How algorithms are being used to deny health insurance claims in bulk

Read the Full Transcript

• John Yang:

  When health insurance companies deny claims, an appeal is not very likely. Federal data shows that companies issued more than 49 million denials in 2021, but customers appealed only about two tenths of 1 percent of them.

  While some denials come with specific justifications, most explanations are vague. Ali Rogin explores how at least one major health insurance company is using artificial intelligence to assess and often deny claims in bulk.

• Ali Rogin:

  Investigative journalists at the nonprofit ProPublica found that the health insurance company Cigna uses an automated system that allows it to instantly reject claims on medical grounds without even opening the customer's file. Some are accusing Cigna of using the system to help cut costs at the expense of the patients, a claim which Cigna denies.

  Joining me now is Maya Miller, one of the reporters who worked on the story. Maya, thank you so much for joining us. Can you explain to me how this system works?

• Maya Miller, ProPublica:

  Yes, absolutely. When you go to a doctor or see your healthcare provider, they'll send in a claim to your health insurance plan and they'll list the diagnosis that think that you have and then a list of tests or procedures that they want to run.

  We learned that there is a process that was developed at Cigna about a decade or so ago in which they created this code, this computer code, this algorithm that says, we're going to prove them if they match certain conditions. But if they don't match another condition, let's say a diagnosis that they don't think that is worthy of that treatment, then it will be sent to the desk of a medical director, which is a company doctor, and that doctor is going to be able to quickly sign off on rejecting that claim and saying, we're not going to cover it in two months.

  Last year, that happened in an average of under 2 seconds. These medical directors were essentially saying for 50 patients at a time, they wouldn't even open a patient file, but they were signing their name off and saying, actually, this claim is not medically necessary, and so we're not going to cover it.

• Ali Rogin:

  I want to read to you from a statement that Cigna has provided. First, they say that this automated process allows us to pay providers for claims quickly and automatically and allows medical directors time to look at more complex reviews. It also says that even a denial should not result in any additional out of pocket costs for patients who are using in network providers.

  What did you find related to those two statements that Cigna makes?

• Maya Miller:

  Yes, so on the first statement, it's true that some of the claims that are processed through the system, which is called Pxdx, if they match the diagnoses and the treatments on the list, if the code sees that there's a match there, they will be automatically approved.

  However, then there's a batch of them that get automatically denied. And that's really what we're focusing on. On the story, is the batch of claims which we found in two months last year, exceeded 300,000 claims. So it's not an insignificant amount.

  And to the second point about, you know, that this shouldn't result in any out of pocket costs for patients. Unfortunately, the reality is oftentimes when we go to a doctor, a nurse, or to see a healthcare provider, we often have to sign off on a form. And in that form we're saying that, you know, we'll take responsibility, financial responsibility, for this outcome, of this care. And it does often get shunted onto the patient.

• Ali Rogin:

  What kind of procedures are we talking about that are getting routinely denied?

• Maya Miller:

  Yes, so unfortunately, were trying to look for a comprehensive list of all of these claims that process through the system. We know that not all of the claims that Cigna has end up going through the system. But of those that do, we learn that vitamin D testing is labeled as part of the system.

  We learned that autonomic nervous system testing to test, like whether your nerves are working well, oftentimes if you have diabetes or other autoimmune disease, you can have some nerve damage in your fingers or your toes and so, testing to kind of figure that out, among other conditions. Unfortunately, we could not find a comprehensive list of all of the different procedures that are tagged for the system. Those were two that were able to conclude from our reporting.

• Ali Rogin:

  Cigna and all insurers, frankly, in claims denial issues will say, well, there's always the appeal process. What does that look like in reality?

• Maya Miller:

  Yes. One in seven claims are denied across the country, and health insurance claims. So it affects a lot of people. And yet one study found that 0.1 percent of all people who experience those denials actually took it to a step to do a formal appeal. When you're going about your day to day, maybe you're facing an illness or you're facing a chronic disease, it's hard to find the time and wherewithal to be able to figure out how to appeal these claims. And the reality is that very few people do.

  And, you know, we actually found through our reporting that Cigna had a presentation in which they were putting forward the idea to put autonomic nervous system testing through the system. And in the cost benefit analysis assessing whether to do this, they assumed only 5 percent of patients would end up appealing this, and that was part of the calculation of whether to put this test into this process or program or to leave it out.

• Ali Rogin:

  Now, Cigna is not the only insurer with some kind of automated system exactly like this. Do you have a sense of how prevalent automated systems are right now in the health system writ large?

• Maya Miller:

  It appears to be pretty prevalent. And, you know, insurance companies are dealing with millions and millions of claims. And so, in order to stay on top of those and to maintain efficiency, which is what they've said, they've, you know, turned to these algorithms and computer coding programs to be able to process claims as quickly and efficiently as possible is what they say.

  So, it seems that this kind of technology is pretty prevalent across the industry, although we're still reporting that out for other insurance plans and trying to confirm that. But, yes, it seems like it's pretty pervasive.

• Ali Rogin:

  Lastly, Maya, there was a robust debate happening about the hazards of using AI. Where does this conversation fit into the broader debate about the utility and the hazards of using artificial intelligence?

• Maya Miller:

  Yes, that's a great question and one that we're actually posing to regulators right now. The former Insurance Commissioner of California, Dave Jones, was telling us that, you know, in the statutes, you need to have a thorough, fair, and objective assessment of these patient claims. He was saying that if an insurer — if a medical director or a company doctor for an insurance plan is really looking at a claim in less than two seconds and an average of less than two seconds, can you achieve that result?

  So, yes, I'd pose that question really to the lawmakers, regulators and to see where they would stand on that.

• Ali Rogin:

  Maya Miller with ProPublica, thank you so much for your time.

• Maya Miller:

  Thanks for having me.