How Artificial Intelligence Controls Your Health Insurance Coverage

Ultra-Processed Foods Can Drive Your Colorectal Cancer Risk (10:35)

Health Insurers use algorithms to decide whether to pay for health care treatments and services that are recommended by a given patient’s physicians

Broadcast Retirement Network's Jeffrey Snyder discusses the use of Artificial Intelligence in the Health Insurance Claims Process with Indiana University Bloomington's Jennifer Oliva

Jeffrey Snyder, Broadcast Retirement Network

This morning on BRN, how artificial intelligence controls your health coverage and joining me now to discuss this and a lot more, Jennifer Oliva is a professor of law at Indiana University. Jennifer, it's so great to see you. Thanks for joining us on the program this morning.

Thanks for having me, Jeff. Artificial intelligence, it's so pervasive now today. It's in everything, or at least you hear it's in everything that we do, from manufacturing to looking at your financial information.

It's also incorporated, Jennifer, in healthcare.

Jennifer Oliva, Indiana University Bloomington

Widely, actually. There are sort of many tracks of this, and folks have read about it, I'm sure, but you don't probably even know when you go in to see your physician how often they're using AI, ranging from typical diagnostic tests where AI is saying, hey, I'm looking at this x-ray and I can see something, or I'm looking at this blood work and I can see something, to copying and assisting doctors with just visual information. Basic patient-client conversations where the AI is copying the notes and pointing out to the doctors, here's things that the patient said versus what their chart reads or what their medical records show.

Again, an area that I'm interested in, in addition to this, is healthcare insurers, other parties in the healthcare system outside of clinicians, also use AI to make determinations about whether your claims should be paid for, how long you should get to stay in the hospital after surgery, how long you should be in a skilled nursing facility for rehabilitation. Anything you can imagine, there's some kind of AI or AI manufacturer in the background pushing products, some of which are in robust use and some of which are coming online every day, as you and I are just talking right now.

Jeffrey Snyder, Broadcast Retirement Network

I can see a lot of the benefits of AI, especially for repetitive tasks. You mentioned some of the case notes and things like that. I can certainly see the benefits there, but let's drill down a little bit more into insurance coverage.

I can see the need for efficiency, but does it present challenges for you and I and other patients, for lack of a better word, in terms of us getting our critical care?

Jennifer Oliva, Indiana University Bloomington

Yeah, I think so, and I think you hit the nail on the head here, is that there's this competing stuff where, for a long time, if insurers or payers would deny a claim and take a deep look and that kind of stuff on it, the patient's health could be degrading during that time. They could be in an acute situation where they really need to answer fast and for the physicians to be able to move. There's this huge push for them to make decisions quickly, and that's on the one hand.

On the other side, it's like, if you're going to use AI to make those decisions, what kind of data and information does the AI care about to make quick decisions? Are they making good decisions or are they just making quick decisions that still benefit the insurer? For example, denying and delaying care saves the insurer money because whatever that care cost stays in their pocket during that timeframe while that's going on, and ultimately, sadly, if a patient dies or doesn't challenge it, that money remains in the insurer's pocket.

Quick decision-making and the federal government and many insurance regulators have pressed these insurance companies to be faster, stop dragging out the process, but on the other hand, even a quick decision that's built on faulty data or problematic criteria or criteria that benefits profit-driven operations may not be improving the situation just because we're gaining quickness.

Jeffrey Snyder, Broadcast Retirement Network

I come from the retirement industry, so I think about how AI is being deployed in the retirement industry. It's making recommendations. It's saying, hey, administrator, operator, this is what you need to be thinking about.

Is that not the case? Isn't there still the human element that would take these recommendations from AI to move a claim forward or to provide long-term care? Is that still the case or is it just immediately saying, no, sorry, based on this criteria that it was programmed by a human being, by the way, no, we're not going to allow this claim?

Jennifer Oliva, Indiana University Bloomington

Well, as a general rule, the best practice is to have a medical professional who's in the kind of same specialty as the decision pertains to review it carefully before the insurer says yes, no, or more information or other things. However, even when you have a law that says that, so California just passed a new law last year that requires this. The federal government requires this in Medicare, to your point.

There's been all this sort of investigative reporting and all these lawsuits recently with inside operators, so doctors and nurses that work for the insurers saying, we just get a list and here's what the AI is saying and we just push buttons, deny, yes, no, yes, no, more information. It's really quickly based on this list of multiple patients that are coming up. They're not going back looking, oh, here's the AI's recommendation, let's do a thorough file review.

And you can imagine why that happens, even though the law says there must be an individual patient assessment by a qualified physician or nurse, you can imagine why that happens because that's a time intensive, resource intensive process that costs a lot of money from fairly expensive inside operators, trained medical professionals. And so the incentive with this kind of stuff is always to say the AI probably knows best or has a high rate of success here, and I have to get as many of these done during the day as quickly as I can because of this quickness issue, and to just push the button. And it's been unfortunate that I've always been skeptical that the insurers are probably doing this, but like I said, several whistleblowers have come forward over the last year or so and said, yeah, we just hit click on 200 claims one way or the other and move on.

It takes me 30 seconds, like direct quotes from people who do this.

Jeffrey Snyder, Broadcast Retirement Network

Is there an audit process? Again, I come from the retirement world where we have to do an annual plan audit. Not we, but the fiduciary brings in a CPA or some type of audit firm to do a deep dive.

They test transactions and they determine whether or not they were processed in accordance with the plan rules. Is that the case with some of these, like I can understand efficiency and AI is constantly learning and evolving, but is the work being audited by an independent entity in order to make sure that it's doing what it should be doing and doing it correctly?

Jennifer Oliva, Indiana University Bloomington

Yeah, and that's sort of what I'm arguing for here is either a front end look at this, testing it against claims data or some kind of back end audit or both. The answer is state insurance commissioners for the insurance plans that they manage have the authority to do this kind of thing, and some do. And that the federal government and the federal public health care programs like Medicare, Medicare Advantage and this have the authority to do this.

CMS, the Centers for Medicaid and Medicare Services at the federal level was concerned about this because, as I told you, all these sort of whistleblowers are coming forward. And a few years ago, they did a claim. They just took one week, one week out of all the Medicare Advantage claims data and said, we're just going to look at one week.

And they found fairly large errors, you know, 13 percent in one area, 18 percent, and said these decisions violate the federal rules on how these claims decisions have to be. One week, seven days of data as millions and millions of claims found these sort of very high error rates. And so when audits have been conducted, that's, again, raised concerns for us because we're like much more oversight is needed for this kind of stuff because the data looks bad.

Another way to do not an audit, but something that in law we would view as important is to say, if you're going through the appeals process, how often are these things overruled later on when another decision maker is looking at. So like in the Medicare space, this is egregious and really troubling. I hate to say it, but when they look at the administrative law judges, your last sort of appeal out of that five step Medicare appeal.

How often does the administrative law judge sitting there taking time looking at the claim reverse the denial when it gets to that stage? And it was 90 percent the last study that was done. So imagine being wrong 90 percent of the time and making like an efficiency argument.

That's very troubling because those folks in that plan are senior citizens who often have multiple health care issues. We're not talking about a 20 year old that didn't get some, you know, simple health care intervention and they're going to be fine. These are people that where the health care intervention and quickly the health care intervention was material to their health outcomes.

Jeffrey Snyder, Broadcast Retirement Network

Yeah. And I mean, you know, baseball players are wrong are off two thirds of the time and that's considered right. They bat 300.

That's considered a great batting average. Getting it wrong 90 percent of the time. I'm like, actually, the weather people might might be in that category, too.

But in all seriousness, let's talk a little bit about because, you know, I think this is a topic that is we're not going to solve in one day, but I think it's we're raising awareness here. What what is the role? How can the federal regulators and the state regulators?

You mentioned California and mentioned the federal government. What do they need to do? Sounds like they're a little behind the curve when it comes to the use of these technologies.

What do they need to do to make this better, not only for the businesses, because we want the businesses to be profitable, but you also want people to be able to get the care that they need so they can live longer.

Jennifer Oliva, Indiana University Bloomington

That's right. So I make a very simple argument and it's actually unfortunate that this isn't what's happening. I'm asking the states to vet these algorithms before they're used to make life and death decisions.

It's that simple as a matter of consumer protection. So state insurance commissioners have some authority. State Medicaid agencies have some authority here.

And it's complicated because different insurers are regulated by different entities. The federal government, instead of having 50 states and four territories having different rules and making it hard on the insurers, which often operate across those jurisdictions. I'm like an alternative here is go to the agency with the inside medical expertise that regulates the clinical algorithms, the algorithms that tell you, look, let's look at their heart rate.

Let's look at the like I said, the clinical notes, the FDA, the Food and Drug Administration, and ask them to take a look at these algorithms before they're allowed to be used by insurers to make sure that we have an error rate that we can accept. And we're transparent about that to the public. The unfortunate thing here with that proposal is that the insurers argue that the Food and Drug Administration doesn't have authority over them because they're not actually making a medical decision.

When what they mean by that is we're deciding whether or not we're going to pay for what you're asking for. But we're not deciding exactly what care you can proceed with because you can always go forward and pay out of pocket yourself. And my pushback against that is health care is extravagantly expensive in the United States.

And a ton of people on these public health programs and even privately insured individuals, ordinary everyday Americans just going to work on their employer-sponsored clinic, couldn't pay for a whole host of these things. Many people couldn't afford to pay for post-pregnancy care. It's so expensive.

And simple things like that. Imagine having to have a complex bypass medications and rehabilitation package. You're in hundreds and hundreds of thousands of dollars.

So it's just not a real world answer to say, well, you can just pay out of pocket and litigate this against us through the appeals process later. That does not make sense for a huge percentage of us in the United States.

Jeffrey Snyder, Broadcast Retirement Network

I would say the great majority of us.

Jennifer Oliva, Indiana University Bloomington

Almost everybody who's not extremely well-resourced and wealthy.

Jeffrey Snyder, Broadcast Retirement Network

Real quick, I had to wrap up our talk, but we talked about what the government's role is. Obviously, we know the insurer's role. I think there's an ethical role there to make things better.

My own personal opinion, you may disagree. But in terms of the patient and the doctor, so the treating doctor, not the doctor that works for the insurance company. Are there roles for both of those parties in order to, I don't know, not regulate, but ensure that they're getting the result that is intended?

But also a positive result, meaning the AI is actually doing what it should be.

Jennifer Oliva, Indiana University Bloomington

Yeah, I think that patients have to be their own advocates here and press their providers. Providers have seen these statistics now and these audits and reports and must be pressing back against the insurers. One of the interesting caveats here, why this is such a difficult area to write research in, is that the providers now are trying to use their own counteracting AI to try to get these claims through better and advocate for their patients.

So there's a much faster way to submit. But I would argue that patients have to be their own advocates here. And so the providers have to push back here because you will be successful a huge percentage of the time on appeal.

And we know that for a fact. So I would just say to patients, don't give up, press your providers, be your own best. And it's terrible to have to ask people to do that when they're unhealthy or they're not feeling unwell.

So, again, a family member that can be an advocate for you, but it's often worth it. The majority of the time, it's worth it to fight here to the extent that you have the ability, capability and support to do so.

Jeffrey Snyder, Broadcast Retirement Network

I lied, not last question, just real quick. To that point, as you're talking, I think of things that I would think are important. Let's talk about advocacy groups, because if you can't, if you're sick, are there not-for-profits or NGOs that can work on behalf of these individuals, these patients to help them?

I mean, there's got to be a huge support mechanism that exists today to help patients.

Jennifer Oliva, Indiana University Bloomington

Yeah, there are people who are claims advocates and patients advocates. Often your own insurance company will have somebody designated for that. And they're not taken advantage of nearly enough.

Your provider also can help you with that. But you need to call your insurer and find out if they have in-house patient advocates. They often do.

If that doesn't work out, you can obviously get an outside person. There are folks whose entire career is in patient advocacy. They'll even come and visit you if you're like an intensive care or an acute care facility.

So you have to take the time. But what I would tell people is when you get a new insurance plan or whatever insurance you're on, while you're healthy or feeling well right now, find out who all those contact informations are. People keep a little one-page Word document on your computer or a little card, whatever works for you.

Put it on your phone. That's the best. And that you have those numbers and can reach out when you get in a crisis situation and are not feeling as well as you did when you did that basic research.

Jeffrey Snyder, Broadcast Retirement Network

Yeah, really good advice. Jennifer, we're going to have to leave you there. So pervasive in our society is AI.

This is just another area where, you know, I think the regulation has to catch and the advocacy has to catch up. Thanks so much for joining us. We look forward to having you back on the program again very soon.

Jennifer Oliva, Indiana University Bloomington

Thank you very much and have a wonderful day.

Jeffrey Snyder, Broadcast Retirement Network

Don't forget to subscribe to our daily newsletter, The Morning Pulse, for all the news in one place. Details, of course, at our website. And your subscription helps support all this great BRN content each and every day.

We're back again tomorrow for another edition of BRN. Until then, I'm Jeff Snyder. Stay safe, keep on saving, and don't forget, roll with the changes.