Democrats on a Senate subcommittee are accusing UnitedHealthcare of using artificial intelligence (AI) to deny claims from Medicare Advantage patients.
In a report released in October, the U.S. Senate Permanent Subcommittee on Investigations (PSI) said that UnitedHealthcare’s rate of denying claims for post-acute care jumped from 10.9% in 2020 to 22.7% in 2022. Denials for skilled nursing care increased especially, with the number of denied claims in 2022 being nine times higher than in 2019.
During this period, UnitedHealthcare introduced several automation efforts, including using AI to help process claims faster. The report also pointed out that United processed more home health service authorizations, which cost less, suggesting the company was pushing patients toward cheaper alternatives rather than post-acute care facilities.
UnitedHealthcare responded to the report by saying it misrepresents the Medicare Advantage program and its clinical practices, and that it follows rules set by the Centers for Medicare and Medicaid Services (CMS) for reviewing post-acute care.
The report highlighted a decision made by UnitedHealthcare in April 2021, where the company approved using a system named “Machine Assisted Prior Authorization” to speed up the review of medical requests.
This system helped doctors and nurses review claims 6-10 minutes faster, but it also led to more denials as the system caught mistakes that were missed before.
In December 2022, UnitedHealthcare tested how AI and machine learning could predict which denials would likely be appealed and overturned. Nearly a year later, in November 2023, UnitedHealthcare faced a class-action lawsuit.
The lawsuit accused UnitedHealthcare and its subsidiary NaviHealth of using AI to wrongly deny claims for elderly patients in nursing homes, claiming the AI model replaced doctors’ decisions on what care was necessary.
The lawsuit said this AI model has a 90% error rate but that UnitedHealthcare continues to use it because only a small number of patients (about 0.2%) appeal the denials. Most patients either pay out-of-pocket or give up on getting the care they need.
How it works:
The nH Predict algorithm, developed by NaviHealth, works by analyzing a patient’s medical records and predicting their future care needs. The algorithm considers factors such as the patient’s age, diagnosis, and functional status to estimate how long they will need post-acute care.
Data Collection: The algorithm collects a patient’s medical records, including their diagnosis, age, functional status, and other relevant information.
Data Analysis: The algorithm analyzes the patient’s data and compares it to a large database of similar patients.
Prediction: Based on the analysis, the algorithm predicts how long the patient will need post-acute care and sets a target discharge date.
Decision Making: The insurance company, such as UnitedHealthcare, uses the algorithm’s prediction to make decisions about approving or denying coverage for additional care.
