BY Blinn E. Combs, Esq and Michael Alexander, Esq., Brown & Fortunato, P.C.
On November 14, two plaintiffs lodged a class action suit against UnitedHealth Group, Inc. (“United”) in the U.S. District Court in Minnesota, where United is headquartered. The complaint alleges that United used artificial intelligence (“AI”) to illegally deny coverage to potentially thousands of United customers who had purchased Medicare Advantage plans from the insurer. The suit is a fresh reminder of AI’s risks.
The complaint follows a string of investigative reports from STAT, a Boston-based news organization covering U.S. healthcare. The reports detail United’s turn to predictive AI technology as a tool to reduce post-acute care costs. STAT’s initial March investigation has already made waves, resulting in a May 17 Senate Homeland Security and Governmental Affairs Committee hearing where Senators warned Medicare Advantage insurers to abide by Medicare coverage rules.
In 2020, United, through its Optum division, purchased naviHealth, a technology vendor specializing in post-acute care management, for a reported $2.5 billion. NaviHealth developed the nH Predict AI Model (“nH Predict”). nH Predict promised savings, in part by reducing medically unnecessary stays in post-acute care facilities. nH Predict relied on a compiled database of 6 million patients, whose treatment data was used to predict how long patients would need post-acute care before returning home. Because of the prospective payment system used by Medicare Advantage plans, decisions about the length of stay are frequently made before a patient has moved to post-acute care.
According to STAT, naviHealth manages post-acute care for several large providers, including United, Humana, and several regional health plans. These insurers manage the Medicare Advantage plans of over 15 million beneficiaries—over a quarter of all Medicare recipients. In late October, United announced plans to discontinue the naviHealth brand and roll the company into a general rebrand of post-acute care companies under the new name “Home & Community Care,” beginning in the first quarter of 2024.
The suit alleges that United used nH Predict as a tool of unjust enrichment, in effect forcing its staff, on threat of termination, to follow the AI’s recommended length of stay within 1%, resulting in increased denial of medical claims for continued care. According to patients’ families and former United staff, these denials frequently flew in the face of physician recommendations and obvious continued impairment. On its website, NaviHealth shows reports indicating the required length of stay based on health data. STAT reports that United has repeatedly denied doctor and patient requests to review individual nH Predict records on the grounds that the information is proprietary.
Although coverage denials are not new, physicians and medical administrators have raised concerns about increasing denial rates by Medicare Advantage plans for care routinely covered by traditional Medicare. Doctors and administrators allege that these denials amount to the subversion of clear Medicare treatment guidelines mandating medical coverage for Medicare Advantage beneficiaries. Medicare Advantage plans now account for more than half of the medical coverage of U.S. seniors.
The suit alleges that by using nH Predict to deny claims, United breached insurance contracts with affected beneficiaries and violated its duty of good faith and fair dealing. The suit also alleges multiple violations of state insurance bad faith laws and the breach of additional common law duties. Plaintiffs have requested damages and injunctive relief. United has defended its claims review process and denied all allegations. To date, no general review of nH Predict’s performance has been made public.
Along with the increase in denials, appeals have risen. According to the Appeals Decision Search, an online database run by the Centers for Medicare and Medicaid Services (“CMS”), appeals increased 58% between 2020 and 2022. CMS tracks only a portion of coverage appeals. The complaint against United alleges that a review of administrative appeals conducted since the institution of nH Predict shows an error rate of 90%, the rate at which appeals result in reversing United’s coverage denials.
Appeals, however, frequently require time and resources not available to medically fragile seniors, who often experience these post-acute care denials at or near the end of their lives. According to a February report from KFF.org, among all 48.3 million denied health insurance claims in 2021 –17% of all claims—only 0.2% were appealed.
While less expensive than hospitalization, out-of-pocket costs of post-acute care can be financially devastating to families. Medicare Advantage benefits vary according to plan, but out-of-pocket costs without coverage can run to hundreds of dollars a day. Medically compromised patients are sometimes forced to spend down their life savings until they are eligible for Medicaid to cover the costs of care.
AI poses several unique ethical and policy challenges, especially in contexts like health care, where policymakers have sought to insulate independent professional judgment from the improper influence of outside corporate and financial actors. On October 30, President Biden issued an Executive Order seeking to establish new standards for the responsible development and utilization of AI-driven technology.
It is too early to predict how the suit will play out but it merits consideration. Healthcare providers and insurers alike need to remain cautious about the often exaggerated promise of AI. Remember that AI is a tool. It should help organizations fulfill their obligations in ways that providers can clearly explain and document. As a tool, it should never be used as a final arbiter of treatment decisions.
Regardless of the functions, AI is used to perform, organizations have a responsibility to both patients and regulators to demonstrate that the tool functions as intended in a way that complies with basic legal and contractual requirements. Overriding the medical advice of physicians through opaque technology is a recipe for costly litigation. United, the U.S.’s largest private health insurer, operates in all fifty states and holds roughly 15% of the private health insurance market share, insuring over fifty million beneficiaries. If the suit succeeds, United’s desire to predict costs could cost the company dearly.