The Stark Law and Pathologist-Owned Labs...and IHC Orders

We all know that the Stark law prohibits doctors from referring lab tests to labs they have an ownership in.   How about the daily occurence of a pathologist owning his ab business, and every hour referring something to it - IHC stains on a breast cancer case, special stains for unexpected possible organisms, and so on.  Why don't these orders "to a lab the referring doctor has an ownership interest in" trigger Stark?

It may seem like a puzzle, and the answers are buried in Stark law definitions and details.

NOTE: The essay below is AI-generated, and should NEVER be construed as actual legal opinion or advice.  Take it as the current ability of AI to understand a situation and write about it.  For legal answers to actual situations, get a lawyer.

###

The Stain Order That Wasn’t a Stark Referral

A case study in pathology, ownership, and the meaning of “referral”

Dr. Dan Michaels had barely settled into Oak Street Pathology Group when a colleague brought him a question from the golf course.

“Everyone knows Stark prohibits doctors from sending lab work to a lab they own,” the colleague said. “So how can we own this laboratory and order stains on our own cases all day long? We do the work, bill Medicare, and Medicare pays. Isn’t that exactly what the law was meant to stop?”

The example was concrete. Michaels had received a breast biopsy for consultation. After examining the tissue, he ordered additional levels, special stains, and immunostains. Oak Street’s laboratory performed the work. The group billed Medicare, which paid the claims. Seven pathologists owned the practice and shared equity in the lab.

The apparent contradiction seemed hard to escape:

1. Stark generally prohibits a physician from referring Medicare patients for designated health services to an entity with which the physician has a financial relationship, unless an exception applies.
2. Clinical laboratory services are designated health services, and pathologists commonly own or invest in the laboratories where pathology work is performed.
3. Pathologists routinely order additional work on referred specimens, and their laboratories bill for it.

Michaels tried Google. The broad summaries he found repeated the general prohibition but did not explain the pathology workflow. 

His nephew Brad, a law student writing an honors thesis on physician self-referral, took the question to two health-law professors. Two months later, Brad returned with a short answer and a long regulatory history.

The answer was in the word referral. The Stark statute contains a specific rule under which certain orders by a pathologist do not count as referrals at all. That is different from saying the pathologist’s ownership interest is irrelevant, and it's also different from relying on the familiar in-office ancillary services exception.

First, sharpen the premise

Stark does not simply say, “A physician may never order a laboratory test from a lab the physician owns.” It generally prohibits a physician from making a referral for a Medicare-covered designated health service to an entity with which the physician or an immediate family member has a financial relationship, unless a statutory or regulatory exception applies. Ownership can create the financial relationship; the prohibited referral is the other essential part of the analysis.

The law defines clinical laboratory services broadly enough to include pathological examinations. Thus, the issue cannot be dismissed by saying that pathology stains are outside Stark. They may be designated health services. The question is whether the pathologist’s order is a “referral” within the law’s definition.

The statute answers that question for a defined class of specialist services. Under Social Security Act § 1877(h)(5)(C), a pathologist’s request for clinical diagnostic laboratory tests and pathological examination services is not a referral when the services are furnished by or under the supervision of the pathologist pursuant to a consultation requested by another physician. The statute names parallel provisions for diagnostic radiologists and radiation oncologists. www.govinfo.gov

The implementing regulation is 42 C.F.R. § 411.351, definition of “Referral,” paragraph (2). In the current text, the request must result from a consultation initiated by another physician; the consultation may be requested from a particular pathologist or from an entity with which the pathologist is affiliated. The tests must be furnished by or under the pathologist’s supervision, or under the supervision of a pathologist in the same group practice. www.law.cornell.edu

So, in the ordinary referred-specimen case, the sequence is:

Treating physician requests pathology consultation → pathologist evaluates the case → pathologist orders stains or other pathology work needed for the examination → the pathologist or another pathologist in the group supervises the services.

When those conditions are met, the pathologist’s order is excluded from the definition of a Stark referral. The fact that Oak Street’s laboratory performs and bills for the stains does not, by itself, turn that excluded order into a prohibited referral.

“Consultation” is a legal term here

The rule does not require a pathologist to see the patient or conduct a conventional bedside consultation. For Stark purposes, § 411.351 defines a consultation as a professional service in which another physician requests the pathologist’s opinion or advice about a specific medical problem; the request and need are documented in the record; and the pathologist prepares a written report for the requesting physician. www.law.cornell.edu

That definition was shaped in rulemaking partly because the ordinary Medicare consultation model fit pathology poorly. A pathologist usually evaluates a specimen rather than taking a patient history or performing a physical examination. CMS ultimately said that the Stark consultation definition does not require either step. It is a Stark-specific definition and does not change Medicare coverage or payment rules for services billed as consultations. www.federalregister.gov

The original physician’s role matters. CMS described the protected specialist as responding to a consultation while the requesting physician remains involved in deciding how to proceed with the patient’s care. If the specialist instead takes over care, or if a patient simply walks in without a consultation initiated by another physician, the special definition may not apply. The pathologist’s subsequent orders then need to be analyzed under the ordinary Stark rules and any applicable exception. www.govinfo.gov

How the rule reached Oak Street

The statute came first

The pathologist language is not a CMS-created loophole. Congress placed the specialist provision in the Stark statute. The 1995 rulemaking was the first major regulatory step that incorporated the clinical-laboratory provisions into 42 C.F.R. § 411.351. Later statutory amendments expanded the corresponding language for radiology and radiation oncology, and CMS incorporated those changes through subsequent rulemaking. www.federalregister.gov

The 1995 final rule confronted the ownership problem directly

In the 1995 final rule, commenters asked CMS about a pathologist who provided professional services to one laboratory but sent specimens to a second laboratory in which the pathologist had a financial interest. They also asked whether the rule should protect a pathologist who recommended additional laboratory work to the attending physician.

CMS drew a practical line. If the pathologist performed or supervised the laboratory testing at the second lab, the pathologist’s request was not a referral under the statutory provision. If the pathologist sent the work to a lab in which the pathologist had a financial interest but did not perform or supervise the services, CMS said the referral could be prohibited unless another exception applied. CMS separately said that a pathologist’s recommendation for more testing was not itself a referral: the attending physician ultimately decides whether the testing is needed and where to order it. www.govinfo.gov

The rulemaking therefore addressed the kind of two-laboratory arrangement that can look superficially like Oak Street’s. The deciding fact was not simply that the pathologist owned part of a lab. It was whether the request arose from another physician’s consultation and whether the pathologist performed or supervised the service as the law requires.

The 2001 final rule defined the consultation boundary

CMS proposed a broader Stark framework in 1998 and received comments that the consultation concept was unclear. Commenters questioned whether a consultant had to take a history and examine the patient, and how to distinguish consultation from a transfer of care.

In the 2001 Phase I final rule, CMS adopted the three central criteria now reflected in § 411.351: another physician requests an opinion or advice about a specific medical problem; the request and need are documented; and the consultant provides a written report. CMS expressly rejected a patient-history or physical-examination requirement for Stark consultations, recognizing that pathologists rarely see patients. www.federalregister.gov

CMS also emphasized that the statutory protection was narrow: it applied to the named specialists and named categories of services. It did not extend the carve-out to all physicians who order diagnostic tests.

The 2007 rule clarified group supervision and walk-in cases

The 2007 Phase III final rule clarified that qualifying services could be supervised by another pathologist in the same group practice. It also rejected a request to extend the carve-out to walk-in patients. CMS reasoned that the consultation provision assumes another physician has already determined that the patient likely needs the specialist’s services. That rationale is absent when the patient initiates the encounter. www.govinfo.gov

CMS further confirmed that the consultation exclusion can cover technical components ordered by hospital-based pathologists when the other requirements are met. But it declined to extend the exclusion to every physician employed by the same hospital. The agency pointed instead to other Stark exceptions, such as the bona fide employment exception, where appropriate. www.govinfo.gov

Why this is not just the in-office ancillary exception?

The pathologist provision and the in-office ancillary services exception can lead to similar practical outcomes, but they work differently.

The pathologist rule is in the definition of “referral.” A qualifying request is not a referral under Stark in the first place. The in-office ancillary services exception, at § 411.355(b), assumes there is a referral and protects it if detailed requirements are met, including requirements concerning who furnishes or supervises the service, location, and billing.

CMS also distinguished the supervision standards. The pathologist provision requires “supervision” as defined by applicable Medicare coverage and payment requirements or relevant health-and-safety standards. The in-office ancillary exception has its own supervision requirements. In other words, one cannot assume the pathologist carve-out is merely an application of the ordinary in-office exception, or that every arrangement qualifying under one automatically qualifies under the other. www.govinfo.gov

What Brad’s memo says about Dr. Michaels

On the facts as presented, Michaels receives a specimen for pathology consultation from another physician. He determines that additional stains are needed to complete the pathologic examination. Oak Street’s laboratory performs them under his supervision, or under the supervision of another pathologist in the same group. The group bills Medicare for the services.

The pathologist’s request fits the statutory and regulatory exclusion from “referral,” assuming the consultation and supervision conditions are documented and satisfied. The answer is not that Stark has no application to pathology, nor that ownership is irrelevant. The answer is that the qualifying pathologist order is specifically excluded from the Stark definition of a referral.

The boundary cases remain important. A walk-in patient, testing outside the pathology consultation, an order sent to a lab without the required supervision, or a referring physician’s own financial relationship with the lab may change the analysis. The pathologist provision also does not resolve separate questions under the Anti-Kickback Statute, state self-referral laws, Medicare billing rules, or payer contracts.

Sidebar: Is the carve-out limited to “diagnostic” tests?

The statute uses the phrase “clinical diagnostic laboratory tests and pathological examination services,” and the regulation refers to clinical diagnostic laboratory tests and pathological examination services. But the provision is not written as a special exemption only for conventional histology, special stains, or immunostains, nor does it separately distinguish diagnostic tests from prognostic tests or genomic profiling.

The relevant categories are the services named in the statute and regulation. The Stark regulation defines clinical laboratory services to include pathological examinations and ties the category to the applicable CPT/HCPCS code list. Therefore, a prognostic or comprehensive genomic profiling service is not automatically in or out merely because it is called “prognostic” or “CGP.” The precise service and code, whether it is a designated health service, and whether it is part of the pathologist’s qualifying consultation must be examined. The label alone does not resolve the Stark question. www.law.cornell.edu

A cold corner of Stark law

This is a technically important but relatively quiet corner of Stark. The statutory carve-out dates to the law’s early history; CMS incorporated it into regulation in 1995, clarified “consultation” in 2001, and addressed group supervision and edge cases in 2007. The rule has not suddenly appeared to excuse a new business model. It reflects a long-standing distinction between a physician steering a patient to a financially connected lab and a pathologist completing the examination requested by the treating physician.

For Dr. Michaels, the golf-course puzzle is solved: the stains are not “referrals” under Stark when the case arises from a qualifying consultation and the pathologist or a pathologist in the group furnishes or supervises the work as the rule requires.

Primary sources

• Social Security Act § 1877, 42 U.S.C. § 1395nn — statutory referral definition and specialist provision.
• 42 C.F.R. § 411.351, Definitions — current definitions of “clinical laboratory services,” “consultation,” and “referral.”
• 42 C.F.R. § 411.355(b), In-office ancillary services — separate in-office ancillary services exception.
• 1995 final rule, 60 Fed. Reg. 41914 (Aug. 14, 1995) — initial regulatory incorporation of the clinical laboratory referral rules and CMS responses to comments on pathology arrangements.
• 1998 proposed rule, 63 Fed. Reg. 1659 (Jan. 9, 1998) — proposed Stark framework and discussion of consultations and supervision.
• 2001 Phase I final rule, 66 Fed. Reg. 856 (Jan. 4, 2001) — final consultation definition and responses to comments.
• 2004 Phase II interim final rule, 69 Fed. Reg. 16054 (Mar. 26, 2004) — next stage of Stark implementation.
• 2007 Phase III final rule, 72 Fed. Reg. 51012 (Sept. 5, 2007) — group-practice supervision, walk-in patients, and hospital-based technical components.

AI Reviews: Three-Day Value Based Cancer Care Conference (VBCC), NYT October 2026

 

AVBCC 2026 and the work of making cancer care deliver value

The 2026 summit of the Association for Value-Based Cancer Care, scheduled for October 7–9 at the New York Athletic Club, presents an oncology system confronting a widening gap between scientific possibility and practical delivery. Its detailed agenda connects drug pricing, community practice economics, artificial intelligence, patient access, and the difficult task of measuring benefit. Across three days, a recurring question emerges: what must change so that better science reliably produces better care? This report examines the program’s themes and inferred priorities, then considers how the conference might evolve by 2029 as today’s proposed solutions face tests of implementation and accountability. AVBCC

GUEST AUTHOR:  Chat GPT 6 Sol 6.1 High.

This analysis uses the supplied October 2 agenda. The meeting is a no-media event; the report describes scheduled topics and draws inferences from their organization, without reporting speakers’ remarks or session conclusions. File: AVBCC 2026 Agenda 10.2

Payment reform reaches into the foundations of oncology practice. Drug economics runs through the program, connecting Wednesday’s sessions on Medicare negotiation, rebates, and outcome contracts with Thursday’s discussions of price transparency, biosimilars, supply channels, and prescribing incentives. Friday’s examination of practices’ dependence on drug margins brings these issues back to care delivery. Read together, these sessions suggest a concern broader than the price of any particular medicine: how to finance oncology services when the economics surrounding drug acquisition and reimbursement change.

That distinction matters. A lower drug price and a sustainable care system are related objectives, but achieving one does not automatically achieve the other. The agenda repeatedly raises the possibility that changes intended to improve affordability could destabilize the revenue supporting clinical infrastructure. The inferred priority is to make those dependencies visible and develop payment arrangements that support the actual work of care. The program does not establish that an era beyond average sales price payment has arrived; it asks how stakeholders should prepare for substantial changes to familiar economics.

Outcome contracts require agreement on who can deliver the outcome. Sessions on pharmaceutical contracting, legal structures, data limitations, and commercial use cases recur across Wednesday and Thursday. Their accumulation suggests frustration with the distance between announcing a value-based agreement and operating one. A contract needs a measurable endpoint, reliable data, an attribution method, and parties able to influence the result. The agenda’s separate examination of who owns the outcome makes this problem explicit.

The underlying difficulty is shared responsibility. A manufacturer supplies a therapy; clinicians select and manage it; patients need access, support, and the ability to continue treatment. A payment arrangement that assigns responsibility without accounting for these dependencies can create disputes rather than improvement. The scheduled contracting examples, including a private-sector application of an Oncology Care Model approach, offer a useful counterweight to broad forecasts. Their presence suggests a program interested in operational experience, although the agenda alone cannot establish whether any arrangement succeeded or produced net savings.

AI is being considered as infrastructure across the care journey. The breadth of AI programming is striking. Wednesday covers strategy, investment, payer applications, clinical pathways, and the movement from individual tools to practice infrastructure. Thursday extends the discussion into revenue cycle operations, specialty pharmacy, human judgment, and longitudinal patient data. Friday combines a demonstration showcase with the question of whether better measurement can make cancer care genuinely value-based.

This distribution suggests that AI’s perceived role is expanding beyond assistance with isolated tasks. The program explores whether it can connect treatment planning, monitoring, reimbursement, and evidence generation. That is a larger organizational proposition than automating documentation.

It also creates a demanding test of value. Faster billing, more complete records, and better symptom control are distinct achievements; evidence for one does not establish the others. The agenda’s attention to implementation, return on investment, security, and human partnership points toward scrutiny of the whole workflow. A reasonable inferred priority is to evaluate what changes after a tool is deployed, including the work it creates for clinicians and staff.

Measurement connects the technology discussion to payment reform. Wednesday’s review of the Oncology Care Model and Enhancing Oncology Model and Friday’s session on AI and digital quality measurement form a thread across the summit. The latter explicitly asks whether electronic health record data, electronic patient-reported outcomes, and oncology interoperability standards can make performance clinically meaningful, auditable, and actionable.

The important distinction is between observing expenditure and understanding care. Claims can describe paid services, but the questions raised elsewhere in the program concern treatment toxicity, delayed access, adherence, appropriate testing, and patient experience. Those issues require more clinical context.

Better measurement would have to guide a response: identifying a deteriorating patient early enough to intervene, detecting a missed testing opportunity, or determining whether a care model improves outcomes. The agenda presents measurement as a possible explanation for the limitations of earlier approaches. That remains a hypothesis, rather than a demonstrated account of why value-based oncology has struggled. Payment design, staffing, and responsibility still need to align with whatever the measures reveal.

Community oncology is the recurring test of whether innovation can scale. The program returns to community delivery through cell and gene therapies, radiopharmaceuticals, bispecific antibodies, oral medicines, molecular testing, and medically integrated pharmacy. The recurring concern is that scientific capability can advance faster than the infrastructure needed to use it safely and consistently.

Across these sessions, access becomes a practical question involving site readiness, staffing, financing, monitoring, distribution, and coordination. Approval and coverage are important milestones, but they do not complete that chain. A therapy must reach an appropriate patient through a service capable of managing it.

The sessions on technology-supported toxicity management and anticipatory monitoring connect directly to this problem. Moving treatment closer to home requires dependable escalation and clinical response, alongside data collection. Advanced-practitioner reimbursement and Friday’s physician compensation discussion add the workforce dimension: expanding access depends on paying and organizing the people who perform that work. The inferred priority is to build delivery capacity alongside therapeutic innovation.

Precision medicine faces both an implementation test and an evidence test. Wednesday’s session on biomarker testing for clinical trial eligibility, Thursday’s genomics-first care discussion, and Friday’s molecular medicine case studies identify failures along the route from specimen to treatment. Testing must be ordered, financed, completed, interpreted, and returned in time to affect a decision. Tissue insufficiency and fragmented workflows can interrupt that route before a promising result becomes useful.

The agenda also questions whether increasingly sensitive information necessarily improves care. Its examination of molecular residual disease surveillance asks about clinical utility and spending. The screening session sets out an evidence chain connecting detection to changed management, patient benefit, and possible harms.

These topics belong together. Appropriate testing can be underused while other applications remain insufficiently supported. The useful question is which test, for which patient, at which decision point, with what consequential action. Clinical trial access and routine-care evidence generation extend that inquiry beyond testing itself. The program’s separate attention to international research competition suggests concern about the system’s capacity to generate future evidence as well as apply existing knowledge.

Transparency is also a debate about power. Employers, benefit consultants, pharmacy benefit managers, group purchasing organizations, distributors, management services organizations, and risk-bearing care partners receive sustained attention. Sessions on ownership, direct contracting, intermediary value, and utilization management suggest a changing map of who controls cancer care and who is accountable for its results.

The tension is that additional organizations may supply capabilities practices need while also creating additional costs, contractual restrictions, or administrative layers. The agenda raises both possibilities. Its discussions of prior authorization alternatives and pathway organizations sharpen the question: if one form of utilization control recedes, which decisions move elsewhere, and who can challenge them?

Data adds another dimension. Wednesday’s pathway-to-value session explicitly considers whether the ability to demonstrate outcomes could favor larger networks over smaller practices. An inference follows: the same infrastructure that enables value contracts could make scale a condition of participation. Shared standards and accessible services may therefore matter to competition and practice independence, as well as technical interoperability.

The patient perspective broadens what counts as value. Patient-focused sessions appear throughout the program, from Wednesday’s assessment of the patient’s situation and unequal access to insider knowledge to Friday’s examination of the gap between institutional ratings and lived experience. Second opinions, financial assistance, survivorship, geriatric assessment, and site-of-care decisions make that perspective concrete.

Together, these topics challenge a definition of value confined to drug selection or episode spending. Timely expertise, manageable symptoms, continuity after treatment, and care suited to functional status can materially shape the patient’s experience. The prevention session on obesity and GLP-1 therapies also extends the program’s horizon toward future cancer burden, while posing scientific questions rather than settling them.

Patient-centeredness becomes most consequential when it changes an operating decision: which service receives funding, what triggers intervention, how treatment intensity is chosen, or whether a proposed cost saving introduces a new access barrier. The agenda’s repeated return to these issues suggests that affordability and patient benefit need to be assessed together.

Policy implementation receives attention alongside policy ambition. Scheduled CMS leadership discussions and multiple sessions on Medicare arrangements, drug pricing, and commercial insurance place government policy throughout the summit. The Medicare Part B effectuation working-group town hall is especially revealing because it addresses the machinery of implementation. Related 340B sessions examine claims identification, verification, financial transactions, and reconciliation.

These subjects show why policy design and policy delivery need separate examination. A pricing framework can leave difficult questions about acquisition costs, payment timing, information exchange, and disputes. The agenda’s inclusion of a working-group report and a regional collaborative readout suggests interest in organized work beyond the annual meeting. It does not establish that recommendations were adopted or problems resolved. It does indicate that the program makes room for the operational details through which policy becomes a daily reality for practices and patients.

A 2029 conference could put results under greater scrutiny. The following possibilities are projections from the 2026 agenda, rather than announced plans or predictions of particular policy outcomes.

AI could move from demonstrations to comparative evaluation. A more mature program might ask which deployments reduced treatment delays, improved toxicity management, or saved staff time after accounting for oversight and integration. Sessions could compare sustained results across settings and examine failures, model changes, and the ability to audit recommendations. Demonstrations would remain useful, but would sit beside evidence of clinical and operational performance.

Outcome contracting could be judged through completed experience. Instead of concentrating on contract design, the conference might examine settlement results, attribution disputes, administrative costs, and patient outcomes. A crucial question would be whether savings survive after paying for the data, navigation, monitoring, and clinical infrastructure used to produce them. Contracts that endure would provide more informative examples than announcements alone.

Community delivery could become a measurable system capability. The discussion might move toward the proportion of eligible patients actually receiving complex therapies, geographic access, treatment delays, and safety across sites. That would allow scrutiny of whether community expansion improves access and lowers total costs while maintaining quality. Readiness assessments could become more specific about staffing, response capacity, and financial exposure.

Diagnostic debates could become more indication-specific. Molecular surveillance and screening sessions might distinguish applications supported by evidence of useful changes in management from those still requiring validation. Genomics-first care could be evaluated through testing completion, turnaround, treatment selection, and trial access. Stronger evidence could justify expansion in some settings and more selective use in others.

Practice sustainability and patient benefit could be evaluated together. A 2029 agenda might compare payment arrangements that explicitly finance navigation, monitoring, survivorship, and multidisciplinary care. It could also examine whether shared data infrastructure enables smaller practices to participate or whether the cost of demonstrating value increases concentration. Patient-facing measures would help assess what these organizational changes accomplish.

The strongest inference from the 2026 program is that oncology’s value problem spans several connected systems. Useful measurements need a clinical response; that response needs staff and financing; contracts need dependable data and defensible responsibility. By 2029, the conference could have a more concrete basis for judging whether these pieces work together. Its most informative reports would show who received better care, what made the improvement possible, and whether the arrangement can last.

Source and organization links: This report is based on AVBCC 2026 Agenda 10.2.docx, dated October 2, 2026. See the Association for Value-Based Cancer Care, the 2026 conference page, and the online agenda. The supplied agenda is the source for the session analysis; the online version may differ.

Winds of Change: Triangulating the Roles of Pathologists, Radiologists, Oncologists in an AI World

(Guest Author: Chat GPT 6, using Sol 6.1 High.) 

We previously did a deep dive article on Socarraz & Lennerz, a vision for fundamental changes in pathology (here).

Today, we're revisiting Socarraz & Lennerz but in a new context: it joins a triangle populated by pathologists, oncologists, and radiologists, using very recent articles predicting their winds of change.

##

Woodcock et al: A Plan to Turbocharge National Precision Oncology

 https://pubmed.ncbi.nlm.nih.gov/42842867/

(Chat GPT; article is open access.)

 

Former FDA Leader Woodcock Calls for National Effort to Close Precision Cancer Medicine’s Prediction Gap

New commentary argues that matching tumors to drugs is only a starting point—and that federal investment is needed to determine which patients will actually benefit.

Precision cancer medicine has delivered some remarkable treatments. But identifying a tumor’s molecular abnormality still too often leaves a crucial question unanswered: Will the drug selected for that abnormality help this particular patient?

A group led by former Food and Drug Administration official Janet Woodcock is calling for a national effort to close that gap. In a commentary published October 7 in the Journal of Clinical Oncology, the authors urge Congress to support a coordinated infrastructure linking cancer trials, molecular testing, patient specimens and treatment outcomes.

Brief Blog: How Whole Slide Imaging Codes Fared in PAMA

Header: I spot-checked 9 PLA codes that include whole slide imaging (sometimes naming computer interpretation) and only 1 of 9 has PAMA repricing data.  (2 were ADLTs).  

###

This is a quick-and-dirty blog.  I was wondering how whole-slide-imaging slides fared in the PAMA pricing process this fall.  I had a list of about 9 codes I'd tracked as WSI codes.  (In some cases, it's hard to tell if the biomarker slides are newly produced for the coded service or were pre-existing (88342).

Two of them are ADLT codes (Castle 0108U and Prelude 0295U), which don't show up in PAMA data at all.

Of the other codes, 0376U dropped from $706 today to $692 in PAMA data. 

That leaves 6 of 9 codes.  None of the remaining tests had PAMA data, so CMS simply leaves the PAMA median price blank.   (Such codes may get crosswalked or gapfilled, TBD).

Click to enlarge.

I gave the above table to Chat GPT and asked it to fill in the empty rows with test names. Click to enlarge.

Can AI help discover new quality measures for oncology?

In some circles, you hear that we need new better quality measures in oncology, but it's not always clear how to go about discovering them.

In this exercise, we try to use "guided creativity" to see if AI can help (Chat GPT).  Since good measures begin with the patient, we asked Chat GPT to list 15 problems that cancer patients face.  THEN, list an intervention or remedy for each, if possible. THEN, take each problem-solution pair, and sketch it as a "quality measure."   This is essentially guided brainstorming for the AI, and some of the 15 may be hopeless mistakes, others obvious, but perhaps a few could be "promising."

CMS Wants Your Views on its Fraud Contractors (UPICs etc)

Earlier this year, CMS posted an RFI on its "CRUSH" anti fraud scale-up.  CMS has generated a proposed rule now percolating at OMB.   This week, CMS pivots to its categories of fraud contractors (eg UPIC) and asks for YOUR views on them.

Blog by Chat GPT.

###

CMS Seeks a New Fraud-Defense Contracting Model, Alongside Pending CRUSH Rulemaking

CMS has opened a new request for information on how to structure its fraud-defense contracts, inviting ideas that could substantially reshape the agency’s investigative operations. Published October 5, 2026, the SAM.gov notice, “Request for Information – Fraud Defense Contracts,” carries Notice ID 270536.

Responses are due November 4, 2026, at 11 a.m. Eastern Standard Time.

Kimberly Brandt, CMS Deputy Administrator and Chief Operating Officer, highlighted the initiative in a LinkedIn post. Her message describes a reassessment of a longstanding contracting model in light of new technologies and investigative techniques, with an emphasis on detecting and preventing fraud before payments leave the government.

The notice comes with two substantive attachments: a six-page Draft Fraud Defense Contracts Statement of Objectives and a 41-page Interested Vendor Response form. Both are available through the SAM.gov notice. Together, they offer a detailed view of the operating model CMS is considering.

The CRUSH Context: Regulatory Policy and Investigative Operations

The new RFI should be read alongside CMS’s earlier CRUSH request for information, published February 27, 2026, at 91 FR 9803. CRUSH stands for Comprehensive Regulations to Uncover Suspicious Healthcare.

That earlier request sought feedback on potential regulatory and programmatic changes, including provider enrollment, payment suspensions, ownership verification, medical review, and data analytics. Its comment period closed March 30.

CRUSH has since advanced to a draft proposed regulation under White House review. As of October 7, OMB’s public listing identifies the CRUSH proposed rule, CMS-6098, as pending, with a received date of August 7, 2026. The listing does not disclose the draft text or establish a publication date. An earlier Discoveries in Health Policy discussion reviewed that regulatory progression.

The October contracting RFI addresses a related operational question: how should CMS organize the people, technology, investigative responsibilities, and financial incentives that make program-integrity policy work? It is a procurement-planning exercise and does not itself change regulatory authority or award a contract.

The Central Change: CMS Would Drive the Investigative Agenda

The draft Statement of Objectives gives CMS the primary role in proactive data analysis, lead generation, prioritization, and law-enforcement vetting. Approved investigations would then be assigned to contractors for development.

Contractors would supply investigative, clinical, analytical, and management expertise. Their work could include targeted prepayment or post-payment medical review, onsite investigations, interviews, evidence collection, and support for payment suspensions, billing-privilege revocations, and law-enforcement referrals. They would also support appeals, hearings, testimony, and litigation.

Contractors could still develop leads, but only as capacity permits and through CMS review and direction before opening an investigation. CMS-assigned work would take priority.

The document repeatedly emphasizes timely, consistent, and defensible action. It also requires records in CMS-designated systems and continuity during contractor transitions. The proposed model therefore concerns control of the investigative pipeline as much as the introduction of new technology.

What the Response Form Reveals

The accompanying form asks questions that reach well beyond incremental improvements to existing Unified Program Integrity Contractor, or UPIC, arrangements.

Centralization and specialization. CMS asks whether medical review, law-enforcement assistance, site verification, and other functions should be centralized or assigned separately. It also seeks views on national specialty contractors focused on particular provider types, and on jurisdictions and field offices capable of handling investigations across geographic boundaries.

Responsibility across the contractor network. The form asks where handoffs among program-integrity contractors, Medicare Administrative Contractors, Recovery Audit Contractors, and the Supplemental Medical Review Contractor delay or weaken administrative action. CMS is explicitly inviting suggestions for redrawing responsibilities.

Medicare Advantage, Part D, and Medicaid. CMS asks whether Part C and Part D investigations should be integrated with fee-for-service investigations. Although the draft Statement of Objectives focuses on Medicare, the form separately considers Medicaid strategy, including the consequences of separating Medicare and Medicaid investigations, managed-care encounter data, and coordination with state Medicaid agencies and Medicaid Fraud Control Units. These are questions under consideration, rather than announced decisions to separate the programs.

Technology and AI governance. The questions cover analytics, modeling, AI, open-source intelligence, complaint intake, and information sharing. They also ask about explainability, human review before action, model validation, ownership of tools and work products, and portability when a contract ends. Under the draft, contractors would need CMS approval before using non-approved AI, automation, external data resources, or material process changes.

Performance and provider safeguards. CMS seeks measures of quality, timeliness, preventable financial loss, and enforcement outcomes. It also asks how to limit false positives and unnecessary provider burden, and protect beneficiary access when a provider is suspended or revoked. Another question asks how investigative findings should feed back into prepayment edits, enrollment screening, and policy changes.

Fixed-Price Contracting Is an Explicit Part of the Agenda

The form cites Executive Order 14402, Promoting Efficiency, Accountability, and Performance in Federal Contracting, and asks about fixed-price, fixed-unit-price, and hybrid arrangements.

This raises a practical challenge. Investigations vary in complexity, staffing needs, duration, and dependence on government decisions. CMS therefore asks what workload information, systems access, and service-level commitments vendors would need to price the work.

The agency also asks which activities are poorly suited to fixed-price contracting and how incentives can avoid rewarding the volume of investigations, referrals, or administrative actions over their quality. Similar concerns appear in the Medicaid performance questions.

CMS is seeking a payment structure that supports useful, defensible results without creating an incentive simply to produce more cases.

Reading Between the Lines

Several reasonable inferences emerge from the documents, although none establishes a final procurement decision.

CMS appears to be considering a substantial redistribution of responsibilities. The repeated emphasis on CMS-generated leads, centralized functions, specialty contractors, and revised jurisdictions suggests a reassessment of the existing operating model. The notice specifically invites respondents to think beyond current UPIC statements of work and traditional manual processes.

The questions suggest interest in broadening the vendor market. CMS explicitly asks what prevents commercial special-investigations units and analytics firms from competing. Its questions about tool ownership and portability also suggest concern about retaining government access to capabilities when contractors change. For incumbent vendors, investigative execution and responsiveness may become more important competitive differentiators if CMS assumes more of the lead-development role.

Greater central control could improve coordination while creating new bottlenecks. Question 7 asks what decision timelines, data access, and approvals CMS itself must commit to so contractors can meet performance standards. That is an unusually revealing acknowledgment: contractor speed depends partly on government speed.

Specialty expertise may gain importance. National contractors organized around provider types could offer deeper understanding of complex billing and clinical practices. Laboratories and advanced diagnostics are plausible examples where specialization could matter, but the RFI does not announce a laboratory-specific contract.

These interpretations point toward a broader redesign of investigative operations, while leaving the actual contract structure, budget, awards, and implementation timetable unresolved.

How to Respond

CMS instructs respondents to email only the completed Attachment 3.2, Interested Vendor Response, to contracting officer Jennifer Kuhn by 11 a.m. EST on November 4, 2026.

All organization-related information must be completed, but respondents may answer some or all questions. Each answer is limited to 4,000 characters, including spaces. CMS expressly states that this is not a request for capability statements.

No contract will be awarded from this RFI. Participation or nonparticipation will not affect evaluation of responses to a later solicitation, and CMS will not reimburse response costs. The agency may follow up with individual respondents.

For stakeholders tracking CRUSH, the contracting documents provide a useful companion to the pending rule: they show how CMS is considering reorganizing the investigative machinery that supports fraud prevention and enforcement.

##

Sidebar: 

Why Post at SAM.gov—and Will the Responses Be Public?

The venue reflects the task: CMS is planning contracts. The earlier CRUSH RFI sought input on possible regulatory and programmatic changes through the Federal Register and Regulations.gov. This RFI asks how CMS should buy and manage investigative services. SAM.gov is the normal procurement venue for that kind of industry outreach. FAR 15.201 expressly permits RFIs to collect market information for planning, without awarding a contract.

Email submissions also fit that process. A contracting officer can collect structured vendor feedback and conduct follow-up discussions. In this case, CMS specifies a response form, directs submissions to its contracting officer, and reserves the right to follow up individually.

The responses should NOT be assumed to become a public comment docket. The supplied notice and attachments do not announce plans to publish individual submissions. The reasonable expectation is that CMS will review them internally, rather than routinely post them as Regulations.gov comments. That remains an inference; CMS has not expressly stated its publication policy.

However, email submission does not guarantee confidentiality. FAR 15.207(b) requires safeguarding RFI information from unauthorized disclosure, but that is not a blanket exemption from lawful disclosure. Agency records can be requested under FOIA, with trade secrets and qualifying confidential commercial or financial information potentially protected. FAR Subpart 24.2 discusses these protections. The separate protection for proposals submitted to competitive solicitations should not automatically be assumed to cover this preliminary RFI.

A plausible practical advantage is more candid vendor input: firms can discuss pricing assumptions, operational weaknesses, and technology without routinely displaying those details to competitors. Nothing in the documents establishes that avoiding public scrutiny motivated CMS’s choice.

PAMA: What Happened to Exome, Genome, and NIPT?

 Under PAMA, what happens to Exome, Genome, and NIPT beginning in 2027?

Here's the data, click to enlarge:

click to enlarge

Exome 2018 

In 2018 PAMA data, Exome came in at $4780, sibling exome at $12,000, and exome re-evaluation aka "dry lab" at $320.   This data was based on very few claims and some thought the $12,000 number may have mistakenly been meant as the family price (parent, parent, child = $4000x3 = $12,000).

Exome 2025

Regardless, the new prices (2025 prices reported in 2026) are exome $3680, sibling $6480, and dry lab $230.  About 30 labs reported an exome price.  [Note, CMS CLFS prices won't drop more than 15% per code per year.]

Genome 2018 to 2025

Moving on to genome, the prices were $5031, $2709, and $2337.   Genome drops to $3018, which is a 40% drop.   Sibling genome is $1650 (much less than sibling exome), and dry lab has no PAMA price report (CMS will  gapfill or crosswalk it).  Note CMS also has proposed taking "dry lab" genomics (81417, 81527) off the CLFS.

##

NIPT 81420

For NIPT, there is the general code 81420 $759 and microdeletion 81422 $759.   These drop to $584 and $621, about -20%.  No surprise that 81420 was reported to CMS by 121 labs, much more than the other codes I review in this blog.  (Not shown here; PLA codes for NIPT for a few labs).

##

It's possible to get raw data and map the distribution of prices reported to CMS, of which only the median result is seen in the Excel.  For more background, an example of distribution, and links, see here.

##

###   Click to enlarge.  Distribution of NIPT per CMS PAMA data.

##

##

SIDEBAR: 121 labs and LDT NIPT?

No FDA-cleared or approved U.S. kit was identified for the cell-free DNA prenatal screening represented by 81420. FDA has described these tests as lacking its authorization; its published list of cleared or approved molecular tests contains no corresponding NIPT assay.

However, 121 laboratories reporting prices does not establish there must be 121 independently developed NIPT assays.  PAMA collects payment information, while inter-laboratory billing arrangements can include specimens sent elsewhere for testing. Send-out testing—with the referring laboratory billing the payer—is therefore a plausible explanation for part of the count. But the exact number of operating NIPT labs - whether 5, 10, or 121 - can't be established from the PAMA data. 

##  Try.

A reasonable working estimate is about 10 distinct U.S. NIPT assay families, with a plausible range of 10–20 when separately validated local versions are included. That is an informed estimate, not a verified census.

Five prominent offerings are readily identifiable: Natera’s Panorama, Labcorp’s MaterniT21, Quest’s QNatal, Myriad’s Prequel, and BillionToOne’s Unity. Additional assays and laboratory implementations, including ARUP’s, push the count beyond five. Natera

The counting problem is that a distinct laboratory implementation need not represent a wholly independent invention: laboratories may use shared commercial technology but validate their own clinical assay. Conversely, a laboratory listing NIPT may simply send it out—Mayo’s MaterniT21 listing explicitly names Sequenom/Labcorp as the performing laboratory. aruplab.com

Thus, “roughly a dozen underlying assays, distributed through a much larger network of billing and referring laboratories” is a defensible characterization. The 121 PAMA reporters cannot establish the exact number, and the broader NIPT market also includes tests billed under codes other than 814

PLA Use of "100 genes or more." It's baaaack!

 For several years, there seemed to be a PLA moratorium on the use of "or more" so you'd see the construction "Panel of 85 genes" instead of "Panel of 85 or more genes."   Use of "Panel of 85 genes" suggested you'd start over with the PLA panel if you added a gene ("Panel of   85  ^86   genes.")

I discussed this in some detail today in a blog about new PLA codes.

https://www.discoveriesinhealthpolicy.com/2026/10/cms-issues-new-october-pla-codes-new.html

CMS Issues New October PLA Codes, New October Clin Lab Fee Schedule CLFS. Essay on "Or More."

CMS has released the October 1, or fourth-quarter, Clin Lab Fee Schedule, CLFS.   The next one, on January 1, will be based on PAMA prices.   CLFS here.  PLA codes (new and  unpriced) run up to 0698U.

AMA has issued PLA codes for the October 1 cycle.  This includes "new" codes released now and effective January 1, 2027  PDF here.  0699U to 0715U are released now for January.

The Mysterious Use of "OR MORE"

Labs that run gene panels don't want a PLA code with an exact number ("panel of 73 genes") since it has to be updated with product cycles (83 genes, 93 genes, etc).   From what I can tell, PLA gave out the format   "N genes or more"    regularly up to 0334U in October 2022.   

Then, we don't see the format again until now, where Guardant RNA test 0696U has "350 or more genes."   

This might seem like the ultimate micro-nerd observation, but if you got a code between 2022 and 2026 ,and had to editorially update it from "75 genes" to " 75 85 genes" this will all register as important.

Here's a Chat GPT extraction, which I'm trusting to find "or more."  Note the gap from 0334U to 0696U.

click to enlarge

(The analysis is based on current CLFS, and won't include deleted examples.)

##

For use of "or more" to represent substances rather than genes, we have more examples, although there is still a gap from 0328U to 0511U.   

See also this uses:

At Least 110...

There is also some usage, not much, of "at least." 

Note that the meaning of "at least 85" and "85 or more" is identical.

##

We should thank the PLA committee for  ag ain employing the term "or more" as in "85 or more genes."  This saves editorial revisions for the PLA itself, for the Path Coding Caucus, the Editorial Panel, and CMS, which must decide which editorial revisions trigger a new pricing year.

##

And here should be an open public link to the above AI analysis. of the latest PLA list from AMA and the 10-2026 CLFS.  

https://chatgpt.com/share/6ac3ffc0-eaf4-83e8-8d54-c02212e5cf79

FDA Awards $1M for Approval Methods for AI Radiology Reports. Will AI Pathology Reports be Far Behind?

I've argued that the billion-dollar app in digital pathology will be an FDA-cleared app for draft report preparation.  (The pathologist then supervises the slide results and confirms/edits the report).  (My blog & PDF white paper.)  

• Critical news 1: FDA has already starting approving some specific apps for draft reports in radiology.
• Critical news 2:  FDA gave a $1M contract research award for how it should assess and approve report-generating apps in radiology.   Worth tracking because:  pathology won't be too far behind.

Report below by Chat GPT 6.

###

FDA Funds Research on AI Generated Radiology Reports and Pathology Should Watch

I’ve argued that the billion-dollar app in digital pathology will be an FDA-cleared application for draft report preparation. The software analyzes the slides and prepares a draft; the pathologist reviews the findings, confirms or edits the report, and signs it out. That is the thesis of my September blog and accompanying white paper. [1]

Radiology is already moving in this direction, including FDA clearance of specific applications with automated reporting capabilities. Now comes another development worth tracking: a $1.29 million FDA research contract to study how AI-generated radiology reports should be evaluated. My thesis is that radiology offers a reliable preview of changes that will reach pathology after a lag. This award addresses a problem both specialties will need to solve.

Understanding Sparano via Understanding the "C Statistic"

The adjacent blog about the Oncotype test and Sparano's new IICM+ test, pivots almost every quantitative comparison on "C-statistic.'   If you're like me, you may be asking, "Whazzat!?!"

Here's a Chat GPT explanation of C statistic.

###

The C-Statistic for the Perplexed: What Does “Better Prediction” Actually Mean?

A new cancer prognostic test is reported to outperform an established test. The evidence includes a C-statistic of 0.74 versus 0.58. Those numbers sound consequential—but what, exactly, did the better test do better? Did it correctly predict recurrence in 74% of patients? Did it identify more patients who needed chemotherapy? Did it estimate each patient’s risk more accurately?

The C-statistic answers a narrower question: How reliably does the test distinguish patients who experience an event sooner from those who experience it later? Understanding that question makes the numbers useful while preventing us from asking them to prove too much.

The explanation begins with the familiar ROC curve. We will connect its sensitivity and specificity axes to comparisons between individual patients, extend that idea to cancer recurrence over time, and then consider what a higher C-statistic establishes about a new test.

Most readers remember that a diagnostic test involves a tradeoff between sensitivity and specificity. A blood test might produce a numerical score. Set the threshold for “positive” low, and the test detects more affected patients but also flags more unaffected patients. Set it high, and false positives decline, but more affected patients are missed.

An ROC curve displays that tradeoff as the threshold moves. Its vertical axis is sensitivity; its horizontal axis is 1 − specificity, or the false-positive rate. The AUC, the area under that curve, summarizes how well the scores separate affected from unaffected patients across possible thresholds.

But AUC has another interpretation that is particularly helpful here:

Choose one patient with disease and one without disease. How often does the patient with disease receive the higher test score?

That probability is the AUC, with half-credit for tied scores. The graph and the patient-pair comparison are two ways of expressing the same quantity.

Consider four invented patients:

Patient	Actual status	Test score
A	Disease	90
B	Disease	60
C	No disease	70
D	No disease	20

Compare each affected patient with each unaffected patient. There are four comparisons:

• A versus C: correctly ordered.

• A versus D: correctly ordered.

• B versus C: incorrectly ordered.

• B versus D: correctly ordered.

The test wins three of four comparisons: AUC = 0.75.

Notice what we have counted: comparisons between patients. We have not chosen a positive-test threshold or counted correctly diagnosed individuals. Those require an additional decision about where to put the cutoff.

Now consider cancer prognosis. Everyone in the study may already have cancer. The task is to anticipate a future event, such as distant recurrence.

We could ask a yes-or-no question: “Did recurrence occur within five years?” That can support a five-year ROC analysis. But a study following patients over many years has additional information: when recurrence occurred. Recurrence at year 2 and recurrence at year 9 are both events, but they represent different clinical courses.

The survival C-statistic, also called the concordance index, evaluates whether the model’s ordering agrees with that observed sequence. Higher predicted risk should generally correspond to earlier recurrence.

Suppose Alice recurs at year 2 and Beth at year 9. If the model assigned Alice the higher risk score at diagnosis, their comparison is concordant: prediction and outcome agree. If Beth received the higher score, it is discordant.

A C-statistic of 0.75 therefore means, approximately, that the model correctly orders three out of four eligible patient pairs. A value of 0.50 represents chance-level ordering; 1.00 represents perfect ordering. Unlike ordinary binary AUC, survival concordance can compare two patients who both experience the event, asking which experiences it sooner.

There is one complication that matters enormously in real studies: we do not observe everyone indefinitely.

If Alice recurs at year 2 and Carol remains recurrence-free through year 8, their ordering is clear. Carol did not recur before Alice.

But suppose Diane leaves the study after year 1 without a recurrence. We cannot determine whether Diane subsequently recurred before or after Alice. Her follow-up is censored: we know she was recurrence-free through year 1, but not what happened afterward.

Different survival C-statistics handle incomplete follow-up differently. Sparano and colleagues used Uno’s C-statistic, which weights observed comparisons to account for censoring under specified assumptions. Consequently, its result is an estimated probability of concordance, rather than necessarily the simple percentage obtained by counting observed pairs. A time-specific ROC AUC and an overall survival C-statistic are related, but they need not have the same value.

This gives the Sparano results a more concrete meaning:

Recurrence period	Oncotype Recurrence Score	Full multimodal model, IICM+
Overall	0.578	0.735
Early: ≤5 years	0.722	0.791
Late: >5 years	0.514	0.710

These are the reported results in the held-out validation cohort. The improvement was larger for late recurrence than for early recurrence.

For the overall comparison, a reasonable plain-language interpretation is:

Under the study’s follow-up definitions and statistical adjustment for incomplete observations, the full model had an estimated 74% probability of correctly ordering an eligible patient pair, compared with about 58% for the Oncotype Recurrence Score.

The difference is about 16 percentage points in concordance. It does not establish that 16 additional patients per hundred would receive the right treatment.

Two further distinctions explain why.

A model can order patients correctly while giving them inaccurate numerical risks. Imagine that one model assigns three patients ten-year recurrence risks of 2%, 5%, and 10%. Another assigns the same patients risks of 20%, 50%, and 90%. The ordering is identical, so their C-statistics are identical. Yet the counseling and treatment implications could be dramatically different.

Whether patients assigned a 10% risk actually experience approximately 10 recurrences per 100 comparable patients is a question of calibration. C and AUC measure discrimination—the ability to distinguish outcomes—not calibration.

Better ordering also need not translate directly into better treatment decisions. Correcting the order of two patients who would receive the same treatment may accomplish little clinically. Improving risk assessment around a treatment threshold could matter considerably. The C-statistic alone does not distinguish those situations. Nor does predicting recurrence establish which patients benefit from chemotherapy. Researchers have specifically cautioned against treating survival concordance as a complete measure of clinical usefulness.

For readers assessing the next “better than Oncotype” headline, the practical questions are therefore:

Question	What answers it?
Does the model reliably place earlier-recurrence patients above later-recurrence patients?	Survival C-statistic
Are its numerical recurrence probabilities believable?	Calibration
How many recurrences are detected or missed at a chosen cutoff?	Sensitivity and specificity at that cutoff and time horizon
Does using the result improve treatment choices or patient outcomes?	Clinical utility evidence

A higher C-statistic is meaningful evidence of better prognostic discrimination in the population studied. Establishing a better clinical test requires the remaining questions to be answered as well.

FOR THE STUDENT:  TEST YOUR UNDERSTANDING

1. A prognostic model has a C-statistic of 0.80. A colleague says, “It correctly predicts recurrence in 80% of patients.” What is wrong with that statement?

   The denominator is patient-pair comparisons, not individual patients. The model correctly orders approximately 80% of eligible pairs, assigning higher risk to the patient who recurs sooner, with appropriate handling of incomplete follow-up.

2. How can an area under a sensitivity–specificity curve also describe patient ranking?

   These are mathematically equivalent interpretations of ROC AUC. Moving the cutoff traces the curve; comparing every affected patient with every unaffected patient measures how often the affected patient scores higher. With half-credit for ties, that proportion equals the AUC.

3. Two models have identical C-statistics, but one reports much higher recurrence probabilities. Can both be equally useful?

   Their ability to rank patients may be identical, while their calibration differs substantially. If one systematically exaggerates absolute risk, it could encourage unnecessary treatment despite its respectable C-statistic. Calibration must be evaluated separately.

4. A new test raises the C-statistic from 0.72 to 0.75. What would you want to know before paying for it?

   Is the improvement reproducible in independent patients, and does it change decisions where treatment benefits outweigh harms? A small increase could be valuable if it improves consequential decisions; a larger increase could accomplish little if management remains unchanged.

References

1. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 1982;143(1):29–36. https://doi.org/10.1148/radiology.143.1.7063747.

2. Uno H, Cai T, Pencina MJ, D’Agostino RB, Wei LJ. On the C-statistics for evaluating overall adequacy of risk prediction procedures with censored survival data. Statistics in Medicine. 2011;30(10):1105–1117. https://doi.org/10.1002/sim.4154. Free full text.

3. Hartman N, Kim S, He K, Kalbfleisch JD. Pitfalls of the concordance index for survival outcomes. Statistics in Medicine. 2023;42(13):2179–2190. https://doi.org/10.1002/sim.9717.

4. Sparano JA, Lama N, Gray RJ, et al. An Artificial Intelligence (AI) model integrating multiscale foundation model histopathology representations with molecular and clinical features predicts early and late distant recurrence in TAILORx. npj Breast Cancer. 2026. https://doi.org/10.1038/s41523-026-01022-y.

Is Sparano et al. Better than Oncotype DX? New Publication.

From time to time a test is published as "Better than Oncotype Dx," which has been around since about 2004.

Some have even been based on H&E slides, WSI, and AI.

Here's a complex multi-modal tour-de-force approach to a new test.  Here's some pre-publicity from 12/2025 on the project, from Caris.  

(See supplements at bottom of blog, for a tech critique of Sparano AI (via Dawood), and views about the test and Caris financials.)

##

“Better Than Oncotype”?
A New Breast Cancer Study Combines AI, Genomics,
and Clinical Data

Recent coverage in ASCO AI in Oncology highlights an intriguing result: a new algorithm predicted breast cancer recurrence more accurately than Oncotype DX. The underlying study, published by Sparano and colleagues in npj Breast Cancer, combines digital pathology, gene expression, and clinical information. It offers substantial evidence of improved prognosis prediction—and several questions about what would make that improvement clinically useful. [1,2]

What the investigators built

Case Study: The Border Between Wellness Devices, Medical @Home Devices, and Coding

 What happens when a $299 wristband challenges a medical service built around much more expensive equipment? 

In this fictional business school case, TrackBP has raised $75 million to bring cuffless ambulatory blood pressure monitoring to market. (One part of the large remote physiological monitoring market, RPM.)  Its CEO sees a cleared FDA predicate and an existing CPT code. What could go wrong? Plenty: matching individual readings does not prove accuracy over 48 hours; clearance does not guarantee coverage; and cheaper technology can unsettle reimbursement assumptions. 

The case explores where wellness becomes medicine and asks whether FDA, AMA, and CMS can keep pace as familiar services move into radically smaller devices.

Case generated by Chat GPT Astra 1 (10/2026).  It should not be used as ground truth, but it can be an example of current abilities of AI to research and write a report.

###

Find the report in the cloud HERE.

What Cancer Payment Rewards, and What Gets Left Behind: Shah 2026 in JAMA Oncol.

 In a new article at JAMA Oncology Shah describes two treatments - one superior - and then reveals we pay $100,000 for the less useful thing and $10,000 for the most useful thing.  He also reveals that most useful thing is radiotherapy.  And it is "facing drastic reimbursement cuts, leading some centers to shut down."  (Shah is a radiation oncologist.)  

 

He doesn't really give a clear description of how and why the two prices are so different.  (And why? Power? Malevalence?  Accident? Don't assume evil when inattention will do.)  Let's take a look at his article and write our own.  

As ten-page white paper HERE.

(Essay by Chat GPT 6).

##

##

##