CMS Rationales about Applicable Lab Rules in PAMA (Messy!)

 Messy stuff.  Just recording PAMA output from AI here.

##

Regarding cms pama data collection rules for lab pricing, applicable laboratory. There is a basic rule about the lab having $12500 in revenue in a six month period and some proportion from Part B CLFS and Part B PFS over "all medicare revenue.' I used to understand this but now I just find it confusing. They intended Part B PFS to represent lab payments (PFS pathology, surgical pathology, immunohistochemistry) since the reporting entity was supposed to be "a laboratory" but in a physician cinic the lab revenue would be confounded with all part b revenue, unlike say at Quest. Can you explain the rationale of the applicable laboratory rules and calculations - not just the rule but the thinking behind it.

I’ll separate the mechanics from the policy logic, because the PAMA applicable-lab definition is one of those rules that is simple as a formula but odd in institutional design.

The easiest way to understand the PAMA “applicable laboratory” rule is this:

CMS was not trying to identify every place that performs lab testing. CMS was trying to identify Medicare-billing lab businesses whose private-payer rates would be meaningful inputs for a national market-based CLFS. That design choice explains most of the oddities.

1. The core PAMA idea

PAMA replaced administratively-set CLFS rates with rates based on the weighted median of private-payer rates reported by “applicable laboratories.” So CMS needed a reporting population that looked like the commercial laboratory market: Quest, Labcorp, independent labs, physician-office labs, some outreach labs, and specialty molecular labs.

But Congress did not simply say “all CLIA labs must report.” It defined an applicable laboratory as a lab that gets a majority of its Medicare revenues from the CLFS or the Physician Fee Schedule. CMS explains that the statute points to revenue from sections 1834A / 1833(h) — the CLFS authorities — and section 1848 — the PFS authority.

2. The current rule, mechanically

A lab generally must satisfy these elements:

First, it must be a laboratory under the CLIA-style definition.

Second, it must bill Medicare Part B under its own NPI, or, for hospital outreach labs, under 14x type of bill rules. CMS’s current PAMA resource says the lab bills Medicare Part B either under its own NPI or on Form CMS-1450 under TOB 14x if it is a hospital outreach laboratory.

Third, it must pass the “majority of Medicare revenues” test:

CLFS revenue + PFS revenue must be more than 50% of the lab’s total Medicare revenue.

The current regulation says the numerator is revenue from CLFS subpart G and PFS subpart B, and the denominator is total Medicare revenue, including Medicare fee-for-service payments, Part D payments, and beneficiary deductible/coinsurance amounts for services in the data collection period.

Fourth, it must pass the low-expenditure threshold:

It must receive at least $12,500 in Medicare CLFS revenue during the data collection period.

CMS’s current PAMA page states this directly: the lab must get at least $12,500 of Medicare revenues from the CLFS in the data collection period.

So the simplified calculation is:

Majority test:
(CLFS Medicare revenue + PFS Medicare revenue) ÷ total Medicare revenue > 50%

Low-expenditure test:
CLFS Medicare revenue ≥ $12,500

3. Why include PFS revenue at all?

Your instinct is right: PFS revenue was included because many laboratory-related Medicare payments are not technically paid on the CLFS. In pathology, cytology, surgical pathology, immunohistochemistry, professional/technical pathology services, and physician-office laboratory contexts, substantial lab-adjacent or pathology-lab business may be paid under the Physician Fee Schedule, not the CLFS.

So Congress/CMS did not want to say: “Only entities whose Medicare revenue is mostly CLFS revenue count.” That would exclude a lot of pathology and physician-office lab activity that is functionally part of the laboratory market. Including PFS was a way to capture labs and pathology entities whose Medicare business is still “lab/pathology business,” even if the statutory payment authority is PFS rather than CLFS.

But this creates the problem you identify.

4. The physician clinic problem

In a clean independent lab, the calculation is intuitive:

Quest-like lab:

• CLFS revenue: large
• PFS revenue: maybe limited
• Other Medicare revenue: little or none
• Result: CLFS + PFS is likely >50% of Medicare revenue.

So it passes the majority test easily.

But in a physician clinic, the same TIN or billing organization may have:

• Office visit revenue
• Procedures
• Imaging
• Drugs
• Lab tests
• Pathology or technical lab work
• Other PFS services

Here, PFS is not necessarily “lab revenue.” It may be ordinary physician-practice revenue. That is the conceptual flaw. The statute uses PFS revenue as a proxy for lab/pathology revenue, but in a multispecialty physician practice, PFS is a huge mixed bucket.

CMS originally proposed a TIN-level approach, partly because private payer rates are often negotiated at the TIN/business level. But commenters objected that this would exclude hospital outreach labs and distort the reporting pool. CMS then shifted the definition of applicable laboratory to the NPI level, while keeping reporting at the TIN level. CMS said this allowed hospital outreach laboratories with their own NPI to qualify, while the parent TIN could still do the actual reporting.

That move helped, but it did not fully solve the physician-clinic conceptual issue. It made the calculation more targeted than a whole corporate TIN, but PFS still remains a broad statutory bucket.

5. Why hospitals mostly failed the test

This is the mirror-image problem. Hospital labs perform huge volumes of lab testing, but most inpatient and outpatient hospital lab testing is paid through IPPS or OPPS packaging, not separately under the CLFS or PFS.

CMS explicitly reasoned that hospital IPPS and OPPS revenue would be in the denominator but not the CLFS/PFS numerator. As a result, most hospital laboratories would not pass the majority-of-Medicare-revenues test. CMS said it believed the statute supported limiting reporting primarily to independent labs and physician offices, because hospital IPPS/OPPS revenues would overwhelm the lab-specific numerator.

That was controversial because hospital outreach labs compete with independent labs. CMS eventually tried to accommodate outreach labs by allowing qualification at the NPI level and later through 14x bill type logic for hospital outreach.

6. Why the $12,500 threshold exists

The $12,500 rule is not part of the “majority” concept. It is a burden filter.

CMS did not want every tiny physician-office lab, dermatology office, urology office, or occasional lab biller to gather and report private-payer data. The agency’s stated balancing logic was: collect enough data to produce meaningful weighted medians, but avoid imposing reporting burdens on small-volume entities. In the 2016 final rule, CMS discussed its original $50,000 idea, then finalized a lower threshold tied to the six-month data collection period.

Today the operative number is $12,500 in Medicare CLFS revenue during the data collection period. That is a gatekeeper: even if the lab is “mostly lab” by Medicare revenue, CMS does not want it reporting unless it has at least a modest Medicare CLFS footprint.

7. The policy logic in one paragraph

The PAMA applicable-lab definition is a compromise between market representativeness, administrative feasibility, and statutory awkwardness. Congress wanted private-payer lab rates from entities that are genuinely in the lab market. The CLFS-only test would have been too narrow, because pathology and physician-office lab services often run through the PFS. So PFS revenue was added. But that creates overbreadth in physician clinics, where PFS includes much more than lab/pathology. Conversely, hospitals perform enormous lab volume, but their Medicare revenue is dominated by IPPS/OPPS, so they often fail the test unless the outreach lab is separately identifiable. CMS therefore landed on an NPI/14x-based applicable-lab definition, TIN-level reporting, a >50% CLFS/PFS majority test, and a $12,500 CLFS floor.

Bottom line

Your old understanding was basically right. PFS was included because Congress/CMS were trying to capture pathology and physician-office laboratory business that is paid under the PFS rather than the CLFS. But the proxy is imperfect. In an independent lab, PFS inclusion is usually sensible. In a physician clinic, PFS can swamp the calculation with non-lab revenue. In a hospital, IPPS/OPPS revenue can swamp the denominator and exclude the lab unless outreach billing is separately identified.

So the rule is not “pure lab economics.” It is a statutory revenue-proxy test built from Medicare payment silos. That is why it feels confusing: it is trying to infer “is this a lab business?” from which Medicare payment systems paid the entity, rather than from a clean managerial accounting definition of laboratory revenue.

Plus how does the lab in hospital know how much medicare revenue it gets. The hospital gets a DRG payment, it goes to the CFO, the CFO pays all salaries in the lab department on a budget, etc. it's not like CMS sent a check for $52 to dr Jones.

Exactly. That is one of the central practical reasons the hospital-lab problem was so awkward.

For an independent lab, Medicare revenue is fairly concrete: the lab bills CPT code X on a 1500/837P claim, Medicare pays Y dollars, and that payment can be tied to the lab’s NPI/TIN and to a specific test.

For a hospital inpatient lab, that accounting does not exist in the same way. The hospital receives a DRG payment for the inpatient stay. The CBC, chemistry panel, surgical pathology, microbiology cultures, transfusion testing, etc., are not separately paid line items under the CLFS. They are internal cost centers inside the hospital’s overall DRG economics.

So if someone asks, “How much Medicare revenue did the hospital lab receive for inpatient testing?” the technically correct answer is often:

It did not receive Medicare revenue in the direct PAMA sense. The hospital received bundled Medicare revenue, and the lab consumed part of the hospital’s internal budget.

CMS recognized this problem in the 2016 rule. It said that laboratory services for Medicare hospital inpatients are not paid fee-for-service but are bundled into IPPS, and most hospital outpatient laboratory services are packaged into OPPS. CMS also said it was unclear how hospital laboratory “revenues” for bundled inpatient/outpatient services would be determined if one tried to use a CLIA-certificate-based hospital-lab approach.

That is why CMS did not try to make each hospital lab calculate an imputed Medicare revenue number from DRGs. It would have required an allocation model, such as:

DRG revenue × internal lab cost-center allocation factor = implied lab Medicare revenue

But that would be managerial accounting, not Medicare payment data. Different hospitals would allocate costs differently. Some would use charges, some cost-to-charge ratios, some cost accounting systems, some budget shares. The resulting “revenue” would be artificial and non-comparable.

So CMS largely ducked that problem by saying:

For PAMA applicable-lab status, count Medicare revenues paid under the CLFS and PFS, not bundled IPPS/OPPS payments.

That means:

Hospital inpatient lab work: generally not separately counted as CLFS/PFS revenue.

Hospital outpatient lab work: often packaged into OPPS and therefore not counted, unless separately paid under rules such as certain non-patient/outreach situations.

Hospital outreach lab work: more likely to be countable if billed separately, especially under its own NPI or later 14x outreach-billing logic.

This also explains why hospital labs objected. They said, in effect: “We are real labs. We compete with Quest and Labcorp. We have private-payer rates. But your Medicare-revenue formula makes us disappear because our Medicare lab work is buried inside hospital payment systems.”

CMS’s answer was basically: “Yes, but the statute defines applicable laboratories through Medicare revenue sources, and IPPS/OPPS are not CLFS/PFS.”

So your “CFO pays the lab department budget” example is precisely the point. In a hospital, the lab is often a cost center, not a Medicare-revenue recipient. PAMA needed a clean, auditable reporting trigger. Bundled hospital payments do not provide one.

RUC Resources Are Numerous at AMA

 

AMA RUC Resources: A Guided Map

AMA has a surprisingly large set of public resources on the RVS Update Committee, usually called the RUC. The RUC is the AMA/Specialty Society Relative Value Scale Update Committee, the physician-led process that makes recommendations to CMS on the relative values assigned to CPT codes under the Medicare Physician Fee Schedule. Your blog page already captures the key links: the RUC home page, the RBRVS overview page, historical PDFs, the RUC process booklet and slide deck, the minutes archive, committee composition, and several 2026 physician payment rule summaries.

Home page:
https://www.ama-assn.org/about/rvs-update-committee-ruc

RBRVS overview and historical policy PDFs:
https://www.ama-assn.org/about/rvs-update-committee-ruc/rbrvs-overview

The central conceptual point is that Medicare physician payment is built on the Resource-Based Relative Value Scale. Each service is valued through three components: physician work, practice expense, and professional liability insurance. These RVUs are geographically adjusted and multiplied by a Medicare conversion factor to produce payment. The 2026 RUC process booklet summarizes this elegantly: physician work averages about 50.9% of total value, practice expense about 44.8%, and PLI about 4.3%.

---

1. RUC home page and RBRVS overview page

The AMA RUC home page is the portal for the topic, while the RBRVS overview page is the better entry point for historical and methodological background. These pages link out to the PDFs that explain how Medicare moved away from historical physician charges and toward resource-based payment. Your blog correctly separates the home page from the overview and historical policy PDFs, which is useful because the PDFs are where the deeper policy history sits.

Home page:
https://www.ama-assn.org/about/rvs-update-committee-ruc

RBRVS overview:
https://www.ama-assn.org/about/rvs-update-committee-ruc/rbrvs-overview

---

2. “Development of the Resource-Based Relative Value Scale” — 12 pages

This is the best historical introduction to why RBRVS was created. Medicare originally paid physicians through a “customary, prevailing, and reasonable” charge system. Over time, that system locked in old geographic and specialty payment patterns, became increasingly distorted by freezes and update limits, and did not respond well to changes in clinical practice or technology. The RBRVS was intended to replace charge history with a more systematic estimate of the resources needed to provide each service.

PDF:
https://www.ama-assn.org/system/files/development-of-the-resource-based-relative-value-scale.pdf

---

3. “Legislation Creating the Medicare RBRVS Payment System” — 12 pages

This PDF explains the legislative story behind the 1989 creation of the Medicare RBRVS payment system. OBRA 1989 established a payment schedule based on three resource components: physician work, practice expense, and professional liability insurance. It also created a five-year transition beginning in 1992, eliminated specialty differentials for the same service, required geographic adjustment, set a budget-neutral conversion factor, and retained limits on balance billing.

PDF:
https://www.ama-assn.org/system/files/legislation-creating-the-medicare-rbrvs-payment-system.pdf

---

4. “History of Medicare Conversion Factor Under the SGR” — 8 pages

This is the best short guide to the conversion factor, the dollar multiplier that turns geographically adjusted RVUs into Medicare payment. The initial 1992 conversion factor was about $31.00. For many years, updates were governed by expenditure target formulas, later the Sustainable Growth Rate. The SGR repeatedly threatened large physician payment cuts, leading to 17 temporary congressional fixes before MACRA repealed the SGR in 2015 and replaced it with statutory updates.

PDF:
https://www.ama-assn.org/system/files/history-of-medicare-conversion-factor-under-the-sgr.pdf

---

5. “Physician Work Component” — 22 pages

This is the detailed technical document for the work RVU side of the RUC world. It explains that physician work is not just time. It includes time, technical skill and physical effort, mental effort and judgment, and psychological stress related to patient risk. It also describes the Harvard RBRVS study, the 1992 refinement process, and the ongoing AMA/Specialty Society RVS Update Process.

PDF:
https://www.ama-assn.org/system/files/physician-work-component.pdf

---

6. “Practice Expense Component” — 27 pages

This may be the most important PDF for readers who want to understand the machinery behind practice expense.

You might also consider how the P.E. rules create conflicts or dilemmas when faced with software licensing fees or per click fees, issues CMS and RUC are wrestling this year.

[See this book handled in more detail, below]

 It explains the transition to resource-based practice expense RVUs, fully implemented by 2002, and the key distinction between facility and nonfacility practice expense values. Nonfacility settings include physician offices, freestanding imaging centers, and independent pathology labs; facility settings include hospitals, ASCs, skilled nursing facilities, and similar settings.

PDF:
https://www.ama-assn.org/system/files/practice-expense-component.pdf

The practice expense document also shows why PE policy is so contentious. Practice costs are not naturally measured service-by-service. CMS and AMA-related datasets must allocate costs such as clinical labor, equipment, supplies, office space, clerical payroll, and indirect expenses across CPT codes. The document therefore becomes essential background for current fights over site-of-service payment and indirect practice expense methodology.

---

7. “AMA/Specialty Society RVS Update Committee: An Overview of the RUC Process” — 13 pages

This is probably the best general-purpose RUC explainer. It states that the RUC was formed in 1991 to make recommendations to CMS for new, revised, and potentially misvalued CPT codes. It also explains the coordination between the CPT Editorial Panel, specialty societies, RUC surveys, RUC deliberation, and CMS rulemaking.

PDF:
https://www.ama-assn.org/system/files/ruc-update-booklet.pdf

The booklet is particularly useful because it describes the operational cycle: CPT creates or revises codes; specialty societies indicate interest; AMA staff distributes surveys; specialty societies survey practicing physicians; recommendations are presented to the RUC; and RUC recommendations are forwarded to CMS, which then proposes and finalizes values through rulemaking.

---

8. “RBRVS and RUC Process” slide deck — 32 pages

The 32-page slide deck is the most visual and teaching-friendly version. It illustrates the three RVU components, the physician work elements, practice expense inputs, professional liability insurance, and the formula for calculating payment. It also gives 2026 conversion factors: $33.5675 for qualifying APM participants and $33.4009 for non-qualifying APM clinicians.

PDF slide deck:
https://www.ama-assn.org/system/files/rbrvs-ruc-process.pdf

The deck also gives current talking points on why the RUC matters. It describes the RUC as an expert panel of volunteer physicians petitioning the federal government, with CMS retaining final decision-making authority. It also highlights the “potentially misvalued services” project: the RUC has examined nearly 3,000 potentially misvalued services, recommended reductions or deletions for more than 1,600 services, and reviewed about 95% of the Medicare physician payment schedule.

---

9. RUC minutes archive and committee composition

Your blog also links to the RUC minutes archive and the current committee composition page. These are valuable because they move beyond general description into the actual governance record: what was reviewed, when, and by whom. The RUC booklet states that the committee has 32 members, with 22 appointed by major national medical specialty societies, plus rotating seats and representatives tied to AMA, CPT, HCPAC, and the Practice Expense Subcommittee.

RUC minutes archive:
https://www.ama-assn.org/about/rvs-update-committee-ruc/rvs-update-committee-ruc-recommendations-minutes

RUC members / committee composition:
https://www.ama-assn.org/about/rvs-update-committee-ruc/rvs-update-committee-ruc-members

---

10. 2026 Medicare Physician Payment Schedule resources

The 2026 AMA final-rule summary is the current-policy companion to the historical RUC materials. It explains that 2026 includes positive conversion factor updates — 3.77% for qualifying APM participants and 3.26% for other physicians — but that these increases are partly offset for many specialties by CMS’s finalized 2.5% efficiency adjustment and changes to the indirect practice expense methodology.

AMA Medicare Physician Payment Schedule page:
https://www.ama-assn.org/practice-management/medicare-medicaid/medicare-physician-payment-schedule

2026 MPFS final rule summary and analysis — 16 pages:
https://www.ama-assn.org/system/files/2026-mpfs-final-rule-summary-analysis.pdf

The 2026 two-page AMA issue brief is the advocacy summary of those same concerns. It argues that CMS finalized a 2.5% reduction in work RVUs and intra-service time for nearly 7,000 services based on presumed efficiency gains, and separately reduced practice expense RVUs for services performed in facility settings. AMA warns that the facility PE policy could reduce physician payment for facility-based services by about 7% overall.

2026 physician payment proposals / efficiency adjustment issue brief — 2 pages:
https://www.ama-assn.org/system/files/cms-2026-physician-payment-proposals.pdf

---

In Conclusion...

Taken together, the AMA’s RUC resources are more than a set of committee documents. They are a map of how Medicare physician payment is built: historical reform away from charge-based payment, legislative creation of RBRVS, the physician work methodology, the practice expense machinery, the conversion factor saga, and the ongoing RUC/CPT/CMS cycle that updates the system every year. For anyone trying to understand Medicare physician payment — especially current debates over practice expense, site of service, misvalued codes, and CMS’s new efficiency adjustment — the AMA RUC library is one of the most concentrated public resource sets available.

###########################

The 27 Page Practice Expense Book - In More Detail

The AMA Practice Expense Booklet: Why It Matters for Today’s AI and Software Payment Debates

The AMA’s “Practice Expense Component” booklet is one of the most useful public explanations of how Medicare assigns the practice expense portion of physician payment under the Resource-Based Relative Value Scale, or RBRVS. It is not a booklet about artificial intelligence, software as a service, digital pathology, clinical algorithms, or autonomous systems. In fact, its machinery is largely built around the older world of staff, supplies, equipment, office space, administrative labor, and site of service. But that is exactly why it matters. These legacy rules are the baseline from which future Medicare payment rules for AI and software will have to evolve.

Medicare physician payment is built from three RVU components: physician work, practice expense, and professional liability insurance. This booklet focuses on the second of these: practice expense RVUs, or PE RVUs. PE RVUs are meant to capture the non-physician resources needed to provide a service: clinical labor, supplies, equipment, rent, utilities, administrative staff, and other overhead. Beginning in 1999, Medicare began transitioning from older charge-based PE values to resource-based practice expense values, and by 2002 the PE component was fully transitioned. CMS later implemented a new PE methodology in 2007.

A key distinction in the booklet is facility versus nonfacility payment. A service that can be performed both in a physician office and in a hospital may have two different PE RVUs. The nonfacility setting includes physician offices, freestanding imaging centers, and independent pathology labs. The facility setting includes hospitals, ambulatory surgery centers, skilled nursing facilities, partial hospitals, and similar settings. This distinction remains central today because many new technologies, including digital and software-enabled services, may have different cost structures depending on whether the service is performed in an office, hospital, laboratory, imaging center, or distributed cloud/software environment.

The booklet explains that the original OBRA 1989 approach to practice expense was a rough proxy. CMS used surveys of physician practice costs, but those surveys generally measured total practice costs, not the cost of each individual CPT-coded service. For example, a physician might report total annual spending on rent, wages, equipment, and supplies, but that did not directly reveal how much rent or nursing time was used for a specific service. CMS therefore had to allocate aggregate practice cost information across services, using specialty cost ratios and Medicare claims patterns.

That older method was criticized because it remained partly tied to historical Medicare allowed charges. In other words, practice expense values could inherit distortions from the pre-RBRVS era. The booklet explains that Congress and CMS eventually moved toward a more resource-based method, requiring attention to the staff, equipment, and supplies used to provide services in different settings. CMS contracted with Abt Associates and used expert panels, survey data, and cost inputs to build a more detailed PE methodology.

The modern resource-based PE method divides costs into categories. The booklet describes six major cost pools: clinical labor, medical supplies, medical equipment, office expense, administrative labor, and all other expenses. These are then divided conceptually into direct costs and indirect costs. Direct costs include clinical labor, supplies, and equipment that can be linked to a particular service. Indirect costs include office expense, administrative labor, and other overhead that cannot be tied as neatly to one patient encounter.

This distinction is one of the most important lessons for AI and software. Traditional practice expense methodology is comfortable with tangible resources: a nurse’s minutes, a syringe, a CT scanner, a procedure room, an exam table. It is less naturally suited to cloud infrastructure, cybersecurity, algorithm maintenance, software licensing, model monitoring, version control, data storage, regulatory maintenance, and human oversight of algorithmic outputs. Yet these future costs will have to be mapped somehow onto the inherited categories of direct cost, indirect cost, equipment, supply, labor, or overhead.

The booklet also shows how fragile these allocations can be. CMS’s methodology uses specialty cost pools, physician time, Medicare utilization data, and code-level cost inputs. When direct cost inputs are changed for one family of codes, the dollars do not simply disappear; they may be redistributed across other codes within the same specialty pool. When multiple specialties perform the same service, CMS may average PE values using Medicare frequency data, which can alter payment effects across specialties.

For policy readers, this is a reminder that practice expense is not a simple accounting exercise. It is a constructed payment model. It uses real data, but also assumptions, crosswalks, scaling factors, utilization weights, and budget neutrality. That matters greatly for software because small choices about classification can have large downstream effects. Is software treated like equipment? Like a supply? Like indirect overhead? Like clinical labor substitution? Like a technical component? Like a separately payable service? The legacy PE framework does not answer those questions directly, but it shows the kind of logic CMS will likely start from.

The booklet’s discussion of data sources is also important. AMA’s Socioeconomic Monitoring System data, later practice expense surveys, CMS crosswalks, expert panels, and supplemental data all play roles in constructing PE values. The booklet notes limitations in survey response rates, specialty sample sizes, and whether physician-level surveys adequately capture practice-level costs. Those problems are likely to be magnified for AI and software, where costs may sit outside the physician practice, inside a vendor contract, embedded in a platform fee, or spread across thousands of uses.

One especially useful section describes the expense-per-hour concept. CMS uses practice expense per physician hour, specialty cost pools, physician time for Medicare services, and procedure-level allocation methods to distribute practice costs across CPT codes. This makes sense for conventional services where physician time and practice infrastructure are closely linked. It is less obvious how it should work when a software tool performs analysis asynchronously, when the marginal cost per use is low but development and maintenance costs are high, or when a service is delivered through a cloud platform rather than a room full of equipment.

The booklet is therefore not merely historical. It is a window into Medicare’s inherited accounting vocabulary. It explains why payment for practice expense has traditionally revolved around labor minutes, supplies, equipment cost, utilization assumptions, specialty pools, site-of-service differences, and budget neutrality. Any serious future framework for AI and software payment will have to decide whether to stretch these categories, revise them, or create new categories that better fit digital medicine.

The AI/software dilemma is that the old PE model was designed for a world where “resources” were usually visible: a technician, a room, a machine, a supply cabinet. Software resources are real, but less visible. They include engineering, validation, FDA-related quality systems, monitoring, upgrades, cybersecurity, cloud costs, licensing, customer support, integration with clinical workflow, and sometimes medico-legal risk management. Some of these look like practice expense. Some look like vendor overhead. Some look like capital. Some look like physician work support. Some may justify separate coding only when the software creates a distinct clinical service rather than merely making an old service easier.

That is why the AMA booklet is a useful starting point for current debates over digital pathology, AI-enabled diagnostics, and software-enabled clinical services. It does not solve the new problems. But it shows the inherited architecture: how Medicare thinks about resources, how CMS translates those resources into RVUs, and why every new technology eventually has to pass through a payment logic that was built for older forms of medical practice.

In short, the booklet is a foundational guide to practice expense as Medicare currently understands it. For AI and software, the next policy task is not to ignore this framework, but to understand it well enough to know where it fits, where it fails, and where new rules will be needed.

###

Here is a deeper dive on a section about a .44 factor deflator that many will find puzzling. 

Page 20.

This is CMS doing a global “shrink-to-fit” step for the indirect practice expense pool.

In plain English: CMS first calculates a theoretical “indirect allocator” for each service. For 99213, Step 8 produces 2.30. But that number is not yet a payable PE RVU. It is more like a claim on the indirect overhead pool. CMS then asks: if we applied all these indirect allocators across the whole Medicare Physician Fee Schedule, using utilization for every service, would the total exceed the available pool of indirect PE RVUs? In 2026, the answer is yes. So CMS scales the whole set down by an adjustment factor, 0.4479. For 99213, that turns 2.30 into 1.03.

The reason the factor looks so severe — about 45 cents on the dollar — is that Step 8 is intentionally not a final RVU calculation. It is an allocator, not a final payment value. CMS is creating a relative measure of how much indirect overhead a service should attract, using direct PE, direct percentage, and work RVUs. Then CMS compresses all those claims back into the aggregate indirect PE pool available under the fee schedule.

For 99213, the booklet’s example works like this. CMS first computes direct PE RVUs of 0.33 from labor, supply, and equipment inputs. It then calculates that 99213 has a 25% direct and 75% indirect PE percentage. The Step 8 indirect allocator is:

0.75 × (0.33 ÷ 0.25) + 1.30 = 2.30

Then Step 11 applies the 2026 indirect adjustment:

2.30 × 0.4479 = 1.03

But even that 1.03 is not the end. CMS then applies a specialty-level indirect practice cost index. For 99213, the 2026 index is 1.13, so the adjusted indirect PE becomes roughly:

1.13 × 1.03 = 1.17

Then CMS adds back the direct PE RVU:

0.33 direct + about 1.16 or 1.17 indirect = about 1.48

After a final PE budget-neutrality adjustment and phase-in, the final 2026 PE RVU for 99213 is 1.46.

So the 0.4479 adjustment is not saying, “CMS thinks indirect costs are only 44.79% real.” It is saying, “The preliminary indirect allocators collectively add up to too much relative to the allowed aggregate indirect PE pool, so we have to scale them down before applying specialty-level adjustments and final budget-neutrality steps.”

The easiest analogy is a pie-cutting algorithm. Step 8 lets every code submit a proposed slice size. Step 10 adds up all the proposed slices. Step 11 realizes the proposed slices exceed the pie. The 0.4479 factor shrinks the slices proportionally so the total fits the pie. Then later steps adjust slices by specialty cost patterns and budget-neutrality rules.

###

I gave an example of clinical labor $10 and supplies $15 and amortized CapEx $20, $45, yet the TC is $19. Woah?

Yes — the bottom line is that the AMA/RBRVS “direct PE inputs” are not a promise that those dollars will show up dollar-for-dollar in the final CMS technical-component payment.

They are raw direct cost inputs that enter a national allocation formula. CMS then scales them down — often sharply — so that all direct and indirect practice expense RVUs across the entire Medicare Physician Fee Schedule fit within the available PE pool.

The key thing you have been missing is probably Step 4: the direct PE adjustment.

In the booklet’s 2026 example, CMS first adds up nominal direct costs:

Clinical labor + supplies + equipment = direct costs

For code 99213, the booklet’s example is:

$19.44 labor + $5.54 supplies + $0.34 equipment = $25.31 raw direct cost

But CMS does not carry $25.31 forward as $25.31 of payable PE. It applies a national direct adjustment. For 2026, that direct adjustment is 0.4228. So the raw direct costs are multiplied down:

$19.44 × 0.4228 = $8.22
$5.54 × 0.4228 = $2.34
$0.34 × 0.4228 = $0.14

So the apparent $25.31 of direct inputs becomes only about $10.70 of adjusted direct PE before being converted into RVUs.

That is the “hidden deflator” you are describing.

The conversion factor is involved mechanically, but it is not really the explanation. CMS explicitly notes that the particular CF used in the PE calculation does not change the resulting RVUs, because the same CF is used in the scaling steps and in converting dollars to RVUs; changes in the CF and the related scaling adjustments offset each other.

So, in consultant shorthand:

The AMA direct PE page gives you the unscaled direct inputs. CMS then applies a national direct-cost scaling factor before those inputs become payable PE RVUs.

Then there is a second major adjustment for indirect PE. CMS calculates indirect allocators, compares the total proposed indirect allocation to the available aggregate indirect PE pool, and then applies an indirect adjustment. For 2026, that indirect adjustment is 0.4479. In the booklet example, an indirect allocator of 2.30 becomes only 1.03 after the indirect adjustment.

After that, CMS applies specialty-level indirect practice cost indices, service-level weighting, direct-plus-indirect summation, and final PE budget-neutrality adjustments. The booklet’s example continues from the adjusted indirect allocator to a specialty/service indirect index, then adds direct and indirect PE RVUs and applies final PE budget neutrality.

For pathology technical components, there is another complication. Many technical-component services have no physician work RVU or minimal physician work. Historically, CMS had a nonphysician work pool for technical-component and other no-work codes, but CMS later eliminated that pool and created “proxy work” RVUs for these services, using clinical staff costs or the Medicare conversion factor. That means TC-only services are not simply “direct lab costs plus rent.” They are being pushed through a special version of the PE allocation model.

So your contrived example is directionally right. Suppose the AMA screen appears to show something like:

$10 clinical labor + $15 supplies + $20 equipment = $45 raw direct PE inputs

A naïve reader might expect at least $45, plus some rent and overhead. But CMS may first apply a direct adjustment roughly like 0.42, turning $45 into about $19 of adjusted direct cost before RVU conversion. Then indirect PE is calculated separately, scaled, specialty-weighted, and budget-neutralized. Depending on the code, the TC payment may end up around the same neighborhood as the adjusted direct cost, not the original raw direct-input total.

The conceptual answer is:

The PE inputs are not paid at invoice value. They are used as relative weights inside a budget-neutral national allocation system. Direct inputs are scaled by a direct adjustment; indirect PE is scaled by an indirect adjustment; TC-only services may use proxy work rules; and the final payment is further affected by specialty mix, utilization, geographic PE GPCI, final PE budget neutrality, and the annual conversion factor.

That is why the final technical component can look mysteriously lower than the apparent AMA direct PE dollars. The missing step is not just the conversion factor. It is the direct PE scaling adjustment, plus the broader PE budget-neutrality machinery.

###

AMA has a lot of resources about the RUC (RVS Update Committee, RVS-UC, RUC).

I embarked on this to find again a document on the RUC practice expense process - the best one isthe 27p "practice expense" PDF below.

###

Home page

https://www.ama-assn.org/about/rvs-update-committee-ruc

Overview and historical policy PDFs

https://www.ama-assn.org/about/rvs-update-committee-ruc/rbrvs-overview

12 pages: Development of RBRVS

https://www.ama-assn.org/system/files/development-of-the-resource-based-relative-value-scale.pdf

Legislation history 12p

https://www.ama-assn.org/system/files/legislation-creating-the-medicare-rbrvs-payment-system.pdf

History of conversion factor - the SGR period 8p

https://www.ama-assn.org/system/files/history-of-medicare-conversion-factor-under-the-sgr.pdf

Physician work component - 22p

https://www.ama-assn.org/system/files/physician-work-component.pdf

Practice Expense 27p * * *

 https://www.ama-assn.org/system/files/practice-expense-component.pdf

CMS publication: 2026 Rules, as summarized in detail by AMA in November 2025 for 2026

https://www.ama-assn.org/practice-management/medicare-medicaid/medicare-physician-payment-schedule

2026 Rule by AMA 16p

https://www.ama-assn.org/system/files/2026-mpfs-final-rule-summary-analysis.pdf

2026 The Hated Efficiency Adjustment 2p two pager

https://www.ama-assn.org/system/files/cms-2026-physician-payment-proposals.pdf

Home Page: The RUC Committee

https://www.ama-assn.org/about/rvs-update-committee-ruc/rvs-update-committee-ruc

13 page book on RUC process

https://www.ama-assn.org/system/files/ruc-update-booklet.pdf

32 page deck on RUC process

https://www.ama-assn.org/system/files/rbrvs-ruc-process.pdf

RUC MINUTES ARCHIVE  1993-2028 * * *

https://www.ama-assn.org/about/rvs-update-committee-ruc/ruc-recommendations-minutes-voting

Composition of the Committee [specialties and names]

https://www.ama-assn.org/about/rvs-update-committee-ruc/composition-rvs-update-committee-ruc

History of Cancer - Chart Note Terms Not the Same as ICD-10 "Z Codes" Personal-History-Of

 Please research online as needed.   This is a correct medical coding question and requires agentic research and internet research. There are codes for colon neoplasm such as C18.6 or C18.7. There is also a set of cancer codes beginning with "Z" such as perosnal history of malignant neoplasm of large intesting, Z85.038. Here is the question. Does use of a Z code require that the cancer is no longer present? I would think not. If I have a persona lhistory of diabetes, I had diabetes yesterday and still today. I would think Z85.038 by its text would cover both a history of treated and cured cancer ("a personal history...") as well as being inclusive of someone who might have residual, metastatic, or lingering cancer today. Is there an official correct coding position and can it be documented objectivey. 

Examples. 

C18.6 Malignant neoplasm of descending colon 

C18.7 Malignant neoplasm of sigmoid colon 

Z85.038 Personal history of other malignant neoplasm of large intestine

###

I'll research this carefully. This is a specific medical coding question with an objective answer in the ICD-10-CM Official Guidelines for Coding and Reporting.I'm getting clear signals that the official guidelines treat "personal history" differently than the user's intuition. Let me get the authoritative source — the ICD-10-CM Official Guidelines for Coding and Reporting, specifically Section I.C.21 on history (of) codes.Excellent — I have the definitive answer directly from the ICD-10-CM Official Guidelines for Coding and Reporting. Let me grab one more authoritative source to confirm the exact wording and cite the CDC/CMS document.I have a definitive, citable answer. Here it is:

Short answer

Your intuition is wrong on this one, and there's an objective, citable rule. In ICD-10-CM, "personal history of malignant neoplasm" (Z85.-) explicitly requires that the cancer no longer exists and is not being treated. It is not a parallel to "personal history of diabetes" — in coding, "history of" is a term of art, not the everyday English meaning.

The official rule

The ICD-10-CM Official Guidelines for Coding and Reporting, Section I.C.21.c.4 ("History (of)"), states:

"Personal history codes explain a patient's past medical condition that no longer exists and is not receiving any treatment, but that has the potential for recurrence, and therefore may require continued monitoring."

That's the controlling sentence. It is published annually by CDC/NCHS and CMS and is incorporated by reference into HIPAA code-set rules, so it has regulatory weight, not just guideline weight. The American Hospital Association's Coding Clinic and the AHA's published FAQs cite this exact language when adjudicating disputes.

So for your colon example:

• Patient currently has malignant neoplasm of the sigmoid colon, on chemo, or with known residual/metastatic disease → C18.7 (or the appropriate active C-code, plus secondary site codes like C78.- for mets). Z85.038 is wrong.
• Patient had sigmoid colon cancer, was resected, has completed treatment, NED (no evidence of disease), now under surveillance → Z85.038.
• Patient had sigmoid colon cancer, primary tumor resected, but has known liver metastases still present and/or is on treatment → still active cancer; code the metastatic site (e.g., C78.7) and, per Coding Clinic, the primary may still be coded as active if treatment is directed at the disease. Z85.038 is still wrong.

Why your diabetes analogy fails

In plain English, "personal history of X" can mean "X has happened to me, including right now." In ICD-10-CM, the phrase has been given a narrower technical definition by the guideline above. The same logic is why there is no "personal history of diabetes" Z code — diabetes is chronic and ongoing, so you code the active disease (E11.-, etc.). 

The Z85 family exists precisely to flag a resolved malignancy that still matters clinically (surveillance, screening intervals, risk stratification, family counseling).

The companion concept is the distinction between history codes and status codes:

• History codes (Z85, Z86, Z87, etc.) — condition no longer present, but clinically relevant.
• Status codes (Z89–Z99) — a current, ongoing state (e.g., Z90.49 acquired absence of part of digestive tract after colectomy, Z93.- ostomy status). 
• These are current and can coexist with either an active C-code or a Z85 history code.

The "still has cancer" trap and the AHA position

Coders frequently see notes that say "history of colon cancer" when the patient in fact still has measurable disease, is on maintenance chemo, or has known mets. 

Coding Clinic has repeatedly said: do not take the provider's word "history" at face value [for using a Z code]; review the record, and if the cancer is still being treated or is still present anywhere (including metastases), code it as active. Query the provider if unclear. 

The COVID-19 FAQ from AHA quotes the same I.C.21.c.4 language to make exactly this point in a different disease context — a personal history code is not appropriate when the patient is "clearly receiving treatment" for the condition or its residual effects.

The plain-English summary from a coding education source captures it well: "[In ICD10] History codes apply only after the condition has been fully treated and no longer exists." And the corresponding error: "Assigning Z85 (personal history of malignant neoplasm) while the patient is still undergoing active cancer treatment is incorrect."

How to document this objectively

If you ever need to defend the position in an audit, chart review, or payer dispute, the citation chain is:

1. ICD-10-CM Official Guidelines for Coding and Reporting, current fiscal year, Section I.C.21.c.4, "History (of)" — the binding definition. Available as a free PDF on the CDC NCHS site (cdc.gov/nchs/icd/icd-10-cm) and on cms.gov.
2. AHA Coding Clinic for ICD-10-CM/PCS — quarterly publication with case-level adjudications; multiple issues address active-vs-history coding for neoplasms (e.g., metastatic disease with primary resected, patients on hormonal therapy like tamoxifen or aromatase inhibitors, etc.). Coding Clinic is the authoritative interpretive body recognized by CMS.
3. ICD-10-CM Tabular instructional notes at category Z85 itself, which sit under the Z77–Z99 block "Persons with potential health hazards related to family and personal history."

If you want the primary source verbatim, the current Guidelines PDF is published annually around July–August by CDC/NCHS for the upcoming October 1 fiscal year and is the single best document to cite. Section I.C.21.c.4 has carried the same definition for years.

Bottom line

Z85.038 is not a "covers both" code. It is a resolved-disease code by official definition. If any malignant disease is still present anywhere in the body or under active treatment, the active C-code(s) apply, and Z85 is wrong — even if the chart note casually says "history of."

Gerrard's 2024 Paper on Thought Capital for Computational Pathology Applications and Approvals

There are some special complexities and new thought capital required in the field of computational pathology and both scientific validation and regulatory science approval and validation. I understand this article by Gerrard pushes the thinking forward. I also include an example of an actual 24p FDA review (as context or background). Please discuss Gerrard's paper from an innovation and thought capital perspective. use deep thinking or agentic mode as necessary.

Publication here.   (DEN240068 decision here.)

###

###

Bottom line:Gerrard's 2024 Contribution

Gerrard et al. is not just another “AI works on H&E slides” paper. Its real contribution is regulatory-science thought capital: it tries to answer the deceptively hard question, what does analytical validation mean when the “analyte” is not a molecule, not a stain, not a pathologist-recognizable morphology, but an AI-derived patient-level risk output? That is a major conceptual shift. The paper is valuable because it gives computational pathology a bridge from impressive retrospective AI performance toward something closer to a clinical laboratory test discipline.

The FDA review provides useful context because it shows the regulatory endpoint of this line of thinking: ArteraAI Prostate was reviewed as a software-only device analyzing H&E whole-slide images from prostate biopsies to provide 10-year risk estimates of distant metastasis and prostate cancer-specific mortality, intended to assist physician prognostic decision-making alongside standard clinicopathologic factors.

Why Gerrard’s paper matters

The key move in Gerrard is the reframing of analytical validation. In molecular diagnostics, analytical validation usually asks whether the assay accurately and reproducibly measures the target molecule: DNA sequence, RNA expression, protein epitope, methylation signal, copy number, fusion transcript, and so on. In traditional pathology AI, analytical validation may ask whether software detects a recognizable slide-level feature: tumor focus, mitoses, nuclei, Gleason pattern, HER2 staining, Ki-67 positivity, or some other human-interpretable morphology.

ArteraAI Prostate does something different. It uses routine H&E slides and generates patient-level prognostic and predictive outputs. Gerrard explicitly argues that for an H&E AI test using nonspecific stains, the meaningful “biomarker” is not the H&E input itself. It is the algorithm output. The paper states that for an AI algorithm using nonspecific probes such as H&E, “the output of the algorithm rather than the algorithm measured input is the only meaningful biomarker of interest.” It further distinguishes Artera from Paige Prostate: Paige points to slide-level regions of likely cancer, while Artera generates patient-level outputs associated with oncologic endpoints.

That is the innovation. Gerrard is saying, in effect: the biomarker is no longer a thing under the microscope; the biomarker is the reproducible computational inference from the digitized tissue.

The paper’s central conceptual advance: the “output as analyte” doctrine

The most important thought-capital phrase, even if not branded this way, is output as analyte. In ordinary lab medicine, the analyte is something prior to the assay: glucose, troponin, EGFR exon 19 deletion, PD-L1 staining, HER2 amplification. The assay detects or quantifies it.

In Artera-like computational pathology, the assay is not simply detecting an established entity. It is extracting high-dimensional image features, integrating them through a trained model, and producing a clinically meaningful risk estimate or classification. There may be no single pathologist-identifiable feature corresponding to the result. The “thing measured” is therefore not an object waiting to be found. It is a model-defined risk signal, validated against patient outcomes.

That creates a new regulatory-science problem. You cannot validate the test by asking whether the H&E stain binds its target, because H&E is deliberately nonspecific. You also cannot validate it by asking a pathologist to confirm the AI’s morphology, because the AI is not necessarily reporting a named morphology. Gerrard therefore adapts analytical validation around reproducibility of the AI output under relevant preanalytic, analytic, operator, scanner, and tissue-selection conditions.

This is a subtle but important departure from both molecular pathology and classic digital pathology.

The hybrid validation strategy

Gerrard’s method is not invented from whole cloth. It is a hybrid. The authors looked for analogies in two directions: first, AI pathology devices such as Paige Prostate, because they also analyze H&E whole-slide images; second, prognostic molecular tests such as breast cancer recurrence-risk classifiers, because they produce patient-level risk outputs. Gerrard notes that Paige is methodologically similar but intended for slide-level localization, while gene-expression classifiers are clinically similar because they produce algorithmic patient-level prognostic results.

That hybridization is a major form of innovation. In regulatory science, progress often comes not from discovering a wholly new method, but from identifying which older validation concepts are portable and which are not. Gerrard’s answer is roughly:

From molecular diagnostics, borrow the seriousness of analytical validation, reproducibility, precision, and assay implementation.

From digital pathology, borrow concern for scanners, whole-slide image quality, operators, and tissue workflows.

From prognostic classifiers, borrow the idea that the clinically relevant result may be an algorithmic patient-level risk score rather than a visible feature.

From AI/ML software regulation, borrow attention to locked models, deployment environment, data flow, software controls, cybersecurity, and change control.

The FDA review shows this same convergence. It references software submission guidance, qualitative binary-output performance standards, cybersecurity guidance, off-the-shelf software guidance, ISO 14971 risk management, IEC 62304 software lifecycle processes, and usability engineering standards. This is not just pathology anymore; it is a composite field of pathology + software engineering + clinical prediction + laboratory validation + regulatory risk management.

What Gerrard actually validated

Gerrard assessed two AI biomarkers in a clinical laboratory setting: one prognostic algorithm and one predictive algorithm for likely benefit from short-term androgen deprivation therapy. The paper evaluated analytical accuracy, intra-operator reliability, inter-operator reliability, and biopsy-set completeness reliability. The reported analytical accuracy ICCs were high: 0.991/0.993 range for the prognostic algorithm and 0.934 for the ST-ADT algorithm, with strong intra-operator and inter-operator reliability and somewhat lower but still substantial reliability when comparing one core versus three or six cores.

The “biopsy completeness” experiment is especially interesting. In molecular pathology, one thinks about limit of detection, tumor fraction, nucleic acid quality, and extraction sufficiency. In computational pathology, the analogous issue may be: which tissue did the model see? One core versus several cores is not merely a specimen-handling issue. It becomes a biological heterogeneity issue. The Gerrard paper recognizes this: prostate cancer is multifocal and heterogeneous, and the AI result may depend on which core or cores are selected. This becomes a computational-pathology version of sampling error, but with a twist: the image-based AI may also have an operational advantage because it can often use the original diagnostic H&E slide rather than consuming additional tissue.

The FDA review as the more formal regulatory endpoint

The FDA review is narrower and more regulatory in tone. The device is described as analyzing scanned WSIs of H&E-stained prostate needle biopsies using an AI/ML algorithm to provide 10-year prognostic risk estimates for distant metastasis and prostate cancer-specific mortality. The review describes a workflow in which a pathologist has already diagnosed prostate cancer, identifies the WSI containing the highest Gleason score, verifies image quality, uploads the WSI, and later reviews and releases the report.

This workflow matters because FDA is not simply approving an abstract algorithm. It is reviewing a specific clinical system: specimen type, scanner, magnification, user role, image quality controls, report generation, traceability, intended population, labeling, and physician use context. The FDA indication also specifies use with FDA-cleared interoperable scanners already authorized for the device, or additional 510(k)-cleared scanners qualified through a Predetermined Change Control Plan.

That PCCP component is important thought capital in its own right. Computational pathology depends on scanners, compression, color profiles, image formats, tissue processing, and software updates. A static one-time validation paradigm is poorly suited to a field where the ecosystem changes. The PCCP is a regulatory mechanism for controlled evolution: not “anything goes,” but also not “freeze the whole world forever.”

Innovation perspective: what is really new here?

The first innovation is biomarker abstraction. The paper moves from physical biomarker to computational biomarker. The H&E slide is not the biomarker in the way HER2 protein or EGFR mutation is. The biomarker is the AI-generated risk signal derived from the slide.

The second innovation is patient-level computational pathology. Much digital pathology AI has been assistive: find cancer, count cells, segment tissue, quantify staining, flag regions of interest. Artera-like AI instead asks a broader clinical question: what is the patient’s future risk? That moves the field from computer-assisted pathology interpretation toward AI-derived clinical prognosis.

The third innovation is analytical validation without a human gold-standard morphology. This is a profound shift. If the model is trained on outcomes rather than named morphology, a pathologist cannot necessarily say whether the model “saw the right thing.” The validation anchor becomes reproducibility of output and clinical association with outcomes, not visual agreement with a human observer.

The fourth innovation is scanner and workflow realism. Gerrard does not simply say, “The model works.” It asks whether the output is stable across operator, day, scanner/workflow, and tissue sampling. The FDA review goes further by embedding the device in a defined lab workflow and requiring attention to scanner qualification, image quality, software controls, traceability, and labeling.

The fifth innovation is regulatory translation. A research AI model can be impressive and still not be a regulated medical product. Gerrard is about the crossing of that valley: taking AI from publication and retrospective validation into a form that can live inside a clinical laboratory and, eventually, an FDA device file.

The deepest issue: analytical validity and clinical validity are intertwined but not identical

Gerrard’s paper helps separate two concepts that are easily blurred in AI pathology. Clinical validation asks whether the model predicts or stratifies clinically meaningful outcomes. The paper summarizes prior clinical validation in large randomized trial datasets, including validation for distant metastasis and prostate cancer-specific mortality, and a predictive ST-ADT interaction analysis.

Analytical validation, by contrast, asks whether the test produces the intended result reliably when performed in the real-world laboratory process. For AI pathology, this includes scanner behavior, image quality, operator handling, tissue selection, repeat scans, and reproducibility of the risk output.

The conceptual problem is that the “analyte” and the “clinical meaning” are close together. If the algorithm output is the biomarker, and the biomarker is defined by its relationship to outcome, then analytical validity can feel circular unless carefully handled. Gerrard’s solution is pragmatic: use the clinically validated model output as the thing to be reproduced, then stress-test the laboratory workflow for reproducibility. That is not philosophically perfect, but it may be the workable regulatory-science solution.

Why this matters for the broader field

This paper helps computational pathology mature from “AI can find patterns” to “AI can be a laboratory test.” That transition requires a new vocabulary. The field needs to define:

What is the analyte?
For Artera-like tests, it may be the model output.

What is the specimen?
Not just tissue, but tissue plus digitization plus image quality plus scanner environment.

What is the instrument?
Not just the scanner, and not just the software, but the scanner–image–algorithm–reporting pipeline.

What is analytical accuracy?
Not closeness to a molecular truth, but concordance with a locked or reference implementation under defined conditions.

What is reproducibility?
Not just same sample/same result, but same patient-level computational inference despite ordinary variation in scanning, operator, day, and tissue subset.

What is change control?
Not just reagent lots, but scanner qualification, software versioning, AI model locking, file formats, color normalization, cybersecurity, and potentially PCCP-governed updates.

That is why Gerrard’s paper is important. It gives the field a proto-template for how to think.

A critical caveat

The paper is also company-authored and product-specific. It is not an independent standards document. It provides a strong example of how one company framed and executed analytical validation, but it should not be mistaken for a universal answer. The FDA review is also specific to the authorized device, intended use, scanner conditions, labeling, and risk mitigations. FDA’s benefit-risk summary emphasizes that erroneous results or incorrect interpretation could lead to inappropriate management, while also concluding that analytical and clinical validation plus labeling and special controls mitigate those risks sufficiently for Class II De Novo classification.

So the thought capital is real, but it is still early. The field will need independent replication, cross-platform studies, post-market monitoring, clearer approaches to race and subgroup performance, calibration drift management, tissue-processing variability studies, and standards for when a model update requires new validation.

My synthesis

Gerrard et al. pushes the field forward because it recognizes that computational pathology is not simply “digital pathology plus AI.” It is a new diagnostic category in which the clinical signal may be latent, high-dimensional, nonsemantic, and outcome-trained. That forces a new analytical validation logic. You cannot validate it like HER2 IHC, because there is no specific epitope. You cannot validate it like Paige Prostate, because it is not merely localizing cancer on a slide. You cannot validate it exactly like Oncotype or MammaPrint, because the input is not an extracted molecular specimen but a digitized tissue image. And you cannot validate it like ordinary software, because the software output is itself a medical biomarker.

That is the “new thought capital”: computational pathology needs its own regulatory science, and Gerrard offers one of the first serious attempts to build it.

Artera Prostate 0376U Medicare Billing & MACs and Codes

 Artera Prostate has a PLA code, 0376U, several years old.  At least to 2024, there were no paid claims (December 2024).   I do not know if there are paid claims in 2025, but for this essay I assumed not.   I do understand the Artera Prostate test (validated on many many thousands of slides) has NCCN endorsement.

Below is direct Chat GPT research.

###

ESSAY

ArteraAI presents an interesting distinction between coding/pricing success and actual Medicare payment success. The company is appropriately described as Los Altos, California–based: FDA’s De Novo database lists Artera, Inc., 108 1st St., Los Altos, CA, as the requester for ArteraAI Prostate. FDA granted De Novo authorization for ArteraAI Prostate on July 31, 2025, classifying it as a pathology software algorithm device analyzing digital images for cancer prognosis. (FDA Access Data)

But the performing CLIA laboratory for the ArteraAI Prostate Test appears to be in Jacksonville, Florida, not California. Artera’s own test materials identify the clinical lab address as 6800 Southpoint Pkwy, Suite 950, Jacksonville, FL 32216. That places the performing lab under First Coast Service Options, the A/B MAC for Jurisdiction N, covering Florida, Puerto Rico, and the U.S. Virgin Islands. (ArteraAI) And yes — FCSO is First Coast Service Options.

The test has a PLA-type CPT code, 0376U, and CMS established a Medicare CLFS payment rate effective January 1, 2024. Artera publicly described this as a CMS payment-rate decision for the ArteraAI Prostate Test, a clinical diagnostic laboratory test. (ArteraAI) That is meaningful, but it is not the same thing as coverage. A CPT/PLA code plus a CLFS dollar amount means the test can be billed and priced; it does not necessarily mean that a MAC is paying claims.

That is the concerning point: according to the claims-data observation you are making, there appear to be no paid Medicare claims for 0376U in Florida or elsewhere, despite the existence of the code, the CLFS price, the ability to enroll/bill through an NPI, and the absence of a clear published LCD either covering or non-covering the test. In that sense, ArteraAI is not “unsuccessful” at coding or pricing; it is unsuccessful at converting those assets into actual Medicare payment.

This is especially striking because the test is not a speculative early-stage assay. It has FDA De Novo authorization, and Artera has stated that the ArteraAI Prostate Test was included in the 2024 NCCN Guidelines for Prostate Cancer as a predictive test for therapy personalization, with Level IB evidence under the Simon criteria. (ArteraAI) Artera’s clinician-facing materials also describe use across NCCN risk groups, with reported 10-year risks of distant metastasis and prostate cancer–specific mortality. (ArteraAI)

The likely rationale for the Florida lab location is therefore not mysterious. A California performing lab would fall under Noridian/MolDX jurisdiction, with the familiar Z-code/technical assessment pathway. By contrast, a Florida performing lab submits through First Coast Service Options / Jurisdiction N, outside the MolDX MAC structure. That may have looked like a cleaner path for a PLA-priced AI pathology test: obtain the code, obtain the CLFS price, bill through a non-MolDX MAC, and avoid a California MolDX technology-assessment bottleneck.  (BQ - At least historically, FCSO paid codes 81401-81408 Tier 2, and PLA, fairly easily.).

But if the claims data show zero paid claims, the strategy may have produced only a formal pathway, not a payment pathway. The result is a reimbursement limbo that is highly relevant for innovators: FDA clearance, NCCN recognition, a PLA code, and a CLFS price still may not produce Medicare dollars when no MAC has articulated a coverage position and claims simply do not pay.

Genomic Health: Looking Back at the 2006 Harvard Case Study

 In 2006, there was a Stanford Case Study about the Genomic Health test - before adoption, before coding, before coverage or guidelines.   (You can still buy the PDF via Harvard)

Here's a combination of TLDR and further strategic analysis of then-and-now.

##

TLDR

The 2006 Stanford case is striking because Oncotype DX basically became the thing the case hoped it would become: a high-value, clinically validated, tumor-genomic test that changed adjuvant chemotherapy decision-making in ER+/HER2− early breast cancer. The case correctly identifies almost every major strategic hinge: clinical validation before broad adoption, physician skepticism, payer evidence demands, premium pricing, CPT/reimbursement friction, centralized CLIA laboratory control, and the need to treat diagnostics more like therapeutics than like commodity lab tests. The initial pivotal study already showed that the 21-gene recurrence score outperformed age, tumor size, and grade as a predictor of distant recurrence, with low-, intermediate-, and high-risk groups showing 10-year distant recurrence estimates of 6.8%, 14.3%, and 30.5%.

What feels “frozen in amber” is the 2006 optimism that genomics would rapidly reorganize medicine into a broad information-first, high-margin diagnostic economy. That happened selectively, not universally. Oncotype DX was a canonical success, but many later genomic diagnostics struggled with evidence, reimbursement, adoption, and differentiation. The case also sits before the later era of TAILORx, RxPONDER, NCCN/ASCO embedding, PLA codes, MolDx/Z-codes, ADLT/PAMA, liquid biopsy, MRD, AI pathology, and FDA-LDT turbulence. So it reads both as a remarkably prescient founding document and as a fossil from an era when “personalized medicine” still sounded like a moonshot rather than a reimbursement trench war.

What the case is really about

The surface story is Genomic Health launching Oncotype DX, but the deeper story is the attempt to create a new category: diagnostics as high-value clinical decision tools, not low-margin lab commodities. Kim Popovits’ opening quote frames the entire case: the system had to recognize diagnostics as having a value proposition comparable to therapeutics, or Genomic Health’s whole model would be jeopardized.

That was the right fight. The company was not merely selling “a 21-gene panel.” It was selling a decision intervention at a specific moment: a woman with early-stage ER+ breast cancer, after surgery, facing the chemotherapy decision. The case repeatedly emphasizes that chemotherapy was expensive, toxic, and of limited absolute benefit for many patients; the unmet need was not more molecular information in the abstract, but better triage of who needed chemo and who could avoid it.

The case is also unusually modern in its evidence strategy. Genomic Health chose not to rely on the fact that CLIA allowed market entry as a “homebrew”/LDT. Instead, the team deliberately used something closer to a drug-development evidentiary blueprint: analytical rigor, archived FFPE tissue, blinded validation, prospectively defined endpoints, and high-profile oncology collaborators such as NSABP. That decision looks very prescient. The case says physicians had been “burned” by uncertain new tests and would not adopt a genomic diagnostic without clinical validation.

What was prescient

The biggest prescient point is that clinical utility would be the moat. The case understood that the hard part was not measuring RNA from paraffin blocks; it was proving that the result changed a real clinical decision. That lesson became the central doctrine of molecular diagnostics reimbursement for the next twenty years.

Second, the case anticipated the now-familiar idea that a diagnostic can be worth thousands of dollars if it prevents overtreatment or undertreatment. Genomic Health’s payer research found that once a test crossed the “expensive” threshold, payers were less sensitive to whether it cost $1,500 or $4,500, provided the test had convincing clinical value and validation. That is an early articulation of value-based diagnostics pricing.

Third, the case correctly saw that workflow fit matters. Oncotype DX used ordinary FFPE tumor tissue, required no special collection, could be sent overnight, and returned results within the two-to-three-week post-surgery chemotherapy decision window. That is a huge adoption advantage. The molecular test was radical, but the specimen logistics were almost boring—and that was part of the genius.

Fourth, it saw that physician education and payer education had to be built together. The case describes a reimbursement dossier, medical-director education, private-payer contracting, exception claims, ABNs, and the need to protect physicians from being financially burned by nonpayment. This is basically the modern playbook for high-value molecular diagnostics, only described before that playbook had become standard.

And fifth, the bet that Oncotype DX could become a durable platform asset proved correct. The later TAILORx trial helped settle the troublesome intermediate-risk category in node-negative HR+/HER2− disease, showing that many women with midrange recurrence scores could avoid chemotherapy without inferior outcomes. (New England Journal of Medicine) RxPONDER later extended the clinical story into selected node-positive patients, especially showing that postmenopausal women with 1–3 positive nodes and recurrence scores 0–25 could likely avoid chemotherapy, while premenopausal women appeared different. (New England Journal of Medicine)

What looks frozen in amber

The case is very 2006 in its language of “the genomics revolution.” It imagines a broad transition in which genomic information might come to dominate therapeutics as the highest-value layer of medicine. That was intellectually plausible, and in some niches correct, but it overgeneralized. Genomics became essential in oncology, rare disease, reproductive genetics, infectious disease, and transplant/MRD-style monitoring—but it did not broadly displace therapeutics as the dominant economic engine of biomedicine.

The case also predates the modern reimbursement bureaucracy. There is no PLA code universe, no MolDx/Z-code architecture, no PAMA shockwave, no ADLT pathway, no elaborate LCD evidentiary machinery, no NCD 90.2, no FDA LDT rulemaking drama. Its CPT discussion is charmingly early: should Genomic Health stack existing codes, or use a miscellaneous code and suffer manual review? Today, that same question would immediately branch into PLA strategy, gapfill/crosswalk, MolDx technical assessment, Medicare contractor jurisdiction, private payer policy, NCCN/ASCO status, and possibly FDA status.

The case is also frozen before liquid biopsy and MRD changed the mental model of oncology diagnostics. Oncotype DX is a tissue-based, one-time, adjuvant decision test. The modern frontier is often serial, blood-based, dynamic: therapy selection from ctDNA, recurrence monitoring, molecular response, tumor-informed MRD, fragmentomics, methylation, and AI-enhanced multimodal risk. Oncotype was an information product, but it was still a relatively static one.

Early misfires or overextensions

The most obvious overextension was the belief that the information layer might become more valuable than the drug layer. That is true in a rhetorical or clinical-decision sense in some cases, but not in market capitalization terms. Drugs did not become commodities. Immuno-oncology, targeted therapies, cell therapy, GLP-1s, and biologics all proved that therapeutics remained the giant economic center of gravity.

A second overextension was the implied scalability of the Oncotype model. The story can make it sound as though once you build the proof engine, many similar high-value genomic tests will follow. In reality, Oncotype DX was almost ideally positioned: common disease, painful binary decision, expensive/toxic avoidable treatment, large clinical trial archives, cooperative oncology networks, motivated patients, and a clear economic offset. Many later tests lacked one or more of those ingredients.

A third “misfire,” if that is not too harsh, was underestimating how long clinical adoption and reimbursement normalization would remain bespoke. The case correctly anticipates an 18–24 month reimbursement struggle, but the broader field did not evolve into a clean, rational marketplace for diagnostics. Even today, a great diagnostic can spend years in a swamp of coding, coverage, evidence thresholds, LCDs, private-payer variation, prior authorization, and physician inertia.

Bottom line

This case is not embarrassing in hindsight. Quite the opposite: it is one of those rare business-school cases where the central thesis largely came true. Oncotype DX became a standard-of-care genomic decision tool, later reinforced by large prospective evidence such as TAILORx and RxPONDER, and Genomic Health itself became valuable enough to be acquired by Exact Sciences in 2019. (Exact Sciences)

But it is also a perfect amber specimen from the first genomics boom: optimistic, therapeutic-disruption-minded, and confident that high-value diagnostics would force the system to recognize their worth. The more nuanced 2026 reading is: yes, but only when the test owns a decision, fits the workflow, has unusually strong evidence, and survives the reimbursement machinery. Oncotype DX did. Many others did not.

###

Addendum - Financial Architecture is Destiny

The visible controversy in 2004–2006 was, “How can a lab test cost $3,000?” But the deeper shift was not merely a price increase; it was a transformation in the financial structure of diagnostics. Traditional reference laboratory economics, exemplified by Quest or Labcorp, were built around high-throughput operations: heavy specimen logistics, high variable costs, broad menus, modest margins, and very little product-specific R&D as a percentage of revenue. The lab’s value was scale, efficiency, and fulfillment. Genomic Health was trying to invert that model. It wanted a diagnostic business with the economics of a biotech product company: substantial upfront development, clinical validation, publications, medical education, payer dossiers, a specialized sales force, and enough gross margin to fund the next product. The Stanford case says this explicitly in business-school language: traditional diagnostics were “high-volume, low-margin,” with little room for R&D, while Genomic Health wanted “high-value, information-rich diagnostics” that could command premium pricing and support ongoing research.

In that sense, Myriad’s BRCA test and Genomic Health’s Oncotype DX were not simply two expensive early genomic tests. They were two early attempts to move diagnostics out of commodity lab economics. Myriad had the patent-protected version of the model: old-stack molecular coding could be assembled into a roughly $3,000 service, defended by intellectual property and clinical distinctiveness. Genomic Health mirrored that price point but justified it less through gene patents and more through clinical utility, proprietary validation, brand, evidence development, and physician/payer education. Oncotype’s price was not only payment for the marginal cost of running RT-PCR on a paraffin block. It was payment for a new kind of diagnostic enterprise: one with perhaps 20% COGS and 20% R&D, rather than the classic lab model of 60% COGS and essentially 0% R&D. The test price therefore carried the burden of funding the whole innovation system around the assay.

That point was easy to miss because the payer and coding systems were still asking a primitive question: what technical steps were performed? Genomic Health was asking a different question: what clinical decision did the information change, and what treatment costs or toxicities did it help avoid? The case captures this tension in the CPT discussion: the company could stack existing codes and reach only about $1,700, which it believed failed to reflect the test’s clinical value, production cost, and development investment; or it could use a miscellaneous code and defend the value claim manually on each claim.

The historical significance, then, is not just that Oncotype DX was “expensive.” It was an early assertion that some diagnostics should be financed like innovation products, not like commodity lab procedures. The market eventually accepted that argument in selected cases, but only under demanding conditions: a defined clinical decision, strong evidence, trusted guidelines, payer education, and a credible story that the price funds real clinical innovation rather than merely exploiting coding arbitrage.