AI Death Panels coming to Medicare

[Gfunk6]

Advertisement - Members don't see this ad

Medicare Will Require Prior Approval for Certain Procedures

A pilot program in six states will use a tactic employed by private insurers that has been heavily criticized for delaying and denying medical care.

www.nytimes.com

Summary:

1. Pilot program, starting in 2026 - lasts six years
2. Includes traditional Medicare only in six state: Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington State
3. Focuses only on a dozen procedures that have "low value" to patients including: devices for incontinence control, cervical fusion, certain steroid injections for pain management, select nerve stimulators and the diagnosis and treatment of impotence.
4. AI makes the initial determination and human signs off
5. The feds contract out this service and companies are incentivized to deny

Welcome to your future.

 

[xrt123]

Please take on SpaceOAR AI. Tell us the value.

 

[medgator]

Please take on SpaceOAR AI. Tell us the value.

Small potatoes

Protons in PCa

 

[yesmaster]

It’s not a bad idea honestly. A lot of grift and low value procedures getting reimbursed these days.

Wake me up when they start denying screening colonoscopies or tPA.

 

[TheWallnerus]

diagnosis and treatment of impotence

AI tackling impotence?

That’s not gonna be hard.

 

[deleted1217978]

The IRS has been publishing their "dirty dozen" tax schemes for years. Not surprising to see CMS doing it now too given that our govt is broke. We are using a shot glass to bail water out of a sinking boat going after fraud like this but i guess it's better than doing nothing at all because of the "muh social security" third rail.

Anybody who is treating a lot of OA with LDRT can tell you what a scam steroid injections are.

Protons for prostate also pale in comparison to the billions med onc diverts from CMS to pharma with a small vig for people with weeks to live.

 

[communitydoc13]

This initiative is limited and in and of itself is not that concerning. The above interventions really are likely to be low value IMO.

However, for us to remain human, we must preserve a meaningful space for human judgement in opposition to AI recommendations.

I recently had a P2P regarding SBRT for a single oligometastases in pCa. I referenced the ESTRO-ACROP Delphi consensus paper from 2022. I asked the reviewer the pointed question, "In this clinical scenario, what is your professional opinion regarding best next clinical measure? Do we believe that there is a very low likelihood of clinical benefit (e.g. prolonged biochemical control or avoidance of symptomatic progression in the future) to radiating this lesion?"

He refused to answer the question. He had abdicated his role as a human expert.

I eventually got it approved.

Efficiency ultimately becomes the enemy of humanity. We want highly inefficient ratios regarding many things. Teacher to student ratio being an obvious example.

 

[yesmaster]

I would say that has less to do with AI and more to do with how the insurance and UM companies operate. A pure AI might be able to go through the reasoning steps with you whereas the dumb human component in your anecdote is the annoying part.

 

[communitydoc13]

AI might be able to go through the reasoning steps

Call me skeptical.

As far as I can tell, AI has no better tethering to the elusive "truth" than humans do. It can be systematically manipulated (as humans can be). It certainly can reference enormous amounts of data in real time for presentation.

It is not clear to me that any "reasoning" goes on at all with AI, although AI is typically very "reasonable"...sort of how it works.

An AI algorithm that demonstrates susceptibility to clever humans in terms of approval will likely be tweaked.

I prefer people making decisions about people.

 

[OTN]

This initiative is limited and in and of itself is not that concerning. The above interventions really are likely to be low value IMO.

However, for us to remain human, we must preserve a meaningful space for human judgement in opposition to AI recommendations.

I recently had a P2P regarding SBRT for a single oligometastases in pCa. I referenced the ESTRO-ACROP Delphi consensus paper from 2022. I asked the reviewer the pointed question, "In this clinical scenario, what is your professional opinion regarding best next clinical measure? Do we believe that there is a very low likelihood of clinical benefit (e.g. prolonged biochemical control or avoidance of symptomatic progression in the future) to radiating this lesion?"

He refused to answer the question. He had abdicated his role as a human expert.

I eventually got it approved.

Efficiency ultimately becomes the enemy of humanity. We want highly inefficient ratios regarding many things. Teacher to student ratio being an obvious example.

I stopped trying to have a conversation with P2P docs years ago. Let's just get through this as quickly as we can, we all know you're going to say "I can't approve this due to guidelines", then I'm going to appeal again, and it will then get approved. I make sure they know that's what's going to happen.

 

[Gfunk6]

I stopped trying to have a conversation with P2P docs years ago. Let's just get through this as quickly as we can, we all know you're going to say "I can't approve this due to guidelines", then I'm going to appeal again, and it will then get approved. I make sure they know that's what's going to happen.

Precisely. There was an era some years ago where P2P meant legitimate discussion amongst peers. Now it is just deny, deny, deny. They won't debate you, render their own opinon, or comment on the merits of the case. All responses will be a variant of, "this piece of paper says you can't do it." The only thing the P2P docs are hoping to avoid is verbal abuse from you.

Eventually, you appeal and it will usually get overturned. But this takes many days or even a few weeks. During this time, the insurance company will hope one of the following happens:

1. You get frustrated and decide to switch to a cheaper treatment or (better yet) not treat the patient at all
2. The patient gets a denial in the mail, panics and wonders "why is this doctor treating me with experimental or non-standard of care" and opts to cancel treatment
3. The patient dies during the appeal/auth period of cancer or *anything* else

That's it. It is a cold calculus. I agree with the posters above who said that AI will at least take the indignity of talking to an unsymphatetic human out of the equation.

 

[radiaterMike]

Call me skeptical.

As far as I can tell, AI has no better tethering to the elusive "truth" than humans do. It can be systematically manipulated (as humans can be). It certainly can reference enormous amounts of data in real time for presentation.

It is not clear to me that any "reasoning" goes on at all with AI, although AI is typically very "reasonable"...sort of how it works.

An AI algorithm that demonstrates susceptibility to clever humans in terms of approval will likely be tweaked.

I prefer people making decisions about people.

Conceivably AI has the ability to learn and adapt. Not sure if that applies to some of the P2P folks I've encountered.

 

[Mandelin Rain]

Call me skeptical.

As far as I can tell, AI has no better tethering to the elusive "truth" than humans do. It can be systematically manipulated (as humans can be). It certainly can reference enormous amounts of data in real time for presentation.

It is not clear to me that any "reasoning" goes on at all with AI, although AI is typically very "reasonable"...sort of how it works.

An AI algorithm that demonstrates susceptibility to clever humans in terms of approval will likely be tweaked.

I prefer people making decisions about people.

Current administration's AI "tweaks".... medicine doesn't work at all on black or brown people, those who've aged out of the workforce, and/or can no longer become pregnant. Also, vaccines cause mass shootings, not guns. Mitochondrial dysfunction visible to the naked eye, is the biggest health threat in America.

 

[Gfunk6]

Conceivably AI has the ability to learn and adapt. Not sure if that applies to some of the P2P folks I've encountered.

LOL, do you think that AI will actually be used to reason out if the auth is legitimate? AI is just window dressing to say "denied" to everything but with the legitimacy of a LLM.

Here is a sample coding script insurance will use:

WHEN a claim arrives:
SAY "Denied." [Default setting: No.]

IF the patient appeals:
SAY "Still denied." [No, but with confidence.]

IF the doctor writes a long letter full of evidence:
REPLY "Please submit more documentation."
LOOP back to start [Endless paperwork circle.]

IF photos show an obvious emergency (e.g., bone saying hello):
REQUEST "Itemized receipt, proof of gravity, notarized X-ray."
LOOP back to start [Delay is the spice of life.]

IF the claim is from an in-network hospital:
SAY "Out-of-network." [Network is a feeling, not a map.]

IF the claim is from an out-of-network hospital:
SAY "Should've been in-network." [See above.]

IF the patient used a generic drug to save money:
SAY "Brand required." [Savings are suspicious.]

IF the patient used the brand:
SAY "Generic required." [Consistency is key.]

IF the appeal reaches an independent reviewer:
CALL THIS "Processing error."
APPROVE a small part (bandage).
DENY the rest (everything that costs money).

IF a regulator asks questions:
SWITCH MODE to "We care deeply™."
SAY "Approved pending review."
AFTER regulator leaves:
SWITCH MODE back to DenyEverything [System restores factory settings.]

ON patient exhaustion (no more calls, no more letters):
MARK claim as "Resolved."
THROW a tiny confetti emoji internally. 🎉 [Morale is important.]

END PROGRAM

 

[radiaterMike]

well AI is just a tool (as are some P2Ps). But if AI ever becomes truly autonomous (and not window dressing for the insurers) I think it could learn to be objective and incorporate guidelines and published data. Unfortunately denials, appeals and P2P will continue so long as people are submitting daily IGRT and sims for skin cancers, IMRT for palliation of simple bone mets in patients with widespread disease, routine use of proton therapy for prostate cancer, 20+ fraction proton therapy for bone metastases, special treatment procedures when already bundled, etc

 

[Gfunk6]

well AI is just a tool (as are some P2Ps). But if AI ever becomes truly autonomous (and not window dressing for the insurers) I think it could learn to be objective and incorporate guidelines and published data. Unfortunately denials, appeals and P2P will continue so long as people are submitting daily IGRT and sims for skin cancers, IMRT for palliation of simple bone mets in patients with widespread disease, routine use of proton therapy for prostate cancer, 20+ fraction proton therapy for bone metastases, special treatment procedures when already bundled, etc

The fallacy in your statement is the assumpiton that the insurance company is trying to do something of worth to the patients (e.g. incorporate guidelines and/or be obective). They. Do. Not. Care. It is all about delaying for as long as possible.

 

[NotMattSpraker]

The IRS has been publishing their "dirty dozen" tax schemes for years. Not surprising to see CMS doing it now too given that our govt is broke. We are using a shot glass to bail water out of a sinking boat going after fraud like this but i guess it's better than doing nothing at all because of the "muh social security" third rail.

Anybody who is treating a lot of OA with LDRT can tell you what a scam steroid injections are.

Protons for prostate also pale in comparison to the billions med onc diverts from CMS to pharma with a small vig for people with weeks to live.

I have had some traction getting some referrals from orthos finally. They will inject the patient then send to me literally in the same visit. Kind of goofy. It seems like this has become an alternative to PRP, which is cash pay and expensive outside the knee. Maybe an overall good through that lens.

Small N, but I've had success with LDRT in these patients.

 

[medgator]

. It is all about delaying for as long as possible.

The float

 

[evilbooyaa]

I stopped trying to have a conversation with P2P docs years ago. Let's just get through this as quickly as we can, we all know you're going to say "I can't approve this due to guidelines", then I'm going to appeal again, and it will then get approved. I make sure they know that's what's going to happen.

Precisely. There was an era some years ago where P2P meant legitimate discussion amongst peers. Now it is just deny, deny, deny. They won't debate you, render their own opinon, or comment on the merits of the case. All responses will be a variant of, "this piece of paper says you can't do it." The only thing the P2P docs are hoping to avoid is verbal abuse from you.

Eventually, you appeal and it will usually get overturned. But this takes many days or even a few weeks. During this time, the insurance company will hope one of the following happens:

1. You get frustrated and decide to switch to a cheaper treatment or (better yet) not treat the patient at all
2. The patient gets a denial in the mail, panics and wonders "why is this doctor treating me with experimental or non-standard of care" and opts to cancel treatment
3. The patient dies during the appeal/auth period of cancer or *anything* else

That's it. It is a cold calculus. I agree with the posters above who said that AI will at least take the indignity of talking to an unsymphatetic human out of the equation.

I guess I haven't had to do a p2p in sometime because most of my stuff just gets approved, but generally I've beena ble to get approval by doing the p2p. Or if I can't, it's genuinely a difference of opinion and I can see where that person is coming from in that the intrepretation of data is slightly different than mine. Not saying I agree with them and we will proceed with appeal PRN, but, as frustrating as it is to have that rare p2p that is "I am reading from this piece of paper that I am paid to not interact with you", most p2ps I've done have the authority to actually approve the thing that is being requested...

 

[HolySmokes]

I guess I haven't had to do a p2p in sometime because most of my stuff just gets approved, but generally I've beena ble to get approval by doing the p2p. Or if I can't, it's genuinely a difference of opinion and I can see where that person is coming from in that the intrepretation of data is slightly different than mine. Not saying I agree with them and we will proceed with appeal PRN, but, as frustrating as it is to have that rare p2p that is "I am reading from this piece of paper that I am paid to not interact with you", most p2ps I've done have the authority to actually approve the thing that is being requested...

This has not been my experience. Not all, but about half of my peer to peer conversations are "I can document more information but I cannot approve XXX because of my company guidelines"

 

[A_DeMichele]

This has not been my experience. Not all, but about half of my peer to peer conversations are "I can document more information but I cannot approve XXX because of my company guidelines"

what p2p’s are allowed to approve as exception varies from one UM company to another, and often a UM company will change approach to that depending on its business situation at the given time

 

[evilbooyaa]

This has not been my experience. Not all, but about half of my peer to peer conversations are "I can document more information but I cannot approve XXX because of my company guidelines"

I would ask for name and credentials and document as such in the chart.

Haven't had to go to this in the post-Luigi era, but heard through the grapevine that they're more 'protective' about it.

 

[HolySmokes]

I would ask for name and credentials and document as such in the chart.

Haven't had to go to this in the post-Luigi era, but heard through the grapevine that they're more 'protective' about it.

The reviewers I have spoken with are happy to provide. They don't seem phased or dissuaded at all, unfortunately.

 

[evilbooyaa]

The reviewers I have spoken with are happy to provide. They don't seem phased or dissuaded at all, unfortunately.

That's great!

If you feel strongly enough that patient has had negatively influenced care by the decisions of a physician who has not seen nor evaluated the patient and has a bad outcome, I would encourage them to pursue a malpractice claim against the p2p physician.

 

[A_DeMichele]

That's great!

If you feel strongly enough that patient has had negatively influenced care by the decisions of a physician who has not seen nor evaluated the patient and has a bad outcome, I would encourage them to pursue a malpractice claim against the p2p physician

Lawyers won't take the case, because appeal to the health plan is the next regulatory step after denial, not the lawsuit

 

[drowsy12]

Lawyers won't take the case, because appeal to the health plan is the next regulatory step after denial, not the lawsuit

Agree

 

[evilbooyaa]

Lawyers won't take the case, because appeal to the health plan is the next regulatory step after denial, not the lawsuit

Yes and if an appeal was filed and the patient had a significant negative medical event while the appeal was in process?

 

[ramsesthenice]

Yes and if an appeal was filed and the patient had a significant negative medical event while the appeal was in process?

One of my colleagues wife is a malpractice layer and I asked what she thought. These are usually a hard sell. Insurance companies have armies of aggressive layers that sound very smart and convincing. Cases where they deny any care and something bad happens sometimes have a shot but even then it’s tough. If they offer something you don’t agree with (like 3D when you want IMRT or protons) and something happens while you wait for appeal, forget it. It doesn’t matter if you are technically right. Civil juries (like criminal juries) are made up of average people with no medical knowledge. Their lawyers only have to make an argument that sounds convincing and they are very good at it. You call them a greedy company, they call you a greedy doctor who is doing unnecessary things, and it can go either way.