dating extensions under 814.39(a)(7).
Labeling / software follow-up
Within 6 months: modify the test results report as agreed, and submit software V&V (executed protocols, regression, risk assessment, unresolved anomalies). This is a post-approval commitment; failure can trigger withdrawal (814.82(c), 814.46(a)(2)).
Within 6 months: provide results + software validation from regression testing on configuration that masks Level 3 SNVs <5% VAF (to mitigate false positives) and show no adverse impact on other reporting.
Specimen-type expansion (CNB/FNA)
No specific post-approval specimen study is required in the approval letter. (Standard PMA reporting still applies.)
Required: support core needle biopsy (CNB) and fine-needle aspirate (FNA) as acceptable tumor profiling specimen types with: (a) a precision study across variant classes (SNV/indel/CNV/fusion) on adequate CNB/FNA across tumor types, and (b) a specimen equivalency study vs matched resections spanning all variant classes. Final protocols due within 1 month; final data, conclusions, reports, and labeling within 12 months.
PCCP (Predetermined Change Control Plans)
Not referenced as an approved PCCP in the approval letter. (Standard 814.39 supplement pathway applies.)
Multiple PCCPs are approved (ADF aggregation, pre-approved software change, variant re-categorization). Changes consistent with PCCPs do not require a supplement; deviations do (per section 515C and 21 CFR 814.39). Annual Reports should include a separate section for PCCP-implemented changes.
Administrative reminders
Submit final labeling before distribution; all the usual MDR/UDI/QS/recall obligations reiterated.
Same; plus explicit instruction to include PCCP changes in Annual Reports.
Practical implications for labs & RA/QA teams
ODxET has actionable, time-bound postmarket studies that matter operationally: you’ll need to plan CNB/FNA accrual across multiple tumor types, cover all variant classes, and structure a matched-pair study design vs resections. Expect labeling updates upon successful completion.
ODxET’s approved PCCPs reduce friction for keeping software/variant knowledge current, but they shift rigor to internal V&V and documentation; ensure your Annual Report cleanly segregates PCCP-implemented changes.
Illumina’s commitment is narrower (report formatting + software V&V within 6 months). Ensure your software validation, regression, and risk files track precisely to the commitments called out in the letter.
Oncomine Dx Express (P240040): CNB/FNA postmarket program (protocols in 1 month; data/labeling in 12 months) plus a 6-month software masking validation; PCCPs approved and must be tracked in Annual Reports.
If helpful, I can turn this into a one-page RA checklist with owners/timelines for each deliverable.
