February 2024—Artificial intelligence and Medicare Advantage contracts were at the center of the Jan. 2 Compass Group virtual roundtable led by CAP TODAY publisher Bob McGonnagle.
“If you want to get into the AI world, there are many lanes you can swim in,” said Michael Feldman, MD, PhD, of Indiana University School of Medicine. And on the matter of money, Stan Schofield, Compass Group VP and managing principal, said almost 60 percent of hospitals in the U.S. remain in the red, “and the shortfalls in volume can’t be made up with Medicare Advantage.”
The Compass Group is an organization of not-for-profit IDN system lab leaders who collaborate to identify and share best practices and strategies. Here’s more of what they said at the start of the new year.
The hot-button issues of 2023 were artificial intelligence, consolidation of testing, and laboratory-developed tests. Becky Simmons, which of those do you think will continue to be a hot topic this year?
Becky Simmons, chief operating officer, NorDx and MaineHealth: Artificial intelligence, because the staffing headwinds we’ve been experiencing are not waning. We have to be aggressive in how we’re going to compensate for that.
Tony Bull, what do you think the surviving buzzwords will be?
Tony Bull, system administrative officer, pathology and laboratory medicine, Medical University of South Carolina: Laboratory-developed tests. The FDA is going forward with its rule. I expect it to go to court, but we’ll be dealing with it for a while.
Tom Johnson, what will the hot-button issues be this year?
Thomas Johnson, PharmD, MBA, VP of hospital pharmacy and laboratory services, Avera Health, Sioux Falls, SD: We’ll see what the federal government does with LDTs. I can’t imagine they’re going to move forward without so many lawsuits that they won’t know what to do next.
Many entities are trying to cancel their contracts with Medicare Advantage organizations. Several have done analyses that show they’re losing money by accepting Medicare Advantage contracts. That’s close to the heart of pathologists and laboratories because it’s hard to get paid in some of these Medicare Advantage situations. Are you seeing that as well?
Tom Johnson (Avera): Yes, in general. As more restrictive Part C plans come out, many people aren’t sure what they’re signing up for and are then surprised by what is covered. Those of us on the other end trying to get paid are surprised by what isn’t covered and by the amount of work it takes to get covered. It leaves the provider with a couple of options—deal with it or give up and then see if you can put the pressure back on people to select different plans. The problem is it’s not easy to get back into the other plans once you’ve moved to a Part C. The patient gets the short end here. I hope our managed care colleagues can work with our systems to make sure the best rates have been negotiated for to ensure we all stay in business.
Dan Mumm, what issues do you think will remain hot, and what are your thoughts about Medicare Advantage?
Dan Mumm, president/group VP, ACL Laboratories, Advocate Health, Racine, Wis.: LDTs are going to stay hot as will AI. It’s a lot of hype, but buckle in because we’ll turn around and it will be in more places than we anticipate or see now.
We’re struggling with the same issues with Medicare Advantage that Tom described. The negotiation between payers and how we get paid after it’s sent to us is a tough road.
Do you have tips for someone who is dealing with this problem?
Dan Mumm (Advocate): Make sure you’re prepared to respond to it. We have groups of people who negotiate with payers that are separate from our laboratory. Collaborating closely with them is important because sometimes, at least from a laboratory perspective, they’ll negotiate something off the laboratory piece to get a concession somewhere else. They don’t understand how devastating that is overall.
Do they tell you?
Dan Mumm (Advocate): We usually find out after.
Medicare Advantage has had good PR with the general public for a long time. It’s marketed as being cost-effective and better than traditional Medicare. Its image may slip this year given these pressures. Stan Schofield, do you agree that Medicare Advantage may be in the crosshairs for the first time?
Stan Schofield, VP and managing principal of the Compass Group (formerly of NorDx/MaineHealth): Absolutely. Everyone went into Medicare Advantage saying, I need an easy solution, I don’t have chronic health conditions, and I don’t travel often to multiple states, so I can stay home and have a solid plan with an AAA or A+ provider and a payer system that works. Everyone signed up on these agreements, and then hospitals and health systems have had massive cost increases and what they’re getting paid isn’t covering the bill. Health care system total cost expenses are up approximately 14 percent over the past two years. These agreements are usually two years and they’re coming up on the short end because insurance companies are saying, Your costs are more. We’re capping this. Hospitals are fighting back and patients are getting caught in the middle. Having a good Medicare Advantage is not as easy as it once was with the macroeconomic dilemma we continue to see in health care.
Nearly 60 percent of hospitals are still in the red in the United States, and the shortfalls in volume can’t be made up with Medicare Advantage. You need high-ticket, full insurance procedures to swing the beds and the economics. No new revenue streams are coming into hospitals, so it’s all about cutting expenses and cutting the back office through mergers and acquisitions. The infrastructure to the billing system and billing and computer staff are all on the block to try to reduce operating expenses.
There’s not enough money to go around and the population isn’t as healthy or procedure free as early data had indicated. Most data on utilization and cost per member per month have turned out to be inaccurate and woefully short of what is required to properly care for a patient with medical necessities.
What advice do you have for a laboratory or pathology operation that is seeing large shortfalls in reimbursement for their services to patients in a Medicare Advantage plan?
Stan Schofield (formerly of MaineHealth): Cut your losses. They don’t understand unless you walk away.
Wally Henricks, what will the buzzwords be this year?
Walter Henricks, MD, vice chair, Pathology and Laboratory Medicine Institute, and laboratory director, Cleveland Clinic: AI will continue to be one that many are interested in. Hopefully there will be shifts in how it is managed. Number one would be an increasing emphasis on the validation of AI’s meaningfulness and value as it moves into clinical use. Pathologists and laboratory people are in a good position and have an expertise base to judge that because in many ways the same questions are asked about the validation of our own tests.
Organizations also need to look at how they govern the introduction of AI and machine learning algorithms into the clinical space and make sure it’s not the Wild West. How do you get an algorithm development project approved to be started? How do you show it’s validated? How do you introduce it into clinical use? What are the parameters for those things? AI governance overlaps with traditional data governance, but there are also unique features.
The less exciting buzzwords are staffing, the expenses that go with staffing, and consolidation, including the integration that goes with consolidation and its hidden costs.
The Cleveland Clinic does a lot of research on technologies and in IT. Would you feel comfortable sitting on a panel to come up with protocols for testing AI-based software, devices, and decision support systems, or help someone on an institutional review board do so? In other words, is this an extension of expertise that exists now on your campus? Or will it require new people?
Dr. Henricks (Cleveland Clinic): It can be an extension of what’s here. It’s important to ask developers the nontechnical, fundamental questions: What’s the source of truth? How do you know that what you’re doing works for what you claim it does? How different is your test environment from your real-life validation? How will you look at performance drift over time? Ask good questions, as you would before introducing any intervention or laboratory test. It’s about not the programming but the outcome and what it will do for decision-making.
Steve Carroll, can you comment on AI and how you define studies for it?
Steven Carroll, MD, PhD, chair, Department of Pathology and Laboratory Medicine, Medical University of South Carolina: We’re wrestling with this because we see a need for it. It’s being driven by shortages in the hospital laboratory and the growing shortage of pathologists. We need ways to extend their work capabilities until we get more people in the pipeline, which will take years. The question then becomes how do you know you’re getting what you really need out of the AI? What validation studies do we have to run to tackle that problem? We’re looking at breast AI algorithms that seem to be performing well.
I see a lot of discussions of AI in journals and magazines but not many studies of it, of the type in which samples are split into two groups and one has the benefit of AI mediation and the other doesn’t, and here are the results. Do you think some of that is to come this year?
Dr. Carroll (MUSC): Yes. We need those analyses so we can have confidence moving forward. I’m asking my informatics division to look at this.
Dan Mumm, is AI going to be essential in a big system like yours?
Dan Mumm (Advocate): Yes. It’s health care, so we have to prove it. If we’re going to introduce a different concept or a way of doing something, we have to validate that it’s at least as good as or better than what we do. Of the three types of AI—narrow, general, or super—we’re still at narrow. That may advance quickly. We have to be nimble in how we look at it and how the basic research goes to applied research and then is implemented clinically.
Rick Vander Heide, what is your take on what we’ve been discussing today?
Richard Vander Heide, MD, PhD, MBA, medical director of pathology and laboratory medicine service line, Marshfield Clinic Health System, Marshfield, Wis.: As far as AI, one of the biggest things we, as pathologists, don’t yet understand is what needs to go into the learning phase. What do we need to put into this AI box to get a more predictive output on the other side? Anatomic pathology-driven prognoses are largely retrospective; histologic features are observed and categorized over the course of many studies and outcomes are reexamined when new immunomarkers are discovered. We don’t know what we don’t know, if you will. The same is true with whole genome sequencing testing, in my opinion. Some features that we currently consider unimportant might be more predictive if combined with other data in AI-driven models. The challenge with AI will be to determine if the output is greater than the sum of the inputs.
Mike Feldman, you’re standing up an important component of computational pathology at Indiana University. Do these developments in computational pathology, AI, and machine learning accelerate what you might call this division of pathology and laboratory medicine into large centers that then serve a huge geography, as yours does?
Michael Feldman, MD, PhD, chair, Department of Pathology and Laboratory Medicine, Indiana University School of Medicine: If you want to get into the AI world, there are many lanes you can swim in, but you need to have proprietors on your team who speak the language, understand the mathematics, the statistics, and how to evaluate this, and can partner with the pathologists. We need to come up with IQCPs but geared toward AI as a test. We’ve started a division of computational pathology to focus on the research, and they’re partnering with the biostats group, the genetics group, mathematicians, and physicists. We’re thinking about what the next grants are and what projects could be valuable. At the same time, Dr. Spyridon Bakas, who’s running the division, works with us clinically when we look at vendors and technology companies. He’s not a physician or a pathologist, but he has a PhD in mathematics so he understands the inner workings of generative AI and how to evaluate these tools. He’s the first hire and we’re going to hire one or two more. Pathology departments that want to make a concerted effort in the science of this space need to go down this route, and we’re seeing other departments doing that.
Are the medical students who are interested in pathology also interested in AI and digital pathology and machine learning? Or is that still a minority self-selection as they move into their pathology selection?
Dr. Feldman (IU School of Medicine): There are more people coming into pathology who have backgrounds in mathematics, engineering, and physics, so when you find someone like that, you have to bring them in and show them that their science plus these clinical opportunities is a future career for them plus 10 of their closest friends.
Medical laboratory science is another focus for AI. The opportunity to make your technologists more efficient will be easier than the barrier of more hires. We have installed a device that can digitize a blood smear and provide a whole slide digital view of that smear, which is a huge benefit over single field of view systems. Our first device was a whole slide imager for blood smears with decision support. The technologists can read a blood smear now in seconds as opposed to minutes. If you have thousands of those a day that have to be reviewed, it’s a big win.
Are people comfortable with adopting those applications?
Dr. Feldman (IU School of Medicine): Yes. Five minutes at the microscope or 30 seconds at the computer. You choose.