Very Brief Blog; Keeping up with ARPA-H, the ARPA-H Dash

 Remember ARPA-H, the health innovation funding institute created by recent legislation?

It's running a contest called ARPA-H DASH.  Enter your idea for health care innovation, and it will be debated and ascend a giant bracket-format context to a winning position of $15,000.

Submit March 29-April 7.  Context runs April 17 to May 16.   That is, the winner will be announced on May16.

ARPA-H has four concentration areas:  

Read all about it:

https://arpa-h.gov/news/dash/

To enter, you give a title, "the transformation you propose" (200 words), how it's done today (200 words), and your new approach (200 words).   Add a citation published after 1/1/2018.  Click the "submit" button.

Brief Blog: Judge Strikes Down ACA Preventive Care Mandates; Likely to be Stayed

 In a case that was running through 2022, a Texas judge has "nationally struck down" the preventive care mandates inside the Affordable Care Act of 2010.   The Act required that private insurers cover services endorsed by the US Preventive Services Task Force.   

In general, this case (Braidwood v HHS/Becerra) has involved arguments as to whether the ACA law could point outwards to the USPSTF decisions and give them, in effect, the force of law.   One workaround could be to have the individual decisions of the USPSTF panelists certified and restated by HHS to achieve statutory compliance in a legalistic way. (This judge believes that would fail; p 26).

I am not an attorney but I would be "very unsurprised" if the national scope of the ruling is not "stayed" or postponed by a further court action soon.  (FN1) In this case, the plaintiffs were concerned about contraceptive benefits and HIV prevention benefits, but the legal arguments about the ACA were general and not specific to those particular situations.

See an open access article at Politico here:

https://www.politico.com/news/2023/03/30/texas-judge-obamacare-00089641

See a PDF of the judge's 28-page decision here:

https://storage.courtlistener.com/recap/gov.uscourts.txnd.330381/gov.uscourts.txnd.330381.113.0_2.pdf

The wide spectrum of argumentation includes the USPSTF legality argument, the religous freedom arguments (which are specific to contraception and PREP), and whether a remedy should be universal or narrowly tailored (striking down the ACA at one extreme, and providing relief only to the present litigants for PREP at the other extreme).

Also at Fierce Healthcare here.  The judge refers to the USPSTF as "the preventive service's task force's constituionally infirm preventive care mandates." (p 20).  The USPSTF became illegal to the extent that the ACA gave it "significant authority pursuant to the laws of the United States" (p. 23).  

FN1

Supporting my guess that this will get postponed, Exact Sciences stock moved only half a percent today.

##

In other legal news today, a shareholder suite against Biogen, alleging it hyped prospects for Aduhelm, was thrown out.   Judge.   Suit.  

Brief Blog: Lab Owner "Pleads Guilty," $89M in Genomic Tests, and Surprise: In Florida

On March 28, 2023, DOJ issued a press release that a lab owner had "pled guilty" in an $89M fraud and kickback scheme.   DOJ here.   The DOJ press release links to a 21 page indictment here. Coverage at Genomeweb.

The indictment refers to several people by name, but the lab is just "Laboratory 1," making it harder to find CMS records of what was billed and how  much money was spent.   

However, I did find a name matching the indictment on a Linked In page (here).  This Linked In entry refers to these lab entities: Clinical Lab Solutions, DBA, myTest Diagnostics.  There is also a remark, "looking for partners" for genomic testing in 2/2021.  

Since CMS data at the level of individual labs has so far only been released up to CY2020, we'll have to see if there was activity at these labs in 2020.  But we can infer from "partners" remark that the lab was still active and growing in 1H2021.

MyTest Diagnostics has an active (or once-active) Facebook page. It gives an address as 9671 Gladiolus Drive, 102, Ft Myers FL, and we found the NPI is 1417407727.   

Using the NPI, we looked at the CMS lab-by-lab data for 2020.  This NPI was paid for about 66 tests, about $8M that year, but 96% were funds were a series of genetic tests, and about half of that were spent onjust a few higher-cost Tier 2 costs (81403-408), with almost $2M in code 81408 alone.  (FN1)  See cloud data here.  

click to enlarge

In this snapshot, every patient, one by one, and 892 patients in a row, individually needed all 6 of the Tier 2 codes 81403-08.  Presumably, each needed these tests to work up his or her distinct signs and symptoms for each elderly patient up to the 892nd.  

AMA defines Tier 2 codes as rarely used codes (they don't have enough usage to be Tier 1 codes).  Often these labs have 75% plus billing in Tier 2 codes.  This lab would have  done so, too, had it billed Tier 2 codes like 81408 in pairs per patient (x2).  

Leaving this lab behind, let's look at national Part B lab industry data for CY2021.  By my calculations, for 2021, assuming near all Tier 2 billing including 81408 is improper in Medicare patients (billed by mills specializing in these codes, billing all Tier 2 codes for each elderly patient, and only billing in states with no edits), about a third of genomic payments in 2021 were fraudulent.  See bar chart below.  This  pattern continued from 2018 and 2019 into 2021, even though the patterns that allowed this literally impossible and absurd billing at the First Coast and Novitas MACs were evidence by mid 2019. 

Click on graphic to enlarge.

click to enlarge (FN2)

__

Since Medicare billing was only $8M in 2020 from the lab we discussed, and since DOJ asserts that $89M was involved, it's a reasonable inference that CMS did nothing to pick up the truly absurd billing pattern we see in 2020.  So, much more money was lost to the same sinkhole in 2021 likely expanding on the pattern in 2020.   This isn't fraud you need AI and supercomputers to find: you could train a 14 year old to recognize this fraud pattern in two minutes.

##

FN1.    Somewhat atypically, the family tier 2 codes were billed here, at least in 2020, in units of 1 per patient. Some would pay 2 under MUE edits, a trick which many labs specializing in Tier 2 codes used. For example, 2 x code 81408 = $4000.     

FN2.  These bar charts were prepared from CMS Part B spending data in 2021.  I used simple blocks of codes; all the 80,000 series codes are laboratory; those 88,000 and up are physician pathology, and so on.  The data won't be precise to the decimal point but is close.  I added all the PLA codes as human genomic, although a slim fraction (especially when monetized) are proteomic.

Very Brief Blog: New CLFS Released for April 2023; And Reminder about Full PLA Code Names Earlier

CMS releases its Clinical Lab Fee Schedule with quarterly updates, to accomodate new quarterly PLA codes.

CMS has released the listing for April 2023 - find it here (Q2-2023).

Short and Long Descriptors

CMS releases the CLFS with short code descriptors (about 20 letters) and a longer code descriptor which is kind of quirky, but can run up to about 10 words.   This is what we see in the Q2 April release.

Still on the CMS website is the Q1 January release.  This release of the CLFS had an extra column, at right, which seems to have the full PLA descriptors (however long).  It's nice to have these in Excel format.  It makes the January Q1 release a collector's item, since the PLA descriptors disappear by the April release.

Note that without too much work, you can ask ChatGPT for formula instructions how to take the CLFS (with columns for code, price, and descriptor) and merge it with the annual release of utilization data (with columns for code number, + number of services, and + total dollars paid).   I don't verify accuracy but I think I have done this in the cloud here.  I pulled columns for 2021 Part B # services and total payments into the existing 2023 CLFS fee schedule showing prices and code names.

click to enlarge. Importing UTIL and PAYMENTS into CLFS.

Transmittal CR12612

The PDF transmittal that announces the new fee schedule is CR12612, January 27, 2022, here.  

Summer meetings?

Last year, the transmittals announcing dates of the Summer CLFS meetings appeared around April 18, 2022, so if they match that, look for a mid April announcement.  Then you can plan summer vacations!

Very Brief Blog: New ADLT Test Granted by CMS - DCISionRT by PreludeDx

On March 28, 2023, PreludeDx (Laguna Hills, CA), released a press release that its DCISionRT test had received Advanced Diagnostic Laboratory Test (ADLT) status from CMS.

DCISionRT test is a gene expression test that helps women with DCIS evaluate the need for additional radiotherapy.   )See also the version of Oncotype Dx which is tailored for DCIS cases.)

• A test that is rated an ADLT by CMS gets 9 months pricing at list price, followed by annual pricing at its median market price (but see FN1).   
• The test must be a sole-source test, and may either be an FDA cleared or approved test, or a MAAA test with conditions.  
• The conditions on the MAAA test are that it provides unique information not available from any other test (42 CFR 414.502 definition of ADLT).
• The webpage where CMS posts ADLT tests is here.   
• The webpage also includes various instructions and application packets regarding ADLT rules.  

The PreludeDx test Press Release is here.  

CMS updated the ADLT one-pager on approved tests concurently, with a "new ADLT" price of $5435 for DCISionRT, applicable in 2Q 3Q 4Q 2023.

• Previously, such as on the 2023 CLFS, 0295U had a price of $1897.  
• This was a crosswalk to 0067U, the 4-slide BBDRiskDx test from Silbiotech.   
• That test had 7 Part B payments in 2021 (totaling $13,000).

click to enlarge

##

Coverage rules?

I couldn't easily find a CMS web page or MolDx article or other article specifying coverage terms and conditions for the PreludeDx test (nor the Oncotype DCIS test.)  However, as an IHC test, the PreludeDx test is outside the scope of MolDx coverage decisions which only address DNA-RNA tests.

The DCISionRT test is a 7-protein immunohistochemistry test, with CPT code 0295U.   Markers include Her2 and Ki67, and inputs include age, size, margin status, palpability.  

As an IHC test, the test would probably be adjucated by Noridian since it is a non-MolDx, non-DNA-RNA test.   

> MolDx DEX registry

The DEX web page for Oncotype DCIS says it's "covered" (a binary answer).   

The similar DEX web page for DCISionRT exists, but doesn't show a tab for covered or not covered, which I learned in the past, means (1) it's being updated but also could mean (2) it is not eligible for MolDx coverage or non coverage because it is out of scope as an IHC test.

> Billing Articles at CMS

The CMS billing article for regular Oncotype DX says it is a different test than DCIS but does not point to coverage rules for Onco DCIS.  A53105.   Keyword searches, for me, and a code search for 0295U, didn't yield a PreludeDx article.

###

FN1.  The cycle for ADLT price resetting is annual, but slow -- with data one year, and then data reporting next year, and then a new "annual" price in the next year after that.  

Very Brief Blog: ProPublica Article on Insurance Denials, with Webinar

 The website ProPublica covers policy topics each week.   In a lengthy March 25 article by Rucker et al., they cover denial issues at a major payor, using "internal documents and former company executives."

https://www.propublica.org/article/cigna-pxdx-medical-health-insurance-rejection-claims

The article is paired with a webinar on March 28, Tuesday at 2 ET.   I don't know if the video will be archived or not.  The webinar page is here:

https://www.propublica.org/events/medically-necessary?utm_source=website&utm_medium=article&utm_campaign=event_20230328

Very Brief Blog: White House 60-page Report on Biotechnology, Genomics Goals

The White House has released an inter-agency, 64-page new report on its goals for the biotechnology industry as a whole, including genomics.

• Find the PDF report here.
• See also a review at Genomeweb, which may be open access, here.
• Coverage also at Forbes, here.

The report has been summarized as including, "next-generation biomarkers of health, advancing cell therapies, improving AI-driven drug development, enhancing gene editing, sequencing the genomes of a million microbial species...[and] standards, benchmarks, data security, and privacy standards for multiomics, [and] integrate multiomics data into EHR systems." 

Recall that the Biden administration had an executive order reflecting its enthusiasm for biotech last fall in 9/2022 here.   New White House fact sheet here.

 

A GPT4 summary of the press release states,

• The Biden-Harris Administration has announced new goals and priorities to advance American biotechnology and biomanufacturing, in response to the President's Executive Order to advance biotechnology and biomanufacturing innovation for a sustainable, safe, and secure American bioeconomy. The announcements aim to strengthen America's bioeconomy, rebuild supply chains domestically, and support the innovation ecosystem while ensuring all that is invented in the United States can be made in the United States. 
• The announcements include 
• releasing a new report outlining a vision for harnessing biotechnology and biomanufacturing research and development (R&D) to achieve societal goals; 
• establishing biomanufacturing priorities for the Department of Defense; and 
• assessing the economic value of the nation's bioeconomy. 
• Other deliverables, such as a plan to expand training and education opportunities and a national strategy for expanding domestic biomanufacturing capacity, are in development.

Very Brief Blog: CMS Clever Two-Page Diagram on Crosswalk/Gapfill Lab Test Pricing

CMS's regulations on new lab test pricing are somewhat dry and painful to decipher.  But, CMS has now produced a graphic two-page description of the crosswalk process and the gapfill process.  Download it at the link:

https://www.cms.gov/files/document/mln9034855-clinical-laboratory-fee-schedule-annual-payment-determination-process.pdf

On a light note, earlier this month I asked CHAT GPT4 to give us some advice on how to maneuver and survive the same process.  Here.

Look for an announcement from CMS in April on the dates of the June/July summer pricing meetings. CMS posts the codes to be discussed 30 days, or almost 30 days, before the first meeting, which takes place usually in late June.

Very Brief Blog: Did You Know? New AMA Cat III Code for Biomarkers Analysis & Rx Recommendations

In January 2023, I missed a press reslease from CUREMATCH, Inc, that they were successful in getting a Category III code for a service that takes in biomarker data (from prior tests) and organizes algorithmic drug recommendations for oncologists.  The code, when priced, will include both the software service and physician time.

• Press release here.  
• CUREMATCH web site here.  
• See CureMatch news here.

The code, 0794T, is on a list of new Category III (temporary) codes from the AMA CPT.   Find the list here.   

• The code text is: Patient-specific, assistive, rules-based algorithm for ranking pharmaco-oncologic treatment options based on the patient's tumor-specific cancer marker information obtained from prior molecular pathology, immunohistochemical, or other pathology results which have been previously interpreted and reported separately.

In earlier years, it was rare if ever that CPT had preamble text before the words of the CPT code.  That type of preamble is now more common for Cat III codes, and there is a long description as a "preamble" before the code text.   See the AMA CPT link above.

This sounds similar to Clinical Decision Support (CSD), for which FDA has regulatory guidance.  A CDS Coalition has demanded that FDA rescind its position - here.

##

GPT 4 summarizes the press release.  

"CureMatch uses a proprietary curated knowledge base and an artificial intelligence platform to quantify and qualify therapy options based on each patient's molecular data. Oncologists and other providers can use the new CPT code indicating the need and demand for this precision oncology tool for a five-year period starting July 1, 2023.  CureMatch's therapy matching system has been shown to improve progression-free survival in patients treated using combination therapies with higher molecular matching scores, as demonstrated in a 2019 I-PREDICT clinical study."   And, "[The new code] will facilitate wider adoption and availability of precision medicine treatment options for cancer patients. CureMatch's proprietary AI platform helps oncologists make the best treatment decisions by analyzing massive amounts of data to quantify and qualify therapy options based on each patient's molecular data."

Fun fact.  Lab industry experts know that PLA codes are company-specific.   In contrast, although Category III codes may be lengthy and tailored, they are not actually company-specific but apply to the described services.  In this case, the core service is, "algorithm for ranking pharmaco-oncologic treatment options based on the patient's tumor-specific cancer marker information."

Fun fact.  I think this code might not be billable by a lab as a CLFS service, because it includes Physician Work and it would have to be classified as a CLIA code, I think, by CMS, to be billable by an entity enrolled as a lab.  I'm not sure what it going to happen, though.  

Flatiron Assist.  See a service called Flatiron Assist, from Flatiron, a nice creative video at Vimeo.

Very Brief Blog: Breakthrough Coverage Introduced as 2023 Legislation

During the Trump administration, there was proposed legislation to trigger CMS coverage of breakthough devices, and there was a corresponding rule-making and program at Medicare, which didn't require Hill action.

The clouds darkened.  The bill never passed.  And the Trump program was nixed or deep-sixed by the Biden administration. End of MCIT - Medicare Coverage for Innovative Technology.

The lawmaking approach to MCIT is back.

• See legislation from DelBene, Wenstrup, and others, press release here.
• March 22, 2023
• Find the legislative text here.  
• Find the introduced bill as 2023 HR 1691 here, https://www.congress.gov/bill/118th-congress/house-bill/1691?s=1&r=2 
• The bill applies back to FDA approvals on 3/15/2021 or later.
• BT devices are "Deemed" to be reasonable and necessary for 4 years.
• The four years are timed from "update of the payment system" for the device.
• Payment mechanisms may include NTAP (hospital add-on).  
• CMS has 3 months to produce a new code for the devices.

Very Brief Blog: Is GPT-4 Better at Medical Boards? Yes, Of Course It Is.

This past winter, we learned that CHAT GPT 3.5 performed pretty well on law school exams and medical boards. Does the newer, larger GPT4 perform substantially better? Yes. GPT4 passed the US Medical Licensing Board Exam by a 20-point margin. 

 Find the Microsoft news article here:

https://www.microsoft.com/en-us/research/publication/capabilities-of-gpt-4-on-medical-challenge-problems/

Find the Microsoft detailed publication-ready article here:

https://www.microsoft.com/en-us/research/uploads/prod/2023/03/GPT-4_medical_benchmarks.pdf

March 20, 2023

Large language models (LLMs) have demonstrated remarkable capabilities in natural language understanding and generation across various domains, including medicine. We present a comprehensive evaluation of GPT-4, a state-of-the-art LLM, on medical competency examinations and benchmark datasets. GPT-4 is a general-purpose model that is not specialized for medical problems through training or engineered to solve clinical tasks. 

Our analysis covers two sets of official practice materials for the United States Medical Licensing Examination (USMLE), a three-step examination program used to assess clinical competency and grant licensure in the United States. We also evaluate performance on the MultiMedQA suite of benchmark datasets. Beyond measuring model performance, experiments were conducted to investigate the influence of test questions containing both text and images on model performance, probe for memorization of content during training, and study calibration of the probabilities, which is of critical importance in high-stakes applications like medicine. 

Our results show that GPT-4, without any specialized prompt crafting, exceeds the passing score on USMLE by over 20 points and outperforms earlier general-purpose models (GPT-3.5) as well as models specifically fine-tuned on medical knowledge (Med-PaLM, a prompt-tuned version of Flan-PaLM 540B). 

In addition, GPT-4 is significantly better calibrated than GPT-3.5, demonstrating a much-improved ability to predict the likelihood that its answers are correct. We also explore the behavior of the model qualitatively by presenting a case study that shows the ability of GPT-4 to explain medical reasoning, personalize explanations to students, and interactively craft new counterfactual scenarios around a medical case. Implications of the findings are discussed for potential uses of GPT-4 in medical education, assessment, and clinical practice, with appropriate attention to challenges of accuracy and safety.

Very Brief Blog: FDA Issues Draft Guidance for PGx on Drug Labeling

As posted by RAPS "Regulatory Affairs," Mary Ellen Schneider notes that FDA has released for comment a new draft guidance on the use of pharmacogenomics in drug labeling.   (The emphasis is on uses in drug labeling, not the genetic testing itself)

Find the RAPS article here.  Find the 13 page draft document at FDA, here.

It explains how FDA expects applicants to present and use PGx findings and data in regulatory submissions for investigational new drug applications (INDs), new drug applications (NDAs), and biologics license applications (BLAs).  Topics include format and level of detail, and this new guidance (when finaled in a year or two) will replace the prior guidance from 2005.  The draft guidance applies to IND submissions and requires detailed reports to be submitted as part of the IND under the Previous Human Experience section, in meeting packages, and in clinical study protocols or reports. 

NDA and BLA sponsors must generally submit the same types of pharmacogenomic data and studies as IND sponsors, though the level of detail varies. 

Comments on the draft guidance can be submitted until 20 June 2023.

SHORT POST: GPT4 Gives Advice for (1) A "Stuck" FDA Review, (2) A "Hostile" Medicare Coverage Challenge

Readers will have seen a number of recent blogs where I post real-life examples of GPT4 responses to Medicare policy questions. 

That series probably topped out by producing a 20-page white paper on Medicare & AI based on...interviewing the AI.

Here, I provide Google Doc Links to two more HHS examples. 

First, ChatGPT4 gives advice for a company whose 510(k) review has gotten bogged down with difficulties.

https://docs.google.com/document/d/1ykm9-qbVhQJIGBj6s-t_wIFHF5_wlhWemwWJY7d2fHA/edit?usp=sharing

Second, Chat GPT gives advice for stakeholders who are fighting a policy that gives overly restrictive rules for a brain tumor therapy. It's fictionalized, but based on a real case. I also, for example, ask GPT to predict opposing attorneys' arguments, and our rebuttals to them.

https://docs.google.com/document/d/1_cPp1uSbZ6tyuZ1whUBYiVzhkNMvvZQQoIAUC39dxbo/edit?usp=sharing

Illustrations from CANVA

Short Post: My White Paper Interviewing GPT4 on CMS Policy for Artificial Intelligence

This weekend I've produced a wide-ranging 22-page white paper on AI and CMS health policy - generated almost entirely by GPT 4.  I asked it 12 questions, and then a 13th question: "Is there anything else you'd like to add?"

GPT4 also generated title, subtitle, summary, promotional tweets, and a two minute video script to raise awareness of the white paper.

White paper online

https://drive.google.com/file/d/1efmn4V4M2dfZw9vbWdUhmILaqUVLX2O7/view?usp=sharing

Video promotion

https://www.youtube.com/watch?v=bKxLnpBDodo

_____

Thanks to Richard Coda Design for design and graphics.  

Thanks to Galileo and Plato for the idea of presenting books in dialog format.

Although I did not use it here, a shout-out to Jason West's video on making brilliant prompts.

AI Exploration Series: GPT-4 Weighs In: Does CMS Hold the Keys to AI Adoption in US Healthcare?

Unlocking AI's Potential in Healthcare:

How CMS Can Make or Break the Revolution

In a 12-question, 5000-word interview, we get the views of Chat GPT-4 on how CMS may be promoting or retarding the adotion of impactful uses of artificial intelligence in healthcare.

The interview includes an AI-generated summary at top.  

• I have clipped the whole interview (which would equal a 30-minute webinar) into this long blog.  
• I have also clipped the whole interview in an open access Google doc, which some readers may prefer.
• https://docs.google.com/document/d/1EXF5Tg2O3AAPKJ7EWU4q8zUegN6cWkC8h23AJS28GlA/edit?usp=sharing
• For the official 22-page white paper version,
• https://drive.google.com/file/d/1efmn4V4M2dfZw9vbWdUhmILaqUVLX2O7/view?usp=sharing 
• For a two-minute video overview:
• https://youtu.be/bKxLnpBDodo

The interview is entirely unedited except for minor wordsmithing of the summary.  The title and subtitle are AI-generated.   The two-minute video overview is from an AI script.

Thanks to Richard Coda Design for design and graphics.  Thanks to Galileo and Plato for the idea of presenting books in dialog format.

____

Summary (AI generated)

 

This interview between Dr Bruce Quinn and GPT-4 discusses various aspects of AI in healthcare, including its applications, challenges, and the roles of different stakeholders. AI has the potential to revolutionize healthcare by improving diagnostics, treatment planning, patient monitoring, and workflow efficiency. Radiology and pathology have been particularly impacted by AI, with innovations that were previously impossible without AI technology.

However, the adoption of AI in healthcare faces challenges, particularly in the Medicare reimbursement system. The current reimbursement models, such as the Medicare RVU payment system, may not account for the unique costs and benefits associated with AI technologies. This can create financial barriers for healthcare providers and organizations, limiting the integration of AI-driven solutions.

The interview explores strategies for engaging with policymakers, focusing on simplifying the message, quantifying the impact of AI, and connecting the issue to broader public health goals. Collaboration with various stakeholders, including healthcare providers, AI developers, professional organizations, and patient advocacy groups, is crucial for driving meaningful change in reimbursement policies and fostering a supportive environment for AI adoption in healthcare.

Also discussed: the roles of organizations like the American Medical Association, the National Institutes of Health (NIH), and the Centers for Medicare & Medicaid Services (CMS) in promoting AI in healthcare. Cross-agency collaborations between the NIH and CMS, as well as efforts to address ethical considerations and integrate AI into medical education, can contribute to the successful integration of AI into the healthcare system.

The take-home lesson is that AI has the potential to significantly improve patient care and outcomes, but its impact largely depends on addressing financial, regulatory, and technical barriers. Collaborative efforts among stakeholders are essential to promote AI adoption in healthcare and ensure its benefits are fully realized.

 

 

Q1. What activities, programs, or policies is the Medicare agency using to Promote or to reimburse for uses of artificial intelligence in healthcare?

 

Q2. Are there any major features of Medicare payment policy that are barriers to expanded use of AI in healthcare?

 

Q3. How does the cost architecture of the Medicare RVU payment system discourage use of AI?

 

Q4. Is the American Medical Association doing much to encourage use of true AI (not just telemedicine) in US healthcare?

 

Q5. What are the most impactful uses of AI in radiology?

 

Q6. What are the most impactful applications of AI in pathology? Focus on things that require AI and we’re not possible without it.

 

Q7. What is the NIH doing to encourage breakthroughs in AI that will improve healthcare in new ways?

 

Q8. Are there AI focused collaborations between NIH and CMS?

 

Q9. Help me compare two things. (1) There is a well known problem that biopharmas, especially new venture funded ones, have difficulty getting good ROI for novel antibiotics and may even go bankrupt after FDA approval. (2) Similarly, I am concerned that healthcare coding and coverage and reimbursement is lacking for breakthrough or out of the box AI innovations, and limiting progress in healthcare. Can you comment ?

 

Q10.  I would like to focus on your point which begins, “for breakthrough AI innovations in healthcare…” How would you explain the problem to a senior CMS policymaker who has limited knowledge about AI in healthcare? How would you help him understand the problem, prioritize the problem, and engage in solutions?

 

Q11. Let’s assume you have been meeting with the senior CMS policymaker, and your presentation has covered these five points. He seems a little disengaged and you are concerned that he won’t do much once the meeting is over. What would you do next? Would you focus on this person, or seek higher level meetings, or turn to other government bodies, what is missing?

 

Q12. Your help is excellent. I find that higher level policymakers should take a broader view of public health goals, yet, they are farther from details like coding or RVU rules where the problems for reimbursements lie. How can we overcome this dilemma?

 

Q13.  Can you provide a single summary (200-300 words) of our discussion.  

########

Q1. What activities, programs, or policies is the Medicare agency using to Promote or to reimburse for uses of artificial intelligence in healthcare?

 

As of my knowledge cutoff date in September 2021, the Centers for Medicare & Medicaid Services (CMS) had already started exploring ways to incorporate artificial intelligence (AI) into healthcare programs and policies. While there may have been developments since then, the following are some notable examples of AI integration within the Medicare system up to 2021:

1. Virtual Care: CMS expanded telehealth services during the COVID-19 pandemic, which included remote patient monitoring and virtual consultations. AI plays a significant role in these services by assisting with diagnostics, patient triage, and care management.
2. AI Health Outcomes Challenge: CMS, in collaboration with the American Academy of Family Physicians and the Laura and John Arnold Foundation, launched the AI Health Outcomes Challenge. This initiative aimed to accelerate the development and adoption of AI-driven solutions to predict patient outcomes and improve care quality.
3. Chronic Care Management (CCM): CMS has recognized the potential of AI-driven remote monitoring and care management tools to improve the management of chronic conditions. They have introduced reimbursement codes for remote patient monitoring, allowing healthcare providers to bill Medicare for these services.
4. Fraud, Waste, and Abuse Detection: CMS has started using AI and machine learning algorithms to detect patterns of fraud, waste, and abuse in the Medicare system. This can help identify and prevent inappropriate billing practices and protect the program's integrity.
5. Electronic Health Records (EHRs): CMS has been promoting the use of EHRs and interoperability to improve care coordination and reduce errors. AI-powered tools can analyze large amounts of data from EHRs to identify patterns and trends that can inform better clinical decision-making.
6. Quality Payment Program (QPP): The QPP focuses on improving the quality of care and encouraging the adoption of innovative technologies, including AI. The Merit-based Incentive Payment System (MIPS) and Advanced Alternative Payment Models (APMs) within the QPP financially reward healthcare providers for using technology to improve patient outcomes.

It is essential to note that the Medicare landscape is continually evolving, and CMS may have implemented new programs or policies since September 2021. It is recommended to visit the CMS website or consult with healthcare experts for the most up-to-date information on AI integration in the Medicare system.

Q2. Are there any major features of Medicare payment policy that are barriers to expanded use of AI in healthcare?

 

There are several aspects of Medicare payment policy that can act as barriers to the expanded use of AI in healthcare. Some of these include:

1. Fee-for-service reimbursement model: Medicare's traditional fee-for-service (FFS) reimbursement model might not adequately incentivize the adoption of AI technologies, as it mainly focuses on the quantity of services provided rather than the quality or efficiency of care. This system may not provide appropriate incentives for healthcare providers to invest in AI solutions that could improve patient outcomes and reduce costs.
2. Limited coverage and reimbursement: Medicare's coverage and reimbursement policies for AI-driven technologies are still in the early stages of development. Many AI-based tools and services might not yet have established billing codes or specific reimbursement pathways, making it challenging for healthcare providers to adopt these innovations due to financial uncertainty.
3. Regulatory and approval processes: The development and adoption of AI technologies in healthcare can be slowed down by lengthy regulatory and approval processes. In the United States, AI-driven medical devices and software must be approved by the Food and Drug Administration (FDA) before they can be reimbursed by Medicare. This can result in delays and added costs for innovators and healthcare providers.
4. Data privacy and security concerns: Medicare payment policies need to address data privacy and security concerns related to AI technologies. As AI systems often rely on large datasets, including sensitive patient information, there is an increased risk of data breaches and misuse of information. These concerns can slow down the adoption of AI in healthcare.
5. Fragmentation of care: The Medicare system is fragmented, with various parts (Part A, Part B, Part C, and Part D) catering to different healthcare needs. This fragmentation can make it challenging to coordinate and integrate AI-driven solutions across different parts of the system.
6. Lack of standardized data and interoperability: AI systems require standardized, high-quality, and interoperable data to function effectively. However, the lack of standardized data formats and interoperability between electronic health record (EHR) systems can hinder the adoption of AI in healthcare.

To overcome these barriers and encourage the adoption of AI in healthcare, Medicare payment policies should focus on creating incentives for providers to invest in innovative technologies, streamlining regulatory and approval processes, and addressing data privacy and security concerns. Additionally, fostering collaboration between various stakeholders, including healthcare providers, payers, technology developers, and regulatory agencies, can further help accelerate AI's integration into the healthcare system.

Q3. How does the cost architecture of the Medicare RVU payment system discourage use of AI?

 

The Medicare Relative Value Unit (RVU) payment system is based on the Resource-Based Relative Value Scale (RBRVS), which assigns a relative value to each medical service or procedure. This value is determined based on three components: Physician Work RVUs, Practice Expense RVUs, and Malpractice RVUs. The total RVUs are then multiplied by a monetary conversion factor to determine the payment amount for a specific service or procedure.

The RVU payment system can discourage the use of AI in healthcare in the following ways:

1. Limited incentives for innovation: The RVU payment system mainly focuses on compensating healthcare providers for their time and expertise, practice expenses, and malpractice costs. It may not account for the potential benefits of adopting AI technologies, such as improved patient outcomes, reduced costs, or increased efficiency. This lack of incentives for innovation can discourage healthcare providers from investing in AI solutions.
2. Inadequate reimbursement for AI-driven services: The current RVU payment system may not adequately reimburse AI-driven services, as it is not explicitly designed to account for the unique costs and benefits associated with AI technologies. This can result in healthcare providers being hesitant to adopt AI solutions due to financial concerns.
3. Fee-for-service model: The RVU payment system is a fee-for-service model, which rewards providers based on the volume of services rendered rather than the quality or effectiveness of care. This model may not create incentives for healthcare providers to adopt AI-driven solutions that could improve care quality, as they may not be directly rewarded for the improved outcomes resulting from the use of AI.
4. Slow to adapt: The RVU payment system may be slow to adapt to new technologies, such as AI, as it requires updates to incorporate new billing codes and reimbursement pathways for AI-driven services. This slow adaptation process can hinder the integration of AI into the healthcare system.
5. Lack of specific billing codes: AI-driven services may not have specific billing codes within the RVU payment system. Without well-defined billing codes, healthcare providers may face difficulties in receiving reimbursement for these services, discouraging their adoption.

To encourage the use of AI in healthcare, the Medicare RVU payment system needs to evolve and adapt to the unique costs and benefits associated with AI technologies. This could involve developing new billing codes and reimbursement pathways, as well as creating incentives for healthcare providers to invest in AI-driven solutions that can improve patient outcomes and reduce healthcare costs.

Very good.

Thank you! I'm glad you found the information helpful. If you have any more questions or need further assistance, please don't hesitate to ask.

Q4. Is the American Medical Association doing much to encourage use of true AI (not just telemedicine) in US healthcare?

 

The American Medical Association (AMA) has shown interest in promoting the responsible adoption of AI in healthcare to improve patient care and outcomes. Although the AMA's efforts are not limited to true AI, they have taken several initiatives to encourage the use of AI and related technologies in the US healthcare system:

1. AI policy development: The AMA has been working on developing policy recommendations related to AI in healthcare. In 2018, the AMA adopted a new policy that provided a broad framework for AI's role in healthcare, focusing on ensuring that AI-driven tools are safe, effective, and evidence-based, while prioritizing patient interests and addressing liability concerns.
2. Ethical guidelines: The AMA has emphasized the importance of developing and implementing ethical guidelines for AI in healthcare. They encourage AI developers, healthcare providers, and other stakeholders to follow ethical principles, such as transparency, accountability, privacy, and fairness, when designing and using AI systems.
3. Education and training: The AMA recognizes the need for medical professionals to be well-equipped to integrate AI into their practice. They support education and training efforts that can help physicians and other healthcare providers understand AI's potential benefits, limitations, and ethical considerations.
4. Collaboration with other organizations: The AMA has been involved in collaborative efforts with other organizations to promote the responsible use of AI in healthcare. For example, they have partnered with organizations like the American Heart Association (AHA) to develop certification programs for AI-driven digital health tools, aiming to establish standardized criteria for evaluating these technologies.
5. Advocacy for research and development: The AMA has been advocating for increased research and development in AI for healthcare applications. They support federal efforts to fund AI research and encourage collaboration between the public and private sectors to drive innovation in this field.
6. Participation in AI-focused events: AMA representatives have participated in various AI-focused conferences, workshops, and panel discussions, contributing their insights and promoting discussions around AI's role in healthcare.

While the AMA's efforts may not be exclusively focused on true AI, they play an essential role in shaping the landscape for AI adoption in healthcare. By developing policies, ethical guidelines, and educational resources, the AMA is working to ensure that AI-driven technologies are responsibly integrated into the US healthcare system to improve patient care and outcomes.

What are the most impactful uses of AI in radiology?

 

AI has made significant strides in radiology, leading to various impactful applications that can improve diagnostic accuracy, efficiency, and patient outcomes. Some of the most notable uses of AI in radiology include:

1. Image analysis and interpretation: AI algorithms, particularly deep learning techniques like convolutional neural networks (CNNs), can automatically analyze medical images (e.g., X-rays, MRIs, CT scans) to detect and classify abnormalities, such as tumors, fractures, or lung nodules. These algorithms can achieve high accuracy and sometimes even surpass human radiologists in certain tasks, resulting in faster and more reliable diagnoses.
2. Computer-aided detection and diagnosis (CAD): AI-powered CAD systems can assist radiologists by flagging potential abnormalities and providing diagnostic suggestions. By serving as a "second opinion," these systems can help reduce diagnostic errors, improve the consistency of interpretations, and alleviate the workload for radiologists.
3. Automated measurements and quantification: AI algorithms can automatically measure and quantify various imaging features, such as the size, shape, and volume of tumors or other structures. This can help radiologists monitor disease progression, assess treatment response, and provide more accurate prognoses for patients.
4. Radiation dose optimization: AI can help optimize radiation doses in imaging procedures, ensuring that patients receive the minimum necessary dose while maintaining image quality. This can reduce patients' exposure to ionizing radiation and minimize potential risks associated with radiation exposure.
5. Image reconstruction and enhancement: AI techniques can help improve the quality of medical images by reducing noise, correcting artifacts, and enhancing image resolution. This can lead to better visualization of anatomical structures and pathological findings, enabling more accurate diagnoses.
6. Workflow optimization: AI can streamline radiology workflows by automating tasks like image sorting, prioritization, and report generation. By identifying urgent or critical cases and prioritizing them for review, AI can help radiologists focus on the most important cases, thereby reducing turnaround times and improving patient outcomes.
7. Predictive analytics: AI can analyze large datasets of radiological images and associated clinical data to identify patterns and correlations that can inform clinical decision-making. These insights can help radiologists predict disease progression, treatment response, and patient outcomes more accurately.

These impactful AI applications in radiology have the potential to revolutionize the field by improving diagnostic accuracy, reducing human errors, enhancing efficiency, and ultimately leading to better patient care and outcomes.

Q6. What are the most impactful applications of AI in pathology? Focus on things that require AI and we’re not possible without it.

AI, particularly deep learning techniques, has shown great potential in the field of pathology, enabling the development of various impactful applications that were not previously possible without it. Some of the most significant AI applications in pathology include:

1. Digital pathology and whole-slide image analysis: AI-powered algorithms can analyze digitized whole-slide histopathology images at high resolution to detect and classify various features, such as cancerous cells, tissue architecture, and biomarkers. The ability to efficiently and accurately analyze large-scale, high-resolution images is a task that was not feasible without AI, given the substantial amount of data involved and the level of detail required.
2. Computer-aided diagnosis and prognosis: AI algorithms can assist pathologists in making more accurate and consistent diagnoses by identifying and quantifying specific features in histopathological samples. These algorithms can also predict patient outcomes, such as the likelihood of cancer recurrence or metastasis, based on the analysis of complex patterns within the tissue samples that may be challenging for human pathologists to discern.
3. Personalized medicine and treatment planning: AI can analyze large amounts of data from pathological samples, including genetic, molecular, and imaging data, to identify patterns and correlations that can inform personalized treatment plans. By considering the unique characteristics of a patient's tumor or disease, AI can help pathologists and oncologists recommend the most effective therapies tailored to individual patients.
4. Biomarker discovery and validation: AI can efficiently analyze large datasets to identify potential biomarkers associated with specific diseases or treatment responses. This process can help researchers discover new diagnostic, prognostic, or predictive biomarkers that were previously hidden in the vast amount of data, leading to new insights and breakthroughs in disease understanding and treatment.
5. Computational pathology and multi-modal data integration: AI can integrate and analyze multi-modal data, such as histopathology images, radiology images, genomic data, and clinical information, to provide a more comprehensive understanding of diseases. This integrative approach was not possible without AI, as it requires the efficient processing and analysis of vast and heterogeneous datasets to generate meaningful insights.

These AI applications in pathology have the potential to revolutionize the field by enhancing diagnostic accuracy, enabling personalized medicine, and uncovering new biomarkers, ultimately leading to better patient care and outcomes. The ability of AI to process and analyze large-scale, complex data has opened up new possibilities in pathology that were not feasible without its advanced capabilities.

Q7. What is the NIH doing to encourage breakthroughs in AI that will improve healthcare in new ways?

 

The National Institutes of Health (NIH) has been actively involved in fostering research and innovation in AI to improve healthcare. The agency has undertaken several initiatives to promote breakthroughs in AI that can lead to better healthcare outcomes, such as:

1. Funding AI research: The NIH provides funding opportunities for researchers focused on AI and its applications in healthcare through various grants and programs. These funding opportunities support AI-related projects across a wide range of areas, including diagnostics, therapeutics, medical imaging, and genomics.
2. Collaborations and partnerships: The NIH collaborates with other government agencies, academic institutions, and private organizations to promote AI research in healthcare. These collaborations help to facilitate knowledge exchange, share resources, and accelerate the development of AI-driven solutions for healthcare.
3. Initiatives for data sharing and accessibility: The NIH supports initiatives that promote data sharing and accessibility for AI research, such as the National Library of Medicine's National Center for Biotechnology Information (NCBI) and the National Cancer Institute's Cancer Imaging Archive (TCIA). These resources provide researchers with access to large-scale datasets that are critical for developing and validating AI algorithms.
4. AI-focused workshops and conferences: The NIH organizes and participates in AI-focused workshops, conferences, and panel discussions to share insights, promote discussions, and identify new research directions in AI for healthcare.
5. Training and educational programs: The NIH supports training and educational programs aimed at developing the next generation of AI researchers and healthcare professionals. These programs help to build a skilled workforce capable of harnessing AI's potential to improve healthcare.
6. Strategic Roadmap for AI in Biomedical Research: In 2021, the NIH released a strategic roadmap that outlines the agency's vision for AI in biomedical research. The roadmap identifies key priorities and strategies for promoting AI research and applications in healthcare, such as fostering interdisciplinary collaboration, addressing data challenges, and ensuring the ethical development and use of AI.

These initiatives by the NIH demonstrate a commitment to advancing AI research and innovation in healthcare. By providing funding, promoting collaborations, and supporting data sharing and training initiatives, the NIH is actively working to encourage breakthroughs in AI that can lead to improved healthcare outcomes.

Q8. Are there AI focused collaborations between NIH and CMS?

 

While there might not be any direct AI-focused collaborations between the National Institutes of Health (NIH) and the Centers for Medicare & Medicaid Services (CMS), these agencies work together in certain areas where AI plays a crucial role. The primary goal of these collaborations is to improve healthcare outcomes, reduce costs, and promote the adoption of innovative technologies, including AI. Some indirect collaborations include:

1. Data sharing initiatives: Both the NIH and CMS have taken steps to make healthcare data more accessible to researchers and developers. By promoting data sharing and the use of standardized data formats, these agencies can help AI researchers access the datasets they need to develop and validate AI-driven solutions for healthcare.
2. Quality improvement programs: CMS runs various quality improvement programs aimed at enhancing patient care and reducing costs in the Medicare and Medicaid systems. Some of these programs, such as the Quality Payment Program (QPP), encourage healthcare providers to adopt innovative technologies, including AI, to improve patient care. The NIH, through its research and support for AI in healthcare, can indirectly contribute to these quality improvement efforts.
3. Interagency collaborations: Both the NIH and CMS participate in interagency collaborations that involve AI in healthcare. For example, they may work together in initiatives led by the Office of the National Coordinator for Health Information Technology (ONC), which focuses on promoting the adoption of health IT and fostering innovation in healthcare. In this context, both agencies may contribute to AI-related discussions, policy development, and research directions.
4. Support for AI research and development: While the NIH primarily funds and promotes AI research in healthcare, CMS can play a role in adopting AI-driven solutions through reimbursement policies and coverage decisions. By working together, both agencies can create an environment that encourages the development, validation, and integration of AI in healthcare systems.

Although there might not be direct AI-focused collaborations between the NIH and CMS, these agencies play complementary roles in promoting AI research and adoption in healthcare. Their joint efforts can help foster a supportive environment for AI-driven innovations, ultimately improving patient care and outcomes.

Q9. Help me compare two things. (1) There is a well known problem that biopharmas, especially new venture funded ones, have difficulty getting good ROI for novel antibiotics and may even go bankrupt after FDA approval. (2) Similarly, I am concerned that healthcare coding and coverage and reimbursement is lacking for breakthrough or out of the box AI innovations, and limiting progress in healthcare. Can you comment ?

 

Certainly, both situations involve challenges related to market dynamics and financial incentives, which can hinder the development and adoption of novel solutions in their respective fields. Let's compare the two scenarios:

1. Difficulty in achieving ROI for novel antibiotics:

• The development of new antibiotics is costly and time-consuming. However, the return on investment (ROI) for these drugs tends to be lower compared to other therapeutic areas, such as oncology or rare diseases. This is because antibiotics are typically prescribed for short durations and are often saved as a last resort to avoid the development of antibiotic resistance.
• Moreover, generic antibiotics are widely available at low prices, making it difficult for novel antibiotics to compete in the market.
• This financial disincentive has led to a decline in antibiotic research and development, as biopharmaceutical companies, particularly new venture-funded ones, may struggle to recoup their investments or even go bankrupt after FDA approval.

2. Limited coding, coverage, and reimbursement for breakthrough AI innovations in healthcare:

• Healthcare coding, coverage, and reimbursement systems, such as the Medicare RVU payment system, can be slow to adapt to new technologies like AI, leading to inadequate reimbursement for AI-driven services.
• The current reimbursement models may not account for the unique costs and benefits associated with AI technologies, and there may be a lack of specific billing codes for AI-driven services.
• This situation can create financial barriers for healthcare providers and organizations to adopt breakthrough AI innovations, as they may not receive sufficient reimbursement to cover the costs and justify their investments.

In both scenarios, the lack of appropriate financial incentives and market dynamics can limit the development and adoption of novel solutions. To address these challenges, it is essential to create a supportive environment that fosters innovation and provides appropriate financial incentives:

• For novel antibiotics, this could involve implementing new reimbursement models, such as subscription-based or value-based pricing, offering incentives for antibiotic R&D, and providing regulatory support to expedite the development and approval process.
• For breakthrough AI innovations in healthcare, this could involve updating reimbursement systems to account for the unique costs and benefits of AI technologies, creating new billing codes for AI-driven services, and incentivizing healthcare providers to invest in AI-driven solutions that can improve patient care and outcomes.

By addressing these financial barriers and aligning incentives, it is possible to encourage the development and adoption of novel antibiotics and breakthrough AI innovations in healthcare, ultimately leading to better patient care and outcomes.

Q10.  I would like to focus on your point which begins, “for breakthrough AI innovations in healthcare…” How would you explain the problem to a senior CMS policymaker who has limited knowledge about AI in healthcare? How would you help him understand the problem, prioritize the problem, and engage in solutions?

 

When explaining the problem of limited coding, coverage, and reimbursement for breakthrough AI innovations in healthcare to a senior CMS policymaker with limited knowledge about AI in healthcare, it's essential to convey the information in a clear and concise manner. Here's a suggested approach:

1. Introduce AI in healthcare: Begin by briefly explaining what AI is and how it has the potential to revolutionize healthcare by enhancing diagnostics, treatment planning, patient monitoring, and workflow efficiency. Highlight a few key examples of AI applications that have demonstrated success in improving patient outcomes and reducing costs.
2. Explain the problem: Describe how the current coding, coverage, and reimbursement systems may not account for the unique costs and benefits associated with AI technologies. Explain that healthcare providers and organizations may be hesitant to adopt AI-driven solutions if they cannot receive sufficient reimbursement to cover the costs and justify their investments. Emphasize that this situation can limit the adoption of AI innovations in healthcare, which in turn may hinder improvements in patient care and outcomes.
3. Discuss the importance of prioritizing the problem: Help the policymaker understand that addressing this issue is essential for promoting AI-driven healthcare innovations, which have the potential to significantly improve patient care, reduce healthcare costs, and increase efficiency. Emphasize that, as AI technology continues to advance, it is critical to create a supportive environment that fosters its adoption in healthcare systems.
4. Engage in solutions: Suggest potential solutions to the problem, such as: a. Updating reimbursement systems to account for the unique costs and benefits of AI technologies, which could involve exploring alternative payment models, like value-based or bundled payments. b. Developing new billing codes specifically for AI-driven services, ensuring that providers can receive appropriate reimbursement for AI-enabled care. c. Collaborating with stakeholders, including AI developers, healthcare providers, and professional organizations, to develop guidelines and standards for the integration of AI in healthcare. d. Encouraging pilot programs and studies to evaluate the real-world impact of AI innovations on patient outcomes and healthcare costs, providing evidence to support future reimbursement decisions.
5. Offer support and collaboration: Offer your expertise and knowledge in AI in healthcare to help the policymaker navigate this complex issue. Suggest working together to explore solutions, gather input from stakeholders, and develop policy recommendations that can help create a supportive environment for AI adoption in healthcare systems.

By explaining the problem clearly and emphasizing the potential benefits of addressing it, you can help the senior CMS policymaker understand the importance of tackling this issue and engaging in solutions that promote the adoption of AI innovations in healthcare.

Q11. Let’s assume you have been meeting with the senior CMS policymaker, and your presentation has covered these five points. He seems a little disengaged and you are concerned that he won’t do much once the meeting is over. What would you do next? Would you focus on this person, or seek higher level meetings, or turn to other government bodies, what is missing?

 

If the senior CMS policymaker seems disengaged and you are concerned about the potential lack of follow-up, you can take several additional steps to ensure the issue remains on the agenda and gains momentum. Here are some suggestions:

1. Recap and address concerns: Before ending the meeting, briefly recap the main points discussed and emphasize the urgency of addressing the issue. Ask the policymaker if they have any concerns or questions that you could address, and make sure to clarify any misunderstandings or provide additional information as needed.
2. Offer follow-up materials: Provide the policymaker with follow-up materials, such as a one-page summary of the key points, relevant research articles, or case studies that demonstrate the impact of AI in healthcare. This information can help reinforce the importance of the issue and serve as a reference for future discussions.
3. Establish a follow-up plan: Suggest scheduling a follow-up meeting or call to discuss progress and any updates on the issue. This can help keep the momentum going and demonstrate your commitment to addressing the problem.
4. Engage other stakeholders: Consider reaching out to other stakeholders within CMS, such as other policymakers, program managers, or department heads who might have a vested interest in AI adoption in healthcare. Share your concerns and insights with them and explore potential collaboration opportunities.
5. Seek higher-level meetings: If you believe that your message is not being effectively communicated or prioritized, consider requesting a meeting with higher-level officials within CMS or the Department of Health and Human Services (HHS). Present your case to them and emphasize the importance of addressing the issue for the future of healthcare.
6. Collaborate with external organizations: Partner with professional organizations, AI developers, healthcare providers, and patient advocacy groups to build a coalition that can raise awareness of the issue and advocate for policy changes. By working together, you can create a stronger and more unified voice in support of AI adoption in healthcare.
7. Engage with other government bodies: Explore opportunities to collaborate with other government agencies that have an interest in AI and healthcare, such as the NIH, the FDA, or the ONC. By working together, you can leverage their expertise and resources to address the issue and promote the development and adoption of AI innovations in healthcare.

By taking these additional steps, you can increase the likelihood that the issue of limited coding, coverage, and reimbursement for breakthrough AI innovations in healthcare remains on the agenda and gains the attention it deserves. Building support from various stakeholders and keeping the conversation going are crucial to driving meaningful change in policy and practice.

Q12. Your help is excellent. I find that higher level policymakers should take a broader view of public health goals, yet, they are farther from details like coding or RVU rules where the problems for reimbursements lie. How can we overcome this dilemma?

 

To overcome the dilemma of higher-level policymakers having a broader view of public health goals while being farther from the details like coding or RVU rules, you can take the following steps to bridge the gap and effectively communicate the importance of addressing reimbursement challenges for AI-driven healthcare solutions:

1. Simplify the message: Translate complex issues related to coding and RVU rules into simpler, more relatable messages that emphasize the impact on patient care, healthcare costs, and overall public health goals. Use clear language and focus on the broader implications of addressing reimbursement challenges for AI adoption in healthcare.
2. Quantify the impact: Present data and real-world examples that quantify the potential impact of AI innovations on patient outcomes, cost savings, and healthcare efficiency. Use compelling statistics and case studies to demonstrate the value of investing in AI-driven solutions and the risks of not addressing reimbursement challenges.
3. Connect to broader policy goals: Link the issue of AI reimbursement to broader public health goals, such as improving patient outcomes, reducing healthcare disparities, and promoting the adoption of innovative technologies. Explain how addressing reimbursement challenges for AI can contribute to these larger objectives and align with the policymakers' priorities.
4. Develop targeted policy recommendations: Develop specific, actionable policy recommendations that can address the reimbursement challenges for AI-driven healthcare solutions. By presenting clear and feasible solutions, you can help policymakers understand how changes in coding and RVU rules can lead to better healthcare outcomes without getting lost in the technical details.
5. Build a coalition of support: Engage with various stakeholders, including AI developers, healthcare providers, professional organizations, and patient advocacy groups, to create a unified voice advocating for policy changes. A strong coalition can help communicate the importance of the issue to policymakers at various levels and increase the likelihood of driving meaningful change.
6. Encourage cross-functional collaboration: Facilitate collaboration between policymakers who focus on broader public health goals and those who are more involved in the technical aspects of healthcare reimbursement. Encourage cross-functional discussions and working groups that can bridge the gap between the high-level goals and the specific coding or RVU rules.
7. Ongoing communication and education: Maintain an ongoing dialogue with policymakers to keep them informed about advancements in AI, the impact of AI-driven healthcare solutions, and the evolving challenges related to reimbursement. Provide educational resources, host workshops, or organize briefings to ensure they stay informed and engaged in the issue.

By simplifying the message, connecting the issue to broader policy goals, and facilitating collaboration and communication, you can help higher-level policymakers understand the importance of addressing reimbursement challenges for AI-driven healthcare solutions and encourage them to take action to support innovation in the field.

 

Q13.  Thank you. Considering the whole range of 12 questions we have discussed above, do you have any concluding comments for me? Is there anything I should have asked you, but didn't?

 

Over the course of our discussion, we have covered various aspects of AI in healthcare, including its applications in radiology and pathology, the role of the American Medical Association, NIH and CMS collaborations, and the challenges related to Medicare reimbursement for AI-driven services. We have also explored strategies for engaging policymakers and driving meaningful change in this area.

As a concluding comment, I would like to emphasize that AI has the potential to revolutionize healthcare, but its adoption and impact largely depend on creating a supportive environment that addresses the financial, regulatory, and technical barriers to its integration. In this context, it is crucial to work collaboratively with various stakeholders, including policymakers, healthcare providers, AI developers, and professional organizations, to ensure that AI-driven innovations can effectively contribute to improving patient care and outcomes.

As for any questions that were not asked, you might consider exploring the following topics in future discussions:

1. Ethical considerations related to AI in healthcare, such as privacy, data security, and algorithmic bias.
2. The role of AI in addressing healthcare disparities and promoting equitable access to care.
3. How AI can be integrated into medical education and training to prepare healthcare professionals for the future.

These additional areas of inquiry can provide further insights into the broader implications of AI adoption in healthcare and help guide your efforts to address the challenges and opportunities associated with this rapidly evolving technology.

 

 ##

Bonus link.  GPT4 also provided a 90 second script for a teaser video promoting this white paper interview.  Here.

Bonus link.  You may have heard GPT can help you with Excel functions or other computer tasks . I posted some of my own Q&A in this area in a Google Doc.