We're going to have now, Dr. Kan come back to the podium and answer the little question about who is exactly going to pay for AI as you bring into the medicine field. Thanks. For all the vendors in the room, I guess I want to say I will pay for AI. Anybody out there will. Radiology actually has, you know, kind of a model in some sense for who will pay for AI. And for a long time, it was the federal government. The MQSA, Mammography Quality Standards Act, actually provided an add-on payment for the use of computer-aided detection in mammography. The challenge with that is that if you ask most mammographers, would you use CAD, otherwise they would answer no. And it's kind of telling. Part of it is, you know, unlike sort of the example we saw earlier of, you know, using the little green bounding box to draw your attention or to help guide your fellows to the lesion, most of the time the mammographers found that they had to remove the little CAD marks from the images, the things that CAD was detecting. But I'm just going to talk a little bit about some of these things like CMS has this NTAP, the new technology add-on payment. We have our current fee-for-service healthcare model. And I think I like the discussion that you had about some of the opportunities we have, because truthfully, a lot of it really hinges on improving our productivity as physicians or in our health system. A large part of where we're seeing the work by companies and kind of part of the pitch is to drive patients to procedures that are higher margin. And then looking beyond that to value-driven healthcare, value-based care, trying to figure out where these things add value. So we had a, I'm not an author of this one, but here's, these folks actually are, we're all members of the RUC, the group that sets the RVUs and represented radiology in that at one point or another. So they know a whole lot more about payment mechanisms than I do. But their conclusion was sustained adoption of AI with our current reimbursement framework may be challenging in fee-for-service. And then as these value-based payment models mature, it becomes increasingly important for us to be able to deliver care and high-quality care at decreased costs, and AI may be a valuable tool in that going forward. So with many of these, as you may know, we have the fee schedule and there are kind of two opportunities. You can either assign a new CPT code to the particular use of AI, and that was done for the tool that was used for diagnosis of diabetic retinopathy, where they actually created a new CPT for the use of this tool. Or within the inpatient prospective payment system, there's this NTAP, new technology add-on payment. And that's the one you may have heard in the news that Viz AI, which was the first one, got that for large vessel occlusion in stroke. Now, the way they did that was not by proving that their system detected stroke better than human or anything like that, but that it sped the patient to diagnosis, and in stroke, time is brain. So what you have to be able to show is that the tool that you have is going to yield better outcomes. Now, there's a little bit of a trap with this NTAP. It's designed as a temporary fix. It basically says you're adopting some new technology that improves care, and Medicare CMS decides that they will support the payment of it. But basically, it's a kind of a temporary bump until they revalue the procedure in terms of RVUs. So you could imagine, at some point, once a majority of institutions have adopted some form of AI, then CMS, rather than kind of putting a bump on it for using, is just going to fold that into the calculation of the resources and other things that go into it. I'm not going to go into this one so much, but just to kind of give you the, you know, these are the mechanisms that are available, certainly, within the federal payers, and the thing that's important to note about it is that a lot of the private payers will follow the lead of Medicare Medicaid when they can in terms of reimbursement of these technologies. So for the Medicare fee schedule, that is all hinged on current procedural technology CPTs. For the hospital outpatient system, we have ambulatory payment codes and this transitional pass-through payment. Most of that, we haven't seen any changes with AI tools. I mentioned the inpatient system. That's the one that has the DRGs, and then this new technology add-on payment that people can, they can accommodate the use of it. And then other ones, this Medicare coverage of innovative technology, they can, FDA, and I'm sure our FDA expert can probably provide more detail about that than I can, but notion of covering innovative technologies. One of the other questions in this is, where does the money go? Who actually should get paid? This was, this is a commentary, actually, I think it's going to come online on, oh no, it came out, yeah, that came online on the beginning of this month. One of my deputy editors, Ronnie Sebro, who will get paid for it. And the question is, who actually owns a stake in AI systems and therefore should get some reimbursement out of it? Is it the patient who, you know, hard to argue when you have tens or hundreds of thousands of patients, right, who all they had was a chest X-ray or CT scan or a flood sample that got folded into some AI system? Is it the healthcare professionals? The AI systems that have been built have all been done off of readings of gastroenterologists and radiologists. Is it the healthcare systems that own the data? Is it the health insurance companies that paid for the acquisition of those clinical studies? Or is it just going to be to the AI companies and the developers? And that is a question in itself as well. And then this whole question of payment gets tied up, I think, into an even broader question, and that's this question of legal liability. One of the differences between you all and us in radiology is if you ask a radiologist, why do you keep all the images, they'll say, for legal reasons. You ask a gastroenterologist, you know, why did you only keep those three stills from your 30-minute endoscopy, and they say, for legal reasons. But different expectations of what it is that we're supposed to preserve and keep, right, so that we can get sued. You know, as somebody pointed out, you know, here's your nightmare scenario. Three patients in your practice had missed lesions on their mammograms. So now an enterprising attorney comes in and offers up a class action lawsuit that basically says we're going to come in, we're going to retrospectively apply AI to your last five years of mammograms. We're going to find all the cancers that you missed, and then those people will become part of that class to that lawsuit. This is what, you know, keeps all of us from, you know, we live in the dark anyway in radiology, so our sleep-wake cycles are highly disrupted. But this is, you know, one of the things that kind of keeps us awake. But the challenge for us is how do we address these things in terms of liability? If AI misses something, was it AI, was it me? The usual, and it's actually rather fascinating to get into the whole law of this because it involves product liability law and malpractice law, and I'm not an attorney, but it really poses quite a number of challenges. This is one thing, and I've been looking to put together actually an op-ed piece to kind of raise this as a question, but this was fascinating. I came across this article, and Mercedes, and this I believe is just in Germany, is going to take lead, they are going to indemnify drivers when you're driving one of their cars that's in full auto mode, which kind of goes, you know, this is sort of different from the way we are, right? If I miss something on an image, and whether AI told me to look for it or not, I'm still responsible for that, and I'm the one who's going to get sued, and, you know, it's kind of, you know, who gets sued, and then you make a mistake, well, yeah, the AI company and the radiologist and the hospital, they're going to go where the money is, right? But what's fascinating about this is, is this the model that we have to adopt for AI truly to come into clinical practice, in other words, if the AI companies effectively indemnify the use of their product, recognizing that just as with driving a car, there are going to be accidents, and there are going to be damages, and maybe the way to deal with that is just to create an insurance model around it. I don't know. I don't know the answer to that. But I think it's something for those of us who are looking to adopt AI into clinical practice want to think about anyway. So who will pay for AI? Unlike in mammography, the federal government, I don't, I wouldn't count on it. You know, there's a lot of interest, but for in clinical practice, it ain't going to happen. For those of us who practice, the hope is that we can use these tools to improve our productivity, to reduce our malpractice risk, and that's a significant savings, and there is the opportunity that it allows us to move to higher value procedures by eliminating some of the kind of more routine stuff. And again, for health systems, if they're vertically integrated, that they have the opportunity to recognize those earnings and to use that, and as well to decrease their system-wide costs. So with that, lots to think about. It's all kind of an unknown landscape, and thanks very much.

payment mechanisms

AI in medical field

radiology

payment models

AI adoption