UnitedHealth Accused of Using Inaccurate AI Model with 90% Error Rate to Deny Care: Lawsuit

UnitedHealthcare, the largest health insurance provider in the US, has been accused of using a faulty AI algorithm to deny essential health coverage to elderly patients.

This algorithm, developed by NaviHealth, a UnitedHealth subsidiary, allegedly leads to patients being prematurely discharged from rehabilitation programs and care facilities, forcing them to exhaust their life savings to access necessary care that should be covered under their government-funded Medicare Advantage Plan.

UnitedHealth Accused of Using Inaccurate AI Model with 90% Error Rate to Deny Care, Lawsuit Says

A lawsuit filed this week in the US District Court for the District of Minnesota highlights the plight of two deceased individuals who were denied health coverage by UnitedHealth.

The lawsuit seeks class-action status for potentially tens of thousands of similarly situated individuals. The allegations are supported by an investigation by Stat News, which obtained internal documents and communications and interviewed former NaviHealth employees.

According to the investigation, the nH Predict AI algorithm prioritizes profits over patient care. Former NaviHealth case manager Amber Lynch, an occupational therapist, stated, “By the end of my time at NaviHealth, I realized that I was not an advocate for patients, but rather a moneymaker for the company. It’s all about money and data points. It takes the dignity away from the patient, and I hated that.”

The lawsuit and investigation shed light on the harmful consequences of UnitedHealth’s alleged use of a flawed AI algorithm to deny critical health coverage to elderly patients.

The case highlights the need for health insurance providers to prioritize patient care over profits and ensure that AI algorithms are used ethically and with transparency.

Inaccuracy In United Health’s AI

UnitedHealth began using the nH Predict algorithm in November 2019, and it is still in use today.

The algorithm is designed to estimate the amount of post-acute care a Medicare Advantage patient will need after an acute injury, illness, or event.

This can include therapy, skilled care from home health agencies, skilled nursing homes, and inpatient rehabilitation centers.

The exact workings of nH Predict are unclear, but it is believed to use a database of medical cases from 6 million patients to make estimates.

NaviHealth case managers input information about a given patient, such as age, living situation, and physical functions, and the AI algorithm generates estimates based on similar patients in the database.

The algorithm provides estimates for medical needs, length of stay, and discharge date.

However, the algorithm has been criticized for not taking into account important factors that can affect a patient’s health and recovery time, such as comorbidities and complications that may arise during their stay, like pneumonia or COVID-19.

As a result, the estimates generated by nH Predict are often overly conservative, according to the Stat investigation and the lawsuit.

For example, patients on Medicare Advantage plans are typically entitled to up to 100 days of covered care in a nursing home after a three-day hospital stay. However, with nH Predict, patients rarely receive more than 14 days of coverage.

When patients or their doctors have requested to see the reports generated by nH Predict, UnitedHealth has denied their requests, citing proprietary information.

Additionally, when prescribing physicians disagree with UnitedHealth’s determination of the amount of post-acute care their patients need, their judgments are overruled.

UnitedHealth Sees this as a Feature Instead of a Bug

The healthcare industry’s history with flawed AI isn’t new, particularly in the US, where issues like algorithmic bias in patient care have long persisted.

While current concerns revolve around attention-grabbing AI chatbots and image generators, UnitedHealth’s acquisition of NaviHealth in 2020 brought a distinct shift in priorities.

Former employees revealed a transition from patient advocacy to a focus on performance metrics, notably emphasizing swift post-acute care.

UnitedHealth’s executives mirrored this change, displayed by Patrick Conway’s statement in a company podcast: “How do we expedite discharges from nursing homes?” The contentious nature of nH Predict’s estimates, seemingly intentional rather than accidental, sets the stage for the lawsuit against UnitedHealth.

The lawsuit emphasizes the glaring inaccuracies in nH Predict’s estimates, evident from the remarkably high reversal rate of coverage denials when challenged—over 90 percent. This discrepancy highlights the algorithm’s wrongful denial of coverage, underscoring its flawed nature.

Despite this, rather than rectifying the situation, NaviHealth’s directives increasingly bind employees to adhere strictly to the algorithm’s predictions.

Over the past two years, case managers faced mounting pressure to align patient stays in nursing homes within a mere percentage of the algorithm’s projections.

What initially began as a 3 percent margin in 2022 tightened to a strict 1 percent target by 2023.

These mandates aren’t mere suggestions; they’re mandatory. Case managers failing to meet these stay-length targets risk disciplinary action or termination. Lynch’s termination, cited in the case, stemmed from missing the length-of-stay target and falling behind on required documentation for daily caseloads.

Statement from UnitedHealth on this Issue:

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.

Ultimately, NaviHealth’s physician medical reviewers, rather than case managers, make decisions on coverage or denials. However, case managers play an advisory role and are held to a target of 1 percent.

Case managers are trained to defend the algorithm’s estimates to patients and their care providers. A training document obtained by Stat revealed that case managers are instructed to use blunt tactics when patients and caregivers dispute denials. The document stated:

If a nursing home balked at discharging a patient with a feeding tube, case managers should point out that the tube needed to provide “26 percent of daily calorie requirements” to be considered as a skilled service under Medicare coverage rules.

If a nurse took a broader tack, and argued a patient was unsafe to leave, case managers were instructed to counter, in part, that the algorithm’s projections about a patient’s care needs, and readiness for discharge, are based on a “severity-adjusted” comparison to similar patients around the country. “Why would this patient be any different?” the document asks.

Manager: What’s the Point of Winning?

Even when patients successfully appeal AI-backed denials, their victory is short-lived. UnitedHealth will often send new denials within days of the initial appeal, according to a former case manager who spoke to Stat.

The case manager claimed that their supervisor instructed them to immediately restart the case review process for any patient who had won an appeal, with the intention of issuing another denial.

As one former case manager put it, “You won, but OK, what’d that get you? Three or four days? You’re going to get another [denial] on your next review, because they want you out.”

The proposed class-action lawsuit includes the family of Gene Lokken, who passed away on July 17, 2022. In May 2022, the 91-year-old fell at home and was hospitalized for six days before being transferred to hospice care.

Doctors believed he was well enough to begin physical therapy, but UnitedHealth only covered 19 days of therapy, despite his doctors and therapists describing his muscle functions as “paralyzed and weak.”

The family appealed the denial, but their appeal was rejected. The rejection letter from UnitedHealth stated that additional physical therapy was unnecessary because there were no acute medical issues, and he was able to feed himself and perform basic hygiene tasks with minimal assistance.

The family was forced to pay out of pocket for his therapy, spending around $150,000 until his death.

The other plaintiff in the lawsuit is the family of Dale Tetzloff, who suffered a stroke on October 4, 2022, and was admitted to a hospital.

After being deemed in need of at least 100 days of post-acute care by his doctors, Tetzloff was transferred to a skilled nursing home. However, after just 20 days of coverage, UnitedHealth denied further payment for his care.

Despite two unsuccessful appeals, Tetzloff’s family continued to fight for coverage but were ultimately unsuccessful.

During the 10 months he spent at the skilled nursing home, they paid out of pocket a total of $70,000. In June 2023, Tetzloff was moved to an assisted living facility, where he passed away on October 11.

The lawsuit alleges that UnitedHealth and NaviHealth breached their contract, acted in bad faith, and unjustly enriched themselves at the expense of Tetzloff and his family.

The plaintiffs are seeking actual damages, damages for emotional distress, disgorgement, and/or restitution, as well as an end to the use of AI-based claims denials.

It is estimated that UnitedHealth saves hundreds of millions of dollars annually by using nH Predict, but the exact amount is unknown.

In 2022, UnitedHealth Group’s CEO was paid $20.9 million in total compensation, while four other top executives received between about $10 and $16 million each.

Related Stories:

Help Someone By Sharing This Article