Has Artificial Intelligence Arrived at Clinical Hematology? The Case of AML.

Artificial intelligence has arrived at the hematology clinic?

A new review in JAMA Oncology (Ansarian et al., 2025) surveys how AI systems are being woven into the management of acute myeloid leukemia (AML)—from image-based diagnosis to genomic risk modeling. 

The numbers are striking—up to AUROC 0.97 for AML detection and >99% accuracy for transcriptome classification.  But the authors also indicate that these tools are beginning to mimic clinical reasoning, by linking bone-marrow morphology with genetic signatures such as NPM1 status. 

The authors also highlight the growing role of federated learning, which lets hospitals train shared models without sharing patient data—an advance that could democratize access to high-performance AI even in resource-limited settings. Below is the abstract of Ansarian et al, positioning AML as a proving ground for AI in the cancer clinic.

### ABSTRACT

Acute myeloid leukemia (AML) is a severe hematologic cancer with complex genetic heterogeneity necessitating personalized treatment approaches. Artificial intelligence (AI) technologies may revolutionize risk stratification, diagnosis enhancement, and treatment planning in addressing critical gaps in AML management, particularly in low-resource health care environments.

Observations  This narrative review synthesizes existing AI applications in 3 primary areas of AML management. 

1. Machine learning algorithms integrating clinical, cytogenetic, and molecular data demonstrate greater prognostic accuracy than conventional European LeukemiaNet (ELN) guidelines. 
2. Deep learning approaches to image analysis yield excellent results for AML subtype identification from bone marrow smears (area under the receiver operating characteristic curve [AUROC]: 0.97) and genetic variant prediction (eg, NPM1 status [AUROC: 0.92]). 
3. AI-driven genomic analysis reveals novel prognostic signatures and therapeutic targets through advanced pattern recognition, with high-dimensional machine learning achieving greater than 99% accuracy in AML classification from transcriptomic data. 

Federated learning approaches enable multi-institutional collaboration with 96.5% accuracy in leukemia classification on heterogeneous datasets.

Conclusions and Relevance  AI technologies hold potential to improve AML treatment through enhanced risk stratification, early detection capabilities, and individualized treatment optimization.

##
AI CORNER
##

Chat GPT 5 writes,

• The review makes a persuasive case that AI could one day unify genomic, morphologic, and clinical data into a single risk framework for AML—but it also reveals how far the field still has to go.

Most cited studies remain retrospective, often trained on small or single-institution datasets, with limited external validation and uncertain generalizability across patient populations. Reported accuracies above 95% may partly reflect closed research settings rather than real-world complexity, where staining variation, scanner differences, and incomplete data routinely confound algorithms. Even the most “explainable” models still need prospective trials, workflow integration, and regulatory clarity before they can influence treatment decisions.

The technology’s direction is unmistakable—but don't overlook the gap between high-performance prototypes and practical, reproducible clinical tools. In that sense, this review captures a turning point for today: AI in leukemia care is not yet routine medicine—but it is no longer mere speculation either.

Podcast: MedTech Talk. Valuation bubble; Reimbursement challenges.

Recently I ran across the podcast, MEDTECH TALK, and I'll mention a few that caught my attention.  (I'm citing a couple from 2023, 2024, but the channel has podcasts right to October 2025).

Find the home page here:

https://medtechmvp.com/media/medtech-talk-podcast

###

Hear about the career path and priorities of Liz Kwo MD MBA, who is now Chief Commercial Officer of Everly Health.   July 2024.  Find it here. #189.

###

There was a huge valuation bubble in biotech, genomics, etc, during 2020/2021.   In this podcast, recorded in July 2023, a panel looks back on the very recent bubble, and predicts the next year.   Find it here. #179.

###

Justin Klein, an MD JD who went straight into venture capital, discusses his career and priorities, also from 2023.  The focus is "the Medtech Ecosystem," how important it is.  There's a substantial section about "reimbursement" as being an opaque unpredictable barrier, a weak link in that ecosystem.  (Around minute 20.-30)  Find it here.  #182.  (See similar reimbursement points at Linked In here.)

• See an extract (summary + transcript) from the Klein interview re: reimbursement:
• https://bqwebpage.blogspot.com/2025/11/jordan-klein-md-phd-medtech-talk.html

###

Valuation bubble.  Biodesix (genomic testing) IPO at $250, peak $480, rapid decline over 90% to $30 and under ($6 today).   Many diverse genomic companies had 2020/2021 valuation spikes, even if they had no connection to Covid testing or therapies.  Invitae rapidly went from $15 to $50 to $15 to 0.

CAP's Newest Podcast - CIPI Council on Informatics and Pathology innovation

 CAP launches a new podcast, CIPI CONNECTIONS.
###

CIPI Cup

One of the several Councils of the College of American Pathologists (CAP) is CIPI - The College on Informatics and Pathology Innovation.   Home page here.  They have five committees:

• Artificial Intelligence
• Cancer
• Digital and Computational Pathology
• Informatics 
• Pathology Electronic Reporting (PERT0

The CIPI podcast is subtitled, 'Insights, Updates, and the People Behind the Innovation."

Find a podcast hosting page here;

https://podcast.ausha.co/cipi-connections/insights-updates-and-the-people-behind-the-innovation

Find the "debut episode" August 8, 13 minutes.  The link, above, also offers a transcript. Here's the tag line:

• Welcome to the debut episode of CIPI Connections, the official podcast of the College of American Pathologists' Council on Informatics and Pathology Innovation (CIPI). Co-hosts Dr. Giovanni Lujan and Dr. M.E. de Baca introduce the podcast’s mission and discuss Dr. de Baca’s inspiring journey from ophthalmology in Germany to leading innovation in pathology. Learn how CIPI and its five committees are shaping the future of digital pathology, AI, cancer reporting, and more—and what listeners can expect from future episodes.

See also a listing of "CAP Podcasts" which includes CIPI Connections and a new series under the banner, "Horror Stories in Pathology Informatics."

Here's an entry point from the CAP / Publications / Podcasts index page:

https://www.cap.org/member-resources/podcasts

https://www.cap.org/member-resources/podcasts/cipi-connections-pathology-meets-ai-translating-ai-frameworks-into-practice

##

AI CORNER

##

Here are AI summaries of two recent podcasts.

##

CIPI Connections Podcast 1 — “Pathology Meets Artificial Intelligence”

This episode of CIPI Connections from the College of American Pathologists explores how existing laboratory validation frameworks can guide the safe implementation of artificial intelligence in pathology. Hosted by Dr. M.E. de Baca, the discussion features Drs. Matthew Hanna, Nick Spies, and Larissa Furtado from the CAP AI Committee, who examine parallels between AI model validation and familiar laboratory processes such as immunostain verification and molecular assay quality control. They emphasize that AI should be treated as another diagnostic test, requiring analytical and clinical validation, documentation, and continuous performance monitoring. The speakers identify gaps in current CAP checklists—particularly around data quality, explainability, and model updating—and highlight the emerging role of AI implementation specialists to ensure safe, effective deployment. Their consensus: existing laboratory rigor provides a strong foundation for AI adoption, but evolving guidance and education will be essential to maintain patient safety and trust as laboratories integrate machine learning tools into clinical workflows.

---

CIPI Connections Podcast 2 — “Horror Stories in Pathology Informatics: Unflagged and Overlooked”

In this inaugural episode of the Horror Stories in Pathology Informatics series, Drs. Alexis Carter and Omar Baba dissect a real-world case involving a missed low cortisol result that went unflagged in an electronic health record, delaying diagnosis of adrenal insufficiency. The case—stripped of identifiers but based on actual events—illustrates how informatics design decisions can compromise patient safety. Because cortisol has time-dependent reference ranges, the lab displayed both ranges in a free-text comment, preventing the system from automatically flagging the abnormal value. The physicians discuss how structured data fields, discrete order codes for morning and afternoon tests, or “ask-on-order-entry” prompts could have prevented the oversight. They also caution against overreliance on visual flags as shortcuts for reviewing lab results, emphasizing the need for clinician education, structured data standards, and interdepartmental communication. The episode concludes with practical takeaways: build structured informatics solutions wherever possible, ensure reference ranges are machine-readable, and reinforce the clinician’s responsibility to review all results—not just those flagged as abnormal.

CMS Releases PFS Final Rule Despite Shut-Down. OPPS Final Rule Pending.

Each fall, CMS normally releasing the final PFS rule and final OPPS rule on November 1, giving them 60 days to take effect (January 1).   CMS got the CY2026 PFS rule out the door on Friday October 31.  The OPPS rule isn't out yet.

See the PFS press release here - "CMS Modernizes Payment Accuracy."  

The main headline is a 2.5% cut in reimbursement for surgical and similar services that CMS believes have "become more efficient."  Stakeholders argued the opposite, that patients had become more complex (a greater and greater percent are over 70).  See MedPage Today here.

• This cut continues an effort, which finally bloomed after years of debate, of shifting some dollars from specialists to primary care.  A couple years ago CMS introduced a G-code which adds about $16 to primary care visits, a perturbation that decreased RVUs for non-E&M visits.  In an effort to increase the number of Infectious Disease doctors, CMS also introduced some special coding for them a couple years ago.

See the PFS Fact Sheet here, which is more detailed and granular than the press release.

CMS notes that much of its practice expense information dates to 2008, but declines to accept a 2014 update from an AMA survery.

CMS may expand its method of using Medicare hospital outpatient rates as a proxy for office-setting rates.   Otherwise, CMS uses an extremely detailed accounting method with pennies and dollars for supplies, capital equipment, staff time, physician time, overhead, and other factors.  

While a telemedicine "cliff" is still in the air, as COVID legislation expires, CMS continues to tweak with rules for virtual services which do exist, such as "virtual direct supervision" rather than "present in the office suite" supervision, for some incident-to services.

Responses to public comments solicited for "software as a service" (in general) at page 447ff, merely remarkng that "comments were appreciated."

The Fed Reg publication will be Wednesday November 5.  A typescript copy is available here (2375pp!)

PLA Codes at Novitas: Vastly Less Controlled than at MolDx

The Novitas MAC seems to go through one payment disaster after another, with notoriously high and bizarre-looking payments for 81408 (rare gene, full sequence) and adjacent codes in 2019-2021.  The OIG's report cited a "billion dollars" in misspending.   Now, in 2023, 2024, the Novitas cash seems to be directed to some otherwise rare gene panels (e.g. for mitochondrial disorders, or for inherited diseases of childhood like cystic fibrosis), and also for code 87798, a microbiology code that is uncontrolled at Novitas.  (See also blog.)

Rare Gene Panels (81419, 81430, 81433, etc); 87798 

The typical labs billing for these in the US are in Texas or Florida, and they recently obtained NPI numbers as LLCs.   The same codes (81408; the rare-gene panels; 87798) are very rarely billed to Medicare by labs like LabCorp, GeneDx, Ambry, Quest, etc.  

In each case, billing in Texas and Florida (but not elsewere) exploded such as multiples of 100X over a couple years.

Novitas and PLA Codes - Striking Lack of Controls Compared to MolDx

I looked at 2023 billing to Medicare Part B for all "U" codes - the PLA codes.  Total billing was $519M.  The distribution was 64% MolDx, 14% Novitas, and 22% NGS MAC.  

However, had NGS MAC not been paying for FMI codes which are under an NCD (0037U, 0239U), NGS MAC PLA payments would drop to just $36M or 7% of national.

In each case, the PLA codes were highly concentrated, with 90% to 97% of PLA payments to the top 10 providers in each MAC system.  (And, as shown below, most PLA payments went to the top 3-4 codes).

• We noticed something strange when we looked up each PLA code in the Medicare Coverage Database.  
• All of the top 10 PLA codes at MolDx fall under NCDs or LCDs.  
• In the Novitas MAC, I could identify only 2 of 10 PLA codes that fell under an LCD or NCD, using the same Medicare Coverage Database.

In both Novitas and NGS MAC, many of the payments (not under LCDs) were for 0240U, 0241U, microbiology codes.  Although MolDx had 64% of all PLA payments, none of the top 10 at MolDx were microbiology codes.

Caveats

• This is first-pass work; I haven't double checked everything or had anyone replicate it.  
• We compare 2023 payments to current 2025 policies.  
• I entered CPT codes one by one in the Medicare Coverage Database and took the result.

Click to enlarge. MAC index = 1, MolDx, 2, NOV/FCSO, 3, NGS MAC.

click to enlarge

##

The CMS cloud database is here.  It has all 2023 Part B claims.  You can filter for codes ending in "U" = PLA codes.

Chat GPT Asks: How Do Commercial, Medicare, and Medicare Advantage Rates Compare?

The point of this blog is just to show how Chat GPT 5 performs, to a request for a medical policy essay.

Issue:  Now and then I see an article comparing Medicare FFS prices with some other price list.  But I don't record and file those articles.  So, below you can see instant, AI-generated research on this topic.  

###

Prompt  Can you please search the internet looking for recent articles that survey what Medicare fee for service pays versus what either regular commerical or medicare advantage pay? There is probably the most data on relative pay rates for office visits. I work in the lab industry, so any published comparative pricing (against medicare part B as a baseline) for pathology or laboratory medicine would be ideal. Please be agentic or thorough.

###

Absolutely—here’s a tight, source-backed roll-up of what the best recent evidence says about Medicare fee-for-service (FFS) versus commercial and Medicare Advantage (MA) payment levels, with a spotlight on office visits and lab/pathology.

###

EXECUTIVE SUMMARY

In recent analyses, Medicare fee-for-service (FFS) payment levels for physician services—including office visits—provide a stable baseline against which commercial and Medicare Advantage (MA) payments can be compared. The predominant finding is that commercial insurers pay roughly 30–40% more than Medicare FFS for similar physician services, with the most recent data showing commercial PPO rates averaging around 140% of Medicare. In contrast, MA plans tend to negotiate provider payments that sit much closer to Medicare FFS levels rather than commercial rates, particularly in hospital outpatient settings and other facility-based services. This dynamic reflects MA plans’ reliance on Medicare benchmarks and regulatory constraints that tether payment growth more tightly to FFS than to the wider commercial market.

In the laboratory and pathology sector, the pricing landscape shows further nuance. For independent clinical laboratory tests covered under the Clinical Laboratory Fee Schedule (CLFS), Medicare rates are essentially derived from the weighted median of private-payer rates by statute, thereby narrowing the gap between commercial and Medicare payments for those services.  [Historically, before 2014, at least some commercial lab rates were below Medicare. BQ]

However, for pathology professional services that fall under the Medicare Physician Fee Schedule (MPFS), commercial-to-Medicare differentials resemble those seen in other physician specialties—commercial payments are commonly 20–50% higher than Medicare. At the same time, when services are rendered in hospital-outpatient settings, commercial payment multipliers may be substantially higher, given the elevated mark-ups typical in hospital-facility billing contexts.

###

DATA AND LINKS

Big picture (latest, credible sources)

• Physician services, overall (incl. office visits): Commercial PPO payment rates now average ~140% of Medicare FFS (up from 136% in 2022; long-run trend rising from ~122% in 2011). (medpac.gov)
  CBO’s broader review is directionally consistent (~25% higher for physicians), and KFF’s lit review shows a typical range of ~118%–179%. (cbo.gov)

• Medicare Advantage (MA) provider prices: For hospital services, MA negotiated prices are much closer to traditional Medicare than to commercial; same-insurer/same-hospital comparisons show commercial prices ≈ 2–3× MA. For physician services in MA, empirical work finds MA rates are generally tethered to Medicare FFS, not commercial. (Mark K. Meiselbach)

• Hospitals (useful context for outpatient labs/E/M): Employers/private plans pay hospitals ~254% of Medicare on average (2022), with wide state variation. (rand.org)

Office visits (E/M) as an anchor

• E/M office visits sit inside the physician trend above; think commercial ≈ 1.3–1.4× Medicare on average, with variation by market and insurer. That average comes directly from MedPAC’s PPO data series (not code-specific but dominated by E/M volume). (medpac.gov)

Pathology & Laboratory medicine

Two different payment systems matter here:

1. Clinical lab tests (CDLTs) under the CLFS (Part B):
   Since PAMA, Medicare sets CLFS rates from the weighted median of private-payer rates. In other words, for most CDLTs, Medicare amounts are pegged to the commercial market median (repriced on a multi-year cycle). This means “commercial vs Medicare” is intentionally narrow for independent lab test codes—though contract-to-contract variation still exists. (Centers for Medicare & Medicaid Services) [In some reports, some commercial rates were below Medicare in 2012. BQ.]

2. Pathology professional services (e.g., 88xxx under the MPFS):
   These behave like other physician services, so commercial ≈ 1.2–1.5× Medicare on average, with specialty- and market-level variance (MedPAC overall ~1.40×; Urban Institute shows wide specialty dispersion). Specialty-specific breakouts for pathology are sparse in public summaries, but the directional physician-category uplift applies. (medpac.gov)

Hospital-based lab pricing: When the same lab test is billed as hospital outpatient, commercial prices often rise with hospital outpatient markups. RAND’s latest round shows commercial outpatient prices several multiples of Medicare—important if your client work involves hospital outreach labs or site-of-service shifts. (rand.org)

Handy ratios you can cite (and where they come from)

• Commercial vs Medicare (physician/clinician services overall): ~140% (2023 average PPO). (medpac.gov)

• Commercial vs Medicare (historical context): 122% → 140% from 2011→2023. (medpac.gov)

• MA vs Commercial (hospitals, same insurer/hospital): Commercial ≈ 2–3× MA (2022 machine-readable data). (Mark K. Meiselbach)

• Commercial vs Medicare (hospitals, all-payer view): ~254% in 2022 (national mean, inpatient+outpatient). (rand.org)

• CLFS lab tests vs commercial: Medicare ≈ weighted median of private rates by rule (thus generally close to commercial medians for independent labs, with timing/refresh caveats). (Centers for Medicare & Medicaid Services)

Nuances & caveats (worth knowing for client work)

• Market power & site of service drive variation. Commercial-to-Medicare gaps widen in concentrated hospital markets and for hospital-outpatient services; office-based settings are closer to Medicare baselines. (rand.org)

• MA pricing is policy-tethered. Because MA benchmarks tie to local FFS spending, MA provider prices tend to cluster nearer FFS than commercial—especially on facility services. (PMC)

• Pathology exposure to MPFS updates. Year-to-year conversion-factor moves (and RVU tweaks) hit pathology like other specialties (e.g., 2024–2025 adjustments), even as CLFS amounts for CDLTs follow separate PAMA cycles. (medpac.gov)

---

Sources you can quote in decks

• MedPAC (Dec 2024 & Jan/Jun 2025): Private PPO ≈ 140% of Medicare for clinician services; time trend from 122% → 140% (2011–2023). (medpac.gov)

• CBO (2022): Commercial physician prices ~25% higher than Medicare on average. (cbo.gov)

• KFF literature review (2020): Private physician payments average ~143% of Medicare across studies (range 118–179%). (KFF)

• Health Affairs (2023): Commercial hospital prices two–three times MA in same hospitals. (Mark K. Meiselbach)

• RAND Round 5 (2024): Employers/private plans pay hospitals ~254% of Medicare (in/outpatient combined). (rand.org)

• CMS CLFS/PAMA (2024 page & guidance): Medicare sets CDLT rates from private-payer weighted medians. (Centers for Medicare & Medicaid Services)

If you want, I can also:

• Pull specific 99213/99214 localities (Noridian/Jurisdiction E etc.) and pair them with Transparency-in-Coverage E/M rates from a target insurer in LA for an apples-to-apples snapshot, or

• Build a pathology-focused mini-table (e.g., 88305, 88184) comparing CLFS vs MPFS vs sample commercial (independent lab vs hospital outpatient), using recent public files.

AMA Releases PLA Applications for 4Q2025

 Each quarter, AMA receives and processes a couple dozen PLA code applications.  The ones for this quarter are posted.  There will be an AMA meeting to vote for the codes next week, November 5.  Codes were accepted til September 9, the PLA committee met on October 23.

Find the PLA page here:

https://www.ama-assn.org/practice-management/cpt/cpt-pla-codes

Find the agenda here:

https://www.ama-assn.org/system/files/november-2025-pla-public-agenda.pdf

There are 28 agenda items, of which about 5 are revisions or deletions.  

##
AMA is currently collecting CPT Category I applications, and those will be available for comment about November 15-30.  They will be voted at the editorial meeting in February in Palm Springs.

AMA Publishes Summary of Panel Actions for CPT September 2025

You may have attended AMA CPT September 2025 for its infamous "Tab 94," the proposal to create a novel coding appendix (sort of a SAMD registry) with novel codes.  The novel codes would be called "CMAA," for Clinically Meaningful Algorithmic Analyses.   This would be exclusively for software-dominant services that do not require physician work.   It's caught the attention of the lab community because AMA PLA has categorized several software-based whole slide imaging tests as PLA tests, but recently began rejected further such applications in PLA, on what I understand to be the theory that "those are dry lab services which are not eligible for PLA codes."  

This is the home page for CPT panel actions here.   This is the report for September 2025 here.  

Tab 94

There were 95 agenda items or "tabs."  Regarding Tab 94, we read,

• The Panel engaged in a robust discussion of this proposal for a framework to report algorithmic and AI-enabled clinical services that do not include traditional interpretative physician or other qualified health care professional (QHP) work. The CPT Panel heard testimony for one hour from over 20 stakeholders representing nearly every sector of the health care industry. This proposal was for discussion only and no vote was taken at this meeting

Labs

Tab 44, genetics of inherited renal conditions, withdrawn.  Tab 45, genetics of inherited immune conditions, accepted.  Tab 45, genetics of Alport syndrome, accepted.  Tab 46 Bladder MAAA, rejected.  Tab 47 Code 0007M deleted. 

Withdrawn!   Rejected!

A rule of thumb is that applications informed they are likely to fail (they are given formal feedback) are usually withdrawn rather than voted and rejected.

There were 34 "withdrawn" items and 18 were rejected.   These 52 codes were over half of all agenda items.

Legislation Watch: The Health Technology Investment Act of 2025; S. 1399

It's become a cliche' that most action occurs in the Executive Branch, with a sleepy agenda on the Hill.

But one topic of interest is the Health Technology Investment Act, S.1399, introduced in April 2025 by Senators Mike Rounds (R SD) and Martin Heinrich (D NM).   

• Find the Senate press release here.
• Find the legislative language here.
• See support by AdvaMed here.

• Sidebar: Funding.  
• A new article in Healthcare Dive writes that, "Digital health funding outpacing last year as huge rounds increase.  Investment in 2025 has reached $9.9 billion."   
• Sidebar: CMS Policy Struggles
• CMS wrestling with paying digital services, the topic of S.1399.  
• This past summer, CMS sought public advice on pricing SAAS-SAMD; entry point here.
• Sidebar:  No Money!
• The "Investment Act" does not actually contain any money.  Rather, it instructs CMS to set up payment systems that are friendlier to software-dominant diagnostics and therapies.

PBS Runs Documentary on the Life of Secretary of Health, RFK Jr

This week, I watched the two-hour documentary on PBS Frontline about his career and controversies. 

See the video (currently streaming) here.
See a review of the film online at MedPage Today - here.  

• The LA Public Library had several books by or about RFK Jr (*):

Here's a Chat GPT 5 summary of the documentary, after it listened to the whole transcript.

###

AI CORNER

###

Senate Letter Criticizes the "AMA Monopoly" on the U.S. Coding System

October saw a sharply worded letter from Senator Cassidy to the AMA, criticizing the "monopoly" (his term) on the US main coding system.  That monopoly, though, was granted by Congress in the U.S. HIPAA law and subsequent CMS regulations, to ensure a uniform coding system for communications between providers and payers.   

click to enlarge

Sources:

• Press release from Senator Cassidy, with full letter link (October 8).  Here. 
• Direct to letter PDF.
• Report at Inside Health Policy (subscription).  Here.
• At Modern Healthcare (subscription).  Here.
• Article at MedPage Today.  Here.
• Report at Fierce Healthcare.  Here.
• At Medical Device Mfgrs Assoc.  Here.
• At Politico Pro (subscription.)  "Cassidy and RFK Jr have something to disagree on."  Here.

SIDEBAR - Going Without AMA?

Interestingly, for Medicare claims, MolDx could run genomic test transactions without the AMA CPT. 

A huge part of MolDx payments are now coded as 81479, "other molecular test," and processed with the Z-code.  CMS could create a G code like - G1479 = "other molecular test."  This G-code wouldn't involve AMA and would be royalty-free for labs and payers to use.   As it chooses, MolDx (CMS) might use "G1479" for current 81479 claims only.   Or MolDx could use "G1479" for all types of molecular tests (whether or not they have Cat I or PLA codes from AMA), adjucating via the Z code.   This could pull out several billion dollars a year of services now coded with AMA CPT codes.

Science Policy; Banned Words at NIH; Is This the Time to Do a PhD?

Two interesting science policy articles this week.   I think both are firewalled, so I'll include short summaries.  Here are the titles:

At STAT, Anil Oza provides a long-format, deep-dive article, on the vocabularies of "banned words" at NIH and the need to change grant titles and abstracts to avoid "banned words" which trigger a fatal withdrawal of funding.   Find it here:

https://www.statnews.com/2025/10/29/nih-banned-words-analysis-grant-title-changes/

At CHRONICLE of Higher Education, an article has the surprising or paradoxical title, "There has never been a better time to start a PhD."  Find it here:

https://www.chronicle.com/article/there-has-never-been-a-better-time-to-start-a-ph-d

Peer Review Is Too Slow; How Impactful Can AI Be?

In Annals of Internal Medicine, Kieran Quinn et al. ask, how can we re-think how we disseminate medical research?   Can we expand the status and role of preprint archives?   And see their Citation 8, Liang et al., in NEJM-AI.  They empirically study how LLM contributes to peer review, and the percentage of reviewers who are favorable is high.  80% of reviewers said LLM review was more helpful than at least some of the available human reviews.

###

Alzheimer tests and FDA: Location, location, location.

HEADER:  FDA’s 2025 Alzheimer’s test approvals reveal a new split in strategy—Fujirebio’s plasma ratio test is indicated for centers for diagnosis, while Roche’s Tau181 test is limited to “rule-out” use, but in primary care.

###

There are some surprising nuances to recent Alzheimer test approvals at the FDA.

 In May 2025, FDA approved the first-ever PLASMA test for Alzheimer's disease, a combined Tau217 and Amyloid42 test from Fujirebio.  (Fujirebio  previously had a CSF test for Alzheimer's.)   The test has a double cut-off; below the lower cut, Alzheimer's is unlikely; above the higher cut, Alzheimer's is likely, and for a few patients in-between, the test is indeterminate.

Recall that the "intended use" is from the viewpoint of the physician, and "indicated use" is the patient condition or situation for use.   The Lumipulse Plasma Ratio is "intended to aid healthcare providers to identify patients with amyloid pathology associated with Alzheimer's disease."  The Plasma Ratio is "indicated for adult patients, age 50 and older, presenting at a specialized care center with signs and symptoms of cognitive decline."  This FDA wording on the label foots to the title of the July 2025 guideline for Alzheimer blood testing, which also uses the phrase "in a specialized care center."  (Palmquvist 2025).

Typically, a sign is objective (fever), a symptom is subjective (headache).  See Fujirebio's detailed FDA summary here.

In contrast, we have a headline in Genomeweb, October 2025, "Roche Targets Primary Care Setting With Newly FDA-Cleared Alzheimer's Rule-Out Test."  This plasma Tau181 test is only intended to "rule out" (which is driven by a one-sided low value), while the indicated setting is "primary care" rather than "a specialized care center."

The new October Genomeweb article on Roche tracks back to a May 2025 article on these debates in the field (here), and a JAMA IM op ed by Wider and Covinsky on settings for blood testing. 

AMA Upgrades Its RVU Database - RBRVS Data Manager

AMA creates CPT codes, an obligatory code set for communications between providers and payers in the US, under HIPAA law.  It also values the codes, through an elaborate, data and committee driven process (the relative value unit or RVU system).   These codes and values are generally adopted by CMS for Medicare.

For years, AMA has had a product called RBRVS Data Manager - originally on CDs! - for years now, on the internet with a subscription (mine as an AMA member is $240/yr).

They did a big update & modernization to RBRVS Data Manager a couple years ago.  I *think* they've done some extra updating this year, too.

###

For example, here's I've searched on 95165 - preparation and provision of allergens - and you can see that there are no less than 9 different tabs, each pretty rich in information.

The tabs are

• General information
• Work RVU history
• Vignette/Service
• RUC rationale
• Geographic adjustment
• Physician time
• Claims data
• Billed together
• PE Inputs

The system has lab codes (8nnnn) but not PLA codes (nnnnU).  

For lab codes - like 81455, tumor gene panel >50 genes - you don't have valuation or RUC data but you do still get the tab for Claims Data, with voume (going back to 2013, over ten years), ICD10 codes billed, and other demographics.

You get to see the Patient Vignette and Service (work) description, which might help you if you are writing a CPT application yourself.   Some Cat I lab codes have no vignette or service, some do.

• 81455 Vignette   A 65-year-old male presents with lung and liver lesions. Pathologic evaluation of biopsies of these lesions reveals a poorly differentiated neoplasm of uncertain primary origin. Tumor tissue is submitted for a targeted genomic sequence analysis of a panel of 250 genes known to be informative in a broad array of cancers.
• 81455 Service     Isolate high-quality DNA from the patient's tumor tissue and perform massively parallel sequencing on the tumor DNA, looking for mutations in 250 genes, which may be genomic targets for therapeutic management. Send the analytical results to a pathologist or other QHP for identification of mutations, interpretation, and preparation of a written report that specifies the patient's mutation status, which may contain information about diagnosis, prognosis, and patient management, to include information about targeted drug therapy.

And here's for Cologuard (original):

• 81528 Vignette:   A 55-year-old asymptomatic male presents for a routine annual examination. A colonoscopy for screening was recommended however the patient declined. The patient is offered a non-invasive screening with a multi-target DNA colorectal cancer screening test.  
• 81528 Service:     The test is comprised of a stool collection kit, fecal occult hemoglobin determination, and quantitative real-time target and signal amplification of 10 DNA markers. The quantitative biomarker results are analyzed through a logistic regression algorithm to produce a qualitative, single dichotomous patient result of negative or positive.

And here's for ROMA ovarian cancer test:

• 81500 Vignette: Patient is a 64-year-old white female who presents with complaints of pelvic pressure and pain. The pelvic exam find an 8 cm firm fixed right adnexal mass. The patient is sent for a pelvic ultrasound which shows a complex cystic and solid adnexal mass The physician and patient agree to surgery. Prior to scheduling the surgery, the physician requests a ROMA test so as to determine whether the case should be triaged to a physician with special expertise in gynecologic malignancy management. 
• 81500 Service: The patient's serum is analyzed for CA125 using chemiluminescent microparticle immunoassay (CMIA) methodology and for HE4 using an enzyme immunometric assay (EIA) methodology.  The results of the CA125 and HE4 assays along with the patient's menopausal status are entered into computer software which uses an algorithm to combine the results of the assays and menopausal status into a numeric score.

And for Tau217:

• 81493 Vignette:  A 65-year-old female presents for neurologic evaluation of gradual cognitive decline. Measurement of phosphorylated Tau is requested. 
• 81493 Service:  Test the patient's sample for phosphorylated Tau by quantitative immunoassay and report the results.

I think current CPT committees edit back the vignette and service pretty short, more like Tau 81493 and less like Tumor 81455.  The longer examples above are maximums, and your experience with today's committees may look more like 81493.

To take another example, 87798 (other pathogen, amplified probe) was used 8,429,396 times in 2023, 57.5% in patients over 75, 0.95 eligible per ESRD, 36.45 male, and the top Dx code was N39, other urinary disorders, which was used 40% of the time.

SEE ALSO - RUC Records Too

Separately from the subscription RUC database above, if you sign up for a AMA logon, you can get voluminous RUC records for code valuations.

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

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

##

AI CORNER

##

What is known about the AMA software, RBRVS Data Manager?

##

Direct AI output below - not proofed or vetted.

##

Here’s the short version:

What it is. The AMA’s RBRVS DataManager Online is a web-based reference for CPT® and HCPCS Level II codes with the RVUs, payment logic, and valuation background those codes carry in the RBRVS/Medicare Physician Fee Schedule. It’s aimed at coding, auditing, reimbursement policy, and internal education. (American Medical Association)

What you get.

• Code content & search: >19,000 CPT/HCPCS code descriptors plus ~5,500 AMA-exclusive CPT clinical scenarios that spell out pre/intra/post service work. (American Medical Association)

• Payment details: tools to calculate geographically adjusted Medicare payments (GPCIs/CF), plus direct practice-expense inputs (clinical staff, supplies, equipment) for each code. (American Medical Association)

• Rationale & RUC material: proprietary valuation insights and RUC rationales for work and PE recommendations—useful when you need to see why a code is valued the way it is. (American Medical Association)

• Utilization & crosswalks: >10 years of Medicare utilization per code and an ICD-10-CM ↔ CPT crosswalk (top five related codes). (American Medical Association)

Typical use cases. Revenue-cycle/audit, payer policy analysis, utilization review, and training physicians/analysts on how codes are valued and paid. (American Medical Association)

Pricing & access (historical reference). AMA lists it as an online subscription (the public store has shown list ~$320/yr; AMA member ~$240/yr in past listings). Exact current pricing may vary by year and member status. (American Medical Association)

Context & adoption. The product is cited by regulators and in workers’ comp rules as the source for RVU data/methodology, underscoring its use beyond provider offices. (Michigan.gov)

Recent note. In early 2024 the AMA expanded/modernized the online RVU database experience (widely noted by industry watchers), but the subscription concept and purpose remained the same. (Discoveries in Health Policy)

If you want, I can compare it feature-by-feature to alternatives (e.g., Optum’s Essential RBRVS data files or EZ-Fees fee-schedule tools) to clarify when AMA’s product is the better fit for a given task. (OptumCoding)