Summarize this content to 2000 words in 6 paragraphs in Arabic Unlock the Editor’s Digest for freeRoula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.When 91-year-old Gene Lokken fell in his home three years ago and fractured his leg and ankle, he was admitted to a local hospital in Wisconsin, US, and then — still weak and in constant pain — to a nursing home. He began to recover with physical therapy, but after a few weeks his insurer refused to reimburse any further care in the facility.Despite recommendations from his doctor and carers, and an appeal he launched, UnitedHealth would not pay continued costs in the facility for Lokken. That left his family to cover bills of more than $12,000 a month until his death one year later.His family, and seven others, are at the centre of one of several ongoing class action lawsuits and probes which claim health insurers have used artificial intelligence (AI) to override physicians’ judgments and deny coverage — despite high error rates in assessing patients’ needs by algorithm.The actions highlight that while AI is becoming increasingly widespread and has considerable potential benefits in healthcare — such as rapid and reliable clinical diagnosis and accelerated drug development — it also brings significant risks and downsides.Defendant UnitedHealth is the nation’s largest insurance company, with its UnitedHealthcare business providing health insurance plans for 53mn Americans. The criticisms also affect other large insurers including Humana, although the issue is particularly sensitive for UnitedHealth, following the killing of Brian Thompson, chief executive of its UnitedHealthcare division, last December in New York. That unleashed a fierce debate over unfair denials of coverage, even though the accused shooter and his family were not clients of the company.The insurer owns and sells to competitors a computer model called nH Predict, developed by its subsidiary NaviHealth. The litigants say this AI determined inaccurately the number of days for which post-care should be reimbursed, and limited employees’ discretion to override it. UnitedHealth and its subsidiary UnitedHealthcare deny the claims, and are scheduled to provide their response to the lawsuit at the end of March.Growing concerns over the uses of AI were sufficient for the administration of former US president Joe Biden to introduce voluntary operating agreements with insurers, payers and providers in 2023. This was reinforced in 2024 by an executive order on standards, guidance and guardrails.The Centers for Medicare & Medicaid Services (CMS), which oversees the US federal Medicare health insurance system, in late 2023 introduced guidance requiring Medicare Advantage plans to make determinations “based on the circumstances of the specific individual . . . as opposed to using an algorithm or software that doesn’t account for an individual’s circumstances”. Its assessments “must be reviewed by a physician or other appropriate healthcare professional”.Separately, a US Senate investigations committee last October highlighted rising rejection rates under the Medicare Advantage programme for “pre-authorisation” reimbursement of “post-acute care”, such as Lokken’s rehabilitation treatments. “Many of the issues that most frustrate patients and providers remain cloaked in uncertainty. This is particularly true of insurers’ use of automation and predictive technologies,” it concluded.UnitedHealth has fought the Lokken-led lawsuit and rejected its claims, which it said in a statement are “based on unfounded allegations and mischaracterises the incredibly valuable work of our experienced and compassionate clinicians. Coverage decisions are only made by medical directors — not by AI — in accordance with CMS Medicare coverage criteria.”It says it pays 98 per cent of all valid claims, while just 0.5 per cent are rejected on the basis of clinical evidence and patient safety. It calls the Senate report “completely partisan” and says it mischaracterised “the Medicare Advantage program and our clinical practices, while ignoring CMS criteria demanding greater scrutiny around post-acute care”.Others have also raised broader concerns about AI’s deployment in healthcare reimbursement, however. As authors of the Journal of the American Medical Association Health Forum argued: “Use of machine learning methods exacerbates some problems with coverage denials, especially the opacity of decision-making. Even the developer may not know why its AI algorithm made a particular recommendation.”A report from the University of Chicago’s NORC research centre for consumer representatives to the National Association of Insurance Commissioners, a regulatory group, said: “AI proponents cite reductions in administrative burden and expedited approvals. However, there are risks that must be considered, such as the exacerbation of existing biases, prioritisation of misaligned incentives, and use of technologies outside their original use case, or design, leading to unintended harm.”The NAIC issued a Model Bulletin on the Use of Artificial Intelligence by Insurance Companies in December 2023 reminding insurers that decisions or actions made or supported by AI “must comply with all applicable insurance laws and regulations”.It is now surveying health insurance carriers on their use of AI and machine learning to better understand and appropriately monitor health insurers’ implementation of this technology.A NORC submission to the regulatory group in November noted that a number of US states have begun to regulate AI in health insurance but added that most “have not been able to keep up with the proliferation of AI’s use”, and called for both transparency and embedding human insight into the use of the technology.The political battle is far from over. In January, the new Trump administration revoked Biden’s AI order, replacing it with a call for a new action plan by mid-year. As an analysis by the law firm Maynard Nexsen put it: “The AI landscape continues to develop, and the regulations appear to be loosening — at least at the federal level. These changes have led to uncertainty among organisations using AI technology.”
