Visual artists fight back against AI companies
A lawsuit filed by three visual artists against artificial intelligence image-generators may serve as an early bellwether of how hard it will be for all kinds of creators to stop AI developers from profiting off their work. (Aug. 31)
For years, vital decisions about who got medical care coverage took place in back offices at health insurance companies. Now, some of those life-altering decisions are being made by artificial intelligence programs.
At least that’s the contention of the two families who sued UnitedHealth Group this week, saying the insurance giant used emerging technology to deny or shorten rehabilitation stays for two elderly men in the months before they died.
They say that UnitedHealth’s artificial intelligence, or AI, is making “rigid and unrealistic” determinations about what it takes for patients to recover from serious illnesses and denying them care in skilled nursing and rehab centers that should be covered under Medicare Advantage plans, according to a federal lawsuit filed in Minnesota by the estates of two elderly Wisconsin patients. The lawsuit, which seeks class-action status, says it is illegal to let AI override doctors’ recommendations for these men and patients like them. The families say assessments like that should be done by medical professionals.
The families note in the suit that they believe the insurance company is denying care to elderly patients who won’t fight back even though evidence shows the AI is doing a lackluster job of assessing people’s needs. The company used algorithms to determine coverage plans and override doctors’ recommendations despite the AI program’s astonishingly high error rate, they say.
More than 90% of patient claim denials were overturned through internal appeals or a federal administrative law judge, according to court documents. But in reality, few patients challenged the algorithms’ determinations. A tiny percentage of patients − .2% − choose to fight claim denials through the appeals process. The vast majority of people insured by UnitedHealth’s Medicare Advantage plans “will either pay out-of-pocket costs or forgo the remainder of their prescribed post-acute care,” the lawsuit says.
Attorneys representing the families suing the Minnesota-based insurance giant said the high rate of denials is part of the insurance company’s strategy.
“They’re placing their own profits over the people that they are contracted with and then legally bound to cover,” said Ryan Clarkson, a California attorney whose law firm has filed several cases against companies using AI. “It’s that simple. It’s just greed.”
UnitedHealth told USA TODAY in a statement naviHealth’s AI program, which is cited in the lawsuit, isn’t used to make coverage determinations.
“The tool is used as a guide to help us inform providers, families and other caregivers about what sort of assistance and care the patient may need both in the facility and after returning home,” the company said.
Coverage decisions are based on the Centers for Medicare & Medicaid Services’ criteria and the consumer’s insurance plan, the company said.
“This lawsuit has no merit, and we will defend ourselves vigorously,” the company said.
Lawsuits of this type are not new. They are part of a growing body of litigation.
In July, the Clarkson law firm filed a case against CIGNA Healthcare alleging the insurer employed AI to automate claims rejections. The firm also has pursued cases against ChatGPT maker OpenAI and Google.
Families pay for expensive care that the insurer denies
The plaintiffs in the suit this week are the relatives of two deceased Wisconsin residents, Gene B. Lokken and Dale Henry Tetzloff, who were both insured by UnitedHealth’s private Medicare plans.
In May 2022, Lokken, 91, fell at home and broke his leg and ankle, requiring a brief hospital stay followed by a month in a rehab facility while he healed. Lokken’s doctor then recommended physical therapy so he could regain strength and balance. The Wisconsin man spent less than three weeks in physical therapy before the insurer terminated his coverage and recommended he be discharged and sent to recover at home.
A physical therapist described Lokken’s condition as “paralyzed” and “weak,” however his family appeals for continued therapy coverage were rejected, according to the lawsuit.
His family opted to continue with treatment despite the denial. Without coverage, the family had to pay $12,000 to $14,000 per month for about a year of therapy at the facility. Lokken died at the facility in July 2023.
The other man’s family also raised concerns that necessary rehabilitation services had been denied by the AI algorithm.
Tetzloff was recovering from a stroke in October 2022, and his doctors recommended the 74-year-old be transferred from a hospital to a rehab facility for at least 100 days. The insurer initially sought to end his coverage after 20 days, but the family appealed. The insurer then extended Tetzloff’s stay another 20 days.
The man’s doctor had recommended additional physical and occupational therapy, but his coverage ended after 40 days. The family spent more than $70,000 on his care during the next 10 months. Tetzloff spent his final months in an assisted living facility, where he died on Oct. 11.
10 appeals to rehab a broken hip
The legal action comes after Medicare advocates began raising concerns about the routine use of AI technology to deny or reduce care for older adults on private Medicare plans.
In 2022, the Center for Medicare Advocacy examined multiple insurers’ use of artificial intelligence programs in rehabilitation and home health settings. The advocacy group’s report concluded that AI programs often made coverage decisions that were more restrictive than what Medicare would have allowed and the decisions lacked the level of nuance necessary to evaluate the unique circumstances of each case.
“We saw more care that would have been covered under traditional Medicare denied outright or prematurely terminated,” said David Lipschutz, associate director and senior policy attorney for the Center for Medicare Advocacy.
Lipschutz said some older adults who appeal rejections might win a reprieve only to be shut down again. He cited the example of a Connecticut woman who sought a three-month stay at a rehab center as she recuperated from a hip replacement surgery. She filed and won 10 appeals after an insurer repeatedly attempted to terminate her coverage and limit her stay.
Importance of having ‘human in the loop’
Legal experts who are not involved in these cases said artificial intelligence is becoming a fertile target for people and organizations seeking to rein in or shape the use of emerging technology.
Gary Marchant, faculty director at the Center for Law, Science and Innovation at Arizona State University’s Sandra Day O’Connor College of Law, said an important consideration for health insurers and others deploying AI programs is making sure that humans are part of the decision-making process.
While AI systems can be efficient and complete rudimentary tasks quickly, programs on their own can also make mistakes, Marchant said.
“Sometimes AI systems aren’t reasonable and they don’t have common sense,” said Marchant. “You have to have a human in the loop.”
In cases involving insurance companies using AI to guide claim decisions, Marchant said a key legal factor might be how much a company defers to an algorithm.
The UnitedHealth lawsuit states that the company limited workers’ “discretion to deviate” from the algorithm. Employees who deviated from the AI program’s projections faced discipline or termination, the lawsuit said.
Marchant said one factor to track in the UnitedHealth cases and similar lawsuits is how closely employees are required to follow an AI model.
“There, clearly, has to be an opportunity for the human decider to override the algorithm,” Marchant said. “That’s just a huge issue in AI and healthcare.”
He said it’s important to consider the consequences of how companies set up their AI systems. Companies should think about how much deference they give to an algorithm, knowing that AI can digest huge amounts of data and be”incredibly powerful” as well as “incredibly accurate,” he said, and leaders should also keep in mind that AI “sometimes can just be completely wrong.”
Ken Alltucker is on X, formerly Twitter, at @kalltucker, or can be emailed at firstname.lastname@example.org.