Modern Evil
Within a few hours of my father being discharged from the hospital, my mother got an email telling her that none of the costs were covered by his Medicare Advantage insurance. No person reviewed the file. It was an automated message from an AI bot that assessed the case and denied claims for payment. My mother, in her early 80s, was frantic.
from the UnitedHealthcare website, but the link goes to another page now
This, it seems, was a feature of UnitedHealthcare’s automated decision-making for its Medicare Advantage customers, based on statistics supporting that most people will not appeal a denial, and that almost all denials of health insurance coverage are upheld, as analyzed in a U.S. Senate Report on Medicare Advantage claim denials.
The Shock! The Horror!
Looking a storyteller in the face can be difficult sometimes. From Hannah Arendt’s Eichmann in Jerusalem, a book about the trial of the architect of the Nazi Holocaust:
As witness followed witness and horror was piled upon horror, they sat there and listened in public to stories they would hardly have been able to endure in private, when they would have had to face the storyteller. (emphasis added)
One result of Mr. Thompson’s untimely death is a wealth of stories on the suffering caused by UnitedHealthcare practices. Similar to Hannah Arendt’s observation of the observers of the trial of a Nazi war criminal, the leaders of UnitedHealthcare “have to face the storyteller” in the public square. With the sheer volume of painful stories told since Thompson’s killing in first-person videos, it may require a lot of endurance. The public witnessing of their customer experiences show the architects of the algorithm and AI-assisted healthcare company the harm they’ve caused.
How We Got Here
A May, 2020 article explains the purchase of NaviHealth by UnitedHealth Group, the owner of UnitedHealthcare. NaviHealth “provides post-acute care management services.” With the NaviHealth purchase, UnitedHealth Group got “technology tools” to support decisions on the “best next course of treatment for each patient.”
With the introduction of “Machine-Assisted Prior Authorization,” UnitedHealthcare, a processor of Medicare Advantage insurance claims, started using Artificial Intelligence to evaluate pre-authorization of medical treatment. They called it an “auto authorization model” and implemented it in 2021. A U.S. Senate Report on Medicare Advantage claim denials concluded: “the company [UnitedHealthcare] knew from testing that at least one of these automation technologies resulted in an increase in the share of those requests being denied.” (emphasis added)
Evidence previously obtained by the Subcommittee suggests that, before naviHealth became a subsidiary of UnitedHealth Group, its nH Predict algorithm was being used to influence outcomes for patients prior to any evaluation by their own post-acute providers. (emphasis added)
People noticed. In 2023, CIGNA, UnitedHealthcare and Humana were sued for using NaviHealth’s AI “to deny care to elderly Medicare Advantage” clients with “a known 90 percent error rate,” according to the Humana lawsuit. The insurers leveraged algorithmic tools to sharply increase claims denials for Medicare Advantage beneficiaries between 2019 and 2022. From FoxNews5:
The suit claimed the company illegally deployed "artificial intelligence (AI) in place of real medical professionals to wrongfully deny elderly patients care owed to them under Medicare Advantage Plans by overriding their treating physicians’ determinations as to medically necessary care based on an AI model."
Like my Dad.
KFF, a U.S. health policy group, reported “we find that consumers rarely appeal denied claims and when they do, insurers usually uphold their original decision.”
Success for UnitedHealthcare
It seems the automatic denial of an insurance claim by an algorithm or AI bot was a feature of Brian Thompson’s innovations at UnitedHealthcare: cut costs by denying pre-authorization for medical treatment, and then deny most appeals of that automated decision.
It worked. UnitedHealthcare has a claim denial rate of about one in three.
Under the section “Leveraging machine learning and artificial intelligence” UnitedHealthcare wrote in a white paper “Data capture and analysis lay the foundation for innovation.” From UnitedHealthcare’s white paper:
While AI and ML may be an important tool in health care moving forward, developing and implementing it responsibly — and ensuring the privacy and security of member health data is protected — matters. Understanding this, UnitedHealth Group continually assesses the use of emergent technology and advanced data and analytics and is focusing on the areas in which its application can help inform decisions and improve operational efficiencies. (emphasis added)
Modern Evil
Eichmann, while the creator of Hitler’s Final Solution, was not personally involved in the killing of his countrymen. He just set up the policies for others to follow while killing people. “He organized the identification, assembly, and transportation of Jews from all over occupied Europe to their final destinations at Auschwitz and other extermination camps in German-occupied Poland.” (from the Encyclopedia Britannica)
Brian Thompson didn’t kill anyone, we are assured. But, he did.
His goal as CEO of UnitedHealthcare was to lower costs of Medicare Advantage insurance claims. Towards that end, automated claims processing was put in place, and tweaked, to get that result. Lawsuits also accuse UnitedHealthcare of using AI to administer insurance claims for its customers. That meant denying more requests for medical treatment. And he knew it.
Implementing automated processes that “decided” whether to authorize treatment for a sick customer and applying it to a group of people that would mostly not appeal that decision had a predicted impact (a negative one) on how many patients’ bills are paid. And if appealed, then almost certainly that appeal would be denied. Et voila! Costs reduced.
Mr. Thompson may not have decided an individual’s fate, but he know only a certain percentage - a small one - would get the health care they needed by the end of the process he oversaw and set into motion.
Nothing exemplifies Hannah Arendt’s concept of modern evil better than that. Algorithm and AI decision-making resulting in denial of essential healthcare saves UnitedHealthcare money. Creators and overseers of these little bots can rest assured that they haven’t marked anyone for death. They have, but can’t see their faces. With this, they absolve themselves of responsibility.
It’s just going to be a little harder to do now.