Administrative costs now account for more than 40% of hospitals' total expenses for delivering patient care, with a significant portion driven by the rising number of care denials stemming from the growing use of artificial intelligence tools by insurers.
Between 2022 and 2023, claims denials surged by an average of 20.2% for commercial plans and 55.7% for Medicare Advantage plans, according to a Sept. 10 brief from the American Hospital Association shared with Becker's.
"One factor driving this growth is the increased use of machine learning algorithms and other artificial intelligence tools," the AHA wrote. "Poor applications of these technologies can result in automatic denials of care without consideration of a patient's individual clinical circumstances or review from a clinician or plan medical director as required."
Although the 2024 Medicare Advantage final rule provided some guidance around payment denials and practices, the AHA emphasized that AI-driven denials remain a serious issue. In its guidance, CMS raised concerns about the potential for AI tools to perpetuate discrimination and bias.
"We are concerned that algorithms and many new artificial intelligence technologies can exacerbate discrimination and bias," the agency said Feb. 6. CMS advised that before implementing AI tools, insurers must ensure they do not reinforce existing biases or introduce new ones.
Scrutiny of AI use by insurers has intensified in recent years. UnitedHealthcare, Humana, and Cigna are facing lawsuits alleging they wrongfully denied care to Medicare Advantage members using AI algorithms. As software technologies evolve rapidly, CMS has sought to clarify the distinction between algorithms and AI. Algorithms are defined as a "decisional flow chart of a series of if-then statements," while AI refers to a "machine-based system" that can make predictions or decisions based on human-defined objectives.
Payers can use algorithms to support coverage decisions, but they are responsible for ensuring that these tools comply with CMS's coverage requirements. Medicare Advantage plans must base their decisions on individual members' medical histories, physician recommendations, or clinical notes, rather than relying solely on broader data sets. Predictive algorithms, for example, may help estimate a patient's length of stay but cannot be used to terminate coverage without reexamining the patient.
CMS also emphasized that Medicare Advantage organizations may deny coverage for basic benefits only for reasons such as network limitations or noncompliance with prior authorization rules. While algorithms can help ensure compliance with internal criteria, they cannot be used to change coverage criteria over time.