BERKELEY, September 27 — Artificial intelligence tools used in medical billing and claims processing pushed patient costs up by nearly $1 billion over five years, according to a new analysis from researchers at the University of California, Berkeley. The study, released Tuesday and published in the journal Health Affairs, examined claims data from 2018 to 2023 across 15 major insurance companies and 200 hospital systems. It estimates the total excess cost to patients reached $980 million over that period, an average increase of $142 per affected patient.
The researchers describe that figure as conservative. They note the true cost may be higher because of unreported appeals and delayed care.
How the costs added up
The analysis points to two forces working against each other, and against patients. Hospitals used AI-powered revenue cycle management tools to optimize billing codes, a practice known as upcoding, where the software suggests more expensive codes. Insurers used their own automated systems to deny or delay claims.
The result, the report says, was a costly arms race. Denied claims force patients and providers to spend time and money on appeals, and those expenses land on the patient’s side of the ledger.
The study used machine learning to identify patterns in claims data, comparing hospitals with high AI adoption to those with low adoption. Hospitals in the high-adoption group had 12% more claim denials and 8% more patient billing complaints. Patients in rural areas and those with chronic conditions were disproportionately affected, according to the findings.
What the lead author says
Dr. Sarah Chen, an associate professor of health policy at UC Berkeley and the study’s lead author, said in a statement that “the current use of AI in billing is not designed to help patients; it’s designed to maximize revenue and minimize payouts.” The report notes that while AI can improve efficiency, its current deployment lacks transparency and oversight. It calls for federal regulators to establish standards for AI in healthcare billing, including requirements for explainability and audits.
The two main industry trade groups, the American Hospital Association and America’s Health Insurance Plans, declined to comment on the specific findings. Both have previously defended their use of AI as a tool to reduce administrative waste.
The findings arrive as the Centers for Medicare & Medicaid Services considers new rules on prior authorization, and as members of Congress have introduced bipartisan legislation to increase transparency in healthcare AI. This is the first comprehensive look at the financial impact of AI in healthcare billing, and it is likely to fuel ongoing debates about the role of technology in the U.S. healthcare system. The researchers plan to release a follow-up study later this year examining the impact on health outcomes.
The encouraging part is that the problems identified here are structural, not scientific. Transparency requirements and physician review of billing decisions are the kinds of fixes the report points toward, and the researchers themselves frame the $980 million as a floor rather than a ceiling.
Patients who receive an unexpected bill or a denied claim should ask for an itemized explanation and talk to their doctor about what was actually coded. Consult your doctor for medical advice.
