UnitedHealthcare is being accused of “malicious implementation” of an AI algorithm to deny medically necessary health care coverage to elderly patients. On Nov. 14, the estates of Gene Lokken and Dale Tetzloff filed a lawsuit against the health insurance company, claiming that it wrongfully cut off care for their loved ones by deferring to an allegedly error-prone AI algorithm.
UnitedHealthcare, in a statement, maintained that the lawsuit is without merit.
According to the lawsuit, Gene Lokken was 91 years old when he fell while at home, fracturing his leg and ankle. After a month with no physical activity in a skilled nursing care center, Lokken’s doctor determined that physical therapy would help Lokken. For just under three weeks, Lokken received physical therapy and showed signs of improvement. Then, Lokken discovered that his coverage was denied and a safe discharge plan was recommended instead. Although the patients appealed, that appeal was rejected, with UnitedHealthcare pointing to the fact that Lokken was self-feeding and did not need much help with hygiene and grooming. Consequently, Lokken paid out of pocket, roughly $12,000 to $14,000 per month for one year until he died.
Dale Tetzloff suffered a stroke in October 2022. His doctor determined that he needed at least 100 days of post-acute care. Tetzloff was at a skilled nursing facility for a total of 40 days before UnitedHealthcare determined that he was ready for discharge. Although Tetzloff’s doctor informed UnitedHealthcare that Tetzloff continued to need occupational and physical therapy to recover, the coverage was denied. Tetzloff paid more than $70,000 out of pocket over 10 months.
Plaintiffs allege that their claims for post-acute care were denied due to defendants’ “illegal deployment of artificial intelligence (AI) in place of real medical professionals to wrongfully deny elderly patients care owed to them under the Medicare Advantage Plans.” Post-acute care is care that is needed for patients who are recovering from serious illnesses and injuries. It includes time at skilled nursing facilities, therapy, inpatient rehabilitation facilities, and more.
The AI model in question is called “nH Predict.” It purports to predict how much care elderly patients will require based on the model. According to the complaint, the nH Predict AI Model compares the patient’s age, diagnosis, living situation and physical function to patients in a database of 6 million patients compiled over the years, and then predicts how long a patient should need to stay, what their medical needs will be, and gives a target discharge date.
However, according to the plaintiffs, the recommendations are strict and unrealistic as to what care a patient actually requires. The nH Predict model led to many post-acute care coverage denials, according to plaintiffs.
The AI allegedly overrode what the decedents’ treating doctors determined was medically necessary care. According to the plaintiffs, the AI model also has a 90% error rate — a fact they say the defendants are aware of. Still, plaintiffs allege that UnitedHealthcare limited their employees from deviating from the prediction by setting targets to keep stays at skilled nursing facilities within 1% of the projected days. In internal documents reviewed by statnews.com, former employees were told that they could be disciplined, including being fired, if they missed that target.
Plaintiffs maintain that defendants used the AI to deny claims because they counted on a very small number of policyholders appealing claims, with most either paying out-of-pocket or forgoing care. “Defendants bank on the patients’ impaired conditions, lack of knowledge, and lack of resources to appeal the erroneous AI-powered decisions,” plaintiffs state in the complaint.
The plaintiffs allege that post-acute care was denied due to a “fraudulent scheme” carried out by UnitedHealthcare that offers a “clear financial windfall in the form of policy premiums without having to pay for promised care.” They assert that defendants breached their fiduciary duties, including those of good faith and fair dealing, and seek remedy of the conduct by recovering damages and enjoining defendants.
UnitedHealthcare spokesperson Aaron Albright issued a statement via email: “The naviHealth Predict tool is not 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. Coverage decisions are based on CMS coverage criteria and the terms of the member’s plan. This lawsuit has no merit, and we will defend ourselves vigorously.”
Plaintiffs seek class-action status. The class sought to represent would include those who purchased Medicare Advantage Plan health insurance from defendants in the four years prior to the complaint’s filing through the present, with a multi-state subclass of those individuals who reside in 21 states. Plaintiffs estimate that there are thousands of members throughout the United States.
Plaintiffs have demanded a jury trial.