The algorithm behind health insurance denials

Some of the biggest health insurers in the country are turning to an algorithm to help determine if a medical claim will be approved. That’s according to a recent investigation led by ProPublica into EviCore, a contractor used to outsource prior-approval requests for much of the insurance industry.

The investigation found that EviCore tweaks an algorithm to increase the likelihood those claims will be denied, which means lower costs for insurers but more patients losing access to potentially lifesaving care.

Marketplace’s Meghan McCarty Carino spoke to T. Christian Miller, a reporter at ProPublica who co-reported this story. The following is an edited transcript of their conversation.

T. Christian Miller: EviCore is a giant company. Almost anyone who’s ever gone to the doctor and been told, “Nope, the insurance company hasn’t approved this,” has probably interacted at some point in time with EviCore. EviCore covers about 100 million people in the United States, that’s about 1 of every 3 insured people in the country. An insurance company like Aetna or Cigna or Blue Cross, will send your case to EviCore, and EviCore will actually look over your case and decide whether or not they’re going to give approval to your doctor to get paid for the surgery or an MRI or whatever treatment that your doctor has recommended for you.

Meghan McCarty Carino: And as you found, EviCore uses an algorithm, which insiders you spoke to call “the dial,” to manipulate coverage decisions. How does that work?

Miller: Yeah, that’s a really interesting thing. Here is how that works: So, your insurance company will send your file to EviCore, and EviCore first runs that file through an algorithm, and that algorithm decides whether or not to approve the care. So, for example, you see your doctor, your doctor says, “Hey, I think you need an MRI.” That request will then go to the algorithm, the algorithm will say, “We think that this MRI request doesn’t have enough information,” so they kick it out to a team of doctors and nurses who will review your request.

What EviCore knows and what the insurance company knows is that the more doctors and nurses who take a human review of that request, the higher the chances are that there is going to be a denial of that request, because people will bring a closer eye. They will look for nuances. They will look for things that the algorithm wouldn’t necessarily catch. And so that’s how they can adjust the number of requests that are going to get turned down or not, by manipulating how often the algorithm takes a first pass at these requests, which are called prior authorization requests.

McCarty Carino: And that potential for cost cutting, it seems to be an explicit pitch that EviCore is making to health insurance companies, right?

Miller: Yes. What our interviews indicated is that a salesperson will go to an insurance company, and they’ll say, “Hey, are you spending too much this year on cardiology procedures? Well, if you employ us to review your cardiology procedures that come in, we will reduce your costs by as much as 3 to 1.” Where that becomes especially interesting is what they call risk contracts. So, in a risk contract, EviCore actually makes money depending upon the amount of money that they save the insurance company. So, in other words, they share in the savings that they make for the insurance company. And a regulator we talked to was concerned that this kind of incentivizes EviCore in a way that might make them more likely to issue a denial than they otherwise would have, because they have a possible monetary gain in that decision.

McCarty Carino: You wrote about one patient in Ohio named Little John Cupp, who died from a heart attack after his insurance company, using EviCore, repeatedly denied coverage for a heart catheter that his physician recommended. What did this man’s experience say to you about how this system works?

Miller: Yeah, I think the case of Little John Cupp shows how the health care system in this country is really deficient in supplying people what they need to get better. In this case, Little John Cupp’s own doctor said, “You need to have this catheterization procedure to make sure that you’re doing well, that you’re healthy, that your heart is not in danger.” And instead, you have this kind of algorithmically-based analysis, which says, “No, Mr. Cupp’s doctor, we don’t think that you’re right. We think you need to do this other test first.” What happens with Little John Cupp is that he eventually does get a scan to look at his heart. It takes about three months for all of the back and forth to go on between his doctor and EviCore, and eventually he gets that heart scan. It shows he has major problems with the amount of blood flowing from his heart. The day after he gets that scan, he dies of a cardiac arrest. Was it related? We’re not certain, but we talked to cardiologists who said if Little John Cupp had had that left catheter exam earlier, it might have played a big role in staving off a possible heart attack or in getting him treatment that he needed earlier.

McCarty Carino: I suppose, sadly, in this country, few people would be surprised that health insurers are taking steps to cut their costs in this way. But what surprised you the most while you were doing this investigation?

Miller: What surprised me the most was that this is just business as usual. Nothing that EviCore did was out of the ordinary. It’s just kind of a routine practice in this country to conduct this kind of cost-cutting measure. Now, it’s important to note that there is a use in having an insurance company or a third party look at what a doctor is requesting, because there are doctors who commit fraud, there are doctors who recommend scans and things like that that aren’t necessary. So, there is a role for cost savings in this whole system. The question is, how do you do that? What’s the best way to do that? And right now, we have companies like EviCore, which is one of a number of companies that do these reviews, and sometimes the reviews end up being right, but sometimes they end up steering people down a route where their care is delayed or that care isn’t paid for, and the execution of that is what we really need to pay attention to.

McCarty Carino: We talk a lot on the show about algorithmic decision-making, or algorithmically assisted decision-making. This certainly isn’t the only instance that we’ve seen in health care. Earlier this year on the show, we talked about how coverage decisions for a lot of patients on Medicare Advantage plans are being dictated by an algorithm. What concerns come up about the role of this technology in the health care context?

Miller: Well, one thing you should know is that the insurance industry, especially in health care, is running full-speed ahead at artificial intelligence. The insurance industry sees AI as a way to greatly lower costs. And like all things involved with artificial intelligence, there’s some truth to that. It will definitely help in some areas in terms of lowering costs, but when you take the people out of the equation — people like your doctor and yourself — the more chance there is of your case not being given full attention. If your case does not fall into a normal algorithmic interpretation, the algorithm will say, “Yeah, based on most things we’ve seen, you don’t deserve this treatment,” when, if you’re talking to your doctor, there might be something special about your case that doesn’t necessarily get picked up in what the algorithm is looking at, and you really do need this procedure in a way that maybe 95% of other people don’t. Eventually, you can appeal it, of course, you can fight it, but all that does is delay the time when you’re actually going to be able to have a treatment. And in that delay, all sorts of bad things can happen.

Stories You Might Like

The pitfalls of letting an algorithm set the rent

The surprising factors that affect who gets to live where

Barbie, T-Swift and the Fed?

When a senior is ill, can an algorithm decide length of care?

Examining private equity’s growing role in the rental market

TurboTax maker Intuit faces mass arbitration